Magmatic Systems

This is Volume 57 in the

INTERNATIONAL GEOPHYSICS SERIES
A series of monographs and textbooks
Edited by RENATA DMOWSKA AND JAMES R. HOLTON
A complete list of the books in this series appears at the end of this volume.
 The eruption of the Krafla central volcano, northeast Iceland during September4-18, 1984. (Top) 06:00 hours on September
5t1;. Silhouetted against dawn's light, the coalescing plumes from seven en-echelon-arranged eruptive fissures rise over the
Gjastykki fissure system. Fountain heights during these first hours of eruption averaged 50-75 m and had flow rates that tended
to be maximized over an individual fissure's center. Prior to the eruption, magma was stored within Krafla's reservoir located
beneath the caldera floor (right margin of photograph). (Bottoms T I:30 hours on September II tho Within 72 hours the eruption
had localized at this northernmost vent. shown here in the process of building a large spatter cone. In this eastward view. the
fountain heights ranged from 30-40 m above the cone top. Collectively. magma flow at Krafla illustrates a number of dynamic
patterns and processes: lateral irjection at depth along the horizon of neutral buoyancy. laminar flow during crack migration,
turbulent flow above surface fissures. and thermally constricted flow beneath volcanic vents. U.S. Geological Survey photographs
by Michael P. Ryan.
Magmatic Systems

Edited by

Michael P. Ryan
u.s. GEOLOGICAL SCRVEY
RESTON. VIRGINIA

ACADEMIC PRESS
San Diego New York Boston
London Sydney Tokyo Toronto
 Cover photo: Sakurajima volcano, Kyushu, Japan, in eruption at Minami-Dake
(the southern crater) during the early morning hours (03:09) of February 17, 1988. This
remarkable photograph was made possible through the use of an automated system that
continuously records atmospheric pressure. The expanding atmospheric shock wave generated
during the eruption outbreak triggered the camera at the Sakurajima Volcanological Observatory,
5.5 km from the summit. The integrated intensity of the resulting Strombolian bombardment
has been recorded by the time-lapse photographic process. Incandescent volcanic bombs
are seen to follow both steep and low-angle trajectories, blanketing the upper flanks and
rupturing on impact. Bomb rupture produces secondary showers of fragments down paths revealed
by their bright comet-shaped trails on the volcano flanks. High in the ash plume, lightning
strokes discharge buildups of static electricity and usually emanate from, or terminate in,
the higher conductivity incandescent plume core.
Sakurajima is a tightly coalesced pair of pyroxene andesite and dacite stratovolcanoes
that have grown within the Aira caldera during the last 13,000 years. The formation of the
Aira caldera itself was associated with the great eruption of the Ito pyroclastic flow about
22,000 years ago, which covered much of southern Kyushu with some 150 km' of rhyolitic ejecta.
The base of Sakurajima today rests about 125 km above the Wadati-Benioff zone, and the dynamics,
mechanics, thermal structure, and deep hydrology of the subduction zone influences the generation,
the segregation, and the ascent of Sakurajima magma. Photograph by T. Takayama, and courtesy
of the Sakurajima Volcanological Observatory of Kyoto University, Professor Kosuke Kamo,
and Dr. Kazuhiro Ishihara.

This book is printed on acid-free paper. 8

Copyright © 1994 by ACADEMIC PRESS, INC.
All Rights Reserved.
No part of this publication may be reproduced or transmitted in any form or by any
means, electronic or mechanical, including photocopy, recording, or any information
storage and retrieval system, without permission in writing from the publisher.

Academic Press, Inc.

A Division of Harcourt Brace & Company
525 B Street, Suite 1900, San Diego. California 92101-4495

United Kingdom Edition published by

Academic Press Limited

24-28 Oval Road, London NWI 7DX

Library of Congress Cataloging-in-Publication Data

Magmatic systems / edited by Michael P. Ryan.

p. ern. -- (International geophysics series : v.57)
Includes index.
ISBN 0-12-605070-8
I. Magmatism. I. Ryan, Michael P., [DATE]. II. Series.
QE46l.M237 124 1994
552' .1--dc20 93-48074
CIP

PRINTED IN THE UNITED STATES OF AMERICA

94 95 96 97 98 99 QW 9 8 7 6 5 4 3 2
To

Herbert and Virginia Ryan

OF EATON RAPIDS, MICHIGAN

and

Tsuneo and Ayako Eguchi

OF KAMAKURA-SHI, KANAGAWA-KEN, JAPAN
 Contents

Contributors xi
Acknowledgments xiii
Prologue xv

Chapter 1 Mechanics of Flood Basalt Discussion .............................. 28

Creep and Melt Migration .............. 28
Magmatism Attenuation and Melt Migration ......... 32
Peter Olson Application of the Experimental Observations
to the Earth's Mantle ................ 34
Overview I
Summary and Conclusions ................. 35
Introduction ............................. 2
References .............................. 35
Continental Flood Basalts. Oceanic Plateaus, and
Mantle Plumes ......................... 2
Constraints from Seismic Tomography and
Chapter 3 Influence of Basaltic Melt on
Geochemistry .......................... 3
Time Variations in Hotspot Activity .......... 4 the Creep of Polycrystalline Olivine under
Hotspot Activity and Geomagnetic Reversal Hydrous Conditions
Frequency ............................. 4 David L. Kohlstedt and Prame N. Chopra
Partial Melting in Thermal Plumes and Diapirs 7
The Numerical Model ..................... 8 Overview . 37
Melting in Solitary Diapirs ................. 10 Introduction . 38
Melting in Continuous Plumes ............... 12 Experimental Procedures . 38
Conclusions ............................. 17 Deformation Results . 41
References .............................. 17 Microstructural Observations . 42
Hot-Pressed Material . 42
Deformed Material . 43
Chapter 2 Melt Migration and Related Infrared Analyses . 49
Discussion . 49
Attenuation in Equilibrated Partial Melts Microstructural Characteristics . 49
Tye T. Gribb, Shanyong Zhang. and Reid F. Cooper Rheological Behavior . 50
Overview ............................... 19 Conclusions . 52
Introduction and Theoretical Foundation 20 References 52
Experimental Approach .................... 22
Experimental Specimens: Co-Mg Olivine-
"Basalt" ......................... 23
Chapter 4 The Generation and Migration
Creep Experiments and Specimen Analysis 24 of Partial Melt beneath Oceanic Spreading
Attenuation Experiments .............. 24 Centers
Experimental Results ...................... 25 David W. Sparks and E. M. Parmentier
Flexural Creep Experiments ............ 25
Microstructural Analysis ............... 28 Overview 55
Attenuation Measurements ............. 28 Introduction 56

vii
Vlll Contents

Melt Migration by Porous Flow ............. 57 Appendix 135

The Forces That Drive Porous Flow 57 References 135
Decompacting Boundary Layers 58
Effect of Buoyant Mantle Upwelling on Melt
Generation and Migration ............... 61
Chapter 7 An Observational and
Three-Dimensional Buoyant Flow beneath an Theoretical Synthesis of Magma Chamber
Offset Spreading Center ............ 61 Geometry and Crustal Genesis along a Mid-
Melt Migration beneath Spreading Centers 65
ocean Ridge Spreading Center
Implications for Crustal Thickness and
Gravity .. .. .. 68
1. Phipps Morgan, A. Harding, 1. Orcutt, G. Kent,
and Y. 1. Chen
Discussion ............................. 74
References ............................. 75 Overview .............................. 139
Introduction ............................ 140
The Seismic Record: Seismic Constraints on
Chapter 5 Dike Patterns in Diapirs Magma Emplacement and Crustal Accretion at a
Fast-Spreading Ridge Axis .............. 142
beneath Oceanic Ridges: The Oman
Northern East Pacific Rise ............ 143
Ophiolite Correlation between Seismic and Ophiolite
A. Nicolas, F. Boudier, and B. Ildefonse Observations of Crustal Structure 143
Current Seismic Picture of an Oceanic Magma
Overview .............................. 77
Chamber at a Fast-Spreading Ridge 145
Introduction ............................ 77
Other Seismic Constraints on Crustal
Regions of Diapirism and Sub-ridge Flow 78
Accretion at a Fast-Spreading Ridge 151
Dike Typology in Relation to Diapiric Areas 82
Delivery of Melt from the Mantle to Supply
Dike Patterns in the Southeastern Oman Diapirs 85
the Magma Lens . . . . . .. . .. . . .. . . .. 154
Dikes in the Plastic Flow Reference Frame 89
Spreading Rate Dependence of the Magma
Dikes in the Ridge Reference Frame 89
Lens: Seismic and Gravity Constraints 154
Discussion ............................. 91
Closure on Observational Constraints on the
Conclusions ............................ 94
Structure of an Axial Magma Chamber 158
References ............................. 94
A Theoretical Model for Crustal Genesis 158
Introduction ........................ 158
Magma Ascent and Emplacement 161
Chapter 6 Neutral-Buoyancy Controlled Hydrothermal Cooling ............... 162
Magma Transport and Storage in Mid-ocean Model Formulation .................. 164
Numerical Determination of Accumulated
Ridge Magma Reservoirs and Their Sheeted- Crustal Strain .................... 165
Dike Complex: A Summary of Basic Model Results . . . . . .. . . . . . . . . . . . . . . . 166
Relationships Neutral Buoyancy versus Magma Freezing as
Michael P. Ryan a Fundamental Limit to Magma Ascent 172
Geochemical Variability between Ridge
Overview .............................. 97 Segments ........................ 173
Introduction ............................ 99 Closure .. . .. .. .. .. .. .. .. .. . .. .. .. .. .. .. 173
III Siru Density-Depth Relationships ......... 99 Reference, ............................. 175
Definitions ............................. 100
The Magma Reservoir Environment and the Region
of Neutral Buoyancy ................... 102
Chapter 8 Deep Structure of Island Arc
Structure of the Ridge Magma Reservoir in Relation Magmatic Regions as Inferred from Seismic
to the Region of Neutral Buoyancy 105 Observations
Lithologic Associations in Ophiolite Complexes 110
Akira Hasegawa and Dapeng Zhao
Region of Negative Buoyancy 110
Region of Neutral Buoyancy ........... 111 Overview .............................. 179
Differentiation and Melt Density ............ 112 Introduction .. .. . .. .. .. .. .. .. .. .. .. .. . .. 179
Icelandic and Hawaiian Analogs ............ 118 Three-Dimensional Seismic Velocity Structure 180
Neutral-Buoyancy Control in Lateral Intrusion P-Wave Velocity Structure heneath the Northeastern
Dynamics: The Sheeted-Dike Complex 122 Japan Arc . . . . . . . .. . . . . . . . . . .. . . . . . . .. 182
Notes on Hydrothermal Interactions ......... 131 Three-Dimensional Seismic Attenuation Structure 185
Summary Relationships and Evolution of the Midcrustal Magma Bodies Detected by Reflected
Oceanic Crust ......................... 133 Seismic Waves ........................ 187
Comments at Closure: A World without Neutral Deep Structure of Arc Volcanoes ............ 191
Buoyancy ............................ 134 References ............................. 193
 Contents IX

Chapter 9 Lateral Water Transport across Application to Magma Transport .......... 234
Fracture Properties of Country Rock 234
a Dynamic Mantle Wedge: A Model for Melt Properties ..................... 236
Subduction Zone Magmatism Implications for Magma Ascent Velocities 236
1. Huw Davies Additional Processes ................. 238
Summary .............................. 238
Overview .............................. 197 References ............................. 239
Introduction ............................ 197
Thermal Model .......................... 198
Brief Review of Previous Thermal Models 198 Chapter 11 Accumulation of Magma in
Outline of Thermal Model ............ 199
Induced Flow in the Mantle Wedge 199 Space and Time by Crack Interaction
Frictional Heating at the Slab-Wedge Interface 200 Akira Takada
Ablation ........................... 200
Overview .............................. 241
Trench Migration . . . . .. . . . . . .. . .. .. . 201
Introduction ............................ 242
Theoretical Background and Numerical
Crack Interactions ....................... 243
Implementation . . .. . . . . .. . . . . . .. . . 201
Approach to Crack Interaction ......... 243
Results of Thermal Modeling .......... 203
Summary of Gelatin Experiments 244
Implications of the Results ............ 204
Interaction of Two Collinear Cracks 245
Adiabatic Upwelling ................. 204
Interaction of Two Parallel Offset Cracks 248
Phase Equilibria for Hydrous Peridotite 206
Conditions for Crack Coalescence 251
Basic Mechanism of Lateral Water Transport 207
Applications to Magma Accumulation in Space and
Discussion of the Lateral Transport Mechanism 209
Time 252
Transition from Silica-Rich Fluid to a
Magma Accumulation ................ 252
Hydrous Melt .................... 209
Variation of a Volcano ............... 253
Transport of Water-Dominated Fluid 210
Conclusions ............................ 255
Estimate of the Water Flux That Enters the
References ............................. 255
Mantle Wedge .................... 212
Importance of Hydrous Minerals Other Than
Amphiboles .. .. .. . .. .. .. .. .. .. .. . 213
How Much Amphibole Is Formed? 214
Chapter 12 Generalized Upper Mantle
How Much Free Water Is There? 215 Thermal Structure of the Western United
Chemistry of Amphiboles ............. 215 States and Its Relationship to Seismic
Model Predictions . .. . .. .. .. .. .. .. .. .. .. . 216 Attenuation, Heat Flow, Partial Melt, and
Presence and Location of the Volcanic Front 216
Cncertainties in Geometric Predictions 216 Magma Ascent and Emplacement
Major Element Composition of Primary Hiroki Sato and Michael P. Ryan
Magmas 217
Overview .............................. 259
Slab Trace Element Signature .......... 217
Introduction ............................ 260
Other Predictions ................... 217
Laboratory Seismic Data for Mantle Peridotite 261
Conclusions . . . . . . . . . . . . . . . . .. . . . . . . .. . . 218
Seismic Velocity of Mantle Peridotite 261
References .. .. .. .. .. . .. .. .. .. . .. .. .. . .. 219
Seismic Anelasticity of Mantle Peridotite 263
Generalized Thermal Structure beneath the Western
United States from Seismic Anelasticity Data 265
Chapter 10 Buoyancy-Driven Fracture Upper Mantle Anelasticity Structure 265
and Magma Transport through the Volcanic Setting .................... 266
Lithosphere: Models and Experiments Upper Mantle Temperature and Partial Melt
Fraction ......................... 268
Moritz Heimpel and Peter Olson
Comparison with Previous Temperature
Overview .............................. 223 Estimates ........................ 272
Introduction ............................ 224 Surface Heat Flow in the Western United
Magma Fracture Models .............. 224 States ........................... 273
Experiments on Buoyancy-Driven Fracture 225 Comparison with Heat Flow Temperature-
Experimental Method and Results 225 Depth Profiles .................... 274
Dimensional Analysis of Results 228 Generalized Thermal Structure beneath the Western
A New Model for the Crack Propagation United States from Seismic Velocity Data 277
Velocity ......................... 229 General Temperature Distributions 277
Shape of Fractures with Constant Fluid Regional Temperature Distributions 277
Volume 230 A Remark on Melt Retention and the Effective
Shape of Fractures with Constant Fluid Flux 231 Melt Content of a Source Region 278
x Contents

Comparison with Heat Flow-Derived Observations of Composite Flows, Dikes, and

Temperatures ..................... 279 Conduits . 321
On Discontinuous Magma Ascent and the Basalt and Rhyolite Systems . 321
Mechanics of Basaltic Underplating 279 Rhyodacite and Rhyolite Systems 323
Inherently Discontinous Nature of the Ascent Models of the Origin of Mafic-to-Silicic Zoning 326
Path .. .. . .. .. 280 Sequential Magma Transport . 326
Penetration of the Mantle-Crust Transition Simultaneous Transport of Different Magma
Zone by a Buoyancy-Driven Crack 281 Types . 327
Penetration of the Mantle-Crust Transition Fluid Dynamics of the Simultaneous Flow of Two
Zone by a Constant Fluid Pressure Crack 284 Magmatic Components . 330
Penetration of a Material Transition Zone by a Polymer Coextrusion Experimental Analogues 331
Fluid-Pressurized Crack with Differential Numerical Models of Encapsulation and Self-
Boundary Loading ................ 284 Lubrication . 334
Closing Discussion ....................... 285 Analytical Modeling of Lubricated Magma
References ............................. 286 Transport . 336
A Rational Model for Magma Chamber Withdrawal 346
A Physical Model for the Contemporaneous
Chapter 13 Aspects of Magma Generation Eruption of Three Domes in the Inyo Volcanic
and Ascent in Continental Lithosphere Chain, Long Valley, California . 349
Summary and Conclusions . 351
George W. Bergantz and Ralph Dawes
References . 352
Overview . 291
Introduction . 291
Melt Generation . 294
Modeling the Physical Interaction of Basalt and Chapter 15 Fluid and Thermal
Crust .. 297 Dissolution Instabilities in Magmatic
Rheology of Suspensions . 297
Numerical Model of Basaltic Underplating 299
Systems
Melt Collection and Ascent . 306 1. A. Whitehead and Peter Kelemen
Melt Convection in a Porous Medium 306
Overview . 355
Compaction . 307
Introduction . 356
Diapirism and Diking . 308
Thermal Erosion . 358
Tectonic Regimes and Magma Ascent 309
Transition of One-Dimensional Flows to
References . 312
Oscillations: An Experimental Study 358
Stability of Thermally Eroded Flows 362
Experiments with Paraffin . 369
Chapter 14 Two-Component Magma Magma Flow Instabilities from Chemical Corrosion 371
Transport and the Origin of Composite References . 377
Intrusions and Lava Flows
Charles R. Carrigan
Author Index 381
Overview 319 Geographic Index 389
Introduction 320 Subject Index 393
 Contributors

Numbers in parentheses indicate the pages on which the authors' contributions begin.

George W. Bergantz (291), Department of Geological tion of Earthquakes and Volcanic Eruptions, Faculty
Sciences, University of Washington, Seattle, Wash- of Science, Tohoku University, Sendai 980, Japan
ington 98195
Moritz Heimpel (223), Department of Earth and
F, Boudier (77), Laboratoire de Tectonophysique, Planetary Sciences, The Johns Hopkins University,
Universite Montpellier 2, URA 1370 CNRS, 34095 Baltimore, Maryland 21218
Montpellier, France
B. Ildefonse (77), Laboratoire de Tectonophysique,
Charles R. Carrigan (319), Earth Science Division, Universite Montpellier 2, URA 1370 CNRS, 34095
Lawrence Livermore National Laboratory, Liver- Montpellier, France
more, California 94550
Peter Kelemen (355), Department of Physical Ocean-
Y. J. Chen (139), College of Oceanography, Oceanog- ography, Woods Hole Oceanographic Institution.
raphy Administration, Oregon State University, Cor- Woods Hole, Massachusetts 02543
vallis, Oregon 97311
G. Kent (139), Institute of Geophysics and Planetary
Prame N. Chopra (37), Australian Geological Survey Physics, Scripps Institution of Oceanography, Uni-
Organization, Canberra, ACT 2601, Australia versity of California-San Diego, La Jolla, Califor-
nia 92093
Reid F. Cooper (19), Department of Materials Science
and Engineering, University of Wisconsin- David L. Kohlstedt (37), Department of Geology and
Madison, Madison, Wisconsin 53706 Geophysics, University of Minnesota, Minneapolis,
J. Huw Davies (197), Department of Earth Sciences, Minnesota 55455
The Jane Herdman Laboratories, University of Liver-
A. Nicolas (77), Laboratoire de Tectonophysique,
pool, Liverpool L69 3BX, England
Universite Montpellier 2, liRA 1370 C~RS, 34095
Ralph Dawes (291), Department of Geological Sci- Montpellier, France
ences, University of Washington, Seattle, Washing-
Peter Olson (I, 223), Department of Earth and Plane-
ton 98195
tary Sciences, The Johns Hopkins University, Balti-
Tye T. Gribb (19), Department of Materials Science more, Maryland 21218
and Engineering, University of Wisconsin-
Madison, Madison, Wisconsin 53706
J. Orcutt (139), Institute of Geophysics and Planetary
Physics, Scripps Institution of Oceanography, Uni-
A. Harding (139), Institute of Geophysics and Plane- versity of California-San Diego, La Jolla, Califor-
tary Physics, Scripps Institution of Oceanography, nia 92093
University of California-San Diego. La Jolla. Cali-
E. M. Parmentier (55), Department of Geological Sci-
fornia 92093
ences, Brown University, Providence, Rhode Island
Akira Hasegawa (179), Observation Center for Predic- 02912

xi
xii Contributors

J. Phipps Morgan (139), Institute of Geophysics and Akira Takada (241), Environmental Geology Depart-
Planetary Physics, Scripps Institution of Oceanogra- ment, Geological Survey of Japan, Ibaraki-Ken 305,
phy, University of California-San Diego, La Jolla, Japan
California 92093
J. A. Whitehead (355), Department of Physical Ocean-
Michael P. Ryan (97, 259), U.S. Geological Survey, ography, Woods Hole Oceanographic Institution,
Reston, Virginia 22092 Woods Hole, Massachusetts 02543

Hiroki Sato (259), Institute for the Study of the Earth's Shanyong Zhang (19), Department of Materials Sci-
Interior, Okayama University, Tottori-Ken 682-01, ence and Engineering, University of Wisconsin-
Japan Madison, Madison, Wisconsin 53706
David W. Sparks I (55), Department of Geological Sci- Dapeng Zhao (179), Seismological Laboratory, Cali-
ences, Brown University, Providence, Rhode Island fornia Institute of Technology, Pasadena, California
02912 91125
I Present address: Lamont-Doherty Earth Observatory, Columbia Uni-

versity, Palisades. New York 10964.

 Acknowledgments

The Magmatic Systems project has been a collective effort in several ways. Each manuscript has materially
benefited from the critical commentary provided by two or more reviewers. These participants are given special
acknowledgment for the expertise, thoroughness, energy, and time they have applied. They are:

Keiiti Aki Department of Geological Sciences, Ross C. Kerr Research School of Earth Sciences, The
University of Southern California . Australian National University
Charles R. Bacon Branch of Volcanic and Geothermal Christopher Kincaid Graduate School of
Processes, U.S. Geological Survey Oceanography, University of Rhode Island
David Bercovici Department of Geology and Arthur H. Lachenbruch Branch of Tectonophysics,
Geophysics, University of Hawaii U.S. Geological Survey
David D. Blackwell Department of Geological Brian R. Lawn Material Science and Engineering
Sciences, Southern Methodist University Laboratory, National Institute of Standards and
C. Wayne Burnham Department of Geosciences, The Technology, U.S. Department of Commerce
Pennsylvania State University, and Department of John R. Lister Institute of Theoretical Geophysics
Geology, Arizona State University and Department of Earth Sciences and Applied
Ulrich Christensen Institut fur Geophysik, Universitat Mathematics and Theoretical Physics, Cambridge
Gottingen University
Reid F. Cooper Department of Materials Science and Nobuo Morita Research and Development
Engineering, University of Wisconsin-Madison Laboratories, Conoco Incorporated
Paul T. Delaney Branch of Volcanic and Geothermal Janet L. Morton Branch of Pacific Marine Geology.
Processes, U.S. Geological Survey U.S. Geological Survey

Robert S. Detrick Department of Geology and Peter Olson Department of Earth and Planetary
Geophysics, Woods Hole Oceanographic Institution Sciences, The Johns Hopkins University
John J. Dvorak Cascades Volcano Ohservatory, U.S. John S. Pallister Branch of Volcanic and Geothermal
Geological Survey Processes, U.S. Geological Survey

Terence ~. Edgar Branch of Atlantic Marine Mervyn S. Paterson Research School of Earth
Geology. U.S. Geological Survey Sciences, The Australian National University

Donald W. Forsyth Department of Geological Rishi Raj Department of Materials Science and
Sciences, Brown University Engineering, Cornell University

Wes Hildreth Branch of Volcanic and Geothermal Neil M. Ribe Department of Geology and Geophysics,
Processes, U.S. Geological Survey Yale University

H. Mahadeva Iyer Branch of Seismology, U.S. Frank M. Richter Department of Geophysical

Geological Survey Sciences, University of Chicago

Ian Jackson Research School of Earth Sciences, The Eugene C. Robertson 917 National Center. U.S.
Australian National University Geological Survey

Shun-Ichiro Karato Department of Geology and Allan M. Rubin Department of Geological and
Geophysics, University of Minnesota Geophysical Sciences, Princeton University

xiii
xiv Acknowledgments

John H. Sass Branch of Tectonophysics, U.S. Rob van der Hilst Research School of Earth Sciences,
Geological Survey The Australian National University
Wayne C. Shanks III Branch of Eastern Mineral E. Bruce Watson Department of Earth and
Resources, U.S. Geological Survey Environment Sciences, Rensselaer Polytechnic
Norman H. Sleep Department of Geophysics, Institute
Stanford University Sarah T. Watson Department of Earth Sciences,
Yoshiyuki Tatsumi School of Earth Sciences, Kyoto Oxford University
University Stephen M. Wickham Department of the
Robert I. Tilling Branch of Volcanic and Geothermal Geophysical Sciences, University of Chicago
Processes, U.S. Geological Survey Lionel Wilson Environmental Science Division,
Lancaster University

The volume has been designed to address perceived needs that relate to both research and to graduate teaching
in Earth Sciences. All chapters have been invited to fill specific niches within the overall physical processes scope.
Aspects, however, of selected chapters have also been presented at the Symposium on the Generation, Segregation,
Ascent and Storage of Magma at the 29th International Geological Congress, Kyoto, Japan. In this connection, I
am indebted to Professor Ikuo Kushiro of the Geological Institute of the University of Tokyo for the invitation to
convene the Symposium and for his kind support. Professor Kosuke Kamo and Dr. Kazuhiro Ishihara of Kyoto
University provided logistical support and gracious hospitality as did professors Toshitsugu Fujii and Yoshiaki Ida
of the Earthquake Research Institute of the University of Tokyo, and Yoshiyuki Tatsumi of the School of Earth
Sciences of Kyoto University.
Administrative support within the U.S. Geological Survey has been provided by Bruce Hemingway, Chief,
Branch of Lithospheric Processes and by Rob Wesson, Chief, Office of Earthquakes, Volcanoes and Engineering.
Debbie Pasquale and Mary Woodruff have applied their wordprocessing skills toward typescript preparation. Shir-
ley Brown, Fran Buchanan, Lendell Keaton, and Nancy Polend have applied their graphics skills in making several
of the illustrations of this volume a reality. Lewis Thompson provided exceptional photographic support. Thanks
to Scotty Livingston and Herb Wolford for additional administrative support.
The editorial staff of Academic Press has been a pleasure to work with and special appreciation is extended to
Charles G. Arthur, Jacqueline Garrett, and Michael Remener.
Professor Kosuke Kamo, Dr. Kazuhiro Ishihara, and Mr. T. Takayama of the Sakurajima Volcanological Ob-
servatory of Kyoto University, have graciously provided the figure that is the cover of this volume. Permissions for
additional figure usage have been kindly granted by:

Disaster Prevention Research Institute of Kyoto Nature

University Polymer Engineering and Science
The Geologists Association of London Sakurajima Volcanological Observatory, Kyoto
Professor C. D. Han. Polytechnic Institute of New University
York Society of Plastics Engineers
Journal of Applied Polymer Science Dr. H. W. Stockman, Sandia National Laboratories
Macmillan Magazines Ltd. U.S. Geological Survey
Professor Nobuhiko Minagawa, Niigata University John Wiley and Sons, Ltd.
 Prologue

The full title of this volume is The Dynamics these centers; the three-dimensional magma buoy-
and Mechanics ofMagmatic Systems, but the con- ancy zonation structure of the magma reservoir;
traction to Magmatic Systems serves as a useful the sheeted dike complex; and the accreting oce-
shorthand. This book focuses on the core prob- anic crust. Island arc magmatism is studied from
lems of igneous petrology and volcanology: mag- the perspectives of geophysical constraints on
matic heat and mass transport processes in the melting domains beneath the volcanic front, the
Earth's mantle and crust. In preparing this book, I kinematics and thermal structure of flow in the de-
have tried to combine viewpoints that represent scending slab, and the melt generation-s-Hy) mi-
the three principal scientific perspectives: the theo- gration region of the recirculating mantle wedge.
retical, the experimental, and the observational. Studies of buoyancy-driven magma fracture and
These perspectives represent the three principal melt-filled crack interactions are presented next
means of study, and together they blend into a and have a general applicability that extends
powerful way of understanding Nature's plan for across specific tectonic settings. Magmatism in
the regulatory mechanisms that control the move- the continental interior is studied from the per-
ment of melts and thus the cooling and differentia- spectives of generalized regions of partial melt
tion of our planet. The scope is devoted to phy- within the subcontinental mantle and the mechan-
sical processes and to the physics of magmatic ics of basaltic underplating; assessments of the
systems in the Earth's interior and the subvolcanic relative roles of convective and conductive heat
environment. transfer and the rheology of partial melt in re-
The chapters present new research results gions of silicic magma generation; two-component
within a surrounding framework of review mate- viscous segregation processes during concurrent
rial that summarizes the theoretical approach, the silicic-basaltic magma flow; and the thermal and
experimental technique, or the relevant physical fluid instabilities in flow regimes that accompany
environment. This provides a useful background surficial lava movements and may participate in
for professional research workers, graduate stu- deeper magma migration.
dents embarking on research in magma transport, In Chapter 1, Olson discusses the fundamental
and the general interest reader seeking a greater process of decompression melting in ascending
familiarity with the issues and concepts of this upper mantle thermal plumes and in rising ther-
new field. The primary tools are continuum me- mal diapirs. Univariant melting that includes the
chanics, analytic and computational fluid dynam- effects of latent heat absorption and melt buoy-
ics, and computational heat transfer. Summary ancy is treated in a self-consistent approach. The
results from high-temperature and high-pressure chapter treats both solitary plumes and continu-
experimental geophysics and seismic tomography ously driven plumes that rise beneath and then
complement the theoretical studies. impact upon the base of thick (continental) litho-
The volume begins with studies of decompres- sphere or, alternatively, the relatively thin litho-
sion melting in ascending mantle plumes and ther- sphere of the oceanic basins. The finite-difference
mal diapirs and the rheology of basaltic partial solutions determined r~veal a generally lens-
melts. Next are studies of the kinematics of mantle shaped region of partial melt enrichment con-
flow beneath mid-ocean ridge spreading centers; formable with the basal topography of the litho-
the geologic field relations of intrusives within spheric keel. The length of the terminal portion of

xv
XVI Prologue

the ascent pathway-as allowed by the thickness tends to "dry out" the matrix olivines, reducing
of the lithosphere itself-interacts strongly with the hydrolytic component of aggregate flow regu-
the melting process and thus largely determines lation. As a result, the rheology of partial melts
the overall volume of melt for a given class of will depend on the H 20 partitioning into the grain
plume. Therefore, solitary plumes are able to pro- boundary melt phase as well as the bulk per-
duce = 106 km 3 of magma over a few million meability-both functions of the overall melt
years (Myr) beneath lithosphere of normal thick- fraction.
ness, while thin lithospheric thicknesses promote The fluid mechanical richness of a mid-ocean
correspondingly greater melting path lengths and ridge substructure is partially the result of the
are associated with = 107 km 1 of magma produc- concurrent and differential flow of the deform-
tion over comparable time scales. These results ing peridotite matrix and the intercrystalline melt
accord with current estimates of the continental phase. This flow is inherently three-dimensional,
and oceanic plateau basalt volumes. reflecting the offset or en-echelon arrangement
At the microscopic level, melting commences of spreading center segments in plan view-an
along the grain boundaries within the ascending organizational pattern that further increases the
lherzolite. Here, spreading melt may initiate the complexity of the flow. Sparks and Parmentier
development of microporous networks and epi- (Chapter 4) have applied the mass balance and
sodes of grain-scale fluid flow and thus assist in force balance equations for concurrent matrix
the attenuation of transiting seismic waves. In a and melt flow in a deforming mantle in three-
highly original series of high-temperature, beam- dimensional simulations of the melt migration
bending experiments, Gribb, Zhang, and Cooper process. Combined with the energy equation, the
(Chapter 2) have induced alternating states of determination of characteristic length scales for
tension and compression in samples of synthetic the solidification region has helped identify where
olivine-basalt partial melts. Microporous flow of in the system the permeability reduction due to so-
melt within the samples is thus experimentally in- lidification promotes enhanced melt pressures and
duced via the production of transient gradients in a resulting matrix dilation. An intriguing result
the dilational stress. Application of the linear vis- has been that melt focussing occurs in the decom-
coelastic Burgers model to the rheology of this pacting (dilating) boundary layer that roofs the
two-phase, liquid-solid system provides relation- melting region. Melt enrichment and permeability
ships for the aggregate shear viscosity and for the enhancement in this magmatic canopy thus allows
bulk viscosity. magma to migrate upslope toward the ridge axis,
Both hydrolytic weakening and melt phase en- assisted by positive buoyancy forces. Geophysical
hanced grain boundary diffusion processes affect evidence for this type of focussing in Iceland,
the rheology of rocks in regions of magma gen- where the magma-rich canopy lies at the top of the
eration. Deformation experiments conducted by asthenosphere and dips symmetrically away from
Kohlstedt and Chopra (Chapter 3) on synthetic the neovolcanic zone axis, has been summarized
basalt and olivine aggregates have thrown light by Ryan (1990).
on the important interactions between these two The availability of relatively large volumes of
mechanisms of weakening. In unmelted mantle magma near the top of the asthenosphere, com-
peridotites, distributed water-derived species bined with the steep thermal gradients and rapidly
within the framework of the crystalline silicates varying stress states associated with the divergent
promote hydrolytic weakening and thus reduc- mantle flow below the Mohorovicic discontinuity,
tions in aggregate viscosity. The infrared spectros- promotes diverse modes of magma fracture be-
copy portion of this study has revealed that ample neath an active ridge. Mapping campaigns in the
amounts of a grain boundary melt may deplete the Oman ophiolite conducted by Nicolas, Boudier,
matrix olivine of hydrous species due to the rela- and Ildefonse (Chapter 5) combined with careful
tively strong H20 partitioning into the adjacent petrofabric analyses have defined the nature of
melt phase. Thus, while the presence of melt dra- these modes and their relationships to the flowing
matically affects the rheology via creep strength asthenosphere and the accreting lithosphere. A
and viscosity reductions (e.g., about a factor of 5 first generation of indigenous gabbro and pyroxe-
in strength reduction for =8% basaltic melt), it nite dikes and veins crosses melt-impregnated
 Prologue xvii

peridotites while a second generation of micro- first demonstrated by Ryan (1987a) for Hawaii,
gabbro and diabase dikes has intruded into cooler and then again for Hawaii and Iceland with hy-
crustal rocks and is regionally organized in pat- pothesized extensions to the East Pacific Rise by
terns that mimic the strike of the reconstructed Ryan (l987b). The recognition of the widespread
ridge itself. Diverse intrusive orientations that role of neutral buoyancy states in the shallow sub-
range from dikes to sills reflect the lithospheric volcanic environment has, I think, profound im-
and asthenospheric stress fields respectively, and plications for how magmatic systems work. In
sill-forming injections have been spatially linked these terms, the importance of the neutral buoy-
with the strongly divergent asthenospheric flow ancy concept is three-fold: (a) an internally con-
fields at the very top of subridge diapirs. Thus, sistent existence criterion for shallow magma
detailed studies of the Oman ophiolite open win- reservoirs; (b) a new tool for the quantitative un-
dows into the roots of mid-ocean ridges and illu- derstanding of the upward (and lateral) evolution
minate their magmatic machinery. of the reservoir; and (c) a powerful means of un-
Gravitational equilibrium and the quest for its derstanding magma dynamics in intrusion epi-
periodic reinstatement are fundamental to the op- sodes within stratified reservoirs as well as in the
eration of magmatic systems. Until recently, the sheeted-dike complex. In Chapter 6, I have
functional maintenance of this equilibrium state reviewed several aspects of the neutral buoyancy
in active volcanic systems was thought to gener- structure of mid-ocean ridge magma reservoirs.
ally require the movement of magma completely The compositionally averaged horizon of neutral
through the lithosphere and its emplacement as buoyancy (HNB) for the East Pacific Rise occurs
lava on the Earth's surface. Lithospheric densities at a depth of = 1000 m to =3000 m beneath the
were generally regarded as always greater than rise axis (at 9°N latitude). Fractional crystalliza-
magma density and states of universal positive tion and elastic crack stability criteria suggest that
buoyancy were believed to characterize the oce- this generalized HNB should be subdivided into
anic and continental lithosphere. These views a deeper picritic horizon (=1400 to =3000 m
have been widespread and pervade the geophysi- depth) and a shallow tholeiitic horizon (=600 m
cal and volcanological literature. The assumption to =1400 m depth). During the fractional crys-
of complete fluid continuity from the magma tallization that transforms a picritic melt and ol-
source region to the eruption site (now known to ivine mixture into tholeiitic melt, parcels of
be generally false) complemented the universal magma must elevate themselves to maintain
positive buoyancy assumption and permitted the gravitational equilibrium. In ophiolite complexes,
balancing of lithostatic and magIj1astatic pressures the uppermost isotropic gabbros are correlated
over great columns of virtually lithospheric di- with the picritic HNB, whereas the sheeted-dike
mensions. Existence criteria for mid-ocean ridge complex corresponds to the tholeiitic horizon of
magma reservoirs were ascribed to be almost neutral buoyancy. These relations are consis-
wholly the result of the initially assumed and then tent with the dominantly tholeiitic nature of the
numerically modeled ridge thermal structures: ro- sheeted dikes as well as the paucity of ocean floor
bust development for high spreading rates, negli- picrites. The combined tholeiitic and picritic HNB
gible development for low spreading rates. The is, in addition, in virtually a 1: 1 correspondence
detailed density states of mid-ocean ridges as with the inferred sheeted-dike complex and the
the combined functions of the relevant range in compressional wave velocity minima region (i.e.,
magma compositions, suspended mineral phases, the magma-rich portion of the reservoir) for the
and the vertical succession of the nonlinear in situ East Pacific Rise as determined by seismic reflec-
density structure of the oceanic crust itself were tion surveys and by tomographic inversions of
not of general interest. These views, with few ex- P-wave travel-time residuals. The dynamics of
ceptions, characterized the peripd following the magma injections within the sheeted-dike com-
inception of plate tectonics to at least the mid- plex are suggested to be analogous to their coun-
1980s, and to some extent remain today. That terparts at Kilauea volcano, Hawaii, and to the
states of positive magma buoyancy are, in fact, Krafla central volcano, northeast Iceland, during,
routinely complemented by states of neutral and for example, the 1975-1984 intrusion-eruption
negative buoyancy in active basaltic systems was episodes. [It is important to recognize that the
XVlll Prologue

present erosion levels in Hawaii and Iceland do waves resulting from rock composition, tempera-
not sample the heart of the tholeiitic HNB (cen- ture, regional structure, and the presence and dis-
tered at =3 km local depth) and thus contain com- tribution of included domains of fluid. Hasegawa
ponents of vertical flow.] From an evolutionary and Zhao (Chapter 8) have employed a three-
perspective, as lithosphere is created at mid-ocean dimensional velocity model that explicitly incor-
ridges and spreads laterally, the horizons of neu- porates the locally complex shapes of the Moho-
tral buoyancy and negative buoyancy ride with it, rovicic and Conrad discontinuities to refine the
and they are suggested to significantly influence velocity structure of the northeast Japan arc based
the ascent of off-axis magma to about 30 Myr and on least-squares inversions of P-wave travel-time
provide the means for shallow off-axis storage residuals. The magma-impregnated low velocity
as well. zones revealed in their study are inclined, and they
Recent advances in the seismic resolution of dip parallel to the dip of the Pacific Plate but are
mid-ocean ridge magma reservoirs have refined found at least some 30-60 km above the plate's
their images, considerably tightened their geomet- upper surface. Produced by velocity contrasts of
ric extents, and have laid the basis for a new gen- 2-6%, they are continuously distributed from the
eration of numerical models that yield kinematic upper crust (shallow magma ascent regions) to a
insights on how ridges work. Phipps Morgan, depth of 100-150 km within the mantle wedge
Harding, Orcutt, Kent and Chen (Chapter 7) have (the magma generation and segregation regions).
reviewed the results of seismic reflection experi- The refrigeration of the mantle produced by
ments over the East Pacific Rise, suggesting that the subduction of oceanic plates leads naturally to
the magma reservoir comprises a relatively fluid- the paradox of prolific arc magmatism directly
rich upper chamber (width: = 1 km; thickness above a mantle wedge that is cooled by conduc-
=50-200 m) that crowns a progressively crystal- tive heat transfer to a juxtaposed downgoing slab.
rich mush where melt contents average 3-5%. Building on the theoretical work of McKenzie
Kinematic modeling of the two-dimensional struc- (1969) and Toksoz, Minear, and Julian (1971),
ture of the flow field induced by steady-state lith- the experimental studies from the laboratories of
ospheric lid translation has used the conservation Burnham, Green, and Wyllie, and the geochemi-
equations in conjunction with McKenzie's finite cal studies of Tatsumi (1989),1. H. Davies (Chap-
deformation formulation. The deformation results ter 9) addresses the why and how of this paradox
accord nicely with textural observations in ophio- by combining thermomechanical modeling with
lite crustal sections: steeply dipping foliations high an incremental mechanism for water transfer from
in the gabbros, with progessive increases in strain the slabs' upper surface well into the wedge inte-
and considerable foliation flattening with depth. rior. The model produces finite element solutions
About 439 of the world's 442 active andesitic to the advection-conduction energy equation in
volcanoes occur in regions of oceanic plate sub- response to the kinematically driven flow field
duction (Gill, 1981). Frequently arranged in great of the subducting slab. The steady-state thermal
sweeping island arcs, their magma generation sys- structures of the wedge and slab as functions of
tems, segregation mechanisms, and ascent net- the subduction velocity reflect the slow secondary
works remain shrouded behind a veil of complex wedge flow and set the stage for the H2 0 migra-
rheology, often sluggish magma migration kinet- tion mechanism. Water is released from amphi-
ics, and largely aseismic ascent pathways. Be- bole at depths near 80 km, and in a cyclic series
cause so much of the inner workings of these sys- of dehydration (H20 release) amphibole refor-
tems remains unknown-even in outline-they mation stages, rides the downward streamlines of
comprise a great and relatively uncharted frontier the wedge flow field in steps that, in inchworm
for research in magma transport. Their petrologic fashion, carry it both laterally and downward into
diversity, violent eruptive behavior, and shear melt generation depths. These positions within the
number underscore the great need for intensely wedge are consistent with the roots of island arc
focused and highly coordinated research in this magmas and the location of the volcanic front
area. A parting of this veil and an initial illumina- above.
tion of an arc system interior has made use of the The processes of linear elastic brittle fracture,
velocity dependence of compressional seismic creep rupture, and stress corrosion cracking oc-
 Prologue xix
cupy portions of a continuum of failure phe- however, for a rejuvenation of work within-and
nomena associated with the migration of magma beneath-the continental interiors. Sato and Ryan
by cracks. Within the oceanic and continental (Chapter 12) have estimated generalized tempera-
lithosphere and within the crust above subduction ture profiles and degrees of partial melt in the up-
zones, the fracture mode of magma ascent, em- per mantle beneath the western United States. The
placement, and lateral intrusion is one of the great study makes use of seismic velocity data for par-
workhorses of magma migration, Without it the tially melted peridotites and published anelasticity
galaxy of veins, dikes, sills, sheets, fissure erup- data for the regional scale upper mantle. Experi-
tions and shallow basaltic-to-andesitic magma mental measurements of the compressional wave
reservoirs would not exist. Heimpel and Olson velocity and elastic wave attenuation in spinel
(Chapter 10) have used gelatine gel-based ana- Iherzolites show a homologous temperature de-
logue experiments to study the mechanics of posi- pendence: families of Vp, Qp-I data for a vari-
tive buoyancy-driven, fluid-filled fractures. In ety of pressures plot as a single band in terms of
contrast with dry remotely loaded fracture pro- the experimental temperature normalized by the
cesses which induce large amounts of elastic sample melting temperature (T/Tm ) , Thus, a
stored strain energy just prior to fracture and are knowledge of the solidus as a function of pressure
associated with catastrophic crack growth, the lo- allows the estimation of temperature as a function
cally loaded buoyancy-driven cracks suggest a set of pressure-and hence depth in the mantle,
of dynamical constraints on the fracture process Therefore, both Vp and Qp-I are of use in esti-
when the fluid-filled crack is imbedded in visco- mating T(Z) for various regions of interest. In
elastic media. Thus, the coupled processes of addition, empirical plots of melt fraction versus
time-dependent external stress relaxation and in- (TjTm ) allow in situ melt fractions to be estimated
ternal fluid flow that accompany increments of once Tm and T have been determined. This pro-
crack extension modulate the crack propagation cedure has produced generalized temperature-
velocities and, by analogy, may significantly influ- depth profiles for major regions of the western
ence the magma migration process. United States, and they cover the depth range 50-
The development and function of well-trodden 300 km. At shallow (crustal) depths they are
magma migration conduits, as well as the nuclea- comparable with conductive geotherrns, while at
tion of shallow magma reservoirs, all owe a great depths greater than 200 km they are compatible
debt to the phenomena of magma fracture coales- with the adiabatic temperature gradient. Partial
cence. Takada (Chapter 11) has used Westergaard melt contents estimated for the intermountain
stress functions and complex variables to com- and western margin regions cover the range
pute the nature of the displacements and stress =3%:50 m:5IO+%, when regionally averaged.
fields near two parallel offset fluid-pressurized These melt contents occur over the 120-200 km-
cracks in a search for crack coalescence criteria. depth range, and transitions from porous media
Complementary gelatine gel-based experiments melt flow to vein and dike flow are expected over
on fluid buoyancy-driven cracks have revealed these depth intervals. A review of elastic fracture
that magma-filled cracks can coalesce and un- stability theory within the context of subconti-
dergo a resulting magma volume increase as long nental magma ascent reveals the fundamentally
as there is an available range in crack sizes and disconnected nature of the ascent pathway: the fi-
ascent velocities. The coalescence process is fur- nite strengths of fluid-weakened mantle perido-
ther enhanced by small differential stress states tites at high temperatures mandate, in turn, finite
in horizontal sections and by concomitant large height magma-filled fractures. Deep dike swarms
magma supply rates. are thus suggested to be the ascent mode for the
The continents have been the birthplace of ig- deepest levels of basaltic underplating along the
neous petrology and most subdisciplines of geo- crust-mantle boundary. The crust-mantle inter-
physics, yet after the inception of plate tectonics, face itself, reflecting changes in both mineralogy
much of the energy and attention in these fields and potential melt content, may be treated as a bi-
of research has been understandably directed off- elastic material boundary, and the morphology
shore, toward the problems-and the promise- and mechanics of dikes that penetrate such an in-
of the ocean basins. Perhaps the time is now ripe, terface depend on the ratios of effective elastic
XX Prologue

moduli on either side, as well as the fluid pressure the early stages of underplating appear to be
loading conditions within the fracture. largely conductive, and multiple basaltic intru-
The shear viscosities of single-phase natural sions are required to thermally mature a deep
silicate melts span an extraordinary 13 orders of crustal section and permit widespread regional
magnitude and the melts of the continental inte- melting episodes in the crust above.
riors must routinely encompass this entire spec- Magma reservoirs may be viewed as great me-
trum. For regions of basaltic underplating and of chanical capacitors that, during magma influxes,
the secondary generation of silicic crustal mag- slowly accumulate potential energy. This stored
mas, the single-phase contents of laboratory cru- energy occurs by virtue of the vertical displace-
cibles are but one end-member of a rheological ments of their roof rocks and caldera floors and by
continuum that includes melt-based suspensions, the compression of the immediate surroundings,
melt-weakened country rocks, and highly altered, including the magma. Conceptually, these are
but as yet unmelted, rock on all scales. The high- akin to compressed springs that are coupled in
solids-fraction domains thus raise the aggregate parallel. Rupture of the reservoir walls suddenly
effective shear viscosities to yet greater values. releases magma into the surrounding country
Accordingly, the migration kinetics of ascending rocks in dike-forming injections-a process that
magma in the continental crust cover a time range relaxes the conceptual springs but works against
that extends far beyond the realm of human ex- the environment by virtue of the crack wall dis-
perience and renders much of the generation, as- placements and the work-of-fracture (atomic bond
cent, emplacement, and replenishment process of breaking) at the advancing crack tip. Importantly,
silica-rich magmas aseismic. Like the pathways the energy dissipated in the high shear boundary
for magma above Wadati-Benioff zones, the con- layers of the near-wall dike interior represents a
tinental interior regions of silicic melt generation significant resistance term in the overall process
remain heavily shrouded and thus deserving of of fluid flow. Inherently repetitious, the overall
carefully focused and well-coordinated research. magma recharge-discharge cycle maximizes its
Bergantz (Chapter 13) reviews aspects of conti- efficiency by (1) maximizing the overall potential
nental magmatism around the general theme of energy reductions, and (2) simultaneously mini-
basaltic underplating. Finite volume solutions of mizing all the associated work done in the dike
the conservation equations have employed the formation episodes. Thus a globally integrated
spatial and temperature dependence of viscosity, least work principle lies at the heart of the magma
density, heat capacity, and thermal conductivity in storage and fracture process. In regions of mixed
a model of basaltic underplating that considers the magma storage, Nature also has a trick up her
progress of crystallization in the basaltic substrate sleeve that helps in the effort to minimize the work
as a function of partial melting in the overlying done in moving fluids and thus enhance the effi-
crust. Three questions are of interest: How does ciency of the magma injection process. Carrigan
convection influence the timing of partial melting (Chapter 14) has derived the lubrication equations
in the crust, and what is the style of convection? for the two-component flow of a power-law fluid
How does variable viscosity (asa function of melt in a dike-like geometry. The solutions describe
composition and crystallinity) influence convec- the process of the hydrodynamic encapsulation of
tion? What are the overall heat transfer rates from the high viscosity (interior) phase by the low vis-
the system relative to those expected from con- cosity (exterior) fluid. Importantly, the outer (low
duction only? Nusselt (Nu) number computations viscosity) phase occupies the high shear-and
during the evolution of the melting region com- energy dissipating-boundary layers at the dike
pare the strength of the overall heat transfer com- wall. The overall flow process is analogous with
ponents and suggest that for critical melt fractions the lubricated pipelining process of industrial set-
over O.3::50 m::50.5, the amount of melt generated tings. It describes the means of providing a least
was indistinguishable from that produced during work solution to the problem of mixed magma
conduction only. This is broadly consistent with transport in settings as diverse as Long Valley,
the notion that the rheological conditions associ- California, and southeast Iceland, for example. A
ated with the onset of crystallization control the substantially more realistic magma withdrawal
subsequent dynamic evolution of the body. Thus model has also evolved from the work, wherein
 Prologue XXI

two or more layers of differing composition (and solidification to advection time scales. These
viscosity) can be simultaneously withdrawn with- forms range from pillows through surface ripples
out invoking the rather special set of circum- to flat sheeted flows.
stances inherent in the draw-up-depth parameter In a direct way, this volume complements the
or in the overtaking requirement of other ap- book Magma Transport and Storage (M. P. Ryan,
proaches. Overall, the model represents a signifi- ed.), John Wiley and Sons, Ltd., 1990. While
cant step forward in advancing our understanding some authors have returned to write new chapters
of this important class of problems. for this book, there is no overlap, and each chapter
The simultaneous attempts to achieve both is a fresh, new and complementary effort, quite
thermal and mechanical equilibrium within re- separate and distinct from the work cited above.
gions of high temperature magma flow, lead to a Increasingly, scientists with diverse back-
number of thermomechanical feedback processes. grounds that have traditionally been spread out
Familiar examples relate to the potential for either over several of the subdisciplines of geophysics,
dike wall solidification or meltback, depending on igneous petrology and volcanology have found
the flow rates, magma temperatures, dike widths that the subject area of magma transport offers an
and the ambient thermal environment (Bruce and exciting and unifying process-oriented framework
Huppert, 1990). Other examples include the role for study. Within this framework lies great cohe-
of the evolving conduit geometric aspect ratio in sion yet remarkable room for individual and team
flow localization during the course of an eruption research initiatives. It is my hope that the present
(Delaney and Pollard, 1982). Whitehead and Ke- book will further this process of unification by
lemen (Chapter 15) explore, theoretically and ex- providing fertile points of departure for research
perimentally, a new class of dynamic feedback as well as offering stimulating areas of discussion
phenomena. Gravity-driven laboratory flow ex- for the classroom and for the seminar hall.
periments, for example, reveal the potential for Finally, detailed author, geographical and sub-
interplay between fluid pressure, environmental ject indices complete the volume and promote ac-
and fluid temperatures, temperature-dependent cessibility and cross-referencing.
viscosity fluctuations, and the resulting flow rates.
Michael Ryan
These effects include flow choking, episodic pres-
sure buildups, and transitions in flow rates that
correlate with the radially inward growth of high
viscosity 'walls' during environmental tempera- References
ture drops. Additional effects have been modeled
theoretically, and include oscillatory pressure- Bruce, P. M., and Huppert, H. E. (1990). Solidification and
melting along dykes by the laminar flow of basaltic
time (P-t) and fluid velocity-time histories that magma, in "Magma Transport and Storage" (M. P. Ryan,
may have either sawtooth or sinusoidal P-t sig- ed.), Wiley, Chichester/Sussex. England.
natures, and are correlative with pulsatile fluid Delaney, P. T., and Pollard, D. D. (1982). Solidification of ba-
flow behavior. These types of relationships hold saltic magma during flow in a dike, Am. J. Sci. 282, 856-
the potential for further application to the throt- 885.
Gill, 1. B. (1981). "Orogenic Andesites and Plate Tectonics,"
tling of magmatic and volcanic flows through re- Springer-Verlag, New York, pp. 390.
gions with high spatial gradients in environmental McKenzie, D. P. (1969). Speculations on the consequences
temperature and conduit geometry. For surficial and causes of plate motions, Geophys. J. R. Astron. Soc.
volcanic process considerations, laboratory simu- 18,1-32.
lations by Whitehead and Kelemen (Chapter 15) Ryan. M. P. (l987a). Elasticity and contractancy of Hawaiian
olivine tholeiite and its role in the stability and structural
of horizontally spreading flows have used paraffin evolution of sub-caldera magma reservoirs and volcanic
as the working fluid. These show transitions from rift systems, in "Volcanism in Hawaii" (R. W. Decker,
smooth perimeter radial flow regimes to finger T. L. Wright and P. H. Stauffer, eds.), C.S. Geo!. Survey
flow regimes as a function of the solidification Prof. Paper 1350, v. 2, 1395-1448.
Ryan, M. P. (l987b). Neutral buoyancy and the mechanical
rate and temperature-dependent viscosity. Ex-
evolution of magmatic systems, in "Magmatic Processes:
periments with lateral spreading flows from line Physicochemical Principles" (B. O. Mysen, ed.). The Geo-
sources have charted a morphological continuum chemical Society. Special Publication No. I. Yoder Sym-
of surface textures depending on the ratio of the posium Volume. 259-288.
XXll Prologue

Ryan, M. P. (1990). The physical nature of the Icelandic basalt magmas In subduction zones, 1. Geophys. Res. 94,
magma transport system in Magma Transport and Storage, 4697-4707.
(M. P. Ryan, ed.), Wiley, Chichester/Sussex, England. Toksoz, M. N., Minear, 1. w., and Julian, B. R. (l97\). Tem-
p. 175-224. perature field and geophysical effects of a downgoing slab.
Tatsumi, Y. (1989). Migration of fluid phases and genesis of J. Geophys. Res. 76,1113-1138.
Chapter 1 Mechanics of Flood Basalt Magmatism

Peter Olson

Overview Units

Tm melting temperature
The dynamics of magma generation in a mantle plume
r; surface melting temperature
are investigated using time-dependent numerical cal-
TD diapir temperature
culations of pressure-release melting in axisymmetric
thermal plumes and thermal diapirs in a viscous fluid TB basal temperature
with a rigid lid. The plumes ascend through the fluid Ji. gravitational acceleration m· s
and spread laterally beneath the lid, producing a broad he added crust thickness m,km
surface uplift and a correlated geoid high. Decompres- k thermal conductivity wm ,.oC '
sion melting in the pIume occurs in a lens-shaped region
r radial coordinate m,km
centered near 100 km in depth. The excess temperature
time
in the melting lens is 100-300° C. The calculations
demonstrate that the initial melting pulse from a single u radial velocity m s
plume can produce 106 krn' of magma in a few mil- Ii' vertical velocity m· s
lion years beneath normal lithosphere and more than y depth coordinate m.km
10 7 km' of magma beneath very thin lithosphere, in ac- z vertical coordinate m.km
cord with recent estimates of continental flood basalt
ZL lid thickness m.km
and ocean plateau volumes. After the initial melting
event the plume constricts to a narrow thermal anomaly ", cylinder height m,km
and the melt production rate decreases by nearly two a thermal expansivity °C '
orders of magnitude, consistent with the long-term his- f3 melting expansivity dimensionless
tory of many hotspots. The calculations also indicate y melting point gradient °C·km
that plume formation initiates an episode of enhanced
p mantle density mg·m
heat flow in the source region. This behavior supports
p, crust density mg·m
recent speculations that deep mantle plumes can ther-
mally couple the core to the mantle. X melting progress variable dimensionless
TJ viscosity pa- s
cb melt fraction dimensionless
Notation ,p, saturation melt fraction dimensionless
Units
ljJ streamfunction m"- s I

CP specific heat at constant pressure kJ· kg ,.oC I To_'fa surface and hasal boundary layer
ages
H enthalpy kJ· m
Hm enthalpy at melting kJ· m
, V del, the gradient operator m

H, enthalpy at saturation melt fraction kJ· m

, V' nabla

L latent heat of melting kl - kg

, u velocity of matrix material m- s

R cylinder radius m,km

r, i unit coordinate vectors, m

T temperature °C
T, interior mantle temperature °C

Magmatic Systems Copyright rr. 1994 by Academic Press, Inc.

Edited hy M. P. Ryan All rights of reproduction in any form reserved.
2 Peter Olson

Table 1
Continental Flood Basalts"

Basalt volume
Hood basalt Age (Ma) Hotspot (X 106km')

Columbia River 15±2 Yellowstone 0.3

North Atlantic 59±2 Iceland >1.5
Deccan 66± I Reunion >1.7
Parana 120±3 Tristan da Cunha -I
Karoo 192±3 Prince Edward ;;;.1.0
Siberia 250±? ? -0.9

"From White and McKenzie (1989).

Introduction model, the flood basalt province is interpreted to

be the result of the partial melting of the head of a
In this chapter I describe the sequence and dura- mantle thermal plume or diapir, and the hotspot
tion of magmatic events accompanying the de- track is interpreted to be the result of partial melt-
velopment of a mantle plume, using calculations ing of the plume conduit (Richards et al., 1989;
of decompression melting in axisymmetric ther- Griffiths and Campbell, 1990).
mal diapirs and in axisymmetric thermal plumes. Many large ocean plateaus are now thought
These calculations are motivated by growing evi- to have a plume origin as well (Mahoney, 1987;
dence in support of a mantle plume origin for White and McKenzie, 1989; Mahoney and Spen-
many continental flood basalt provinces as well as cer, 1991; Larson, 1991; Tarduno et al., 1991).
some oceanic plateaus. The calculations are also There are, however, some important physical dif-
intended to establish the temporal relationships ferences between ocean plateaus and the conti-
between near-surface magmatic activity associ- nental flood basalts. One significant difference is
ated with mantle plumes and heat transport into that ocean plateaus tend to be larger than conti-
the hot thermal boundary layer feeding the plume. nental flood basalts. As Table 1 indicates, conti-
This relationship is a key element in the current nental flood basalt volumes are typically 1-2 X
debate on the extent of the thermal coupling be- 106 km", whereas the data in Table 2 indicate
tween the Earth's core and mantle. ocean plateau volumes are typically 10 times that
size and the largest, the Ontong-Java plateau,
may exceed 50 X 106 km'. It is unreasonable to
Continental Flood Basalts, Oceanic Plateaus, explain this difference by postulating either a size
and Mantle Plumes

Many of the largest flood basalt events preserved Table 2

in (and on) the continental crust appear to be as- Large Oceanic Plateaus a
sociated with the point of origin of linear volcanic
hotspot tracks. Table 1 lists several prominent Basalt volume
continental flood basalt provinces and the hotspot Plateau Age (Ma) (X10 6km')
tracks that emerge from them. The association of 90-115
Kerguelen 20
flood basalt provinces with the beginning of a hot-
Manihiki 115-125 12
spot track, together with evidence that they are
Mid-Pacific Mts, 75-130 21
emplaced very rapidly and have a distinct isotopic
Ontong-Java Ll7±2 54
signature (Duncan and Richards, 1991; Carlson,
1991), has led to the embellishment of a "starting Shatsky 130-150 11

plume" model of flood basalt formation originally -From Larson (1991); Tarduno et al. (1991); Mahoney (1987);
proposed by Morgan (1981). In the starting plume Schuhert and Sandwell (1989).
 1. Mechanics of Flood Basalt Magmatism 3
difference or a temperature difference between is controlled primarily by the thickness of the
plumes rising beneath continental and oceanic lithosphere. According to the calculations, melt
lithosphere. Also, it is unlikely that the difference production in a plume can vary by a factor of 40
in size is due to greater longevity of the sources of or more because of differences in the thickness
oceanic plateaus. For example, the mammoth On- of the lithosphere beneath continental crust and
tong-Java plateau appears to have been built in young oceanic crust. The calculations also show
just a few million years, the same formation time that the rise of a leading diapir at the head of a
as continental flood basalts (Tarduno et al., 1991). continuous plume can produce a transient mag-
If we suppose that both continental flood basalts matic event with the magnitude and time scale of
and ocean plateaus are the result of mantle plumes, the largest ocean plateaus.
then it is reasonable to seek an explanation for the
systematic difference in size by appealing to dif-
ferences in the volumes of melt produced within Constraints from Seismic Tomography
plumes. and Geochemistry
The amount of melt produced in a given plume
is proportional to the amount of decompression Figure 1 shows vertical cross sections of the upper
the plume experiences in the region where the mantle at four hotspots from a model of shear wave
temperature is at or above the solidus. The cal- velocity variations derived by Zhang and Tani-
culations presented here show that temperatures moto (1992). Low velocity regions between 100-
above the solidus occur at 200 km depth or shal- and 200-km depths are found beneath the four hot-
lower, so the extent of melting in a mantle plume spots, Hawaii, Iceland, Azores, and Tristan da

Hawaii (20, 205, 105) Iceland (64, 340, 135)

o~;::;';~;:;:;;:;:::-"""'=:;;;;;;:;;;;:;:::::--J
100

200

300

E 400
.:s: -30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30
;- A A' B B'
! 0 Azores (38, 332, 100) Tristan da Cunha (-37, 348, 75)

100 .........:-_-::..-::;.;

200

300

400 +-...,....-,----r-.,......~&r_--+-r-=:;="¥"=,::=-=t
-30 -20 -10 0 10 20 30 -30 -20 -10 o 10 20 30
C C' D D'
Distance (degree)
-5.0_miml:::::::I:::::::1 t::;=:e (JElt 5.0
Shear Wave Velocity Anomaly

Figure 1 Cross sections of upper mantle shear wave velocity structure at four hotspot locations, as determined by Zhang and
Tanimoto (1992). Abscissa scales are the great circle degrees from the hotspot. Latitude, longitude, and azimuth of the great
circle cross section are indicated above each panel. The scale har gives velocity perturbation in percent. Note the low velocity
anomaly beneath each hotspot between 100 and 200 km in depth.
4 Peter Olson

Cunha. The horizontal extent of each region is waii and is probably less than 0.3 km' yr " ' for
500-1000 krn, approximately. Note there is little all hotspots (Duncan and Richards, 1991). This
indication of deep root structures beneath the low activity amounts to only a few percent of the
velocity regions; at least there is no evidence for 13 km 3 yr - I present-day rate of crust production
broad roots on the scale of several hundred kilo- estimated for the mid-ocean ridge system.
meters, the limit of resolution in the Zhang and The present-day level of hotspot activity, how-
Tanimoto study. However, regional-scale tomog- ever, is not representative of longer-term aver-
raphy indicates a deep, narrow root beneath Ice- ages. Figure 2 shows rates of oceanic crust pro-
land to about 375 km in depth (Ryan, 1990), which duction over the past 150 Myr averaged in 5-Myr
is evidently not seen on the scale of Fig. 1. bins, for ocean plateaus and for the Earth as a
Several interpretations of the broad structures whole, as determined by Larson (1991). Over the
beneath hotspots are plausible, but one based on past 10 Myr, oceanic plateaus have contributed
the dynamics of a buoyant plume approaching the about 1.5 krn' yr -I of crust on average, or roughly
lithosphere is particularly simple and appealing. 10% of the global rate for the same period. This
According to the plume model, the low seismic fraction is generally consistent with the fraction
velocity regions represent portions of the upper of the total heat transport at hotspots, compared
mantle intruded by the hot and perhaps partially to the heat transport at mid-ocean ridges. Davies
melted plume material as it spreads out beneath (1988) and Sleep (1990) have calculated that the
the lithosphere. Since each of these hotspots has present-day hotspots transport about 2-4 TW of
a well-dated track and is known to have been ac- heat to the lithosphere, while sea floor spreading
tive for at least 40 Ma, it is unlikely that these seis- transports roughly 10 times as much.
mic anomalies represent newly formed plumes. A The data summarized in Fig. 2 also indicate
more likely interpretation is that the low velocity that hotspot activity has been even greater at vari-
regions represent the structure of fully developed ous times in the past. In particular, the crustal pro-
plumes stagnating beneath the lithosphere. duction rate at oceanic plateaus during the mid- to
Seismic topographic images provide con- late Cretaceous (125-65 Ma) was three to five
straints on the pattern of mantle convection struc- times larger than in the past 10 Myr. The pulse in
tures, but it is hazardous to use images such as hotspot activity accounts for a significant part of
those in Fig. 1 for quantitative estimates of mantle the increase in total oceanic crust production dur-
temperature or melt content. However, there are ing that interval of time. Several particularly large
other ways to estimate the temperature of mantle oceanic events were responsible for the Creta-
hotspots. Schilling (1991) combined geochemical ceous pulse, including the formation of the On-
and topographic data from hotspot swells to infer tong-Java plateau and Mid-Pacific Mountains on
the excess temperature Ll T in the mantle beneath the Pacific plate and the Kerguelen plateau on the
near-ridge hotspots. The term excess temperature Antarctic plate. Larson's (1991) plate reconstruc-
refers to the temperature in the plume material tion for that time shown in Fig. 3 indicates that the
supporting the hotspot swell, relative to normal largest events occurred in two clusters, one near
mantle at the same depth. For three of the hotspots the former Pacific-Farallon plate boundary and
in Fig. 1, Schilling finds Ll T = 162, 263, and 198 the other near the Indian-Antarctic plate bound-
degrees Kelvin at Tristan, Iceland, and Azores, ary. In addition to defining two main centers of
respectively. plume activity, Fig. 3 also suggests that the most
massive magmatic events occur when hotspots
form near spreading centers, where the oceanic
Time Variations in Hotspot Activity lithosphere is thin.

There appear to be large changes in hotspot ac-

tivity with time. The evidence for variability in the Hotspot Activity and Geomagnetic
strength of hotspot magmatism comes from com- Reversal Frequency'
parisons between the present-day rate of hotspot
volcanism and the rates inferred for the past from Evidence that thermal plumes originating from
continental flood basalts and ocean plateaus. The the deep mantle thermally couple the core to the
present-day hotspot activity is dominated by Ha- mantle comes from the correlation between hot-
 35

30

WORLD TOTAL
(EXCEPT TETHYS)

NO MAGNETIC
REVERSALS
,---I
r--- I :
.- 1 I

r--- I :
1 1 1
I 1 1L _
I 1 1
I L I
1
PACIFIC RIDGES:
r -- -I (EXCEPT TETHYS)I
1 ------------L I
1
---I r---I
.-.-.-.~: =: .~ -·-·-·-L._.__ .-.-.-r·-·-l GONDWANA
L._._._~RIDGES
5
L '\. _._ ._....... _._._ ......
;
,
L._ . ._.~·-·-·-·I
..... _._.

o 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150

MILLIONS OF YEARS

Figure 2 Global oceanic crust production for the past 150 Myr averaged in 5-Myr bins, from Larson (1991). Shown is the
world total, plus the contributions from Pacific ridges, Gondwana (Atlantic and Indian) ridges, and oceanic plateaus. The interval
marked "no magnetic reversals" is the Cretaceous normal geomagnetic polarity superchron.

Figure 3 Plate houndary reconstruction near 83 Ma (magnetic anomaly 34) from Larson (1991) showing location of ocean
plateau basalts formed during the mid-Cretaceous plume pulse. Striped polygons represent surviving ocean plateaus with
125-80 Ma age; stippled polygons indicate possible locations of ocean plateaus since subducted. The two clusters of plateaus
suggest the sources were either two superplumes or two groups of smaller plumes. The absolute longitudes are unknown; tic
marks on horizontal axes are 90° apart in longitude.
6 Peter Olson

spot activity and the duration of geomagnetic The model is predicated on the assumptions that
polarity epochs. Larson and Olson (1991) have mantle convection is highly time dependent and
shown that the rate of crust production in flood that plumes are formed and destroyed repeatedly.
basalt and oceanic plateau provinces is inversely In cases where the plume survives for a long time,
correlated with the frequency of geomagnetic po- a linear hotspot track is formed on the overriding
larity reversals during the past 150 Myr. The Cre- plate. In other cases the plume is so short-lived
taceous plume pulse coincides very closely with that only an isolated thermal develops from the
the long Cretaceous normal polarity superchron, a hot thermal boundary layer, and a flood basalt or
40-Myr interval without field reversals. Since the ocean plateau is created, but no linear hotspot
Cretaceous superchron, the length of an average track.
polarity interval has decreased in step with hot- One problem with using this model to explain
spot activity. This correlation suggests that varia- the correlation between hotspot activity and geo-
tions in mantle plume activity somehow affect the magnetic reversal patterns is the uncertainty in
operation of the geodynamo. the temporal relationship between the increased
One theoretically plausible mechanism for heat flow at the base of the mantle and the pulse
long-term coupling of the core and mantle is in near-surface magmatism. The correlation dis-
variations in heat flow at the core-mantle bound- cussed above indicates very little time difference
ary. Thermal coupling is envisioned to act as fol- between the two phenomena, with changes in re-
lows. During periods of high plume activity, heat versal frequency synchronous with changes in
flow from the core to the mantle is high, and con- hotspot activity to within 5-10 Myr (Larson and
vection in the core is vigorous. Some models of Olson, 1991). This picture is quite different from
the geodynamo indicate that vigorous convection an earlier study by Courtillot and Besse (1987),
tends to stabilize the dynamo against frequent re- who considered only continental flood basalts and
versals (Olson and Hagee, 1990) by increasing the concluded that hotspots and polarity epoch length
strength and stability of the dipole field. These variations were 90° out of phase.
models predict an increase in polarity epoch length Plume dynamics do not require that the tem-
when core heat loss increases. During quiet pe- poral difference between basal heat flow and mag-
riods when plume activity is low, heat transfer at matic activity be small. It is even unclear which
the core-mantle boundary is low, and core con- of the two phenomena should lead in time. Mag-
vection is relatively weak. The same dynamo mod- matic activity and basal heat flow occur at dif-
els predict that the magnetic field is then prone to ferent ends of the plume and are governed by
frequent reversals. different physical processes. The onset time for
The concept of thermal coupling between core magmatism is related to the transit time of the
and mantle entails the following sequence of plume through the mantle. This time is difficult to
events. During quiet times, when plume activity is estimate, because the lower mantle viscosity is
low, the plume source region (assumed to be the poorly constrained, and also because the solid-
hot thermal boundary layer in the D" region at state phase changes in the transition are expected
the base of the mantle) grows by conduction. The to interrupt plume development (Liu et al., 1991;
quiet interval ends when thermal instabilities, orig- Weinstein, 1993). The onset of increased core heat
inating in the thermal boundary layer, develop flow represents the time required to substantially
into starting plumes. As the plumes rise through thin the thermal boundary layer in the D"-layer
the mantle, the thickness of the thermal boundary source region. This heat flow increase may be re-
layer is reduced in the region near the plume, and lated to the rise of the developing plume, or to a
the heat flow from the core to the mantle is in- general increase in mantle flow velocities. Since
creased proportionally. The core responds by in- the time scales of each of the processes may con-
creased vigor of convection, changes in the con- ceivably range from a few to several tens of mil-
vective pattern, or both, which alter the reversal lions of years, there is no theoretically based rea-
frequency. Arrival of the plume at the base of son to suppose they will occur in close temporal
the lithosphere results in massive amounts of de- proximity. This point was stressed recently by
compression melting, leading to formation of con- Loper (1992), who used a model of forced con-
tinental flood basalts or ocean plateau basalts. vection to argue that the D" layer would require at
 1. Mechanics of Flood Basalt Magmatism 7
least 100 Myr to respond to plume formation and Fowler, 1990a,b; Mutter et al., 1988; Farnetani
would prevent core heat flow from varying phase and Richards, 1992). All of these techniques in-
with plume-related magmatism. In view of this volve approximations and rely on assumptions of
uncertainty, it is important to establish the causal how the melt is transported relative to the solid
relationship between near-surface and deep-seated matrix. However, there is general agreement on
variations in thermal plumes. several basic points. First, nearly all melting takes
place by decompression, as rising material crosses
the local solidus. Second, partially molten rock is
Partial Melting in Thermal Plumes permeable even for very low melt fractions (po-
and Diapirs rosity), and therefore melt can migrate through the
mantle even when the melt fraction is very low, on
An idealized but conceptually useful model of a the order of 1% (McKenzie, 1985). At higher po-
developing mantle plume supposes a viscous fluid rosities, model studies indicate the melt percola-
layer initially at rest and at a uniform tempera- tion velocity is much greater than the solid matrix
ture, which is heated from below and cooled from velocity associated with mantle convection (Scott
above. Conductive thermal boundary layers de- and Stevenson, 1989), and therefore the processes
velop at the top and bottom of the fluid. At the of melt extraction and crustal growth can be re-
upper boundary a rigid but thermally conducting garded as instantaneous on the time scale of the
lid, representing the mechanically strong portion larger-scale convective motions .
of the lithosphere, caps the fluid layer. Instabilities . The specific problem of melt production in
first develop in the lower thermal boundary layer mantle plumes has been considered previously
and grow into thermal plumes. At the leading edge by Campbell and Griffiths (1990), Watson and
of each plume is a large thermal diapir. If the plume McKenzie (1991), and Arndt and Christensen
is continuous in time, a narrow conduit is estab- (1992). Watson and McKenzie applied relations
lished behind the diapir (Olson et al., 1987; Grif- for pressure-release melting of dry peridotite de-
fiths and Campbell, 1990). As the plume ascends, rived by McKenzie and Bickle (1988) to steady-
its highest temperature portions begin melting state calculations of an axisymmetric plume be-
where the plume material intersects the melting neath a conducting lid to model the Hawaiian
curve. As the in situ melt fraction increases, the hotspot. They determined the combination of tem-
velocity of melt percolation through the mantle perature in the core of the plume, lid thickness,
approaches and exceeds the transport velocity and asthenosphere viscosity that gave the best fit
of the plume. The melt then separates from the to the present-day crust production rate, geoid,
plume and is eventually incorporated into the over- and swell height for Hawaii. They also showed, in
lying lid as added crust. Because the rate of melt the context of their steady-state model, how litho-
production is particularly large in the leading dia- sphere thickness limits the rate of melt production
pir, a starting plume produces a very large initial by limiting access to the zone where decompres-
magmatic pulse. sion melting can occur. The calculations by Arndt
A variation of this model consists of a solitary and Christensen (1992) focused on magma com-
diapir thermal rising from depth in the mantle, positions and specifically sought to determine the
without a trailing conduit. This model is intended fractions of magma produced by melting within
to simulate a transient instability originating from the plume and by remelting of the lithosphere.
the breakup of a lower mantle plume by transition They concluded that at most a few percent of the
zone phase changes (Liu et al., 1991). The soli- total magma is derived from remelted lithosphere;
tary diapir is a useful simplification for numeri- nearly all of the melt comes from the plume itself.
cal experimentation because the only important The study by Campbell and Griffiths applied melt-
source properties are the initial radius, depth, and ing relations to a highly idealized analytical model
temperature of the diapir. of a rising thermal diapir to estimate the magma
Numerous techniques have been used for incor- volumes produced in a plume head. Uniform vis-
porating partial melting in models of mantle con- cosity plumes were assumed in the calculations
vection (Scott and Stevenson, 1989; Watson and by Campbell and Griffiths (1990) and Watson
McKenzie, 1991; Arndt and Christensen, 1992; and McKenzie (1991), whereas temperature-
8 Peter Olson

dependent viscosity was used in the calculations in an infinite Prandtl number, uniform-viscosity
by Arndt and Christensen (1992). In all of these Boussinesq fluid can be written in cylindrical
studies, the melt production is calculated using (r, z) coordinates as
parameterizations of experimentally derived melt-
ing relations for dry peridotites. These relations
(
V2 _ ~i.)2l/J
r iir
= _ rpg
TJ
(a ar + (3 a<fJ)
(J[
ar
(1)
were obtained from batch melting experiments,
whereas the condition assumed in the models, and
with melt continuously extracted from the matrix,
aH
is in fact closer to a fractional (Rayleigh) melting - + u . V H = kV2T. (2)
process. In addition, the effects of latent heat upon at
melting and the additional buoyancy due to melt Here, l/J is the streamfunction of the motion; T is
within the plume were ignored. temperature; <fJ is the in situ melt fraction; u is the
The approach used here differs from these oth- velocity of the matrix material; R is gravity; p is
ers in several important ways. First, a simple uni- reference density; TJ is the viscosity; a and {3 are
variant melting formula with a constant pressure thermal and melting expansion coefficients, re-
derivative is used. Second, the effects of latent spectively; k is thermal conductivity; and V is the
heat in melting and melt buoyancy are included in gradient operator. The energy equation (2) is writ-
the calculation in a self-consistent manner. I find ten using an enthalpy variable H, which in this
that latent heat absorption is important in the en- context is the sum of latent and sensible, heat,
ergy balance of the plume and affects both the
thermal structure and the extent of melting. Also, (3)
I assume in this model that the melt is trans- where C p is specific heat at constant pressure, and
ported with the matrix at low melt fractions and L is the latent heat of melting. The streamfunction
escapes from the mantle when it exceeds a criti- is related to the matrix velocity u = (ru, zw) by
cal value. In these calculations, percolation is ne-
glected when the melt fraction <fJ is less than a tu, w) = ( - ! al/J, ! al/J). (4)
saturation value <fJ" and is assumed to be instan- r ilz r ilr
taneous when <fJ > <fJ,. Calculations of mantle
Melting and solidification are specified as fol-
structure and melt production based on this method
lows. I assume a simple, univariant melting law in
are presented for both solitary thermal diapirs and
which the melting temperature varies with depth
continuous thermal plumes.
according to
In all the numerical models of melting in
thermal plumes, including the calculations pre- (5)
sented here, the effects of plate motion on plume
where Tm is the melting temperature, 'Y is its depth
structure are not explicitly considered. Constant-
derivative due to hydrostatic pressure, y = 20 - 2
velocity plate motion transforms an otherwise
is depth below the top surface 2 = 20, and the
axisymmetric hotspot into a bilaterally symmet-
subscript 0 denotes the melting temperature at
ric hotspot track, elongated in the direction of
surface pressure. The dry peridotite solidus of
plate motion (Olson, 1990). The resulting plume
Takahashi and Kushiro (1983), corrected for the
structure is then fully three-dimensional. It is be,
adiabatic gradient of the mantle, can be closely
yond the scope of this study to incorporate fully
approximated using (5) with TmO = 1100° C and
three-dimensional effects, although this will soon
'Y = 4°C/km.
be possible to do because numerical models of
It is convenient to introduce two reference en-
plumes with moving plates are now being devel-
thalpy functions. The first corresponds to sensible
oped (Ribe and Christensen, 1992).
heat at the melting temperature,
(6)
The Numerical Model and the second corresponds to enthalpy at the
saturation value of the in situ melt fraction, <fJ"
The conservation equations for vorticity and en-
ergy for axially symmetric thermal convection H, = H m + pL<fJ,. (7)
 1. Mechanics of Flood Basalt Magmatism 9
The subscript s refers to saturation conditions. In thermal conductivity as the underlying viscous
addition, I define a progress variable X to denote mantle. The sidewall r = R and the centerline
the volume of melt extracted from a fixed-volume r = 0 are reflecting boundaries (zero heat flow,
element of the mantle and added to the crust. As- impermeable, and zero shear stress).
suming the extracted melt moves vertically, the Initial conditions in both sets of calculations
additional crustal thickness from melt extracted include a cold thermal boundary layer at the sur-
from the mantle is given in terms of X by face and isothermal interior at temperature T,. The
upper thermal boundary layer, representing the
(to thermal lithosphere, is specified by an error func-
he = Jo X dy. (8)
tion diffusion profile with an age 7 o- In the solitary
diapir calculations, an initially spherical thermal
According to this prescription, three states are
anomaly with temperature T D is introduced near
possible: subsolidus, partially molten without melt
the base of the cylinder. The diapir position is
percolation, and partially molten with percolation.
fixed for a short time (typically a few million
Melt fraction, temperature, and the progress vari-
years), allowing it to develop a diffusive thermal
able can be determined from H in each of these
halo, and is then released. In the continuous plume
three states using the following relations:
calculations, a hot basal thermal boundary layer is
1. Subsolidus (H < H m):
included as an initial condition, specified by a
aX H basal temperature T 8 and an error function diffu-
</J =at = 0; T=-.
ic,
(9) sion profile with an age 7 8, A small, long wave-
length perturbation is added to the basal thermal
2. Partially molten, no percolation (H m < boundary layer, in order to initiate a rising in-
H<H,): stability on the centerline.
Equations (1)-(11) have been solved for both
H - Hm solitary diapirs and for continuous plumes using
pL finite difference methods. The enthalpy in Eq. (2)
T = Tm ; (10) is advanced in time using an explicit method with
aX = o. the upstream differences for calculating the ad-
at vection terms. Temperature and porosity are up-
dated at each time step according to the criteria
3. Partially molten, percolation (H > H,):
expressed in Eqs. (9)-(11), and the streamfunc-
</J = </J,; tion is updated by solving (1) as two modified
T = Tm ; (1 I) Poisson equations.
aX = a (H - H,) In these calculations the flow is restricted to a
at at pL cylinder 750 km in depth, representing only the
upper mantle. This restriction is not intended to
We divide the cylinder into a uniformly vis- mean that this model applies only to plumes origi-
cous mantle 0 < Z < ZL and a rigid mechanical lid nating in the upper mantle. I found that a grid size
h < Z < zo, where ZL is the mantle-lid bound- of 10 km or finer is necessary to resolve the ther-
ary. Equations (1)-(11) are applied to the viscous mal structure and the distribution of melt within
mantle; only the heat conduction version of (2) is the plume, and this requirement made calculations
applied in the lid. of the whole mantle depth prohibitively expen-
The boundary conditions are as follows. The sive. Nevertheless, the general behavior of these
bottom surface Z = 0 is isothermal, impermeable, calculations provides some insight useful for in-
and free-slip (zero shear stress), conditions which terpreting the evolution of flood basalts, regard-
permit the plume to draw material from the en- less of the depth of origin of the plumes.
tire hot thermal boundary layer. The top surface In the solitary diapir calculations, the lid thick-
Z = Zo is isothermal, and the base of the lid ness and the initial excess temperature of the dia-
Z = ZL is no-slip (zero velocity). The rigid lid sim- pir were varied. In the continuous plume calcula-
ulating the mechanically strong, elastic portion of tions, only the lid thickness was varied. Values of
the lithosphere has zero velocity and the same the physical parameters common to the calcula-
10 Peter Olson

Table 3
Thermophysical Properties and Model Parameters

Parameter Model value Reference

Cylinder depth, z 0 750km

Cylinder radius, R 1000km
Grid spacing, ~z, ~r IOkm
Interior temperature, TI 1300°C Jeanloz and Morris (1986)
Surface temperature, To DoC
Surface melting temperature, Tmo 1l00°C Takahashi and Kushiro (1983)
Melting point gradient, Y 4°C'km- 1
Takahashi and Kushiro (1983)
Latent heat of melting, L 400kJ'kg- 1 Stebbins et al. (1983)
Saturation melt content, f/J, 0.01 McKenzie (1985)
Specific heat, C p IkJ·kg-l.oC-1 Turcotte and Schubert (1982)
Mantle density, P 3.3 mg m i 1 Anderson (1989)
Crust density, p , 2.9 mg -rn ? Anderson (1989)
Thermal expansion, a 3XIO-5.oC-I Anderson (1989)
Melting expansion, f3 0.12 Turcotte and Schubert (1982)
Gravity, g 9.8 m- S-2 Anderson (1989)
Thermal conductivity, k 3.3 W . m - I • °C I Horai and Simmons (1970)
Viscosity. n 0.5XI02IPa·s Nakada and Lambeck (1989)

tions are listed in Table 3. The diapir calculations in crustal thickness from melt extracted vertically
were typically run for 15-20 Myr, and the plume from the diapir).
calculations were typically run for 42 Myr. In ad- The first stage in the evolution of the diapir is
dition to the temperature, velocity, and in situ melt represented in Fig. 4 by the image at t = 5.2 Myr
fraction, the total crust production and production (top left) and corresponds to the time when the
rate, surface geoid, dynamical surface uplift (due rising diapir first undergoes decompression partial
to normal stress acting on the lid), and isostatic melting, but has not yet flattened out beneath the
uplift (due to crustal thickening) were calculated. lid. This stage is characterized by partial melting
to 200 km in depth, a large Gaussian-shaped dy-
namically supported uplift approximately 3 km in
Melting in Solitary Diapirs amplitude, and correlated with the topography, a
positive geoid anomaly of nearly 70 m in ampli-
Figures 4-6 show a typical result of calculations tude. While the melt content of the diapir is large
on ascent and partial melting in an isolated ther- at this stage, very little melt has been extracted
mal diapir with an initial 200-km radius. and added to the crust. The second stage in the
Figure 4 shows the distribution of temperature evolution is represented in Fig. 4 at 9.1 Myr (top
and melt concentration, along with the distribu- right). Here the diapir has begun to collapse be-
tion of surface observables at various times in the neath the lid, forming a lens of partial melt ap-
evolution of a diapir rising beneath a 40-km-thick proximately 600 km in diameter and centered near
lid and 50-Myr thermal lithosphere. The surface 110 km in depth. Magma extraction and crustal
observables plotted are the geoid anomaly, the dy- formation are most rapid at this stage, as indicated
namical topography (the topography induced by by the large difference between the total topogra-
vertical normal stress on the lid), and the total phy and the dynamically supported topography.
topography (the sum of dynamical topography Due to the lateral spreading of the diapir, both the
and the isostatic topography due to an increase dynamic topography and the geoid subside rap-
 8
1,5 ]'
80~ "....-.,...... _' ~.
6
4
'--"

;8 ~~. -- p
----- 60 1.0 >-
""~ '2
.::.. 0.5
:r:
9:e

:;L--
0_. _ o .J
~.
:-. 0:::
CJ -20 -2
-4 "............
] 0.0 g
-40 /,-,/
"
---.---- -0.5 g;
OpT-.- _' qO I-

-_&
~_:

-----
E 200
.::<.
200
°t @~~ ~---v===:::.0
~O
200
200
f- 400 400
400
0...
Ld MELT FRACTION 400
[) MELT FRACTION
600 .01 600
600 .01 600
.001
_001
-1000 -500 o 500 1000
-500 o 500 1000
-1000

o
E:
'--"

(5
40 ~-

7: r_._ ~ 'i-;i:
" .. ,~~

_.~, ,

....... ...;:.,
1:
o
15
10
5
/ ~./::.,-----_.~ -.".~., .r>:
1.5
1.0
0.5
-----

~
~
'--"
0...
w ~-.#'"
o ,.
CJ
-20 f=- -2 -5
-10
......
'.' . .- ...... /
,./
.. -..,,-./ /
0_0
-0.5
-1.0
Ci.
g
g;
01 iii jiii ~----lO I-

o Pi i:
Ci\\,'; __ , --------- ~i~; ~0
2------ 200~ 200
200
,... ......,. ... :r

200
400
~~ 400l-"
~,~ MELT FRACTION 400 400
() 600 .01 600 MELT FRACTION
• _,oql • L ~ ~ _ _~ 600 .01 600
.001
-1000 -500 o 500 1000
RANGE (km) -1000 -500 o 500 1000
RANGE (krn)
Figure 4 Partial melting of a solitary axisymmetric thermal rising beneath a stationary 40·km·thick rigid lid. The initial diapir temperature and radius are 1800°C
and 200 km, respectively. Other parameters used in the calculation are given in Table 3. The times of each image in million years after release of the diapir are as
follows: top left, 5.2; bottom left, 7.1; top right, 9.1; bottom right, 13.1. The upper plot in each image shows the distribution of geoid height (solid curve), dynamic
topography (dashed curve), and total topography (dotted curve). The lower plot shows the distribution of temperature contours and ill situ melt fractions (shaded)
Lithosphere contours (solid) are 0-1200° C in 200° C intervals; diapir contours (dashed) increase from 1400°C in 100° C intervals. The images are reflected about
the symmetry axis.
 12 Peter Olson

120 ~ 6
..,
~
0.8 .~~~

- ~.

. _ _ _Crust
100 ~ ~ E
.Y
E 0.6
E 80::-
1--<"
4'::="
c-,
a
'-'
~
o
60 t' s:
D-
o
'-
Q)

E
0.4
Q) ~ :::J
o 40 2 g'
,~ - - Tapa D-
o
0
> 0.2
20 ~ ---·~yn Tapa
f-
==
Q)
Diapir
Release Melt
..... . ... G~e~ai~d ~_ 0 :::;: .j.
O'~~ 0.0
o 5 10 15 20 0 5 10 15 20
Time (Myr) Time (lAyr)
Figure 5 Time series of surface observables from the calcu- Figure 6 Time series of mantle melt volume and crust vol-
lations in Figure 4. Shown are the time series of axial (r = 0) ume from the calculations in Figure 4.
geoid height, axial topography. and axial dynamic topography.

idly, and the total topography rapidly decreases gradual refreezing of the melt remaining in the
in elevation during this stage, despite the rapid mantle.
addition of crustal material. Surprisingly, the pe-
riod of most active crust building coincides with
rapid subsidence, particularly in the central region Melting in Continuous Plumes
above the diapir, Note that most of the melt is de-
rived from portions of the diapir where the excess I have also made a series of calculations of par-
0
temperature is 200 C or less, and virtually all of tial melting in plumes initiated by the instability
the melt is derived from portions with less than of a hot thermal boundary layer. The presence of
0
300 C excess temperature. a basal thermal boundary layer in axisymmetric
The third stage in the evolution is represented geometry ensures the existence of a continuous
in Fig. 4 by times later than 13 Myr (bottom right plume and provides a simple way to model the
column). This final stage is characterized by the long-term behavior of the hotspot, as well as its
cooling and resolidification of the partially molten initial development. In the calculations summa-
diapir within the mantle, slow subsidence at the rized in Table 4, the basal temperature is fixed at
surface, and greatly diminished rates of crustal ad- either 2000 or 18000 C, and the thickness of the
dition. The lateral spread of the diapir is checked mechanical lid is varied between 0 and 100 km.
by the formation of a ring-shaped instability in the The values of all the other parameters are given in
cold thermal boundary layer, formed from the ma- Table 3.
terial displaced outward when the diapir first be- Figure 7 shows the distribution of temperature
gan to collapse. The motion induced by the ring and melt concentration, along with the distribu-
instability destroys the thermal anomaly of the
diapir on a faster time scale than that if the diapir Table 4
simply stagnated beneath the lid and cooled only Results of Plume Calculations
by conduction. Within 15 Myr the diapir is essen-
tially gone, and the dynamics are dominated by Crustal Crustal Crustal
the sinking of the ring instability. The three stages volume production production Heat flow
are also seen in the time series of axial topogra- Lid at 20 Myr peak at 40 Myr at 40 Myr
(km) (10' km ') (km'/yr) (km'/yr) (GW)
phy, geoid, mantle melt content, and added crust
volume and volcanic heat, as shown in Figs. 5 0 11.5 un 0.019 560
and 6 for the calculations in Fig. 4. The first stage 10 6.11 0.81 0.067 318
is marked by rapid (1-2 Myr in duration) uplift 20 4.31 0.55 0.031 221
and occurs as the diapir approaches the lid. The 50 2.47 0.306 0.003 215
second stage is the main magmatic episode and is
75 1.29 0.135 0.0046 201
marked by rapid subsidence. The last stage in-
100 0.45 0.048 0.004 200
cludes slow subsidence, little crustal addition, and
 r-
s:
.,'"e-
I")

:l
;;-
'"
g,
...,
8
Co
.,
~

'"
~
.,:::
IJC
Figure 7 Partial melting of a thermal plume rising beneath a stationary 20-km-thick rigid lid. The basal temperature is 2000° C. Other 3
parameters used in the calculation are given in Table 3. The times of each image in million years after plume release started by introducing ~.
a long-wavelength perturbation in the basal thermal boundary layer are as follows: top left, 6.0; bollom left, 12.1; top right, 18.0; bottom 3
right. 36. The upper plot in each image shows the dynamically supported topography (solid curve) and the total topography, which includes
the isostatic crustal component (dotted curve). The lower plot in each image shows the distribution of temperature contours and the ill situ
melt content (shaded). The lithosphere contours (solid) are O-1200°C in 200°C intervals; plume contours (dashed) increase from 1400°C ......
in 100°C intervals. The images are reflected about the symmetry axis. ....,
14 Peter Olson

tion of surface topography at various times in Beyond 40 Myr the plume structure is essentially
the evolution of a thermal plume originating from as shown in the last image of Fig. 7, although the
0
a 2000 C lower boundary, rising beneath a 20- calculation never reaches a true steady state. Small
km-thick mechanical lid. Both the dynamical to- fluctuations in plume strength and melt produc-
pography and the total topography (including iso- tion continue to occur, but the overall structure of
static crustal topography) are shown above each the flow and the temperature fields do not change.
cross section. The image at 6 Myr (top left) is The images in Fig. 7 show the melt zone ex-
equivalent to the first stage in the evolution of the tends from 50 to 200 km in depth and is centered
solitary diapir shown in Fig. 4. The plume head near 100 km in depth, where the excess tempera-
has begun to spread out beneath the mechanical ture at 100 km is 300 0 C at all stages. This high-
lid, and the melt content of the plume is high. The temperature core of the plume, with as much as
region occupied by partial melt is 1200 km in di- 600 0 C excess temperature, extends upward from
ameter and extends from 50 to 200 km in depth. the hot boundary layer into the partially molten
The surface uplift has reached its maximum, ap- lens. However, most of the plume head and most
proximately 4 km at the point directly above the of the melt lens have excess temperatures rang-
0
plume centerline. Only a small amount of melt has ing from 100 to 350 C, generally consistent with
been added to the crust at this time. The plume at excess temperatures deduced from hotspot mag-
12 Myr (Fig. 7, lower left) is in the second stage, mas and topographic rises (Sleep, 1990; Schilling,
characterized by lateral spreading beneath the lid, 1991). The role of latent heat absorption is evident
peak melt concentrations near 100 km in depth, in Fig. 7, where the isotherms in the melt lens are
peak rates of crust production, and rapid subsi- horizontal and evenly spaced at the melting point
dence due to loss of the dynamically supported to- gradient. Without latent heat absorption in the cal-
pography. At this time, the partially molten zone culation, the high-temperature plume core would
is 1600 km in diameter, and the topography above extend closer to the surface, so the "excess tem-
the plume axis approaches 4 km. The images at perature" and the melt volume that would be in-
18 and 36 Myr (Fig. 7, right column) show the ferred for such a plume would be much greater.
evolution of the plume from its initial structure, Accordingly, latent heat is a necessary element in
dominated by the head, to the final conduit struc- modeling mantle plume structure.
ture, dominated by a thin cylinder-shaped thermal Figure 8 compares the time series of basal heat
anomaly and a restricted zone of partial melting. flow and crust production from the calculation in
The zone of partial melting shrinks to a 200-km- Fig. 7. The main pulse of crust production lasts
diameter lens centered near 100 km in depth. about 15 Myr. At the peak of activity, the rate of
The dynamic topography is reduced to a long- production is 0.55 krnvyr. In comparison, the pro-
wavelength Gaussian-shaped swell with only a duction rate at 40 Myr, typical of the long-term
1.0-km elevation at 36 Myr (Fig. 7, lower right). rate, is only about 0.03 kmv/yr, nearly a factor
of 20 lower. This reduction factor is typical for
hotspot tracks originating from continental flood
0.6 . basalts (Richards et al., 1989). The long-term
average basal heat flow of 220 GW is compa-
rable to Sleep's (1990) estimate for the present-
day strength of the Hawaiian hotspot. During the
period of highest activity, the basal heat flow is
about 60% greater than the long-term average
value.
Vl
:J Figure 8 demonstrates that the variations in
0.0 ~
o 10 20 30 40 50
basal heat flow occur in association with the vari-
Time After Plume Initiation (Myr) ations in crustal production, although the two
time series are not closely in phase over most of
Figure 8 Time series from the calculation in Figure 7. Basal
heat is the total heat flow into the base of the cylinder, sup-
the calculation. Initially, the flow associated with
porting the plume. Crustal production is the rate of melt ex- the accelerating plume entrains a large part of the
tracted from the plume. thermal boundary layer. This entrainment accounts
Figure 9 Partial melting of a thermal plume rising beneath a stationary 75-km-thick rigid lid. The basal temperature is 2000°C. Other
parameters used in the calculation are given in Table 3. The times of each image in million years after plume release started by introducing a
long-wavelength perturbation in the basal thermal boundary layer are as follows: top left, 6; bottom left, 12.1; top right, 18; bottom right, 36.
The upper plot in each image shows the dynamically supported topography (solid curve) and the total topography, which includes the isostatic
crustal component (dotted curve). The lower plot in each image shows the distribution of temperature contours and in situ melt content
(shaded). The lithosphere contours (solid) are 0-1200° C in 200° C intervals; plume contours (dashed) increase from 1400° C in 100° C inter-
vals. The images are reflected about the symmetry axis.
 16 Peter Olson

for the initial sharp increase in basal heat flow crust in the main pulse (before 20 Myr), which is
near t = 5 Myr, which is well correlated with the representative of the continental flood basalt vol-
steep rise in magmatic activity. However, the cor- umes given in Table I. The peak production rate
relation between the two signals degrades after is about 0.135 kmvyr, the long-term average rate
the magmatic peak. As magmatic activity drops is about 0.005 kmVyr, and the long-term average
sharply after 7 Myr, basal heat flow continues to plume strength is about 200 GW, about 10%
slowly increase, reaching a maximum near 18 Myr. weaker than the plume in Fig. 7. As in Fig. 8, the
Inspection of Fig. 7 reveals that the heat flow time series in Fig. 10 show the basal heat flow and
maximum occurs when cold sinking material dis- rate of crustal production increase nearly simulta-
placed from the top thermal boundary layer ap- neously, but the heat flow reaches a maximum
proaches the basal thermal boundary layer. In this about 12 Myr after the peak in magmatic activity,
calculation the heat flow maximum lags behind when the material displaced from the cold thermal
the peak in magmatic activity by approximately boundary layer by the plume head reaches the
the time needed for the material displaced by the basal boundary layer.
plume head to sink to the hot thermal boundary Table 4 and Fig. 11 show the total volume
layer. More generally, the peak heat flow corre- added to the crust from several thermal plume
sponds to the time when the large-scale circula- calculations for various lid thicknesses and basal
tion is most developed. temperatures. As expected, increasing the basal
In the mantle, the largest scale of the circulation temperature increases the amount of melt pro-
is probably controlled more by subduction than by duced. More importantly for the plume model,
plumes. Consequently, periods of increased core however, is the fact that the volume of melt pro-
heat loss are likely to be more causally related to duced is quite sensitive to the lid thickness. As in-
increases in sea floor spreading rates than hotspot dicated in Fig. 11, crustal volumes typical of con-
activity. The time series in Fig. 2, which show that tinental flood basalts can easily be produced from
sea floor spreading was also high during the Cre- a plume rising beneath a 40- to 80-km-thick me-
taceous geomagnetic polarity superchron, are con- chanicallid. The same plume produces nearly 10
sistent with this interpretation. times as much melt when it ascends beneath very
Figures 9 and 10 show the results of a calcula- young lithosphere, where the mechanically strong
tion similar to Figs. 7 and 8, but with an 75-km- lid is only 10-20 km thick. Lid thickness strongly
thick lid, representative of a plume rising beneath controls the amount of melt produced in a plume
stable continental lithosphere. The same sequence by controlling the depth at which the leading dia-
of events found in the thin lid case is present in pir spreads out. Melting occurs only in that part of
this case, with the major difference being slower the plume which crosses the solidus before reach-
rates of crustal generation when the lid is thick. ing the lid. The thicker the lid, the smaller is the
This calculation produced 1.29 X IOfi km ' of portion of the diapir that enters the melting zone,

r---
'"E 100 . 00' - . - T - - ~ - -~ ~- - ~-
• T.-2000'C
- - -- •
"1
-~~~0.20~
.x I
.T.=1800'C
: ;;:;-..
<D
0 ~ ••
10.00 ..
0.15.§
(])
.• ..
..
"-'"
~
'" •
1.00 r
o : c: E
'.
i 02 II:~___ Basal_l0.10~
::J
0
>
0.10 ! -"
•
•

~
o
Vl
~0.05]
~ ~
-
0
Ul
::J
L 0.01 ~ ............................. ~.

0.0 1..•..~~. Crus! 0 00 u U 0 20 40 60 80 100 120

o 10 20 30 40 50 Lid Thickness (km)
Time After Plume Initiation (Myr)
Figure 11 Total crustal volume produced in 40 Myr after ini-
Figure 10 Time series from the calculations in Figure 9. tiation by partial melting in a plume rising beneath rigid lids
Basal heat is the total heat flow into the base of the cylinder, of various thickness. The basal temperature is either 2000 or
supporting the plume. Crustal production is the rate of melt 1800°C, as indicated by symbols. Other parameters used in the
extracted from the plume. calculations are given in Table 3.
 1. Mechanics of Flood Basalt Magmatism 17
and the smaller is the portion that melts. The large Arndt, N. T.. and Christensen, U. R. (1992). The role of litho-
spheric mantle in continental flood basalt volcanism: Ther-
difference in the volume of the major oceanic pla-
mal and chemical constraints. J. Geophys. Res. 97,10967-
teaus versus the major continental flood basalts 10981.
can be explained, in the context of the plume Carlson. R. W. (1991). Physical and chemical evidence on the
model, in terms of differences in the portions of cause and source characteristics of flood basalt volcanism.
plumes that undergo partial melting beneath thin Austral. J. Earth Sci. 38, 525-544.
Campbell. I. H.. and Griffiths, R. W. (1990). Implications of
versus thick lithosphere.
mantle plume structure for the evolution of flood basalts,
Earth Planet. Sci. Lett. 99, 79-93.
Courtillot, v.. and Besse, J. (1987). Magnetic field reversals,
Conclusions polar wander. and core-mantle coupling, Science 237,
1140-1147.
Calculations of partial melting in thermal plumes Davies. G. F. (1988). Ocean bathymetry and mantle convec-
and diapirs indicate the following general proper- tion. I. Large-scale flow and hotspots. J. Geophys. Res. 93,
10467-10480.
ties for mantle plumes:
Duncan, R. A.. and Richards. \1. A. (199\). Hotspots, mantle
1. There are three stages in the magmatism plumes, flood basalts. and true polar wander, Rev. Geo-
of a mantle plume. The first stage occurs as the phys. 29, 31-50.
plume approaches the lithosphere from below and Farnetani, C. G.• and Richards, M. A. (1992). Numerical mod-
is characterized by rapid uplift, rapid melt genera- els of oceanic flood basalt events, EOS Trans. Amer. Geo-
phys. Union 73, 533.
tion, but relatively little crust production. The sec-
Fowler, A. C. (1990a). A compaction model for melt transport
ond stage occurs as the plume head spreads out . in the Earth's asthenosphere. Part I. The basic model, in
beneath the lithosphere, and is marked by rapid "Magma Transport and Storage" (M. P. Ryan, ed.), Wiley,
subsidence and rapid crust production. The last Chichester/Sussex, England.
stage corresponds to the contraction of the plume Fowler, A. C. (I 990b). A compaction model for melt trans-
port in the Earth" asthenosphere. Part II. Applications, in
and the melt zone toward a steady state and is
"Magma Transport and Storage" (M. P. Ryan, ed.), Wiley.
marked by slow subsidence and greatly dimin- Chichester/Sussex, England.
ished crustal production. Griffiths. R. Woo and Campbell, I. H. (1990). Stirring and struc-
2. The amount of melt produced in a plume is ture in mantle starting plumes. Earth Planet. Sci. Lett. 99,
very sensitive to the thickness of the mechanical 66-78.
Horai, K. and Simmons, G. (1970). An empirical relationship
lithosphere. The large difference between conti-
between thermal conductivity and Debye temperature for
nental flood basalt and ocean plateau volumes is silicates. J. Geophys. Res. 75, 678-082.
probably due to the differences in lithosphere Jeanloz, R., and Morris. S. (I 9Xti). Temperature distribution in
thickness beneath continents and young ocean the crust and mantle, Annu. Rev. Earth Planet. Sci. 14,
crust. 377-415.
Larson, R. L. (l99\). The latest pulse of the Earth: Evidence
3. Basal heat flow increases during plume for-
for a mid-Cretaceous superplume, Geology 19,547-550.
mation, supporting the suggestion that deep mantle Larson, R. L., and Olson, P. (I 99\). Mantle plumes control
plumes thermally couple the core to the mantle. magnetic reversal frequency. Earth Planet. Sci. Lett. 107,
However, the time of maximum heat flow is gov- 437-447.
erned by the sinking of cold material displaced by Liu, Moo Yuen, D. A., and Honda, S. (1991). Development of
diapiric structures in the upper mantle due to phase transi-
the plume head, and generally lags behind the
tions, Science 252, I83ti-1 X39.
peak in magmatic activity. Loper. D. C. (1992). On the correlation between mantle plume
flux and the frequency of reversals of the geomagnetic
field, Gcophys. Res. Lett. 19,25-29.
Acknowledgments Mahoney, J. J. (1987). An isotopic survey of Pacific, oceanic
plateaus: Implications for their nature and origin, in "Sea-
The author thanks D. Bercovici. U. Christensen. S. Watson, mounts. Islands and Atolls" (B. Keating et al.. eds.), AGU
M. Ryan. and J. Mahoney for helpful comments. This work Monograph 43. pp. 207 -220. Washington D.C.
was supported by NSF Grant EAR 8916152. Mahoney, J. J., and Spencer. K. J. (199 I). Isotopic evidence for
the origin of the Manihiki and Ontong-Java oceanic pla-
teaus, Earth Planet. Sci. Lett. 104,196-210.
McKenzie, D. P. (1985). The extraction of melt from the crust
References and mantle, Earth Planet. Sci. Lett. 74, 81-91.
McKenz.ie, D. P., and Bickle, \1. J. (1988). The volume and
Anderson. D. L. (1989). "Theory of the Earth," Chap. 3, composition of melt generated by extension of the litho-
Blackwell, Oxford. sphere, J. Petrol. 29, ti25-n80.
18 Peter Olson

Morgan, W. 1. (198 l ), Hotspot tracks and the opening of Scott, D. R., and Stevenson, D. 1. (1989). A self-consistent
the Atlantic and Indian oceans, in "The Sea," Vol. 7, model of melting, magma migration and buoyancy-driven
pp.443-487. circulation beneath mid-ocean ridges, J. Geophys. Res. 94,
Mutter, 1. c., Buck, W. R., and Zehnder, C. M. (1988). Con- 2973-2988.
vective partial melting. I. A model for formation of thick Sleep, N. A (1990). Hotspots and mantle plumes: Some phe-
basaltic sequences during the initiation of spreading, 1. nomenology, J. Geophys. Res. 95, 6715-6736.
Geophys. Res. 93,1031-1048. Stebbins, J. E, Carmichael, I. S. E., and Weill, D. E. (1983).
Nakada, M., and Lambeck, K. (1989). Late Pleistocene and The high-temperature liquid and glass heat contents and
Holocene sea-level change in the Australian region and heats of fusion of diopside, albite, sanidene and nepheline,
mantle rheology, Geophys.J./ntl. 96, 497-517. Amer. Mineral. 68, 717 - 730.
Olson, P. (1990). Hotspots, swells and mantle plumes, in Takahashi, E., and Kushiro, I. (1983). Melting of a dry peri-
"Magma Transport and Storage" (M. P. Ryan, ed.), Wiley, dotite at high pressures and basalt magma genesis, Amer.
Chichester/Sussex, England. Mineral. 68, 859-879.
Olson, P., and Hagee, V. L. (1990). Geomagnetic polarity re- Tarduno, 1. A., Sliter, W. v., Kroenke, L., Leckie, M., Mayer,
versals transition field structure and convection in the H., Mahoney, 1.1., Musgrave, R., Storey, M., and Winterer,
outer core, J. Geophys. Res. 95, 4609-4620. E. L. (1991). Rapid formation of Ontong-Java plateau by
Olson, P., Schubert, G., and Anderson, C. (1987). Plume for- aptian mantle plume volcanism, Science 254, 399-403.
mation in the D"-layer and the roughness of the core- Turcotte, D. L., and Schubert, G. (1982). Applications of con-
mantle boundary, Nature 327,409-413. tinuum physics to geological problems, in "Geodynam-
Ribe, N., and Christensen, U. R. (1992). A 3D dynamical ics," Wiley, New York.
model of the Hawaiian plume, EOS Trans. Amer. Geophys. Watson, S., and McKenzie, D. (1991). Melt generation by
Union. 578. plumes: A study of Hawaiian volcanism, 1. Petrol. 32,
Richards, M. A, Duncan, R. A, and Courtillot, V. E. (1989). 503-537.
Flood basalts and hotspot tracks: Plume heads and tails, Weinstein, S. A. (1993). Catastrophic overturn of the Earth's
Science 246,103-107. mantle driven by multiple phase changes and internal heat
Ryan, M. P. (1990). The physical nature of the Icelandic magma generation, Geophys. Res. Lett. 20,101-104.
transport system, in "Magma Transport and Storage" White, R., and McKenzie, D. (1989). Magmatism at rift zones:
(M. P. Ryan, ed.), Wiley, Chichester/Sussex, England. The generation of volcanic continental margins and flood
Schilling, 1.-G. (1991). Fluxes and excess temperatures of basalts, J. Geophys. Res. 94, 7685-7729.
mantle plumes inferred from their interaction with migrat- Zhang, Y.-S., and Tanimoto, T. (1992). Ridges, hotspots and
ing mid-ocean ridges, Nature 352, 397-403. their interaction as observed in seismic velocity maps, Na-
Schubert, G., and Sandwell, D. (1989). Crustal volumes of the ture 355, 45 -49.
continents and of oceanic and continental submarine pla-
teaus, Earth Planet. Sci. Lett. 92, 234-246.
 Chapter 2 Melt Migration and Related Attenuation
in Equilibrated Partial Melts

Tye T. Gribb, Shanyong Zhang, and Reid F. Cooper

Overview Unit;

E" activationenthalpyfor steady- kJ . mol lor eV

Anelasticity associated with the migration of the melt state flow
phase in synthetic (Co-Mg) olivine-basalt partial melts
L length betweenouter load points m
is examined experimentally. Newtonian-viscous, textur- of flexure specimen
ally (quasi)equilibrated partial-melt aggregates are sub-
P load N
jected to four-point flexural loading (U 1MAX = 3.45 MPa)
at elevated temperature (1070-1200° C); the creep re- QE 1
attenuationin Young's modulus dimensionless
sponse is characterized by a substantial, decelerating mode
transient that would be absent but for the presence of R. stiffness(modulus) Pa
the melt phase. The transient strain is created by the T temperature K or 0 C
flow of liquid from the compression side of the flexure L velocityof the solid phase in a m · s 1

specimen to the tension side; the melt flow is driven by partial melt (vector)
the gradient in dilatational stress caused by the flexural a length betweeninner load points m
loading. The long-distance (relative to the grain size) of flexure specimen
migration of the melt phase is demonstrated via elec-
h specimenwidth m
tron microscopy; the anelasticity of the melt-flow pro-
d grain size m
cess is demonstrated by the strain recovery that occurs
during stress-free annealing of the deformed speci- g acceleration due to gravity m . s~ 2
mens. Compression-compression attenuation measure- h specimenheight m
ments (Llul = 20 MPa; T = 1070-1130° C; 10 -0.5 2:: k permeability m?
/2:: 10 3.5 Hz) reveal an absorption band with Qi 1 ~ 1 n exponentrelatingstress and dimensionless
that is temperature and frequency independent. A nu- steady-state strain rate
merical fit of the creep data to a discrete distribution of
P app appliedpressure Pa
linear anelastic (Voigt/Kelvin) elements suggests that
P. effectivepressure Pa
this attenuation behavior is effected by three distinct
losses. A simple D' Arcy flow model for the melt migra- Po pore (liquid-phase) pressure Pa
tion in the flexed beams, however, demonstrates the r radius of triple junction m
same attenuation behavior. The discrete analyses are time
consistent with there being a continuous distribution u velocityof liquidphase (vector) m· S-I
of relaxation times associated with the melt migration
(x, y, z) Cartesiancoordinatesystem m
in these high-strain-amplitude experiments. The com- variablesfor creep specimen
bined results suggest that dilatational anelasticity may
be the principal dynamic signature of texturally equili-
f frequency Hz
brated partial melts. VII pressuregradient (vector) Pa- m
D.p density contrastbetweensolid kg . m '
and liquid phases
Notation compliance Pa 1
Units
Y,I solid-liquid interfacial energy Jv rn 2

E. activationenthalpyfor anelastic kJ . mol- 1 or eV bulk viscosityof partial melt Pa- s

flow aggregate

Magmatic Systems Copyright © 1994 by Academic Press, Inc.

Edited by M. P. Ryan 19 All rights of reproduction in any form reserved.
20 Tye T. Gribb et al.

Units dinger and Anderson, 1989). The effect of partial

7) ; viscosity Pa s melting on the steady-state viscosity of partially
shear viscosity of partial melt Pa- s
molten olivine-basalt aggregates supports the idea
7)
aggregate that grain boundaries in these systems remain melt
7), shear viscosity of the melt phase Pa· s free: the enhancement of deformation kinetics due
¢ volumetric melt fraction dimensionless
to melting, in both Newtonian (Cooper et al.,
1989) and non-Newtonian (power-law) regimes
lT 1,2,1 principal stresses Pa
(Beeman and Kohlstedt, 1993), is distinctly small
U1MAX maximum value of the maxi- Pa
mum principal stress
(less than an order of magnitude change in vis-
cosity compared to the subsolidus value) and fully
E strain, inelastic strain dimensionless
accounted for by the presence of melt on triple
E, anelastic strain dimensionless
junctions only.
Eu maximum anelastic strain dimensionless
The interconnected melt network on triple junc-
e,.. steady-state strain rate tions allows for the porous-media permeation of
T characteristic time for anelastic melt through regions of partial melting in the
transient
Earth's mantle. Beyond this allowance, however,
the physical extraction of melt from a partially
molten zone requires that the crystalline resid-
uum undergo compaction, specifically by a due-
Introduction and Theoretical Foundation . tile process, and thus displace the liquid phase.
Driven by gravity-induced positive buoyancy (i.e.,
The physical processes by which steady-state vis- the density contrast between the liquid and the re-
cosity, seismic wave attenuation,and melt migra- siduum), this melt migration process has been de-
tion and segregation occur in partial melts must be scribed phenomenologically using D' Arcy's law:
understood to develop testable hypotheses con- the relative velocities of the residuum and the liq-
cerning the structure, chemistry, and dynamics of uid are linearly proportional to the inverse gradient
affected regions in the upper mantle. The steady- in pore fluid pressure, with the constant of pro-
state mechanical behavior of partial melts (i.e., that portionality being the permeability of the resid-
unaffected by rupture of the crystalline residuum) uum (e.g., Turcotte and Schubert, 1982, pp. 382,
is uniquely related to the (quasi)equilibrium distri- 413-416),
bution of the liquid phase on the scale of the grain
u - U = - [k(q" d)/1]cJ VII, (1)
size. In silicate systems this equilibrium state,
dictated by the relative energies of solid-solid where u and U are the velocities of the melt and
(i.e., grain boundaries and interphase boundaries) the residuum, respectively, k is the permeability
and solid-liquid interfaces, has been generally (which is a function of the volumetric melt frac-
accepted as having the liquid phase confined to tion, q" and the grain size of the residuum, d), 1]r
three-grain edge intersections ("triple junctions") is the viscosity of the melt, and vn is the pressure
and melt-free grain boundaries (e.g., Vaughan et gradient. A pressure gradient in the melt phase is
al., 1982; Waff and Bulau, 1982; Jurewicz and created by the momentum transferred to the liquid
Watson, 1985; Daines and Richter, 1988). The de- by the gravitational collapse of the residuum and
gree of this restricted interconnectivity has been it is resisted by (i) the buoyancy provided the re-
shown to be affected by volatiles and by a multi- siduum by the liquid phase and (ii) the capillarity
phase crystalline residuum (e.g., Fujii et al., 1986), force that seeks to preserve some amount of liquid
but wetted grain boundaries have not been dem- in the triple junction network (McKenzie, 1984;
onstrated to be an equilibrium structure. A recent Ribe, 1987; Riley et al., 1990):
analysis suggesting that crystalline anisotropy,
VII = 1]V 2U + (g + 1]/3)V(V . U)
specifically of olivine in an ultramafic system, (2)
- (1 - q,)gD..p - H(Y,I> q,)Vq,.
may allow for thermodynamically stable, wetted
grain boundaries (Waff and Faul, 1992) is at odds In Eq. (2), the first two terms on the right-hand
with other experimental and observational studies side represent pressure gradients in the melt due
(e.g., Cooper and Kohlstedt, 1982, 1984; Schwin- to shear deformation and dilatational deformation
 2. Melt Migration and Related Attenuation 21
of the residuum, respectively (1] is the shear vis-
cosity and ~ the bulk viscosity of the partially
molten aggregate), the third term is the buoyancy
(g is the acceleration due to gravity and lip is
the difference in density of the solid and liquid
phases), and the final term is the force due to cap-
illarity (H is a function of 'Y,I, the solid-liquid
interfacial energy, and of cP). Equations (I) and
(2) have been evaluated numerically to character-
ize melt extraction in a gravity field (McKenzie,
1984; Richter and McKenzie, 1984; Scott and
Stevenson, 1986; Ribe, 1987); application of the
models to geologic conditions, however, requires
knowledge of the physical parameters of the sys-
tem, i.e., 1]" 1], ~, and the relationship between cP
and the permeability for the environment of inter- v;
est. These parameters must ultimately be charac-
terized through experiments. Figure 1 Schematic diagram of a melt-filled triple junction.
We have concentrated on flexural deformation The size of the junction is a function of the radius of curvature,
experiments, using model partial-melt aggregates r. 'and the dihedral angle, O. 0 i, a function of the interfacial
that are carefully prepared to be in microstruc- energies of the system. The melt fraction <p can be calculated
from r, e, and the grain size of the crystalline phase, although
tural equilibrium and working in the temperature- one must make assumption, concerning grain shape to do so.
differential stress regime that allows charac-
terization of steady-state behavior. (Phenomena of water from a submerged sponge by squeezing
involving fracture of the crystalline residuum have it and monitoring the subsequent influx of the
been carefully avoided.) The flexural creep ap- expulsed water by the release of the differential
proach so developed is predicated on a first-order pressure while still submerged. The influx of the
analysis of the effect of hydrostatic pressure on liquid into the sponge in this simple example il-
the melt fraction in a microstructurally equili- lustrates the important fact that this effect of Pe
brated partial melt (Cooper, 1990; Raj, 19X2). The on the liquid fraction is completely recoverable
radius of curvature, r. of the melt-filled triple (given access to the liquid reservoir); the "com-
junctions (see Fig. I) is related to the effective paction" process for the sponge is thus an anelas-
pressure in the partial melt system, tic effect, driven by the gradient in the dilatational
(3) component of the stress state, the kinetics of which
are dictated by the variables in Eq. (2).
where Pe, the effective pressure, is given by the Placing a microstructurally equilibrated partial
difference in applied (e.g., overburden) pressure melt specimen in a four-point flexural stress con-
(Parp) and the pressure in the liquid phase (the dition maximizes the gradient in the dilatational
pore fluid pressure, Po): component of the stress state relative to the devia-
p, = Parr - Pli' (4) toric component, thus allowing the greatest reso-
lution of the kinetics of melt migration in a creep
For a fixed value ofthe solid-liquid interfacial en- experiment. In four-point flexure, the portion of
ergy, one sees that an increase in P: should lead to the beam between the inner load points experi-
a reduction of r and therefore of cP (since the melt ences "pure bending"; i.e., there are no applied
fraction can be described simply by a geometric shear forces in this region of the beam and, as
argument for a known value of r [i.e., cP :x r-Id>: such, the beam in this region experiences a maxi-
Waff, 1980]); of course, a decrease in p; assum- mum principal stress (0'1. = O'n) that varies lin-
ing the system has access to a reservoir of liquid, early from compressive to tensile across the speci-
should cause an increase in r (and cP). The poten- men thickness (Crandall et al., 1972, Chap. 7),
tial mechanical behavior of this surface-tension
phenomenon is easily illustrated by the removal O'I(Y) = [6P(L - a)/bh'Jy, (5)
22 Tye T. Gribb et at.

A requires the melt phase to subsequently flow back,

thus recovering the transient strain of the speci-

z~ ifO~, men. One sees, then, that the flexural approach

inherently avoids the problem of needing an ex-
ternal melt reservoir to study the recoverability of
the melt-migration process.
I~'---L ·1 We present within the results and interpreta-
tion of flexural creep experiments performed on
+y
a synthetic olivine-basalt aggregate, a simple
B +h/21--..,.
model system (i.e., homogeneous melt phase in
chemical equilibrium with a single-phase crys-
talline residuum) for a peridotitic upper mantle.
.c..--+-h/2
-y The results are interpreted relative to the ideas
2
discussed above, with an emphasis on the role
()lMAX= ±3P(L-a)/bh of dilatation-stress-induced melt migration on the
()2 = ()3 = 0 characteristic attenuation for an equilibrated zone
Figure 2 The configuration and stress distribution for four-
of partial melt. Preliminary Young's modulus
point flexure (after Cooper (1990)). (A) Specimen dimensions (compression-compression) attenuation experi-
(h and h), load point placement (L and a), loads (P). and co- ments are compared with the predictions of the
ordinate system (x, y. z), (B) Elastic stress distribution for that flexural creep data.
portion of the beam between the center load points (i.e., from
x =![L - a] to x = t[L + all. This stress distribution holds
for creep deformation if the beam has a Newronian-viscous
rheology. Experimental Approach

There are a number of practical requirements in-

where P is the load applied to each load point and volved in the flexural creep approach to the study
L, a, b, h, and yare specimen load and shape pa- of melt migration. Portions of the specimen are
rameters as shown in Fig. 2. Equation (5) is an subjected to tensile stress, and so, to avoid rupture
elastic flexure analysis that holds for Newtonian of the specimen, the thermal and mechanical po-
viscous behavior of the beam and is but slightly tentials-as well as the specimen grain size-
modified for non-Newtonian viscous flow (e.g., must be selected so as to promote a Newtonian/
Hollenberg et al., 1971). In the beam, a z and o , superplastic rheology. Further, a Newtonian solid
are both zero, and the maximum values of a l (i) has a negligibly small intrinsic transient strain,
(a'',IAX) are determined by substituting ± h/2 into which allows the melt migration effect(s) to be iso-
Eq. (5). The state of loading thus creates a linear lated in the creep data, and (ii) allows the simple
profile in Papp given by the mean-normal stress elastic stress distribution for four-point loading
(cf. Dieter, 1986, pp. 46-48): (i.e., Eqs. (5) and (6)) to be directly applicable to
the experiments. The loads P used to produce a
Papp(Y) (a l + a2 + a3)/3
given maximum principal stress in a flexure speci-
(6)
[2P(L - a)/hh']y.
men are very small compared to that required for
A comparison of Eq. (6) with Eq. (3) indicates the same stress in a tension or compression test;
that the flexural loading should force a reduction as such, small amounts of friction in the creep ap-
in ¢> on the compressive side of the specimen and paratus can produce substantial uncertainties in
an increase in ¢> on the tension side. Given that the the actual mechanical potential to which the speci-
forced motion of melt should cease when the lo- men is subjected. These factors, and others, influ-
cal melt fraction is again in equilibrium with the enced the design of both the specimens and the
applied flexural state of stress, the melt motion creep apparatus. For example, an O-ring seal or a
should produce a distinct, decelerating anelastic bellows arrangement 'on the specimen chamber
transient in the creep (strain versus time) curve. could easily produce a 20-to-30% uncertainty in
Of course, removal or reduction of the loads P alMAX of a specimen. The ability to perform ex-
 2. Melt Migration and Related Attenuation 23
periments in air (i.e., without any environmental a sheet of previously cobalt-soaked platinum and
seals on the apparatus) was thus crucial mechani- fired in air first at 400 0 C for 2 h to burn out the
cally, although this approach has, of course, pre- organic binder and then at 12000 C for 10 h to sin-
cluded the use of natural ferro-magnesian olivine- ter the beam; the heating and cooling rates em-
basalt specimens (which are unstable in air). ployed in the thermal schedule were 1000 C . h -I.
The specimens thus prepared had final dimensions
Experimental Specimens: Co-Mg of approximately 40 X 10.8 X 3.2 mrn, Micro-
Olivine-"Basalt" structural analysis of the sintered beams revealed
a mean grain size for the olivine of 4.9 ps», a melt
The specimens employed in this study were an ag- fraction of -12 vol% of the olivine-plus-basalt
gregate consisting of crystalline (COO.74Mgo.Z6h content, and a real porosity (isolated voids of a
Si0 4 olivine in chemical and microstructural equi- diameter similar to the grain size) of -4 vol%.
librium with a liquid that is analogous to a The microstructure of this material was typical
basalt-at least in terms of its polymerization. of fine-grained olivine-basalt materials nearing
Specimens were prepared by the sintering of pre- microstructural equilibrium (cf. micrographs in
viously fused, ground, and compacted powder of Fig. 7): most of the melt phase was confined to
the appropriate bulk composition. Co-Mg olivine well-formed triple junctions with diameters (i.e.,
was used because of its stability in air and its ease of a circle scribed within the triple junction) of
of melting (the liquidus for the 0.74Co-0.26Mg approximately 0.4 ,urn; pools of melt bordered by
composition is estimated as 16500 C). Having no 'faceted olivine grains (i.e., the structures recently
constraints on the solid-liquid partitioning behav- described by Waff and Faul (1992» were addi-
ior in this system, the "basalt" chemistry was es- tionally noted. The chemistry of the olivine and
timated using a direct analogy to Fe-Mg parti- the residual glass phase, as determined by energy-
tioning (Roeder and Emslie, 1970) (Co2+and FeZ+ dispersive x-ray spectroscopy (EDS) on a scanning
having identical ionic radii); the bulk chemis- transmission electron microscope, are presented
try was calculated assuming a melt fraction of in Table 1. The bulk composition calculated from
approximately 10 vol%. Analytical-reagent-grade these analyses is slightly enriched in SiO z and is
(or better) oxide, carbonate and nitrate pow- depleted in CoO and MgO from the original batch
ders (CaCO, was the calcium source; NaNO, the that was fused; this fact can be explained by pro-
sodium source), precisely measured to provide cessing of the cullet in the fluid-energy mill: More
the desired bulk olivine-basalt composition, were of the less-dense, silica-rich glass phase was in-
tumble-mixed and fused in a cobalt-soaked plati- cluded in the "fine" powder used to fabricate the
num crucible for 16 h at 16500 C in air. The melt
thus produced was poured directly into a cold- Table 1
water bath and the semicrystalline cullet that re- Compositions of Co-Mg Olivine-Basalt Material'
sulted was shattered into a powder using a fluid-
energy particle mill. The mean particle size of Olivine Residual glass
Oxide (moltA) (mol'lc)
the powder from which the beam specimens were
produced was 2.0 ,urn, with 90% of the particles SiO, 33.50± 1.42 77.83 ±3.16
less than 5.3 ,urn in diameter. AI.O, 0.45::0.50 13.35± 1.80
Beam specimens for the flexural creep tests
MgO 15.91::0.95 1.17±O.lO
were prepared by sintering. The silicate powder
CoO 49.76 ± 2.45 2.02 ± 0.36
was mixed with a polypropylene glycol binder
CaO 034±0.09 5.46±2.53
(5 wt% binder in the mixture; the binder was first
;>;a,O 0.17±O.lO
dissolved in dehydrated 2-propanol alcohol, suf-
(Co" ,.Mg ".24}zSiO 4
ficient to cause the powder/binder mixture to be
slightly damp after mixing) and a pre-weighed "EDS data from a scanning transmission electron microscope. The
amount of the mixture was placed in a rectangular "metallurgical thin film" data reduction algorithm (a standardless tech-
nique) was used to determine cation contents from the x-ray emission
die and compacted to a (nominally hydrostatic) data; oxygen was calculated stoichiometrically. The errors indicated are
pressure of 20 MPa. These beams were placed on ± I standard deviation of the measured compositions.
24 Tye T. Gribb et al.

beams. We exploited this observation by repro- vature (E1MAX = hlp, where E is inelastic strain and
cessing a small amount of this powder through the p is the radius of curvature). The natural logarithm
mill and then fabricating beams from the powder of the curvature between the outer and inner load
consequently placed by the mill in the "coarse" points should be linearly proportional to the natu-
fraction; a few beams having a mean grain size of ral logarithm of the horizontal distance along the
6.0 JLm and a melt fraction of -7 vol% were thus beam, with the constant of proportionality being
produced (these, too, had -4 vol% real porosity). the stress exponent. As such, creep specimens
were quenched from their deformation tempera-
Creep Experiments and Specimen Analysis ture while still under load. (The quenching of
creep specimens was accomplished by disrupting
Four-point flexural creep experiments were per- the furnace power: the result was that specimens
formed in a dead-weight creep apparatus; the ap- were quenched to 900 0 C at a rate of approxi-
paratus and experimental protocol are described mately 0.5 0 C . s ~ I.) The shape of the crept speci-
by Cooper (1990). The specimen cradle in this men was then carefully measured by a device in
case, however, was fabricated from a-SiC; solid which the specimen was dragged at a controlled
Al z0 3 rods, 3.1 mm in diameter, were used to ap- rate under a gravity-fed displacement transducer
ply the load. The outer and inner span lengths (Jakus and Wiederhorn, 1988). The digitized data
used were L = 30.5 mm and a = 15.2 mm, re- could then be numerically manipulated to deter-
spectively. Temperatures employed in these ex- mine the curvature of the beam, etc.
periments ranged from 1050 to 12000 C; the tem- The anelastic nature of the measured transient
perature was both controlled and monitored by strain was ascertained by a comparison of the
a Pt/Pt-13% Rh (R-type) thermocouple located specimen curvature (strain) in the as-deformed
-7 mm from the specimen center and at the exact (and quenched) state and after stress-free anneal-
vertical location of the undeformed specimen; ing of the specimen at a temperature correspond-
temperature control was accurate to ± 0.5 0 C, ing to that at which it was originally deformed.
while differences within the zone occupied by the In these annealing experiments, specimens were
specimen were ± 20 C of the control temperature. placed such that if gravity were to have an effect,
The maximum principal stress applied to the spec- it would add to the creep strain of a specimen and
imens was ('IMAX = ± 3.45 MPa, corresponding not assist any strain recovery.
to a vertical load on the specimen cradle of 2P = The microstructure of as-deformed-and-
10 ~ (1.02 kgf). Strain (and strain rate) was de- quenched specimens was analyzed using scanning
termined by monitoring the vertical displacement electron microscopy in the backscattered electron
of the top piston (i.e., the displacement of the in- imaging mode. Specimens from the pure-bending
ner load points) through the use of two direct- region of the specimen were sectioned sequen-
current displacement transducers whose output tially along the y direction (Fig. 2) and the micro-
was summed, digitized, and fed into a microcom- structure in the (x, z) plane recorded for each
puter. Strain and strain rate determined by this section on 12 to 15 random micrographs at a
method are accurate, provided the total inelastic magnification corresponding to an image area of
strain does not exceed 0.02 (Hollenberg et al., -1200 JLm z (approximately 100 olivine grains
1971): as such, experiments were terminated at would be imaged in each micrograph). The areal
this strain. percentage of the melt phase (approximately equal
The shape of a beam that is plastically de- to ¢) was determined by digital image analysis
formed in four-point flexure is very sensitive to for each micrograph.
the total strain as well as to the stress exponent
that characterizes the relationship between the Attenuation Experiments
steady-state strain rate and the magnitude of dif-
ferential stress. The curvature between the inner A modest number of Young's modulus attenua-
load points is constant (i.e., the beam has circular tion experiments were performed on the Co-Mg
curvature in this region) with the maximum speci- olivine-basalt aggregates, to test ideas arising
men strain being linearly proportional to the cur- from the creep results. Small rectangular pris-
 2. Melt Migration and Related Attenuation 25
matic specimens (3.0 X 3.0 X 8.0 mm) of the the Burgers solid-analysis predictions of Eo and Ess
¢ = 0.12 material were deformed in a compres- match closely with values determined by a visual
sion-compression mode in a servomechanical- analysis of the data, a result confirmed by many
actuated creep apparatus employing a gravity-fed measurements of flexural creep on similar, equili-
extensometer (the apparatus is fully described by brated glass-ceramic materials (Cooper et 01.,
Meyer et 01. [1993]). Specimens were deformed 1989; Cooper, 1990). The highest temperature
in the temperature range 1100 to 1160° C (con- data (1130° C, ¢ = 0.12 and 1200° C, ¢ = 0.07)
trol to ± 0.5° C; variation across specimen within are an exception: these experiments did not reach
± 1° C) with the maximum compressive stress steady state in the E = 0.02 limit of the experi-
varying sinusoidally from 5 to 25 MPa at con- ment. For example, the measured transient strain
trolled frequencies of 10 -05 ~ f~ 10 3.5 Hz. The (via the strain recovery experiments, see below)
strain-versus-time data thus produced consisted of for the 1130°C, ¢ = 0.12 experiment is 5.8 X
a Newtonian steady-state component correspond- 10 3 , whereas the Burgers solid analysis pre-
ing to the mean value of (J"I (15 MPa) with the dicted Eo = 7.7 X 10 - 3, a prediction consistent
sinusoidal component of amplitude -10 - 3 super- with the Eo results determined for the lower-
posed. The data analysis therefore consisted of temperature experiments that did achieve a steady-
subtracting the steady-state component from the state strain rate.
data and subsequently determining the phase angle Post-experiment measurements of the curva-
between the applied stress and the anelastic strain. ture of creep specimens revealed that all beams
The attenuation (Q E I, where subscript E refers to are characterized by a Newtonian (diffusion-
Young's modulus) for this lossy material is given dominated) rheology (i.e., E" ex: (J"'iMAX and II = I)
by the tangent of the phase angle (e.g., Green for the temperature, stress, and grain-size regimes
et 01., 1990). studied. The results of the measurements are
shown in Fig. 4 and are compared to the line
representing a power-law (dislocation-dominated)
Experimental Results rheology with a stress exponent of n = 3. The
ability of the curvature measurements to discrimi-
Flexural Creep Experiments nate between the rheologies is plainly evident.
The anelastic nature of the transient strain is
The flexural creep (inelastic strain versus time) demonstrated in Fig. 5, which shows the effect
behavior of the Co-Mg olivine-basalt specimens of post-deformation, stress-free annealing experi-
is illustrated by the data presented in Figs. 3A ments on the curvature of the tensile surface of the
and 3B for the ¢ = 0.12 and ¢ = 0.07 aggregates, 1130° C, ¢ = 0.12 beam. In the figure, one sees
respectively. Qualitatively, one sees that each that, consistent with pure bending, the curvature
creep curve is characterized by a pronounced, de- of the beam is quite constant between the inner
celerating strain rate that ultimately approaches a load points. (The curves themselves are the digi-
steady state. Using a regression analysis, the data tally analyzed displacement data; the markers on
have been described in the figures using the linear each curve have been added solely for curve iden-
viscoelastic Burgers solid model (Fig. lOA; e.g., tification.) Further, one sees that strain is recov-
Findley et 01., 1976), ered with annealing, with strain recovery reaching
a limit of 5.7 X 1O~3 after -250 h. One should
E(t) = Eo(1 - e ft T
) + Essf, (7)
note that this recovered strain value matches well
where E denotes inelastic strain, Eo is the total an- with the amount of transient strain measured dur-
elastic strain, Ess is the steady-state strain rate, ing creep (5.8 X 10- 3 , see text above). Equally
and t is time. The variable T is the characteristic noteworthy, though, is the fact that the relaxation
time of the anelastic transient, that is, the time re- time for this strain recovery is greater by a factor
quired to produce 1 - lie (-63.2%) of the total of -102 than that characterized by the fit of the
ane1astic strain or, alternatively, the time at which Burgers solid model to the creep curve. For each
the ane/astic strain rate has decreased to lie specimen, the recovered strain from stress-free
(-36.8%) of its initial value. One can see that annealing was within 10% of that measured for
26 Tye T. Gribb et aI.

A 25

1130°C
'"----
• = 2 . 7 X 10- -1 6
t:•• -3 S
o 20 t:o = 7.7 x1 9s
x T=3.1x10s
'-"
c
0
L... 15
+-'
if)

o
:;:;
en
0 10
(])
c

0
+-'
0 5 Co-Mg Olivine-Basalt
f-
CTn,wc = 3.45 MPa; rp = 0.12

5 10 15 20 25 30 35 40
Time (8 x10- 3)

B 20 ............-.,.....,....,..."T"""T-.-.....-.....,.......-T"""T".....-.,.....,....,..."T"""T..............,.....,..-r-T"""T".....-.....,

1200°C
. -68
t:..- . X 10-
-3
6 -1
s
t:o=2 .1x1 q
T=3.3x10 s

c
oL...
+-'
if)

o 10
:;:;
en
o
(])
c

o 5
+-'
o Co-Mg Olivine-Basalt
f-
CTtt.lAX = 3.45 MPa; rp = 0.07

5 10 15 20 25 30
Time (8 x10- 3)
Figure 3 Four-point flexural creep curves for the Co-Mg olivine-basalt aggregates: (A) ¢ = 0.12 specimens; (B) ¢ = 0.07
specimens. The parameters listed (and the solid curves) are based on a Burgers solid-model regression of the data. The curves
are characterized by a large transient strain, the anticipated effect of the migration of the melt phase across the specimen. Linear
regression analyses of E" and of T versus inverse temperature give activation enthalpies of E" = 500 ± 35 kJ . mol-I and E. =
210 ± 10 kJ . mol " ' forthe steady-state and transient creep processes, respectively.

the transient on the creep curve. The errors were steady-state and the anelastic creep rates follow
always negative (i.e., less strain recovered than Boltzmann/Arrhenius behavior,
expected).
The temperature dependencies of both the i. IX exp( - E/kT), (8)
 2. Melt Migration and Related Attenuation 27

/
I
n = 3/
E /
E
Q)
L...
rp = 0.07
• 11000 e
0
/ .
••
.3 -0 • 1200 e
o
~ /0
::J
U rp = 0.12
'-" /
o 10700 e
0
C
/ o: 11000 e
~-1 / o 1130 e
N

o /
E
L...
/ Co-Mg Olivine-Basalt
o O'lMAX = 3.45 MPa
Z

0.5 1.0 1.5 2.0

In(distance along beam, mm)
Figure 4 A In-In plot of specimen curvature versus distance between the outer and inner load points; the slope of the lines
thus produced indicates the stress sensitivity ofthe steady-state creep response. The aggregates are characterized by a Newtonian-
viscous, diffusional rheology (i.e., /l = I). A line for /l = 3 (dislocation rheology) is shown for comparison.

where E is the activation enthalpy, T is the abso- an activation enthalpy of E ss = 500 ± 35 kJ .

lute temperature, and k is Boltzmann's constant. In mol " ' (5.2 ± 0.4 eV). The activation enthalpy
the case of steady-state creep, a linear regression for the anelastic flow can be evaluated by deter-
of In(E ss ) data versus inverse temperature reveals mining the time necessary to complete a certain

20 'l""T'".,....,.."T""'I"...,........,...,....,...........'l""T'"T""1
r-T'".,....,.."T""T...,............................,....,.......,.....,........,...,....,...........

Co-Mg Olivine Basalt: rp = 0.12; T = 1130°C

_____ 15 Geee€l as crept

'"0
ktJ.~ 50 h
GDi3Cl-El 250 h
~
- ••• 365 h
tJ./tr tJ.-&' tJ."'l!>-.. t a=5.7x 10-3
x I A

I" t-II,~"i.- -- ~ i;:

~
'\
---.i-
E 10 ~~' s,A
E lI" ~
'-"
Q)
~, ~

-'.\...
,
L...
::J
-+-'

l
0 ' ~
5
'.'"
l
~
::J
0
inner loading points

~20 -15 -10 -5 0 5 10 15 20

Distance Along Beam, x (mm)
Figure 5 Relaxation of anelastic strain as revealed by specimen curvature change following stress-free annealing (11300 C,
if>= 0.12 specimen). A relaxation of 5.7 X 1O-3was achieved after annealing for -250 h; this amount matched well the transient
strain recorded in the creep curve (5.8 X 10 -3).
28 Tye T. Gribb et 31.

9.5 Co-Mg Olivine-Basalt; Anelastic Response

O'llAAX = 3.45 MPa; rp = 0.12
eo = 5x10- 3
9.0 Eo = 270 kJ/mol

eo = 4x10- 3
Eo = 247 kJ/mol
~8.5
(/)
'--'
Q) eo = 3x10- 3
Eo = 216 kJ/mol
E 8.0
:;:;
..........
c
7.5 eo = 2x10- 3
Eo = 188 kJ/mol

7.0

7.2 7.3 7.4 7.5 7.6

1/T (K- 1 X 10 4 )
Figure 6 Activation enthalpy of the anelastic transient determined from the time required to achieve a specific anelastic strain,
E,.The mean value of E, is 230 ± 35 kJ . mol "'.

amount of anelastic strain (the calculated steady- shown; the melt phase has indeed migrated across
state strain is subtracted from the total strain, cf. the specimen during flexural creep in a manner
Eq. (7». Figure 6 displays the evaluations of the consistent with the stress-induced gradient in ef-
c/J = 0.12 data of Fig. 3A. Here one sees that fective pressure (cf. Eqs. (3), (4), and (6».
the average of the many evaluations gives E. =
230 ± 35 kJ . mol-I (2.4 ± 0.4 eV), although the
Attenuation Measurements
value of E. apparently increases modestly with in-
creasing anelastic strain. A linear regression on Results of the compression-compression atten-
the values of In(r) versus inverse temperature uation measurements are shown in Fig. 9 (the
gives E. = 210 ± 10 kJ . mol " ' (2.2 ± 0.1 eV). curves marked "Exp."), The measured attenu-
ation is distinctly insensitive to frequency and
Microstructural Analysis temperature: QE I is approximately unity (i.e.,
the phase angle measured between applied stress
It was generally observed in the post-deformation and strain response stayed near 45°) across the
sectional microstructural analysis that the melt frequency range studied regardless of the tem-
fraction had decreased on the compression side perature. (The errors in the measurements shown
of a crept specimen and increased on the ten- in the figure are approximately ± 0.1 log unit
sion side. Image analyses on the c/J = 0.12 speci- in QE"I.)
mens indicated that the mean value of c/J measured
across a specimen displayed such behavior, but as
the sample standard deviations for c/J at a given Discussion
value of y was large, the analyses were inconclu-
sive. Not surprisingly, this result prompted the Creep and Melt Migration
development of the c/J = 0.07 specimens. Lower-
ing the initial melt fraction allowed resolution of A general interpretation of the flexural creep be-
the melt migration effect, as demonstrated for the havior and the deformation-induced microstruc-
1100° C specimen in the micrographs presented ture of these fine-grained, Newtonian-viscous
as Fig. 7 and the melt fraction profile in Fig. 8: partial-melt specimens is straightforward: Consis-
The difference in c/J across the specimen is clearly tent with the thermodynamic argument made in
 2. Melt Migration and Related Attenuation 29
Co-Mg Olivine-Basalt
T=1100°C; U 1MAX = +3.45 MPa

Figure 7 Backscattered electron SEM images from the ¢J = 0.07 specimen deformed at 1100~C. These micrographs were
taken in the (x, z) plane of the specimen (cf. Figure 2A), The values for ¢J indicated are areal percentages determined by digital
image analysis. (A) y = + 1.7 mm (at the surface of maximum compressive stress); (B) y = 0 (at the center plane); (C) y =
- 1,5 mm (near the surface of maximum tensile stress), In the micrographs, the light gray phase is olivine, the dark gray phase
is glass, and black contrast is produced by porosity.
30 Tye T. Gribb et al.

the chapter introduction, the anelastic strain re-

Co-Mg Olivine-Basalt
corded in the data is directly and uniquely related
(TlllAX = 3.45 MPa to the migration of the melt phase, driven by the
a> 1
1100·C
c gradient in the dilatational component of the ap-
o
n: plied stress, from the compression side of the
beam specimen (i.e., y > 0) to the tension side
~
<l>
c-o
<l>
U (y < 0). The final distribution of the melt phase in
the specimens is consistent with the relationship
between the linear gradient in effective pressure
and the inverse relationship between the effective
pressure and triple junction radius (cP ex: pt ? ex:
5 6 789 10 11
y -2; Fig. 8). Further, the steady-state rheology of
Melt Fraction,' 'P (oreal s ) the two-phase, solid-liquid system is revealed in
Figure 8 Melt distribution for the cP = 0.07, 1100°C speci- the steady-state creep rate of the beams. Applica-
men after deformation and quenching. The data points repre- tion of the linear viscoelastic Burgers solid model
sent the mean values of digital analyses of 12 to 15 micro- to the data (Eq. (7); Fig. lOA) allows direct inter-
graphs at a given location; the error bars are ± I standard
pretation for the viscosity parameters employed
deviation of the mean. The heavy vertical line represents the
initial (pre-deformation) condition of the specimen. While not in the melt-migration models: the shear viscosity
a unique description, given the magnitude of uncertainty of the is Tf = !<U1MAX!€SS); the bulk viscosity is g =
data, the curve compared to the final melt distribution is that . (~)(UIMAx'T/EO) (Cooper, 1990; Green and Coo-
expected from a straightforward application of Eqs. (3) and per, 1993).
(6), i.e., cP :x y -2.
The temperature dependencies of the bulk and
shear viscosities provide clues as to the physical
processes that rate limit the flow responses. The
steady-state creep behavior of Newtonian-viscous
olivine-basalt partial melts in this thermodynamic
1.0 , Co-Mg Olivine-Basalt
,, rp = 0.12 and microstructural regime has been demonstrated
,,
,, to consist of a series kinetic process involving
, solid-state grain boundary diffusion and solution-
, ,
0.5 '''- . :
/ ,, liquid-phase transport reprecipitation ("pressure
IW "- solution") of the ionic components of olivine
ao
" , ' (Cooper et al., 1989). The effect of the grain size
8' -0.0 \ , on the steady-state creep rate isolates the rate-
\ ..
limiting step to either grain boundary or liquid-
- - 1070°C Invert phase transport; the activation enthalpy measured
-0.5 - - - 1100·C Invert for steady state indicates grain boundary diffu-
- - - - - 1130·C Invert
~ 1100°C Exp. (lICT=20 MPa) sion as rate limiting. The activation enthalpy for
~D<3 G€J 1130·C Exp.
H~~<> 1160·C Exp. liquid-phase transport of ions should match that
for the liquid's viscosity. The E ss determined for
-1·~5.0 -4.0 -3.0 -2.0 -1.0 0.0
the flow of olivine-basalt was distinctly higher.
I0910(f, Hz)
This solid-state rate limitation occurs not because
Figure 9 Young's modulus attenuation (QE") of the cP = 0.12 pressure solution kinetics are sluggish-on the
specimens in compression-compression experiments. The ex-
contrary, they are very rapid: orders of magni-
perimental data (marked "Exp." in the key) show a band of
attenuation at QE I - I for the frequency range studied that is tude faster than grain boundary diffusion in this
additionally temperature independent. (Errors on the experi- system-but because the microstructural equilib-
mental measurements are approximately ± 0.1 log units in rium, which is characterized by melt-free grain
QE''.) These data are compared with attenuation spectra that boundaries, does not allow access of the solid to
were calculated from the compliance behavior of the creep
the liquid phase except at triple junctions. The
specimens (Figure II), which indicated three discrete linear
loss mechanisms account for the attenuation behavior of the ~500 kJ . mol-I noted here for the activation en-
partial melt. thalpy of the steady-state process in the Co-Mg
 2. Melt Migration and Related Attenuation 31

0"; E, t

Tl2
Anelastic
Voigt/Kelvin
I Viscoelostic
Maxwell
I
Element Element

t/ T
e(t) = (O/R2)(1- e- ) + (CT/17,)t + (CT/R,)
T = 172/ R2 ; ~2= 1/R 2

••• 0"; E, t

()
et =CT LI
{!, " (1- e-tiT.)
I
i-1

Figure 10 Linear viscoelastic models (after Findley et al. (1976)): (A) Burgers solid model; (B) generalized Voigt/Kelvin
model. The creep function of the Burgers model (A) is given by the superposition of the elastic, steady-state plastic and anelastic
responses. Systems containing a distribution of linear losses can be modeled by a number of Voigt/Kelvin elements placed in
series (B). One should note that the magnitude of the stiffness or the viscosity of an element in the series representation allows
for modeling of steady-state flow and elastic responses. If the stiffness of a single element goes to zero, the element then describes
steady-state viscosity; obversely, if the viscosity of a single element goes to zero, the element then describes the elastic behavior.
As such, the entire creep function-and not just the anelastic behavior-can be modeled by the series arrangement of Voigt!
Kelvin elements.

olivine-basalt specimens is logically argued, then, cess: in order for the melt to move, dilatation (ten-
as that characterizing grain boundary diffusion in sile side) and contraction (compressive side) of
the olivine residuum. The activation enthalpy for the crystalline residuum must occur in addition to
the anelastic transient, however; is distinctly dif- (i.e., simultaneously with) the physical flow of the
ferent; the value for E., -230 kJ . mol ' ', is ap- liquid. One of these two processes will be rate
proximately one-half that for the steady-state pro- limiting and thus be responsible for the magnitude
cess and matches well that for the viscosity of a of the activation enthalpy measured for anelastic
poorly polymerized silicate melt (e.g., an olivine- flow. The dilation/compaction of the residuum
saturated basalt; cf. Richet, 1984). must occur via the same-grain-scale creep mecha-
The physical process of melt migration in the nism responsible for steady-state flow; as such, if
specimens is also a coupled, series kinetic pro- it were rate limiting the anelastic flow, E. would
32 Tye T. Gribb et al.

be equivalent to E". Such is not the case for these 2.0

Co-Mg olivine-basalt specimens: the clear dif- Co-Mg Olivine-Basalt
ference between E, and E ss and the appropriate ,......, 0'1MA)(= 3.45 MPa; rp = 0.12
0>
magnitude of E, indicate that the melt migration o
process is rate limited by the ability of the liquid
~ 1.5 '"I
\
to flow. I
o \
Q..
'-' Ijl
I
0& 1.0 I
Attenuation and Melt Migration
Q) ~ 1070·C """
g ,,1>-"'_ 11 OO·C """
The adequacy of the linear viscoelastic Burgers .S? 138888 1130·C
solid model (Fig. lOA) to fully describe the dy- a. "
II
II
E 0.5 II
II
namics of the partial-melt system is brought into o II
u II I
question when one compares the prediction of ::t
dynamic properties via the Burgers model with \'J: ,
I I

the measured attenuation response. The Burgers 0.0 -1 0 3 4 5

model consists of a single, linear energy loss (the s)
anelastic Voigt/Kelvin element of viscosity TJz,
Figure 11 Compliance spectra (<I> versus log r) determined
restorative stiffness R z , and characteristic time by numerical regression to a large number of discrete linear
T = TJz/R z) plus the linear steady-state plastic and anelastic elements (e.g.. the model in Figure lOB). The result
elastic responses (the viscoelastic Maxwell ele- shown, that the compliance behavior represents three discrete
ment of viscosity TJ 1 and stiffness R I)' The dy- anelastic loss mechanisms, was quite robust to the magnitude
of the time step used in the regression analysis. These spectra
namic response for this model for the frequency
have been Fourier transformed to produce the QE I spectra
region explored in the attenuation experiments shown with the experimental attenuation data in Figure 9.
here is an increasing QE I with decreasing fre-
quency, caused by the steady-state creep response,
punctuated by a superposed single anelastic peak positive values of <P j as negative compliances
(Green et al., 1990). The model thus assigns the have no physical meaning. The results of such an
dilatational-stress-induced melt migration to an analysis for the c/> = 0.12 creep data are shown in
anelastic process having a single characteristic Fig. 11: There are apparently three discrete linear
time. The attenuation "band" noted in the experi- losses that characterize the anelastic response of
ments suggests, instead, that there may be many the material. This result is distinctly robust rela-
energy losses that sum to produce the measured tive to the magnitude of the time element used in
behavior, that is, a distribution of loss mecha- the numerical regression. Fourier transformation
nisms (cf. Anderson and Given, 1982). To test this of these compliance spectra results in the attenu-
idea, the creep data were reanalyzed, allowing for ation spectra (the continuous curves) shown with
an extremely large number of discrete linear loss the experimental attenuation data in Fig. 9. One
elements (Fig. lOB) to be applied to the anelastic sees that the three discrete linear losses sum to
creep curve; in this case, the anelastic creep func- create an attenuation band near QE I = 1 that plots
tion becomes quite closely to the experimental data. Similar
results have been noted for flexural attenuation
N
experiments on metasilicate glass-ceramic speci-
EaU) =a ~ <Pj(l - e- tITj ) , (9) mens (Gribb and Cooper, 1991). It should be
n=l
noted that the creep and attenuation experiments
where <P j is the compliance (== R i-I) of a single, were performed on entirely different apparatus.
given anelastic loss element that has a character- As such, the close correlation noted in Fig. 9 is
istic time of Tj. One should note that the Eo of any strong evidence that the attenuation measurements
anelastic loss element is equal to rr<I>. The anelas- represent properties of the olivine-basalt speci-
tic creep data can be regressed against Eq. (9) mens and not any dynamic property of the defor-
to create a retardation spectrum, <P( T) versus T mation apparatus.
(Gribb, 1992; cf. Thigpen et al., 1983). Restric- Given that the material studied here is a New-
tions must be placed on the algorithm to ensure tonian (i.e., linear) solid, it is tempting to ana-
 2. Melt Migration and Related Attenuation 33
Node ,for Numerical Analysis
6 11 16 21 26 31 36 41 46
A B 4

Finite-Difference D'Arcy Flow Model »-r--;

Synthetic Compliance
0.3 UJ
:'=
c
:J 3

c
:3 0.2 compression
side
tension
side
IQ)
0
L
~

:'=
.0
L
o
o ~ ~2
G: 0'
c 0&
'iii
0
e cV
0
o
0.1 I-...-----=-=---""'"=-~~'-?-'-----____. .s c
,g1
a.
E
0
u
0.0 L.-_ _~----:----::__--.....J
rR
positive Y 0 negative Y 00 2
Distance from Center Plane, y (arbitrary units)

Figure 12 D'Arcy's law model for melt flow in pure bending. (A) IjJ versus y. One sees that the melt fraction is predicted to
evolve to the IjJ oc y -2 form characteristic of steady state. In this example, the magnitude of IT'MAX was chosen sufficiently large
to magnify the form of the results. The initial, uniform melt fraction is 0.1. The time interval between the curves is constant, with
the earliest time shown as the fine solid line, the final time shown as the bold solid line, and each intermediate curve having a
dash length proportional to its respective time step. (8) Compliance spectrum determined from the model. Application of the
many-linear-element algorithm to the model indicates that the melt migration, a single-loss process as modeled here, is charac-
terized by four discrete linear elements, a result that matches closely the behavior of the experimental specimens (cf. Figure II).
The analysis indicates that the melt-migration process in the flexure experiments cannot be characterized by a single relaxation
time. This discrete spectrum is consistent with cPr T) having a continuous form, one that is most likely related to the geometry of
the flexure experiment.

lyze the three linear losses revealed for their motion of the liquid phase driven by the linearly
physical significance. For example, while the mi- varying pressure gradient and resisted by capillar-
croscopy results reported here provide compelling ity (i.e., as represented by Eqs. (3), (4), and (6)).
proof of a dilatational anelasticity in equilibrated This "zeroth-order" approach is appropriate for
partial melts, the Young's modulus attenuation ex- the case where D'Arcy flow of the liquid rate lim-
periments do not isolate dilatational from shear its the behavior (Ribe, 1987)-such as the present
attenuation mechanisms. The measured attenu- case for Co-Mg olivine-basalt as evidenced by
ation response thus could contain also (i) an in- the magnitude of Ea. Figure 12A shows the evo-
trinsic, shear-stress-induced effect of melt-free lution of <p with increasing time. While parame-
grain boundaries with a very short 7 (e.g., Raj, ters in the model have been chosen to magnify the
1975) and (ii) a local (i.e., grain-scale) melt effect of bending on <p, one sees that the melt dis-
flow ("squirt"; Mavko, 1980), also shear-stress- tribution assumes the <p ex: y -2 form expected at
induced, with a 7 somewhat larger than the intrin- . steady state. These analyses can be converted into
sic grain boundary effect. While we cannot rule strain versus time "data" by monitoring the vol-
out these effects (they need to be tested for specifi- ume change of the end elements and converting it
cally in a driven-torsion test on an equilibrated to strain; these strain "data" can then be analyzed
partial melt), an alternative approach to the analy- using the linear-element algorithm (Eq. (9)) to
sis suggests that the three discrete losses revealed generate a synthetic compliance spectrum. The
uniquely relate to the dilatational-stress-induced form of such a synthetic spectrum is shown as
melt migration. Fig. 12B: four discrete peaks are noted for what
We present in Fig. 12 the results of a simple is obviously (in this exercise) a single physical
finite-difference model (all units arbitrary) for flow loss process. Fourier transformation of these four
of the liquid phase in a flexed, partially molten peaks into an attenuation spectrum produces an
beam. D'Arcy's law is here applied to describe the attenuation band at Q E'! - 1, a result that is not
34 Tye T. Gribb et at.

influenced by parameters used in the model (e.g., the experimental and synthetic spectra shown in
original <p, O"IMAX, 1]r,etc.). The spacing of the dis- Figs. 11 and 12B, respectively, they are consis-
crete peaks in the compliance spectrum and their tent: the integral in r of the continuous function
increasing magnitude with time match closely in would equal the sum of the compliances shown as
form the compliance spectra determined from real discrete peaks in the figures.
creep data on the Co-Mg olivine-basalt aggre- The result, in both the physical data and the
gates (cf. Figs. 11 and 12B). model, that the attenuation in the beams is nomi-
We can at least conclude from the model re- nally frequency and temperature independent and
sult and its comparison to the experimental data has a value of Qli l - 1 suggests a strong geome-
that the anelasticity created by dilatational-stress- try dependence for the loss mechanism in flexure.
driven melt migration in these Newtonian speci- While the melt experiences a more difficult task
mens cannot be described by a single characteris- of flow with increasing strain, as noted above, the
tic time. Rather, the results are consistent with rate capability of the solid residuum to accom-
there being a continuous compliance distribution, modate the melt may be exceeded simultaneously
<P(r), for the melt-migration process in these ex- (the dilatations required of the tensile side of the
periments. Physically, one can argue that it is the specimen can be quite large, cf. Fig. 12A). The
evolution of <p(y) with increasing strain that pro- phenomenon is such that the QIi I - 1 result is
duces a continuously changing anelastic compli- expected: a phase angle of 45° can occur under
ance during creep. The model (Fig. 12A) indicates conditions where the strain produced by anelastic
that the initial changes in <p(y) occur at the ex- 'flow is equivalent to that produced by the steady-
tremes of the specimen (i.e., y = ± h/2), and with state mechanism (each on the order of 10- 3 ) ; the
increasing time (that is, total specimen strain) the fact that these two components of strain are intri-
distribution effect migrates from these extremes cately linked with increasing specimen deforma-
toward the specimen center. However, the inverse tion provides the result. We thus anticipate the dis-
relationship between p; and r, combined with the covery of a distinct attenuation peak associated
fixed total volume of melt in the system, results in with dilatational-stress-induced melt migration at
a greater portion of the melt migration path (that very low strain amplitudes (:510 ~ 5) that has ap-
is, the y dimension) experiencing a decrease in <p. plication to the dynamics in the geological setting.
As a consequence the melt experiences an ever Whether or not this loss is represented by the
greater resistance to flow, throughout most of the shortest r peak in the discrete compliance spec-
specimen, with increasing anelastic strain. trum (Fig. 11) remains a question to be answered
There is likely a variety of continuous <P(r) with additional experiments.
functions that can be reasonably fit to the creep
and attenuation data collected on the Co-Mg oli- Application of the Experimental Observations
vine-basalt specimens. One of these, perhaps the to the Earth's Mantle
simplest, is derived from a power law description
of the anelastic flow, i.e., There exist a variety of thermally activated mech-
anisms for attenuation of seismic waves in the
(10)
Earth's mantle. As carefully outlined in the re-
where the exponent M < 1. Equating this power views by Jackson (1986) and Anderson (1989,
law expression with the anelastic creep function Chap. 14), these mechanisms include the motion
determined from a continuous distribution of of lattice dislocations, the motion of grain bound-
Voigt/Kelvin elements (that is, the integral ex- ary dislocations and other diffusive motion on
pression of Eq. (9)) produces (Findley et al., grain boundaries, the motion of lattice point de-
1976, pp. 69-70; cf. Anderson and Given, 1982) fects, and the motion of liquid in partial melts.
With this variety, a distribution of relaxation
(11)
times is anticipated, and thus the absorption band
which on a log-log plot would produce a straight thought to be characteristic of attenuation pro-
line with negative slope. One should note that, de- cesses in the upper mantle is explained logically
spite the gross differences in form between the (e.g., Anderson and Given, 1982).
continuous spectra represented by Eq. (11) and Experiments to determine internal friction re-
 2. Melt Migration and Related Attenuation 35
sponse and mechanisms in earth materials and melt fraction is sufficiently large to affect a signifi-
analogs place constraints on the geophysical ar- cant loss of grain boundary area. The latter crite-
guments, however, and thus the capabilities to iso- rion may indeed represent the case in active vol-
late individual mechanisms proves valuable in in- canic zones at shallow depth.
terpreting seismic data. For example, experiments
indicate that the solid-state losses noted above are
known primarily to affect shear waves. Further, in Summary and Conclusions
the case of dislocations, relaxation times may vary
according to the lengths of dislocations and the Flexural creep experiments on microstructurally
magnitude of the thermal barrier to dislocation equilibrated olivine-basalt partial melts display a
motion (e.g., the energy required for the formation distinct anelastic transient that is related to the mi-
of jogs and kinks on individual dislocations). In gration of the melt phase within the specimen.
the case of grain boundary losses, the grain size is This anelasticity is argued thermodynamically
the distance over which a mechanical potential is to be dilatational. The fact that melt-free grain
dissipated (Raj, 1975); thus a grain-size distribu- boundaries characterize the steady-state micros-
tion creates a spectrum of relaxation times. What tructure of partial melts of upper-mantle compo-
can be said, then, of the attenuation effect of par- sition supports the idea that dilatational anelastic-
tial melting? ity may be the unique dynamic signature of partial
Many seismic studies have indicated that dis- melting under conditions that produce a low melt
tinct absorption of S-waves occurs below volcanic fraction and that disallow fracture of the crystal-
zones. For example, Solomon (1973) interpreted line residuum. The melt-migration attenuation ef-
observations of strong S-wave attenuation beneath fect noted in the experiments is sensitive to the
the mid-Atlantic ridge as an indicator of high melt magnitude of applied strain because of the cou-
fraction (-10 vol%), citing the grain boundary pling of the melt migration to the deformation
melt film model of Walsh (1969). Anderson and kinetics of the crystalline residuum. Further ex-
Sammis (1970) applied this model as well to the perimentation into attenuation behavior of partial
low velocity zone to support the idea that its struc- melts at seismic/subseismic frequencies thus re-
ture included ~ 1 vol% melt. The experiments quires working at low strain amplitudes.
used to support these ideas, though, have often
been dynamic melting experiments: measuring
the attenuation response of a material both below Acknowledgments
and above the solidus temperature. As melting oc-
curs initially on solid interphase (grain) bounda- We happily acknowledge many fruitful discussions with
ries, a substantial increase in shear attenuation Drs. Douglas Green and Dallas Meyer on various aspects of
measured in such experiments is not surprising. this work. The fusion of the Co-Mg silicate melts used to fab-
However, the knowledge of equilibrium micros- ricate the creep specimens was done by the Experimental
tructure formation in partial melts, as outlined in Melting Group at Coming, Inc., Coming, New York; Elton
Harris is thanked for his assistance in this regard. The electron
the introduction to this chapter, combined with microscopy work was performed in the Center for Materials
the calculated efficiency of the compaction and Science at the University of Wisconsin-Madison; the instru-
melt extraction process that suggests that less ments there are beautifully maintained by Dick Casper and
than 1 vol% melt could be retained in an actively Rick Noll. This manuscript benefited from careful reviews by
melting region at depth (Riley et al., 1990), dis- Mervyn Paterson, Rishi Raj, and Michael Ryan. We gratefully
acknowledge the National Science Foundation for financial
avows any easy application of the results of dy- support through Grant EAR-9005226.
namic melting experiments to the steady-state
structure of the upper mantle. We have demon-
strated clearly that seismic/subseismic-frequency
References
anelasticity in an equilibrated silicate partial melt
is primarily, if not uniquely, a dilatational phe-
Anderson, D. L. (1989). "Theory of the Earth," Blackwell
nomenon. A substantial shear attenuation effect Scientific, Boston.
due to melting may be limited, then, only to re- Anderson, D. L., and Given, J. W. (1982). Absorption band Q
gions of incipient melting or regions where the model for the Earth, J. Geophys. Res. 87, 3893-3904.
 36 Tye T. Gribb et aI.

Anderson, D. L., and Sammis, C G. (1970). Partial melting in Jurewicz, S. R, and Watson, E. B. (1985). The distribution of
the upper mantle, Phys. Earth Planet. Inter. 3, 41-50. partial melt in a granitic system: The application of liquid
Beeman, M. L., and Kohlstedt, D. L. (1993). Deformation of phase sintering theory, Geochim. Cosmochim. Acta 49,
Fine-Grained Aggregates of Olivine Plus Melt at High 1109-1121.
Temperatures and Pressures, 1. Geophys. Res. 98, 6443- Mavko, G. M. (1980). Velocity and attenuation in partially
6452. molten rocks, J. Geophys. Res. 85, 5173-5189.
Cooper, R. E (1990). Differential stress-induced melt mi- McKenzie, D. (1984). The generation and compaction of par-
gration: An experimental approach, 1. Geophys. Res. 95, tially molten rock, 1. Petrol. 25, 713 - 765.
6979-6992. Meyer, D. W., Cooper, R F., and Plesha, M. E. (1993). High-
Cooper, R E, and Kohlstedt, D. L. (1982). Interfacial energies temperature creep and the interfacial mechanical response
in the olivine-basalt system, Ad\'. Earth Planet. Sci. 12, of a ceramic matrix composite, Acta Metall. 41, 3157-
217-228. 3170.
Cooper, R E, and Kohlstedt, D. L. (1984). Solution-precipi- Raj, R. (1975). Transient behavior of diffusion-induced creep
tation enhanced diffusional creep of partially molten oli- and creep rupture, Metall. Trans. A 6A, 1499-1509.
vine-basalt aggregates during hot-pressing, Tectonophys- Raj, R. (1982). Creep in polycrystalline aggregates by matter
ics 107, 207-233. transport through a liquid phase, J. Geophys. Res. 87,
Cooper, R. E, Kohlstedt, D. L., and Chyung, K. (1989). 4731-4739.
Solution-precipitation enhanced creep in solid-liquid ag- Ribe, N. M. (1987). Theory of melt segregation-A review,
gregates which display a non-zero dihedral angle, Acta J. Volcano Geotherm. Res. 33, 241-253.
Metall.37, 1759-1771. Richet, P. (1984). Viscosity and configurational entropy of sili-
Crandall, S. H., Dahl, N. C, and Lardner, T. 1. (1972). "An cate melts, Geochim. Cosmochim. Acta 48, 471-483.
Introduction to the Mechanics of Solids," 2nd ed., Me- Richter. EM.. and McKenzie, D. (1984). Dynamical models
Graw-Hill, New York. for melt segregation from a deformable matrix, J. Geol.
Daines, M. 1., and Richter, E M. (1988). An experimental 92,729-740.
method for directly determining the interconnectivity of Riley, G. N., Jr., Kohlstedt, D. L., and Richter, F. M. (1990).
melt in a partially molten system, Geophys. Res. Lett. 15, Melt infiltration in a silicate melt-olivine system: An ex-
1459-1462. perimental test of compaction theory, Geophys. Res. Lett.
Dieter, G. E. (1986). "Mechanical Metallurgy," 3rd ed., Me- 17,2101-2104.
Graw-Hill, New York. Roeder, P. L., and Emslie, R E (1970). Olivine-liquid equilib-
Findley, W. N., Lai, J. S., and Onaran, K. (1976). "Creep and rium, Contrib. Mineral. Petrol. 29, 275-289.
Relaxation of Nonlinear Viscoelastic Materials," North- Schwindinger, K. R., and Anderson. A. T., Jr. (1989). Synneu-
Holland, Amsterdam. sis of Kilauea Iki olivines, Contrib. Mineral. Petrol. 103,
Fujii, N., Osamura, K., and Takahashi, E.-I. (1986). The effect 187-198.
of water saturation on the distribution of partial melt in the Scott, D. R., and Stevenson, D. J. (1986). Magma ascent by
olivine-pyroxene-plagioclase system, J. Geophys. Res. porous flow, J. Geophys. Res. 91, 9283-9296.
91,9253-9259. Solomon. S. C. (1973). Shear wave attenuation and melting
Green, D. H., and Cooper, R. E (1993). Dilatational anelastic- beneath the mid-Atlantic ridge, J. Geophys. Res. 78,
ity in partial melts: Viscosity, attenuation and velocity dis- 6044-6059.
persion, J. Geophys. Res. 98,19,807-19,817. Thigpen. L., Hedstrom, G. w., and Bonner, B. P. (1983). Inver-
Green. D. H., Cooper, R. E, and Zhang. S. (1990). Attenua- sion of creep response for retardation spectra and dynamic
tion spectra of olivine/basalt partial melts: Transforma- viscoelastic functions. J. Appl. Mech. 105,361-366.
tion of Newtonian creep response. Geophys. Res. Lett. 17, Turcotte, D. L., and Schubert, G. (1982). "Geodynamics: Ap-
2097-2100. plications of Continuum Physics to Geological Problems,"
Gribb, T. T. (1992). "Low-Frequency Attenuation in Micro- Wiley, New York.
structurally Equilibrated Silicate Partial Melts," M.Sc. Vaughan, P. 1., Kohlstedt, D. L., and WaffH. S. (1982). Distri-
Thesis, Univ. of Wisconsin-Madison. bution of the glass phase in hot-pressed, olivine-basalt ag-
Gribb, T. T., and Cooper, R. E (1991). Attenuation in a model gregates: An electron microscopy study, Contrib. Mineral.
silicate partial melt: Comparison of direct measurements Petrol. 81, 253-261.
and inversions of creep data (abstract), EOS Trans. Am. Waff', H. S. (1980). Effects of the gravitational field on liquid
Geophys. Union 72(Supp.), 508. distribution in partial melts within the upper mantle,
Hollenberg, G. w., Terwilliger, G. R., and Gordon, R. S. J. Geophys. Res. 85,1815-1825.
(1971). Calculation of stresses and strains in four-point Waff, H. S., and Bulau, J. R. (1982). Experimental determi-
bending creep tests, J. Am. Ceram. Soc. 54,196-199. nation of near-equilibrium textures in partially molten
Jackson, I. (1986). The laboratory study of seismic wave at- silicates at high pressures, Ad\'. Earth Planet. Sci. 12,
tenuation, in "Mineral and Rock Deformation: Laboratory 229-236.
Studies," Geophys. Monogr. Series, Vol. 36, pp. 11-23, Waff, H. S., and Faul, U. H. (1992). Effects of crystalline an-
American Geophysical Union, Washington, DC. isotropy on fluid distrib?tion in ultramafic partial melts,
Jakus, K., and Wiederhorn, S. M. (1988). Creep deformation J. Geophys. Res. 97, 9003-9014.
of ceramics in four-point bending, J. Am. Ceram. Soc. 71, Walsh, 1. B. (1969). A new analysis of attenuation in partially
832-836. melted rock, J. Geophys. Res. 74, 4333-4337.
 Chapter 3 Influence of Basaltic Melt on the Creep of Polycrystalline
Olivine under Hydrous Conditions

David L. Kohlstedt and Prame N. Chopra

Overview with water while the single-phase sample was fully

saturated, the hydrolytic weakening effect was larger
Constant displacement-rate and load-relaxation experi- in the latter than in the former. If both samples had
ments were performed at high pressures and tempera- been fully saturated with water, the difference in creep
tures to investigate the rheological behavior of partially strength may have been larger. Hence, the rheology of
molten aggregates of fine-grained olivine with a small partially molten upper mantle rocks in an ascending
amount of included basaltic melt under hydrous condi- mantle plume or a mid-oceanic ridge environment will
tions. Two-phase samples with melt fractions of either depend critically on the partitioning of water between
0.9 or 8.6 vol% basalt and grain sizes of 12 or 15 /Lm the melt and solid phases and on the permeability of the
were fabricated by hydrostatically hot-pressing pow- rock (i.e., melt fraction).
ders of San Carlos olivine plus synthetic basalt. Single-
phase samples with a grain size of 12 /Lmwere prepared
by hot-pressing powders of San Carlos olivine. Scan- Notation
ning and transmission electron microscopy observa- Units
tions demonstrated that the dihedral angle was -30°,
A Materials parameter in the S" • mm . MPa· n
such that the melt formed an interconnected network
strain-rate relationship
along three-grain and through four-grain junctions but
did not generally wet entire grain boundaries. Infrared Q Effective activation energy for kJ· mol"
creep
spectra revealed that the olivine in the sample without
basalt and in the sample with 0.9 vol% basalt was satu- R Gas constant 8.314(3) ± 0.00008
rated with water-derived species, while the olivine in ]·K'·mol·'
the sample with 8.6 vol% basalt was undersaturated be- T Absolute temperature K, kelvins
cause the water partitioned largely into the melt phase. T Temperature °C
The samples were deformed under hydrous conditions d.d, Grain size m
at temperatures of 1300 and/or l200°C, a confining
I::.d Length of the grain boundary m
pressure of 300 MPa, and a strain rate of -10.4 s·'. lost to the melt phase
Stress exponents of 1.0 ::5 n ::5 1.4 for both the two-
m Grain size exponent dimensionless
phase and the single-phase samples combined with a
comparison with published creep results indicated that n Stress exponent dimensionless
the samples deformed predominantly by grain boundary E, Ef Strain dimensionless, %
diffusion creep. For a common grain size at I300°C, the E, Er Strain rate s
partially molten sample containing 8.6 vol% basalt was E, Single-phase strain rate s"
a factor of 5 weaker than the sample without melt. Both s']
E2 Two-phase strain rate
samples were substantially weaker than single-phase
<P Melt fraction dimensionless
aggregates of anhydrous olivine. It is proposed that the
difference in creep strength between the melt-free and a, Stress MPa
the melt-rich samples was a result of two competing ef- U, Maximum principal stress MPa
fects: First, the presence of melt caused the partially un Applied differential stress. with MPa n
molten sample to be weaker than the single-phase oli- stress exponent
vine sample. Second. the presence of water resulted in (J Equilibrium dihedral angle at 0, degrees
hydrolytic weakening of both samples; however, be- melt crystal interfaces
cause the partially molten sample was undersaturated

Magmatic Systems Copyright © 1994 by Academic Press, Inc.

Edited by M. P. Ryan 37 All rights of reproduction in any form reserved.
38 David L. Kohlstedt and Prame N. Chopra

Introduction ture and pressure. Hence, compressive creep ex-

periments have been carried out on fine-grained
Pressure-release melting of ascending mantle ma- (-lO-,um) aggregates synthesized from olivine
terial beneath mid-ocean ridges produces par- powders plus a synthetic basalt. Fine-grain-size
tially molten rocks at depths as great as 100 km. materials were used for two reasons: First, a pri-
Trace element analyses of mid-ocean ridge basalts mary role of the melt phase is to enhance the ki-
(MORBs) and of diopside grains in abyssal peri- netics of grain boundary diffusion (Coble) creep
dotites (Salters and Hart, 1989; Johnson et al., (Cooper and Kohlstedt, 1984b; Cooper et al.,
1990) as well as laboratory studies of the per- 1989); to observe diffusion creep at laboratory
meability of olivine-rich aggregates (Riley et al., strain rates, small grain sizes are essential. The re-
1990; Riley and Kohlstedt, 1990, 1991) indicate sults can be extrapolated to the larger grain sizes
that melt segregation is relatively rapid such that expected for mantle rocks on the basis of consti-
the amount of melt at any instant in the sub-ridge tutive equations for melt-enhanced diffusion creep.
upwelling mantle is less than 1%. The presence of Second, because the distances over which the melt
even this small amount of a melt phase can affect must redistribute are small, the solid-melt micro-
the physical properties of the rock, including not structure reaches, or at least closely approaches,
only the seismic velocity and seismic attenuation textural equilibrium reasonably quickly in these
but also the electrical conductivity and viscosity samples (Cooper and Kohlstedt, 1984a, 1986).
of the aggregate. This point is important because in the mantle the
To date, relatively few experiments have been' rate of melt redistribution is rapid compared to the
carried out to determine the rheology of partially rate of deformation for a rock with a grain size of
molten ultrabasic rocks. On the basis of published 3 mm. In contrast, in the laboratory the situation
results from creep experiments on two-phase ce- is reversed even for a rock with a grain size of
ramic materials such as silicon nitride plus an only 50 ,urn (Kohlstedt, 1992). Consequently, un-
oxynitride silicate glass, large decreases in flow less laboratory samples are allowed to equilibrate
strength might be anticipated. For fine-grained ag- for long time periods at elevated te~peratures
gregates of silicon nitride, the flow strength de- (-100 h for a grain size of 50 ,urn but only -1 h
creases by over two orders of magnitude if a very for a grain size of 5 ,urn), the melt-solid distribu-
small amount of an amorphous phase is added. tion-and possibly also the associated flow be-
For this system, the glass phase forms a thin havior of the aggregate-is unlikely to be repre-
(-l-nm-thick) film along the grain boundaries sentative of that of a partially molten mantle rock.
(for a review, see Kohlstedt, 1992). The melt dis-
tribution in two-phase aggregates of olivine plus
basalt is quite different, however. In this case, the Experimental Procedures
melt forms an interconnected network along triple
junctions (three-grain edges) and through four- The olivine used in this study was obtained from
grain junctions (tetrahedral four-grain corners) but San Carlos, Arizona. Small pieces of olivine, gen-
largely does not wet the grain boundaries (Waff erally <7 mm in their longest dimension, were
and Bulau, 1979; von Bargen and Waff, 1988; washed in acid and water to improve their clarity
Waff and Faul, 1992). Uniaxial hot-pressing ex- and to remove the bulk of other adhered materials.
periments on fine-grained aggregates of olivine Fragments of this material that appeared clear and
plus a few percent basalt under anhydrous condi- free from contamination by minerals other than
tions at 1 atm total pressure suggest that, at least trace amounts of spinel (i.e., notably free from py-
in the diffusion creep regime, the presence of the roxene and layer silicates) were hand-picked over
melt phase results in a rather modest (factor of a light table. As illustrated in Fig. 1, after the
2-5) increase in creep rate (Cooper and Kohl- selected grains were fractured with a steel jaw
stedt, 1984b). crusher, they were ground -15 g at a time for
To extend this earlier work, the present study 15 min in a small steel shatterbox to produce a fine
investigates the influence of a few percent basaltic powder. This powder was then processed through
melt on the flow properties of polycrystalline oli- a cyclone air classifier set to separate out olivine
vine under hydrous conditions at high tempera- grains smaller than 10 ,urn. This fine-grain-size
 3. Basaltic Melt and Polycrystalline Olivine 39
Table 1
Chemical Compositions of Olivine and Basalts

Mid-ocean
Hot -pressed ridge
STEEL
r -_ _--J~SHATTERBOX~---...J San Carlos Synthetic basalt
GRINDER olivine basalt analogue
(wt%) (wt%) (wt%)

SiO, 40.82 51.54 51.0

COARSE AI,03 14.05 17.6
FRACTION
Fe oxides 9.6 11.03 9.1
NiO 0.23
MgO 49.25 8.13 7.4
CaO 12.54 12.1
Na,O 2.71 2.8
FILTERING Total 99.89 100.00 100.0
AND TO COLD
DRYING ~PRESSING

Figure 1 Schematic diagram illustrating the processing steps

used to prepare fine-grained olivine powders. that of a mid-ocean ridge basalt as illustrated in
Table I (see also Cooper and Kohlstedt, 1984a),
made up 7.1 wt% of the two-phase material. This
fraction, however, always contained some large melt content corresponds to 8.6 vol% at standard
"renegade" olivine grains 40 to 60 /Lm in size as temperature and pressure, if the density of San
well as small quantities of small iron particles in- Carlos olivine (Fo n ) is taken as 3310 kg . m' and
troduced during the crushing and grinding proce- that of mid-ocean ridge basalt as 2740 kg . m'.
dures. These contaminants were then separated Olivine and olivine-basalt powders were cold-
from the fine-grained olivine by two processes. pressed at 200 MPa into IO-mm-diametercylindri-
The iron was removed by vigorously stirring slur- cal pellets with a length-to-diameter ratio of I : I or
ries of -60 g of powder in 1 liter of water for less. These pellets were subsequently isostatically
-1 h with a horseshoe magnet. Typically, the iron hot-pressed in 25-mm lengths with 0.2 ml of de-
recovered on the magnet amounted to -0.03 wt% ionized water in iron-jacketed assemblies (Chopra
of the powdered sample. The large renegade oli- and Paterson, 1981; Paterson et al., 1982) at a
vine grains were then removed by adding the slur- confining pressure of 300 MPa and a temperature
ries to plastic tanks filled with water. The large of l200°C for periods between 15 and 60 min.
grains were allowed to settle, and the suspended Three samples were prepared in this manner: oli-
fine-grained material ("fines") was filtered off vine without basalt, olivine with 0.9 vol% basalt,
and dried. These sample preparation procedures and olivine with 8.6 vol% basalt. In the second
produced finely powdered olivine with angular, sample, the melt fraction decreased from 8.6 to
fragmented shapes as illustrated in the scanning 0.9 vol% during hot-pressing due to melt loss
electron microscope (SEM) micrograph in Fig. 1 through a hole in the upper end of the jacket.
of Chopra (1986). The average grain size was less Cylindrical specimens 7 mm in diameter and
than 10 /Lm, with the particle size distribution 16 mm in length for use in deformation experi-
shown in Fig. 2 of Chopra (1986). ments were fashioned from the hot-pressed cyl-
To introduce the basalt phase, the olivine pow- inders by diamond core drilling, combined with
ders were coated with a thin layer of a gel made grinding and lapping of the ends to ensure par-
from tetraethyl orthosilicate and compounds of allelism and flatness. These specimens were then
magnesium, iron, calcium, aluminum and sodium rejacketed in iron with an additional 0.08 ml
by the method developed by Cooper and Kohl- of deionized water and deformed in constant
stedt (l984a). This gel coating, which after fir- displacement-rate and load-relaxation tests us-
ing at 600°C had a composition corresponding to ing a gas-medium high-pressure apparatus (Pater-
40 David L. Kohlstedt and Prame N. Chopra

son 1970, 1977) at temperatures of 1300 and/or In the deformation experiments on water-
1200°C under a confining pressure of 300 MPa. At bearing samples jacketed in iron, the oxygen fu-
the completion of a deformation experiment, the gacity was buffered at a value determined by the
sample was cooled quickly to lOOO°C to preserve Fe-FeO solid state buffer. At this oxygen fugac-
deformation-induced microstructural features. ity, olivine near the jacket was reduced, forming
The deformation testing procedures used were precipitates of an iron-nickel alloy (see Fig. 13of
similar to those reported by Chopra (1986). How- Karato et at. (1986)) and pyroxene (Mackwell
ever, because of the low flow strength exhibited et al., 1985). In the single-phase samples, the py-
by the olivine plus basalt sample at 1300°C, the roxene fixed the oxide activity at the olivine-
strength of the iron jacket had to be taken into ac- enstatite phase boundary. In the partially molten
count in calculating the differential stress on the samples, the basaltic melt fixed the oxide activity
sample from the measured applied load. The frac- at about the same value.
tion of the applied load that was borne by the After the deformation experiments were com-
jacket was calculated using the flow law and data pleted, specimens were encapsulated in epoxy
given by Frost and Ashby (1982). Their Eq. (2.19) resin and sliced longitudinally. From these slices,
was used to solve for the flow stress of iron at each ultrathin sections of < 10 /.Lm in thickness, pol-
instantaneous strain rate in our deformation tests. ished on both sides, were prepared for optical mi-
At 1200°C, the iron accounted for 20% of the load croscopy. Sections of such thinness were neces-
applied during creep of the sample composed of sary to minimize the number of grains of differ-
olivine plus 8.6 vol% basalt; this result is similar· ent crystallographic orientation superimposed in
to that of Karato et at. (1986) who used an em- the light path. Polished planar surfaces were also
pirical approach to estimate the strength of the prepared from these slices for SEM observations,
iron jacket. At 1300°C, the iron supported 55% of and thin foils were produced by ion-beam thin-
the applied load during deformation of this two- ning for transmission electron microsocpy (TEM)
phase sample. The values of stress reported by analyses. Samples for the SEM were etched in
Chopra (1986), which had not been corrected for H3P04/HCI; this etchant dissolves olivine at a rate
the strength of the jacket, were recalculated for in- that depends upon the orientation of the grain and
clusion in Table 2. hence produces relief at grain boundaries. This

Table 2
Summary of Experimental Conditions and Results

T IT, Er E, d,
Material Run (0C) (MPa) (s -1) (%) (urn)

Olivine Hot-press 1200 7

(I h)
Olivine a Deformation 1200 121r b 1.88 X 10- 4 9.8
1300 49r 1.54 X 10- 4 18.9 12
Olivine + basalt Hot-press >1200-
(15 min)
Olivine + 0.9 Deformation 1200 112 1.40 X 10- 4 13.7 IS
vol% basalt
Olivine + 8.6 Hot-press 1200 7
vol% basalt (I h)
Olivine + 8.6 Deformation 1200 63r 1.40 X 10- 4 8.7
vol% basalt 1300 16r 1.42 X 10- 4 15.5 12

"From Chopra (1986).

bThe "r" indicates that a load-relaxation run followed the constant displacement-rate test
'Temperature probably exceeded 1200'C, because melt escaping from the sample moved the thermocouple.
 3. Basaltic Melt and Polycrystalline Olivine 41
treatment greatly improves both the ease and the
Constont Displacement Rote
quality of observations of grain boundaries and '0
0.. Wet Conditions
at the same time leaves the intergranular basalt ~150 T=1200oC
t~10-'s-'
relatively unaffected and higher than the olivine C/)
rn d~10 J.Lm
grains. The etched material was coated with a Cll

gold-palladium alloy to minimize charge buildup ~ 100

olivine + 0.9 vol% basalt
under the electron beam.
To examine the water content in the deformed
samples, Fourier transformed infrared (FTIR) 8.6 vol% bosolt
spectroscopy was carried out on sections that were
-500,um in thickness and polished on both sides. 5 10 15
Room-temperature absorption spectra were col- Strain (10- 2 )
lected in the wavenumber range 220-450 mm-'. Figure 2 Plot of differential stress versus strain for samples of
To remove the contribution resulting from scatter- olivine, olivine plus 8.6 vol% basalt, and olivine plus 0.9%
ing of the infrared beam at grain boundaries, a basalt deformed at 1200°C.
background correction was made by least-squares
fitting a third-order polynomial to the FTIR data 60
outside the range 280-380 mm'. The concentra- Constant Displocement Rate
tion of OH was estimated from the height of the l50 Wet Conditions
~ T=1300oC
OH-stretching bands in the range 300-370 mm' t~l O-'s-' olivine
rn 40
using the relation proposed by Paterson (1982). rn d~10 J.Lm
...Cll
U) 30
0
Deformation Results :g 20
...Cll
Cll

::: 10 + 8.6 vol% basolt

One sample each of olivine without basalt, olivine 0
plus 0.9 vol% basalt, and olivine plus 8.6 vol% 0
0 2 4 6 8 10
basalt was deformed as part of the present inves- Strain (10- 2 )
tigation. The conditions and results of the isostatic
hot-pressing runs, constant displacement-rate ex- Figure 3 Plot of differential stress versus strain for samples of
olivine and olivine plus 8.6 vol% basalt deformed at 1300°C.
periments, and load-relaxation tests are summa-
rized in Table 2. In addition to the temperature, T,
at which each run was performed, the stress, a., the strength of the olivine-basalt aggregate is
strain rate, Er. and total strain, tr, at the end of the about one-fifth that of the olivine specimen. The
constant displacement-rate tests are given. Also sample containing 0.9 vol% basalt exhibited a
included is the mean grain size, d., at the end of strength approximately equal to that of the olivine
the hot-pressing run and the deformation experi- specimen; however, as described under Discus-
ments. The deformation behavior of the single- sion, the strengths of these two samples should
phase olivine sample has been reported elsewhere not be directly compared, because the grain size
(Chopra, 1986). of the former is -25% larger than that of the
The differential stress versus strain results latter.
from the constant displacement-rate experiments The results of the load-relaxation tests on the
at 1200 and 1300°C are plotted in Figs. 2 and 3, olivine and the olivine plus 8.6 vol% basalt speci-
respectively, for all three samples. The specimen mens are plotted in Figs. 4 and 5 as the logarithm
with 8.6 vol% basalt is substantially weaker than of the decaying differential stress versus the loga-
the basalt-free specimen of the same grain size at rithm of the strain rate. The method for converting
both temperatures. At the lower temperature, the the raw load versus time data into differential
strength of the olivine plus 8.6 vol% basalt speci- stress versus strain rate results is described in de-
men is approximately one-half that of the single- tail in Kohlstedt and Chopra (1987). Data from
phase olivine material; at the higher temperature, the later parts of these experiments (i.e., for strain
42 David L. Kohlstedt and Prarne N. Chopra

-3.5 r--r----.----,---,---....----,----, mens can be described by a power law relation of

-4.0
Load Relaxation the form
(Q)
I[/)
T=1200·C
'--'
d~10 f'm (Tn
-4.5 E = A d m exp - RT '
(l)

-0 -5.0 where E is the strain rate, A a materials parameter,

0::::
c -5.5 (T the applied differential stress, d the grain size,
o Q the activation energy for creep at 300 MPa con-
-b -6.0 fining pressure, R the gas constant, and T the
U)
... olivine absolute temperature (in kelvins); in Eq. (1), n
CJl -6.5
a '" olivine + 8.6 vol% basalt
and m are the stress exponent and grain size ex-
- 7. a L..-~-...J---'-:__-"':_-........,,---:-'-:--'
-0.5 0.0 0.5 1.0 1.5 2.0 ponent, respectively. The data from each of the
log Differential Stress (MPa) load-relaxation tests shown in Figs. 4 and 5 define
Figure 4 Plot of log strain rate versus log differential stress straight lines, the slopes of which correspond to n.
for load-relaxation experiments carried out at 1200·C. At 1200°C, the value of n is 1.3-1.4 for both the
single-phase olivine sample and the two-phase ol-
-3.0 r--.----r---,---,.----.----,----, ivine plus 8.6 vol% basalt sample. At 1300°C, n
I[/) -3.5
Load Relaxation is 1.0-1.1 for both samples. Thus there is no sys-
T=1300·C
'--'
d~1 0 f'm
tematic difference in n between the two types of
-4.0 specimen.
(l)

-0 -4.5 With the limited amount of deformation data

0::::
available here and the observed change in stress
c -5.0
o exponent with temperature, it is not possible to
-b -5.5 determine accurately an activation energy. Analy-
U)

CJl -6.0
• olivine sis of the results from the constant displacement-
a o olivine + 8.6 vol% basalt
rate and load-relaxation tests in terms of Eq. (1)
-6.5 L..J1...-_..L-_-l.._ _L-_...L:-_-'-,,--'
-1.0 -0.5 0.0 0.5 1.0 1.5 yields activation energies in the range 300-
log Differential Stress (MPa) 575 kJ . mol-l (assuming that the grain sizes in the
experiments at 1200 and 1300°C are the same).
Figure 5 Plot of log strain rate versus log differential stress
for load-relaxation experiments carried out at l300·C. An examination of the results in Table 2 and
Figs. 4 and 5 suggests that the activation energy
for creep of the olivine plus 8.6 vol% basalt
rates below -10- 5.5S-I) typically were not plotted sample may be somewhat larger than that for the
because of the scatter in the low-stress, low-strain- single-phase olivine sample, but additional ex-
rate data. This noise is the result of two factors: periments are required to verify this observation.
First, the magnitude of background fluctuations in
stress increases relative to the overall differential
Microstructural Observations
stress level as the latter decreases. These fluctua-
tions arise from the expansion and contraction of Hot-Pressed Material
the loading train and specimen resulting from a
pressure cycling of about 1.5 MPa and a tempera- As illustrated by the SEM micrographs in Fig. 6,
ture cycling of about 1°C in the high-pressure de- the basalt in hot-pressed specimens is concentrated
formation apparatus. Second, the magnitude of the in polygonal-shaped pockets at three- and four-
elastic distortion of the apparatus, which is a nec- grain contacts. The olivine grains are equiaxed
essary input to the calculation of the instantaneous with both gently curving and flat grain bounda-
cross-sectional area of the specimen and hence of ries. At this scale, olivine-basalt interfaces are
the stress, becomes increasingly uncertain with frequently straight-edged. Small olivine grains,
decreasing load because of its variable nonlinear- such as the one in the center of the micrograph in
ity from one sample assembly to the next at very Fig. 6a, which appear in cross section to be largely
low loads (Kohlstedt and Chopra, 1987). surrounded by basalt are not uncommon in these
The high-temperature deformation data for isostatically hot-pressed specimens. Voids were
both the olivine and olivine plus basalt speci- not observed.
 3. Basaltic Melt and Polycrystalline Olivine 43

Figure 6 Scanning electron micrographs of olivine plus basalt sample isostatically hot-pressed at 300 MPa and l200°C. The
sample has been etched such that the melt tends to stand higher than the olivine grains. (a) Secondary electron image. At the
center of the micrograph, a small grain of olivine is nearly surrounded by melt. (b) Composition (i.e., mixed secondary electron
and backscattered electron) mode. One melt-filled triple junction and one large, faceted pocket of melt are marked by the small
and large arrows, respectively.

Deformed Material times. Each thin section was approximately 1.2 X

0.8 mm in area and 10 /-Lm in thickness.
Optical Microscopy The grains in the deformed single-phase olivine
Longitudinal thin sections of each of the de- specimen are approximately equant and even-sized
formed samples were observed with an optical mi- with an average size of approximately 12 /-Lm. A
croscope operating at magnifications up to 1000 few pores are located along the grain boundaries.
 3. Basaltic Melt and Polycrystalline Olivine 45
Over the area of the thin section approximately tures with the single-phase olivine specimen. The
100 nearly equant grains with apparent dimen- grain size of the olivine is again bimodal. In this
sions between 30 and 75 ,urn are interspersed ran- case, 1000 intercepts measured parallel to the di-
domly among this fine-grained material. Many of rection of maximum principal stress yielded a
these larger grains contain within them subspheri- mean grain size of 12.0 ,urn, while similar mea-
cal shells of pores. In section, these shells appear surements perpendicular to 0", gave 12.3 ,urn. The
as rings, similar to those described in Figs. 3 and anisotropy in grain shape for the two-phase sample
5 of Cooper and Kohlstedt (1984a). Some of these amounts to ~2.59'0 flattening. The microstructure
rings are nearly equant, while others define elon- of this specimen is also characterized by a general
gated or angular shapes. The pores making up lack of dislocation-associated features. The only
these rings were originally located at the bounda- evidence of dislocations observed was a single tilt
ries of growing olivine grains but became trapped wall with a _4° misorientation in one of the large
in their hosts when the mobility of the spreading grains.
grain boundaries exceeded their own mobilities. There are again approximately 100 large and
These rings of pores thus define upper limits to the generally equant grains of largest dimension be-
sizes of the original grains from which their hosts tween 30 and 75 ,urn. These grains generally con-
grew. The size of the precursors defined in this tain some captured pores, although the concentra-
way is always approximately 20 ,urn in at least one tions of these pores are invariably much lower
direction. than those in the single-phase olivine sample. In
Measurements of the size of the fine-grained the five grains observed in which the distribution
olivine which makes up the vast majority of the of poresdoes define a ring, this ring is -20 ,urn in
specimen were made using the linear intercept its largest dimension. At 1000 times magnification
method of Exner (1972) with a sectioning cor- in reflected light, small melt bodies can be seen
rection factor of 1.5. One thousand measure- along many grain boundaries. However, the clar-
ments made parallel to the direction of the maxi- ity of the observations is restricted by the limits of
mum principal stress, 0"\, yielded a grain size of resolution of the optical microscope.
11.5 ,urn, while a corresponding number of mea- The microstructure of the olivine plus 0.9 vol%
surements made perpendicular to 0"\ yielded an basalt sample is similar to those of the two samples
estimate of 12.6 ,urn. This anisotropy in grain described above. The primary difference, in this
shape is oriented in such a way as to be consis- case, is that the grain size (15 ,urn) is somewhat
tent with an origin by flattening during the defor- larger for this sample. Grain growth likely oc-
mation. However, the extent of this anisotropy, curred during hot-pressing, since the temperature
-9%, is less than one-half the total macroscopic probably increased substantially above 1200°C
strain of the specimen, -19%. The microstruc- when the melt that escaped from the sample moved
ture of this deformed single-phase specimen lacks the thermocouple several millimeters away from
dislocation-associated features (e.g., undulatory the sample.
extinction, deformation bands, and subgrain walls)
although a firm conclusion in this regard is made Scanning Electron Microscopy
difficult by the small grain size and the problem, As illustrated in the secondary electron images
even in this ultrathin section, of superimposed .in Fig. 7, melt is located in all of the three- and four-
grains. grain junctions; however, some larger melt pockets
The deformed olivine plus 8.6 vol% basalt do exist. Except for these large melt pockets, the
specimen shares a number of microstructural fea- melt does not form a continuous film along the

Figure 7 Secondary electron images of olivine plus 8.6 vol% basalt sample deformed at 300 MPa at both 1200 and 1300°C.
The sample has been etched so that the melt tends to stand higher than the olivine grains. A region in (a) is viewed at higher
magnification in (b), and a region in (c) is viewed at higher magnification in (d). In (a), three melt-filled triple junctions and three
large, faceted pockets of melt are marked by short and long arrows, respectively. In (b), three melt-free grain boundaries are
marked by curved arrows and two melt pockets separating neighboring grains are marked by straight arrows. In (c). note that the
grain boundaries between melt-filled triple junctions are all free of melt; this point is made particularly clear with the high-
magnification view in (d).
46 David L. Kohlstedt and Prame N. Chopra

grain boundaries (i.e., two-grain junctions), within melt distribution relative to the direction of O'j was
the resolution of the SEM « 10 nm). observed.
The mean dihedral angle (), determined from
tangents drawn to olivine grains in triple junctions, Transmission Electron Microscopy
is -30°. Hence, the melt is interconnected along The dark-field (DF) and bright-field (BF) TEM
triple junctions (e.g., Bulau et al., 1979), a point images of the microstructure of the sample of oli-
verified by the presence of melt in all of the triple vine plus 8.6 vol% basalt reinforce the optical and
junctions (see also Daines and Richter, 1988). scanning electron microscopy observations made
Careful examination of a large number of mi- above. First, the melt is present in all the triple
crographs revealed no apparent influence of dif- junctions. While some of the melt-solid inter-
ferential stress or finite strain on the melt dis- faces are smoothly curved as expected for a two-
tribution. The melt distribution in the deformed phase system with isotropic interfacial energies,
samples appeared identical to that of the unde- others are flat or faceted (Fig. 8a; see also Fig. 7a).
formed samples. In addition, no anisotropy in The flat interfaces involve low-index crystallo-

Figure 8 (a) Dark-field TEM images of melt-filled triple junctions. The images were formed using electrons diffusely scattered
from the melt, so that the melt phase is lighter than the olivine grains. The dark region near the middle of one of the melt-filled
triple junctions (southwest quadrant) is a hole produced by ion thinning. Both straight and smoothly curved basalt-olivine
interfaces are present. (b) Bright-field TEM image of grain boundary containing very small pockets of trapped melt, several of
which are marked by short arrows. The grain boundary, which runs horizontally in this micrograph, is bounded at either end by
a melt-filled triple junction, marked by long arrows. The fringes running parallel to the boundary indicate that the boundary is
inclined at an angle of -30 0 to the incident electron beam. (c) Dark-field TEM image showing the transition from a triangular-
shaped melt-filled triple junction to a melt-free grain boundary. The transition in the upper left of the micrograph is marked by a
pair of filled arrows, and the triple junction is marked by an open arrow.
continues
 3. Basaltic Melt and Polycrystalline Olivine 47
graphic planes of olivine, such as (010) and {110}, high-resolution analytical electron microscopy ob-
indicative of marked anisotropy in the crystal servations demonstrate that basaltic melt does not
structure of olivine (Waff and Faul, 1992). generally form a continuous film along the bound-
Second, in rare cases small «O.I-fLm) pock- aries between olivine grains (Vaughan et al., 1982;
ets of melt were trapped in a grain boundary Kohlstedt, 1990). Although the transition from the
(Fig. 8b). However, high-resolution TEM and triple junction to the grain boundary is generally

Figure 8 Continued.
Figure 9 (a) Dark-field, diffuse-scattering TEM images illustrating typical microstructure of the deformed sample of olivine
plus 8.6 vol% basalt. Melt-filled triple junctions appear as bright, triangular-shaped regions and are marked by solid arrows. The
dislocations are straight or gently curved lines within the olivine grains. (b) In the lower micrograph, a low-angle boundary,
marked by an open arrow, divides the grain along a line running NE-SW. A trapped pore, marked by a curved arrow, can be
seen along this boundary.
 3. Basaltic Melt and Polycrystalline Olivine 49
6
difficult to identify precisely (e.g., Fig. 8a), it is -250 H/ I0 Si, for the thermodynamic condi-
occasionally clearly delineated, as illustrated in tions used in the present study (Bai and Kohlstedt,
Fig.8c. 1992). Hence, the olivine is apparently saturated
Third, dislocations were observed in most, but with water-derived species, and most of the water
not all, of the olivine grains. In grains which con- in these samples is distributed in the melt and
tained dislocations, the dislocation density was in trapped in the pores. For the olivine plus 8.6 vol%
the range 1 X 10 12 to 5 X 1013 m' (Fig. 9). The basalt sample, the absence of these peaks in the
dislocation density in natural San Carlos crystals 300-370 mm' range indicates that the concen-
is ~ 1010 m', tration of OH-related species within the olivine
grains is relatively small. For this sample, most of
the OH-related defects have presumably gone into
Infrared Analyses the melt, since the olivine/melt partition coeffi-
cient may be as small as 0.01 (Michael, 1988;
As demonstrated by the FTIR spectra in Fig. 10, Dixon et al., 1988).
all three samples contain significant concentra-
tions of OH-related species. In each case, the total
concentration is >2500 H/10 6 Si. The OH-related Discussion
species are present within the olivine grains, along
grain boundaries, in the quenched melt, and in Microstructural Characteristics
bubbles (Karato et al., 1986). For the olivine
without basalt and the olivine with 0.9 vol% ba- The similarity in final grain sizes of our specimens
salt samples, the peaks in the wavenumber range without basalt and with 8.6 vol% basalt from both
300-370 mm' are primarily associated with OH- the hot-pressing experiments and the deformation
related species in the olivine grains and possibly experiments (Table 2) suggests that under hydrous
along grain boundaries. Experiments on single conditions the presence of basaltic melt in the
crystals of olivine demonstrate that the maxi- two-phase specimens does not have a large effect
mum concentration in the olivine grains is only on the kinetics of olivine grain growth. A similar
conclusion can be drawn from the observation
that the precursors to the large olivine grains,
rr>;

'I 3.0 ,.--,-,-,-,-,.--,......,,......,,......,--,--.--.--.--.--, as defined by the size of rings of pores, are simi-
E FTIR Spectra lar irrespective of whether the matrix material is
S2.5 single-phase olivine or two-phase olivine plus
+'
~ 2.0
basalt. However, it must be borne in mind that
o the melt-free sample was saturated with water,
:t:
(J)
1.5 while the melt-rich sample was undersaturated.
o By comparison, grain growth during hot-pressing
U 1.0
c under anhydrous conditions is much more rapid in
o
:;:::; 0.5 samples of olivine plus basalt than in samples of
0..
I....
olivine only (Cooper and Kohlstedt, 1986).
g 0.0 4LO::-:O~~~'--::-':-:-'--'---'---'--::-"'-::-''---~'---:-::-:'
~ The microstructure of the olivine plus 8.6 vol%
350 300 250
Wavenumber (rnrn ") basalt sample reflects a marked anisotropy in the
solid-melt interfacial energy. A large fraction of
Figure 10 FTIR spectra from the three samples after the de- the olivine-basalt interfaces are flat (see also Waff
formation experiments. Clear peaks are present near wave-
numbers of 340 and 350 mm- l in the spectra from the single-
and Faul, 1992). For an isotropic solid, the solid-
phase olivine and the two-phase olivine plus 0.9 vol% basalt melt interfaces should be curved.
samples, indicating that OH-related species are present in the Finally, it should be emphasized that, while
olivine grains. These peaks are absent from the spectrum from melt does not form a continuous film along most
the sample of olivine plus 8.6 vol% basalt, suggesting that most of the grain boundaries, it is present in all of the
of the water partitioned into the melt. (Note that the sinusoidal
oscillations in the spectrum from the sample with 8,6 vol%
triple junctions. This observation, which is con-
basalt are interference fringes that occur because this sample sistent with the measured dihedral angle of -30°,
is quite thin, -300JLm.) demonstrates that the melt forms an intercon-
50 David L. Kohlstedt and Prame N. Chopra

nected network along the three-grain junctions graph, values of n = 1 and m = 3 were used in
and through the four-grain junctions. It should Eq. (I) for this purpose. At 1300°C, the sample
also be emphasized that some grains are separated with 8.6 vol% basalt is a factor of -5 weaker than
on at least one side from a neighboring grain by the melt-free sample. A similar decrease in flow
large, faceted pockets of melt (Figs. 6 and 7). strength (a factor of 2-5) was determined from
uniaxial hot-pressing experiments on fine-grain
Rheological Behavior olivine-basalt aggregates under anhydrous con-
ditions (Cooper and Kohlstedt, 1984b). Note that
The load-relaxation results obtained with both the a similar comparison cannot be readily made for
single-phase olivine specimen and the two-phase the creep results obtained at 1200°C because the
olivine plus 8.6 vol% basalt specimen can be ade- grain size for two of the samples was measured
quately described by a power-law relation be- only after the experiment at 1300°C.
tween strain rate and differential stress, Eq. (I). To describe the creep behavior of partially
The stress exponent, n, determined for both the molten aggregates with 0 < (J ::S 60° by solution-
olivine and the olivine plus basalt specimens is precipitation enhanced diffusion creep, Cooper
1.0-1.1 at 1300°C (Fig. 5). This Newtonian (i.e., et at. (1989) analyzed the effects of introducing
n = I) behavior indicates that diffusion dominates melt into the triple junctions. They concluded
the creep process. However, at 1200°C some con- that, although the rate of deformation is enhanced
tribution to the deformation process due to dislo- by short-circuit diffusion through the melt-filled
cation creep, for which n = 3.5 in olivine (Chopra' triple junctions, the deformation process is rate
and Paterson, 1984; Karato et al., 1986; Bai et al., limited by diffusion through the melt-free grain
1991), is suggested by the fact that the stress ex- boundaries. Their model is valid as long as the
ponents are greater than unity (n = 1.3-1.4, length of grain boundary perpendicular to the
Fig. 4) for both types of samples. This point is triple junction that is lost to the melt phase, !::J.d, is
reinforced by the flattening of the olivine grains. small compared to the total length of the grain
For the fine-grained, single-phase olivine sam- boundary (i.e., as long as (J is not too small and
ple deformed under hydrous conditions as part the melt fraction ¢ is not too large). In this case,
of the present study, the flow stress measured at the strain rate of the two-phase aggregate relative
1300°C and -10.4 s' agrees within 10% with the to that of the single-phase aggregate is given by
value reported by Karato et at. (1986) for samples 4
crept under similar conditions. Likewise, the stress (;2 d
(;1 = [ (d - !::J.d) ] ,
(2)
exponents (1.0 < n < 1.4) obtained here are in
good agreement with that (n = 1.4) determined
by Karato et at. (1986). On the basis of a stress where!::J.d is a function of both (J and ¢, and where
exponent near unity and a grain size exponent E, is the strain rate of the single-phase aggregate
in Eq. (I) of m = 3, Karato et at. (1986) con- and E2 is the strain rate of the two-phase aggregate.
cluded that under hydrous conditions their fine- Within the context of this model for solution-
grained samples deformed by grain boundary precipitation enhanced diffusion creep, the results
diffusion creep (Coble, 1963). Hence, we argue reported here for creep at I 300°C of aggregates of
that the dominant deformation mechanism in our olivine plus 8.6 vol% basalt and of melt-free oli-
high-temperature experiments must also be grain vine are compared in Fig. 11. The reasonably
boundary diffusion (i.e., Coble creep). good agreement between the experimental results
The presence of the 8.6 vol% basaltic melt in and the model suggests that the strain-rate en-
the composite specimen results in a pronounced hancement observed in the partially molten oli-
lowering of the flow strength relative to the speci- vine plus basalt aggregate under hydrous condi-
men without basalt and the specimen with only a tions might indeed be a result of short-circuit
small amount (0.9 vol%) of basalt. To make a diffusion through the melt-filled triple junctions,
quantitative comparison of the data for the three as described by Cooper et at. (1989).
samples, however, the results in Table 2 must be A comparison of the rheological behaviors of
normalized to a common stress and grain size. On the melt-free and partially molten samples may be
the basis of the discussion in the previous para- more complicated, however. The FTIR spectra in
 3. Basaltic Melt and Polycrystalline Olivine 51
some grains from at least one of their neighboring
30 ,-T"""""""""""""''''''''''''''''''''''''''''''''''''-''-'-''''''''''''''''''''''''''''''''''''''''
Diffusion Creep grains (Figs. 6 and 7), rather than in triple junc-
25 0<0;>60°
T=13000C tions as assumed in the model.
20 The suggestion that the single-phase sample
'LV
deformed under fully hydrous conditions while
<, 15 the two-phase (8.4 vol% basalt) sample deformed
N
'LV
10
under more nearly anhydrous conditions is sup-
ported by the observation that the activation en-
5 ergy for creep of the former is smaller than that
for the latter. Such a relationship between water
0.10 0.15 0.20 content and activation energy for creep has been
¢ noted previously for dunite (Chopra and Paterson,
1984; Karato et al., 1986). It should also be noted
Figure 11 Normalized strain rate of a partially molten, two- that many ceramic materials contain a melt phase
phase aggregate (1'2) relative to that of the melt-free aggregate
(1',) versus melt fraction (rf». The solid curves are from the
that forms a thin (1- to 2-/Lm-thick) film that en-
model of Cooper et al. (1986) for a variety of dihedral angles, tirely coats the grain boundaries at high tempera-
e. The solid circle was calculated using the creep results ob- tures. For such fine-grained ceramics, the pres-
tained in the present study for the olivine plus 8.6 vol% basalt ence of the melt phase results in a very large
sample and the single-phase olivine sample. Data were cor- (> 102 ) increase in strain rate, even if the melt
rected to a common stress and a common grain size using
n = I and m = 3, respectively, in Eq. (I). The dihedral angle
fraction is less than 1 vol% (e.g., Hwang and
for the sample of olivine plus basalt is - 30°. Chen, 1990; Kohlstedt, 1992). In this case, how-
ever, the large enhancement in strain rate occurs
because diffusion through the melt-filled grain
boundaries is orders of magnitude faster than dif-
Fig. 10 suggest that in the single-phase sample the fusion through melt-free grain boundaries,
olivine grains and grain boundaries are probably At present, it is premature to extrapolate the
saturated with water-derived species, whereas in creep results on partially molten olivine aggre-
the two-phase sample most of the water parti- gates from laboratory to geological conditions.
tioned into the melt, leaving the olivine under- Substantial uncertainties still exist in the depen-
saturated. Hence a direct comparison of the two dence of viscosity on temperature, melt fraction,
samples may not be appropriate. In fact, for de- and water fugacity. Nonetheless, the present study
formation experiments carried out under anhy- and those of Hirth and Kohlstedt (1992a, 1992b)
drous conditions in the diffusion creep regime, and Beeman and Kohlstedt (1993) demonstrate
partially molten samples containing -8 vol% of that the rheology of partially molten aggregates
either a natural MORB or a synthetic MORB crept both in the diffusion creep regime and in the dis-
about 16 or 50 times faster, respectively, than their location creep regime will depend sensitively on
single-phase counterparts of the same grain size melt fraction and, hence, on the permeability of
(Hirth and Kohlstedt, 1992a). Consequently, the the deforming rock. If melt segregation is rapid
difference between the creep rate of the single- enough that the amount of melt in a partially mol-
phase sample and that of the two-phase sample ten upwelling mantle rock remains less than about
might be significantly larger than that measured in 1%, as suggested by geochemical measurements
our experiments, if the latter were fully saturated (Salters and Hart, 1989; Johnson et al., 1990),
with water (i.e., if the effects of melt and water on then the effect of melt on viscosity will be small.
creep strength were combined). This point was However, if melt is prevented from migrating out
anticipated by Karato (1986). Such a deviation of the rock mass such that several percent melt ac-
from the model of Cooper et al. (1989) might well cumulates, then the viscosity may decrease sub-
be expected once the melt fraction exceeds the stantially more rapidly with increasing melt frac-
equilibrium value, which is -2.3 vol% of the tion than predicted by models for deformation
olivine-basalt system (Riley, 1990). At larger of two-phase aggregates (for diffusion creep, see
melt fractions, a significant portion of the melt is for example Cooper et al. (1989); for dislocation
present in large, faceted pockets that separate creep, see for example Chen and Argon (1979».
52 David L. Kohlstedt and Prame N. Chopra

Conclusions Deformation: Laboratory Studies. The Paterson Volume"

(8. E. Hobbs and H. C. Heard, eds.), pp. 25-31, American
Geophysical Union, Washington, DC.
Under hydrous conditions, the melt in aggregates Chopra, P. N., and Paterson, M. S. (1981). The experimental
of olivine plus basalt is largely restricted to three- deformation of dunite, Tectonophys. 78,453-473.
grain and four-grain junctions, which form an in- Chopra, P. N., and Paterson, M. S. (1984). The role of water in
terconnected network. Melt does not wet most of the deformation of dunite, J. Geophys. Res. 89, 7861-
7876.
the grain boundaries as a thin continuous film, but
Coble, R. L. (1963). A model for boundary-diffusion con-
large pockets of melt do separate some grains trolled creep in polycrystalline materials, 1. Appl. Phys.
from at least one of their neighboring grains. 34,1679-1682.
In the diffusion creep regime at l300°C, the Cooper, R. F.. and Kohlstedt, D. L. (l984a). Sintering of oli-
presence of 8.6 vol% melt reduced the viscosity vine and olivine-basalt aggregates, Phys. Chern. Minerals
11,5-16.
of an olivine-basalt rock by a factor of -5 rela-
Cooper, R. E, and Kohlstedt, D. L. (I 984b). Solution-
tive to that of a single-phase olivine aggregate of precipitation enhanced creep of partially molten olivine-
the same grain size. However, while the olivine basalt aggregates during hot-pressing, Tectonophys. 107,
aggregate was saturated with water, the olivine- 207-233.
basalt sample was undersaturated because the Cooper. R. E, and Kohlstedt, D. L. (1986). Rheology and
structure of olivine-basalt partial melts. 1. Geophys. Res.
water partitioned strongly into the melt phase.
91,9315-9323.
The difference in viscosity would likely be much Cooper, R. E, Kohlstedt, D. L., and Chyung, K. (1989).
larger if both samples were deformed at the same Solution-precipitation enhanced creep in solid-liquid ag-
water fugacity. gregates which display a non-zero dihedral angle, Acta
Metall.37,1759-1771.
Daines, M. J., and Richter, E M. (1988). An experimental
method for directly determining the interconnectivity of
Acknowledgments melt in a partially molten system, Geophys. Res. Lett. 15,
1459-1462.
Dixon, 1. E., Stolper, E., and Delaney, J. R. (1988). Infrared
Support from the National Science Foundation through Grants
spectroscopic measurements of CO, and H,O in Juan de
EAR-8916438, OCE-9200471, and EAR-9220039 are grate-
Fuca Ridge basaltic glasses, Earth Planet. Sci. Lett. 90,
fully acknowledged. The authors are indebted to Mervyn Pat-
87-104.
erson for the use of his high-pressure rock deformation labo-
Exner, H. E. (1972). Analysis of grain and particle-size distri-
ratory and to Martha Daines, Greg Hirth, Shun Karato, Steve
butions in metallic materials, Int. Metall. Rev. 17,25-42.
Mackwell, and Mark Zimmerman for stimulating discussions
Frost. H. J., and Ashby, M. E (1982). "Deformation Mecha-
and insightful comments. The thoughtful suggestions of two
nism Maps, The Plasticity and Creep of Metals and Ce-
anonymous reviewers helped improve this manuscript.
ramics," Pergamon Press, Oxford.
Hirth,1. G., and Kohlstedt, D. L. (I 992a). The transition from
dislocation creep to diffusion creep in partially molten oli-
vine aggregates, EOS Trans. Am. Geophys. Union 73, 529.
References Hirth. 1. G., and Kohlstedt, D. L. (l992b). The effect of melt
on the strength of olivine aggregates deformed in the dif-
Bai, Q., and Kohlstedt, D. L. (1992). Substantial hydrogen fusion creep regime, EOS Trans. Am. Geophys. Union 73,
solubility in olivine and implications for water storage in 517.
the mantle, Nature 357, 672-674. Hwang, C; and Chen, 1. W. (1990). Effect of a liquid phase on
Bai, Q., Mackwell, S. 1., and Kohlstedt, D. L. (1991). High- superplasticity of 2-mol% Y,O,-stabilized tetragonal zir-
temperature creep of olivine single crystals. I. Mechanical conia polycrystals. 1. Am. Ceram. Soc. 73, 1626-1632.
results for buffered samples, J. Geophys. Res. 96, 2441- Johnson, K. T. M., Dick, H. 1. B., and Shimizu. N. (1990).
2463. Melting in the oceanic upper mantle: An ion microprobe
Beeman, M. L., and Kohlstedt, D. L. (1993). Deformation of study of diopsides in abyssal peridotites, J. Geophys. Res.
fine-grained aggregates of olivine plus melt at high tem- 95,2661-2678.
perature and pressures, J. Geophys. Res. 98, 6443-6452. Karato, S. 1. (1986). Does partial melting reduce the creep
Bulau,1. R., Waff, H. S., and Tyburczy, 1. A. (1979). Mechani- strength of the upper mantle? Nature 319, 309-310.
cal and thermodynamic constraints on fluid distribution in Karato, S. 1., Paterson, M. S., and FitzGerald, 1. D. (1986).
partial melts, J. Geophys. Res. 84, 6102-6108. Rheology of synthetic olivine aggregates: Influence of
Cheri.L w., and Argon, A. S. (1979). Steady state power-law grain size and water, J. Geophys. Res. 91, 8151-8176.
creep in heterogeneous alloys with coarse microstructures, Kohlstedt, D. L. (1990). Che'mical analysis of grain boundaries
Acta Metall. 27, 785-791. in an olivine-basalt aggregate using high-resolution, ana-
Chopra, P. N. (1986). The plasticity of fine-grained olivine at lytical electron microscopy, in "The Brittle-Ductile Tran-
high pressure and temperature, in "Mineral and Rock sition in Rocks: The Heard Volume" (A. G. Duba, W. B.
 3. Basaltic Melt and Polycrystalline Olivine 53
Durham,1. W. Handin, and W. F. Wang, eds.), pp. 211- pressure rock-deformation experiments, High Temp. High
218, American Geophysical Union, Washington. DC. Press. 14, 315-318.
Kohlstedt, D. L. (1992), Structure, rheology and permeability Riley, G. N., Jr. (1990). "Liquid Distribution and Transport
of partially molten rocks at low melt fractions, ill "RIDGE in Silicate Liquid-Olivine Materials," Ph.D. Dissertation,
Volume on Melt Production and Migration" Geophys. Cornell Univ., Ithaca, NY,
Monogr. 71 (1.Phipps-Morgan, D. K, Blackman, and 1. M, Riley, G, N" Jr" and Kohlstedt, D. L. (1990). An experimen-
Sinton eds.), pp. 103-121, American Geophysical Union, tal study of melt migration in an olivine-melt system,
Washington, DC. ill "Magma Transport and Storage" (M. P. Ryan, ed.),
Kohlstedt, D. L., and Chopra, P. N. (1987). High temperature pp. 77-86, Wiley, Chichester/Sussex, England.
apparatuses for rock deformation, ill "Methods in Experi- Riley. G. N., Jr., and Kohlstedt, D. L. (1991). Kinetics of melt
mental Physics-Geophysics" (G. G. Sammis and T. L. migration in upper mantle-type rocks, Earth Planet. Sci.
Henyey, eds.), pp, 57-87, Academic Press, San Diego. Lett. 105,500-521.
Mackwell, S. 1.. Kohlstedt, D. L., and Paterson, M. S. (1985). Riley, G. N., Jr., Kohlstedt, D. L., and Richter, F. M. (1990).
The role of water in the deformation of olivine single crys- Melt migration in a silicate liquid-olivine system: An ex-
tals, 1. Geophys. Res. 90, 11319-11333. perimental test of compaction theory, Geophys. Res. Lett.
Michael, P. J. (1988), The concentration, behavior and storage 17,2101-2104.
of H 2 0 in the suboceanic upper mantle: Implications for Salters. V. J. M., and Hart, S. R. (1989). The hafnium paradox
mantle metasomatism, Geochim. Cosmochim. Acta 52, and the role of garnet in the source of mid-ocean-ridge ba-
555-566. salts, Nature 342, 420-422.
Paterson, M. S. (1970). A high pressure, high temperature ap- Vaughan, P. J., Kohlstedt, D. L., and Waff, H. S. (1982). Dis-
paratus for rock deformation, lilt. J. Rock Mech. Min. Sci. tribution of the glass phase in hot-pressed, olivine-basalt
7,517-526. aggregates: An electron microscopy study, Contrib. Min-
Paterson. M. S. (1977). Experience with an internally heated eral. Petrol. 81, 253-261.
gas-medium apparatus to 500 MPa, ill "Proceedings of the Von Bargen, N., and Waff', H. S. (1988), Wetting of enstatite
2nd International Conference on High Pressure Engineer- by basaltic melt at 1350°C and 1.0- to 2,5-GPa pressure,
ing, Brighton, 1975," pp. 209-213, Institution of Me- J. Geophys. Res. 93, 1153-1158.
chanical Engineers, London. Waff, H. S., and Bulau, 1. R. (1979). Equilibrium fluid distri-
Paterson, M. S. (1982). The determination of hydroxyl by in- bution in an ultramafic partial melt under hydrostatic stress
frared absorption in quartz, silicate glasses and similar conditions,1. Geophys. Res. 84, 6109-6114.
minerals, Bull. Mineral. 105,20-25, Waff, H. S., and Faul, U. H. (1992). Effects of crystalline an-
Paterson, M, S" Chopra, P. N" and Horwood, G. R. (1982). isotropy on fluid distribution in ultramafic partial melts,
The jacketing of specimens in high-temperature, high- J. Geophys. Res. 97, 9003-9014.
This page intentionally left blank
 Chapter 4 The Generation and Migration of Partial Melt
beneath Oceanic Spreading Centers

David W. Sparks and E. M. Parmentier

Overview Units

M magnitude of mass anomaly kg' m ?

The narrowness of the zone of crustal emplacement at
V along-layer melt velocity m- S-I
oceanic spreading centers indicates that melt formed
in the mantle beneath a spreading center must be fo- W mantle upwelling velocity m- S-I

cused toward the spreading axis. Focusing can be ac- X degree of melting dimensionless
complished by lateral flow in sloping decompacting Xo maximum degree of melting in upwell- dimensionless
boundary layers at the top of the melting region. These ingcolumn
layers form by decompaction or dilation of the solid T temperature °C
matrix, in response to an excess melt pressure that de- g gravitational acceleration m S-2
velops beneath an impermeable boundary. The rela-
h seafloor topography m
tively high melt fraction and permeability in the bound-
ary layer allow melt to flow upslope to the spreading h, oceanic crustal thickness m
axis. The extraction efficiency of decompacting layers k matrix permeability m2
is determined by the temperature and upwelling distri- ko constant in permeability relation, m2
bution in the mantle. Eg. (4)
Mantle flow and melting beneath an offset spreading n exponent in permeability relation, dimensionless
center has been investigated using three-dimensional Eg. (4)
numerical experiments. The flow consists of the super- p pressure kg·m- l·s- 2
position of passive flow driven by plate-spreading and
time
buoyant flow driven by thermal expansion and compo-
u mantle velocity m- S-1
sitional density gradients due to melt extraction. Buoy-
ant flow affects the distribution of melt production by Ur melt velocity rn- s I

increasing the along-axis variation in upwelling be- u, solid matrix velocity m· S-I

neath the spreading center, and generating melting in z depth m

the upwelling limbs of off-axis thermally driven con- r melt production rate S-I
vective rolls. Approximate paths for the extraction of
a coefficient of thermal expansion °C-I
melt along decompacting layers at the top of the melt-
ing region are calculated. These paths predict the divi- f3 compositional density parameter, dimensionless
Eg. (17)
sion of the melting region into separate melt extrac-
tion regions for different spreading segments. Predicted y constant in melting relationship, °C-I

along-axis distributions of crustal thickness, topogra- Eg. (8)

phy, and gravity have greater variation at slow spread- 8 decompaction length scale m
ing rates, in agreement with observations. s, freezing length scale m
8, compaction length scale m
Notation ? bulk viscosity of solid-melt aggregate Pa·s
Units
() slope of solidus °C· m- I
C, specific heat J. kg-I. C-I K thermal diffusivity m 2 • S-l

G gravitational constant m'·kg- l,s·· 2 /1-r melt viscosity Pa- s

L latent heat of fusion J . kg " ' /1-, mantle shear viscosity Pa- s

Magmatic Systems Copyright © 1994 by Academic Press. Inc.

Edited by M. P. Ryan 55 All rights of reproduction in any form reserved.
56 David W. Sparks and E. M. Parmentier

Units Morgan and Forsyth, 1988; Shen and Forsyth,

degree of melt depletion
1992).
~ dimensionless
The major sources of buoyancy in the mantle
P mantle density kg m :"
v

beneath spreading centers are thermal expansion,

Pc oceanic crustal density kg'm- 3
the presence of low-density melt in pore spaces,
Pr melt density kg m " '
v

and compositional variations caused by the ex-

pw seawater density kg' m ? traction of melt. During the partial melting of pe-
Po reference mantle density kg m " '
v
ridotite, Fe is preferentially partitioned into the
a wavenumber m' ' melt phase. When the melt is extracted, the re-
l/> melt fraction dimensionless maining residual mantle is depleted in iron and is
l/>o melt fraction in decompacting layer dimensionless compositionally less dense. The density change
w slope of decompacting layer dimensionless associated with 25% melt extraction is equivalent
to the density change associated with 200°C ther-
mal expansion (Oxburgh and Parmentier, 1977).
Mantle that upwells beneath the axis experiences
Introduction more melting than mantle that upwells off-axis.
As this mantle is advected through the melting re-
Spreading centers are the largest regimes of melt gion, a gravitationally stable density stratification
generation in the Earth. Spreading is relatively is developed.
continuous in time, and the chemistry of the source The most complete studies of buoyant flow
material as well as the mantle flow is relatively and melting have been two-dimensional. These
uniform compared with melting regimes in hot- studies have shown that thermal and composi-
spot plumes and in continental systems. For these tional buoyancy enhance upwelling rates beneath
reasons, spreading centers are perhaps the sim- a spreading center, but do not qualitatively change
plest systems in which to model the generation the form of flow beneath the axis from that of pas-
and transport of magma. However, most of the sive upwelling (Sotin and Parmentier, 1989; Scott
available observational information about the na- and Stevenson, 1989; Cordery and Phipps Mor-
ture of melt production and migration at spreading gan, 1992). The buoyancy due to the presence of
centers is indirect and is based on seismic and a significant fraction of melt in the mantle can
gravity surveys, and sonar imaging of the seafloor. cause the upwelling to become much more fo-
One of the fundamental observations of spread- cused, due to positive feedback from increased
ing centers is that over 90% of the oceanic crust is rates of melt production. If the permeability of the
emplaced within a 1- to 2-km-wide neovolcanic mantle is low enough that a melt fraction of a few
zone (Macdonald, 1984). Either the upwelling percent is retained in the mantle, then the melting
which produces melt is extremely focused be- region narrows to a region that is only 20-40 km
neath a spreading center or the melt is generated wide (Rabinowicz et al.. 1984; Scott and Steven-
in a tens to hundreds of kilometers wide region son, 1989; Cordery and Phipps Morgan, 1992). If
and then focused into the neovolcanic zone. even higher melt fractions are retained (2: 15%) a
Several studies have investigated mantle flow large reduction in viscosity that can focus upwell-
beneath spreading centers driven by passive rift- ing into a very narrow region, on the order of the
ing and local buoyancy forces. Passive plate- neovolcanic zone (Buck and Su, 1989), may re-
driven flow consists of a broad upwelling region sult. If melt generation were so strongly focused,
and 100- to 300-km-wide melt production region no lateral melt migration would be required since
centered beneath the spreading axis (Reid and all melting would occur essentially beneath the
Jackson, 1981). The width of the melt production neovolcanic zone. However, the interconnectivity
region and the amount of melt produced increases of melt at very small melt fractions « I%, Daines
with increasing spreading rate. Passive upwelling and Richter, 1989), and trace element (Johnson
and melting are relatively uniform along most of et al., 1990) and ophiolite studies (Ceuleneer,
the length of spreading segments and are reduced 1992), indicate that the amount of interstitial melt
beneath offsets (Forsyth and Wilson, 1984; Phipps in the mantle is small.
 4. Partial Melt Beneath Oceanic Spreading Centers 57
Three-dimensional numerical studies of com- mantle. We refer to the dilation of the matrix of
bined mantle flow and melting have only recently grains to accept a higher melt fraction as "decom-
become computationally practical. Buoyant flow paction." Since this boundary slopes away from
increases the amount of along-axis variation. The the axis near a spreading center, it provides a
compositional buoyancy due to melt extraction channel within which melt can migrate laterally
can cause upwelling that is segmented along-axis toward the surface. The existence, position, and
into broad zones of enhanced upwelling and re- efficiency of this boundary at extracting melt will
duced zones of decreased upwelling, in which be determined by the temperature and flow field
there is little or no melt generated (Parmentier in the mantle.
and Phipps Morgan, 1990). Thermal buoyancy In this paper we first review the decompacting
increases these along-axis variations, further en- boundary layer model of melt extraction. We then
hancing and localizing upwelling, but does not discuss three-dimensional numerical experiments
significantly narrow the melting region (Sparks on thermal and compositional convection and melt
and Parmentier, 1993). The major effect of ther- production. Thermal and compositional buoyancy
mal buoyancy on melting is to produce localized enhances the along-axis variations in upwelling
regions of melt production far off-axis in the up- rate beneath the axis, and thermal buoyancy drives
welling limbs of thermally driven convective rolls. off-axis convective rolls that align with the spread-
Initial three-dimensional studies show that several ing direction. Convective rolls can have a large ef-
percent of retained melt will produce narrow up- fect on the shape of the melting region, with melt-
wellings, which are segmented in the along-axis ing in the upwelling limb extending for well over
direction (Jha et al., 1992). However, as in two- 100 km from the axis. Melt extraction in decom-
dimensional studies, the upwelling is still at least pacting layers will be least efficient in the up-
an order of magnitude wider than the neovolcanic welling limbs, where there is little melting and
zone. Some mechanism of focusing melt during the melting region is relatively flat. If extracted,
migration in the mantle is thus required to produce this melt will further contribute to along-axis
the observed narrow region of crustal formation. variations in melt production. We then discuss
The two proposed mechanisms of melt mi- the implied along-axis variations in crustal thick-
gration are porous flow through small channels ness, and the resulting effects on topography and
along grain boundaries (cf. Waff and Bulau, 1979; gravity.
Watson, 1982; Cooper and Kohlstedt, 1986) and
flow through a fracture network (cf. Shaw, 1980;
Spence and Turcotte, 1985; Sleep, 19&8; Ryan,
1988; Stevenson, 1989). Both of these mecha-
Melt Migration by Porous Flow
nisms must operate in some regions of the mantle.
The Forces That Drive Porous Flow
Melt is generated at grain boundaries, so it must
initially be transported through the pore spaces The conceptual framework of magma migration
around grains. However, near the surface, the through a porous mantle was developed in several
mantle is subsolidus. Since melt moving in pore early studies (Frank, 1968; Sleep, 1974; Turcotte
spaces is in thermal equilibrium with the solid and Ahem, 1978). The equations that include the
matrix, melt cannot exist in pore spaces where . role of stresses due to a deformable mantle matrix
the temperature is below the solidus. Therefore, have been derived and discussed by several authors
the melt must move through the lithosphere in (McKenzie, 1984; Richter and McKenzie, 1984;
fractures. .
Ribe, 1985; Scott and Stevenson, 1986; Fowler,
A proposed mechanism for lateral melt trans- 1990a, 1990b; Spiegelman, 1993a, 1993b):
port in the asthenosphere beneath spreading cen-
ters arose out of a study of the behavior of a de- a¢
formable porous medium near an impermeable - + V[¢url = r (I)
at
boundary (Sparks and Parmentier, 1991). Excess
melt pressures near the boundary cause melt to a¢
- - V[(l ¢ )u s ] r (2)
collect in a thin layer beneath the impermeable at
58 David W. Sparks and E. M. Parmentier

/Lr<P pendent on the rate at which melt is generated. Be-

(l <p)(p - Pr)g = T(Ur - Us)
neath a spreading center, about 20% of the mantle
(3) is melted during about 60 km of upwelling (cf.
4=S)V XUs
-( /L,v Phipps Morgan, 1987). The stresses associated
(+"3 V(V . Us) with this slow compaction are small compared to
the buoyancy of the melt (Ahern and Turcotte,
k = ko<P", (4) 1979) so the resistance to the compaction of the
matrix is negligible. Compaction stresses are im-
where <P is melt fraction, { is the bulk viscosity of portant over a length scale defined by
the solid-melt aggregate, r is the rate of melt pro-
duction, g is the acceleration of gravity, k is the ({ + 4/L,/3)k
permeability of the solid matrix, U is a velocity 8c = (5)
/Lr
field, P is density, /L is shear viscosity, and the sub-
scripts sand f refer to the solid matrix and melt (McKenzie, 1984). In the mantle, this length is
phase. respectively. In this formulation, we make typically on the order of tens of meters to I km.
the Boussinesq approximation, neglecting density Ribe (1985) showed that the stresses which arise
differences except in the buoyancy term in Eq. (3). from compaction are important only if a large
The mass balances for the melt and solid phases amount of melting occurs over a length scale com-
are given by Eqs. (1) and (2). The force balance, parable to 8 c • Therefore, porous flow in the melt-
Eq. (3), is expressed in terms of the solid and melt. ing region beneath spreading centers is closely
velocities: vertical melt flow is driven by the approximated by Darcy's law, in which melt mi-
buoyancy of the melt (left-hand side) and opposed grates vertically under the force of gravity. Verti-
by the viscous interaction of the two phases (first cal porous flow is rapid enough so that only small
term on right-hand side). These two terms make melt fractions can be maintained (Ahern and Tur-
up Darcy's law for flow in a rigid porous medium. cotte, 1979).
The other terms on the right-hand side of Eq. (3) Compaction stresses also allow the propaga-
arise from the viscous deformation of the solid. tion of solitary waves in a porous medium (Scott
Equation (4) relates permeability to melt frac- and Stevenson, 1984; Richter and McKenzie,
tion, with n usually taken to be between 2 and 3 1984; Barcilon and Richter, 1986). A local maxi-
(Cheadle, 1989). mum in the melt fraction created by a perturbation
The second term on the right-hand side of in melting will propagate through the mantle as a
Eq. (3) describes the effect of stresses arising from wave in the melt fraction distribution and shed
the incompressible shear deformation of the ma- smaller waves in its wake. Solitary waves mov-
trix. The divergent mantle flow beneath a spread- ing through a slowly melting matrix, such as up-
ing center creates pressure gradients that tend to welling mantle, tend to decrease in amplitude
direct melt toward the spreading axis (Spiegelman (Scott and Stevenson, 1986; Fowler, I990b), so
and McKenzie, 1987). However strain rates be- that small perturbations in the melting rate may
neath a spreading center are small enough that not produce a significant time dependence in the
large mantle viscosities (2: 10 2 1 Pa . s) are required amount of magma reaching the top of the melting
to create stresses sufficient to cause significant region. However, solitary waves are amplified by
melt focusing. Unless some other mechanism al- their passage through a region where melt freezes
lows lateral melt flow, such as fractures controlled (Spiegelman, 1993c).ln the next section, we show
by the local stress distribution (Sleep, 1984) or an that the stresses due to the deformation of the ma-
anisotropic permeability created by accumulated trix are important near a freezing boundary in the
strain (Phipps Morgan, 1987), porous flow in the mantle.
melting region is essentially vertical.
The last term in Eq. (3) describes the stresses Decompacting Boundary Layers
that arise from volume changes in the solid phase.
If melt is extracted from the mantle, then the solid As the temperature of the upwelling mantle drops
matrix must compact. However, the magnitudes below its solidus, melt can no longer exist in pore
of the stresses that arise from compaction are de- spaces. Solidification, or freezing, of melt reduces
 4. Partial Melt Beneath Oceanic Spreading Centers 59
the permeability. The effect of this permeability Equation (3) can be derived in terms of sepa-
reduction was explored by Sparks and Parmentier rate pressures in the solid and melt phases, with
(1991). A narrow layer of increased melt fraction the differential pressure proportional to the com-
forms, bounded by impermeable mantle above paction rate and the bulk viscosity of the system
and a region of small melt fractions and vertical (Scott and Stevenson, 1986). This differential pres-
percolation below. sure drives compaction and decompaction. Within
This result is most easily understood by con- the melting region, where compaction and extrac-
sidering a one-dimensional problem in which tion occur, the pressure in the melt is less than the
freezing is collapsed onto an interface (Fig. 1). A pressure in the solid. Near the freezing boundary,
freezing interface would exist if the melting of pe- the pressure in the melt becomes greater than the
ridotite were a univariant phase change. All melt pressure in the solid, leading to decompaction of
present in the pore spaces will freeze as it is car- the matrix. The existence of overpressured bound-
ried across the interface, so the permeability van- ary layers at the top of partially molten regions
ishes and the melt and solid velocities are equal at was suggested by Fowler (1985) as a mechanism
the interface. Some distance below the interface, for the initiation of lithosphere fracture and sub-
melt is moving faster than the solid, so as the sequent magma eruption. However, if excess fluid
interface is approached, the melt velocity must de- pressures always lead to the fracture of the over-
crease and the solid velocity increase. These ve- lying impermeable medium, then the neovolcanic
locity gradients result in a dilation or "decem- zone at spreading centers would tend to be as wide
paction" of the solid matrix. The melt fraction as the melting region.
increases from the small value th~t is supported by The distance over which decompaction occurs
Darcy flow in the melting region to the maximum is controlled by a balance of the buoyancy forces
degree of melting at the interface (Fig. I). driving the melt upward, and the viscous forces
resisting the deformation of the solid matrix. If the
matrix is very strong, it opposes rapid deforma-
Temperature, meltfraction
tion, and the increase in melt fraction is spread
over a larger region. The thickness of the layer of
thermal boundary layer increased melt fraction is described by the decom-

freezing interface paction length scale, 8,

/
(? + 4,url3)W X o
/ (6)
(p - Pr)g
/
melting region
where W is the upwelling velocity of the melt-
matrix system, and Xo is the maximum degree of
melting beneath the layer. This length scale is de-
- - -x termined by a balance between buoyancy and the
l/l compaction stresses and is identical in form to the
................... T
"reduced compaction length" derived for instan-
taneous melting at an interface (Ribe, 1985).
melting begins
Since the melting temperature of peridotite is
a function of the degree of melting, due to the
primitive mantle changing Fe/Mg ratio (cf. Hess, 1989), freezing
in the mantle does not occur on an interface, but
Figure 1 Sketch of the temperature (T), degree of melting is spread over a distance related to the balance
(X), and melt fraction (rt» in a one-dimensional column of among the advection of heat with the solid mantle,
mantle upwelling at constant velocity. The melting process is conduction to the surface, and the release of la-
approximated as univariant so that the temperature remains on
the solidus until conduction becomes dominant near the sur-
tent heat during freezing. We can estimate the
face. The rate of melting is constant, so X decreases linearly characteristic length scale of freezing from the
with depth. energy balance equation for a supersolidus re-
60 David W. Sparks and E. M. Parmentier

gion. For one-dimensional upwelling at constant along the top of the melting region. Since conduc-
velocity, vv, tive cooling penetrates progressively further with
2T
distance from the axis of a spreading center, the
WdT = K d _ ~y(dT - ()) (7) top of the melting region slopes away from the
dz dz? C; dz ' axis. This is true for all models of upwelling,
where () is the slope of the mantle solidus, and y whether passive or buoyant. Since the melt frac-
relates the stable melt fraction, X, to the tempera- tion in the decompacting layers is relatively high,
ture above the solidus: the permeability is also high. The dec ompaction
stresses act perpendicular to the layer boundary,
X = yeT - Tsolidu,)' (8) and partially counteract the vertical buoyancy
The solution to (7) has the form force. The component of the buoyancy force that
acts in the direction of the layer drives melt later-

T(z) = C\ eXP[~(l + ~:)z] ally (Fig. 2), and melt flows upslope along the
layer towards the spreading axis.
W Ly The flux of melt in the layer is determined by a
+- -()z + C2• (9) balance among the lateral porous flow, propor-
K Cp
tional to the local slope, the melt supply to the
The integration constants, C\ and C 2 , must be de- layer from the melting region, and freezing within
termined by matching conditions on the heat flux the layer. Approximate, two-dimensional steady-
at the top and bottom of the upwelling column state solutions for the behavior of the layer show
(Sparks and Parmentier, 1991); however, the char- that a significant fraction of melt formed over a
acteristic length scale of freezing can be deter- wide melting region beneath a spreading center
mined by inspection. In the melting region the can be delivered to the spreading axis by po-
second term dominates, and the temperature in- rous flow within the layer (Sparks and Parmentier,
creases linearly with depth, with a slope slightly 1991). Lateral melt flow in the layer and subse-
modified from that of the undepleted solidus. In quent extraction at the spreading axis decreases
the freezing region, the first term dominates, and the melt fraction in the layer to a few percent. The
the temperature decays exponentially toward the fraction of melt extracted decreases with increas-
surface. The freezing length scale, OCn is ing distance from the spreading axis, as the up-
welling rate and total degree of melting decrease.
OCr = ~( C
p
~ LY). (10) The formation of decompacting boundary lay-
ers has also been observed in highly resolved
In the univariant melting problem described above, two-dimensional porous flow calculations (Spie-
the freezing length is zero. If OCris large compared gelman, 1993c). An imposed sloping freezing
with the compaction length scale, 0, then melt is boundary causes a layer of relatively high melt
frozen as it approaches the boundary layer, and fraction to form and direct melt flow along the
matrix deformation stresses are not important. boundary. Less prominent layers of increased melt
This is analogous to the behavior in the melting fraction propagate back from the freezing bound-
region, where compaction occurs, but the stresses ary. One-dimensional analysis also indicates that
generated are small because melt generation .is the melt fraction may increase toward the upper
spread over many kilometers. When the two length boundary of a melting region in an oscillatory
scales are equal, then the viscous effects of the im- fashion (Fowler, 1990b, Appendix A).
permeable boundary are felt before much melt has In the one-dimensional upwelling problem de-
frozen, and decompaction occurs. For an upwell- scribed above, matrix shear viscosity and the bulk
ing rate of 3 cm/yr, OCr is about 1 km. Therefore, viscosity of the system appear only in combina-
under most conditions of decompression melting tion. This combination of viscosities was varied
in the mantle, some degree of pooling of melt in over a range of values, but was held to be constant
the decompacting layers is likely to occur. throughout the column. Viscosity is a complex
Decompacting layers are important for melt function of temperature, pressure, composition,
extraction at spreading centers because they form and melt fraction. Probably the most important of
 4. Partial Melt Beneath Oceanic Spreading Centers 61

Figure 2 Schematic cross section of a spreading center with a broad region of melt production (stippled). The solid lines are
mantle streamlines. The arrows show the direction of melt flow. The inset is a blow up of a melt-rich layer at the top of the
melting region, in which melt flows laterally toward the ridge.

these parameters to the behavior near a freezing on basalt composition, since melt segregated into
boundary is melt fraction, since the others vary veins is more likely to be out of chemical equi-
over relatively long length scales with respect to librium than the melt distributed in pore spaces
the compaction length. A reduction in the matrix (Spiegelman and Kenyon, 1992). More work is
viscosity within the layer will result in a reduc- needed on the properties and physics of partially
tion in the layer thickness (Eq. 6), decreasing the molten mantle to assess vein formation. In the
compaction length and the along-layer flux. How- next section we deal with the long-term results of
ever, recent measurements of the effect of melt melting and melt extraction, so we assume the
fraction on viscosity suggest that even for several process of melt extraction is steady state, hoping
percent melt, the reduction of viscosity is only to capture at least the time-averaged behavior.
about an order of magnitude (Kohlstedt and Hirth,
1992). Nonlinear effects associated with vari-
able viscosity could lead to interesting and unpre- Effect of Buoyant Mantle Upwelling on Melt
dieted behavior, but these effects still need to be Generation and Migration
investigated.
The stability of decompacting boundary layers Three-Dimensional Buoyant Flow beneath
in the presence of a realistic distribution of melt- an Offset Spreading Center
ing and freezing in two and three dimensions is a
fundamental problem that has not been addressed Numerical Formulation
to date. If large local melt fractions develop in Three-dimensional numerical experiments
parts of the layer, melt could segregate into veins. were conducted to examine the patterns of buoy-
These veins need to form a connected network ant upwelling in the mantle beneath an offset
in order to focus melt to the axis, since isolated spreading center, and its effect on melt migration.
veins will either be carried into the subsolidus The geometry of the spreading center was taken
mantle, where they will freeze, or propagate up- to be periodic in the along-axis direction, so the
ward through the lithosphere and erupt off-axis. region examined consisted of two spreading seg-
Vein formation would have an important effect ments, separated by a transform fault (Fig. 3).
 62 David W. Sparks and E. M. Parmentier
y
west north ag
- + u . 'ilg =r
>< east
at
(12)

r = D[(T - TSOlidUS)] (13)

Dt 600°C
TsolidUS = 1100 (0C) + 4.0 z (km) (14)

200 km
'il . u = 0 (15)
Z ILs'il 2u - 'ilp - pg =0 (16)
p(T, g) = Po(1 - a'I' - f3g). (17)
Here, t is time, z is depth in kilometers, p is pres-
Figure 3 Schematic representation of the region of mantle in
sure, K is thermal diffusivity, L is the latent heat
the three-dimensional numerical convection experiments. The
of melting, C p is specific heat, p is mantle density,
shaded area is the rigid lithosphere, defined by the I 100°C iso-
therm. Convection is confined to the asthenosphere (unshaded
TJ is mantle viscosity, a is the coefficient of ther-
region). The position of the plate boundary is shown in each
mal expansion, and f3 describes the decrease in
plane by double lines for ridge segments and a single line for
compositional density due to the extraction of
the transform. For convenience in referring to the different
iron-rich melt. The values of the parameters used
parts of the region in later figures, a directional system is de-
fined by the arrows at the upper right. in these experiments are given in Table 1.
The viscosity structure of the mantle is an im-
portant factor in determining the pattern of buoy-
This geometry is symmetric about the middle of ant flow. We approximate the temperature and
the spreading segments, so that the calculation pressure dependence of viscosity by confining
need only consider one-half of each segment. buoyant flow to a uniform viscosity asthenosphere
The flow consists of a passive component and bounded by no-slip boundaries on the top and
a buoyant component. The passive component is bottom (see Fig. 3). The top boundary is pre-
driven by the two plates spreading at an imposed scribed by an isotherm (chosen to be 1100°C) rep-
velocity along the top of the region. This flow resenting the bottom of the lithosphere, and the
drives upwelling of deep mantle into the bottom bottom of the asthenosphere is prescribed at a
of the region, and flow out the ends of the region. depth of200 km. The vertical boundaries are sym-
Buoyant flow is driven by density gradients metry planes for the buoyant flow. Buoyant flow
due to thermal expansion and the compositional is forced to be two-dimensional at the vertical out-
changes due to melt extraction. Melt is assumed flow boundaries by prescribing the temperature
to be extracted rapidly from the mantle, so that the field to be independent of distance from the axis
melt fraction in the melting region is very small. near the boundaries. These boundaries are placed
Melt-rich, decompacting layers on the order of far from the spreading center to minimize their ef-
1 km in thickness are too small to have an effect fects on flow near the melt production region.
on the overall density. Therefore, the melt fraction The region considered in the experiments is
is neglected in the calculation of buoyant flow. 200 km deep, 150 km in the along-axis direction,
The extraction of melt from the mantle will gen- and 1200 km in the spreading direction, with the
erate a dilatational component of flow, but this plate boundary centered on the top of the region.
component is small when the melt production re- The spreading segments range in length from 75
gion is broad, and it is thus neglected. to 225 km, and the transform offset is 75 km. The
The governing equations for temperature, T, temperature is fixed at O°C at the surface and
the degree of melt depletion, g, the rate of melt 1410°C at the bottom of the asthenosphere, the
production, I', and the mantle buoyant flow ve- latter value chosen to produce approximately 6 km
locities, u, are of crust. Finite difference approximations were
used to solve (11 )-(17). The buoyant flow equa-

at Cp
(L)
-et + u . 'ilT = K'ilZT - - r (11) tions (15)-(16) were cast in a streamfunction-
vorticity formulation and solved using a multigrid
 4. Partial Melt Beneath Oceanic Spreading Centers 63
Table 1
Physical Parameters and Values Applicable to the Mantle

Physical Typical mantle

parameters values Description

L 6XI0'J'kg- 1 Latent heat of fusion

Cp 1000 J . kg -I . °C-I Specific heat
K 1O- 6m'"s-1 Thermal diffusivity
ko 10 -7_10 -9 m- Constant in permeability
relation, Eq. (4)
n 2-3 Exponent in permeability
relation, Eq. (4)
P 3300 kg· m- 3 Mantle density
Pr 2800 kg m >'
r Melt density
a s x 1O-,oC"1 Coefficient of thermal expansion
13 0.024 Compositional density parameter
in Eq. (17)
~ 10 19-10"Pa's Bulk viscosity of solid-melt
aggregate
fl., 10 19_10" Pa sr Mantle shear viscosity
fl.r O.I-IOPa·s Melt viscosity

iterative Poisson solver. The numerical methods The axis-perpendicular rolls form in response to
are discussed in more detail in Sparks and Par- along-axis temperature gradients across the frac-
mentier (1993) and Sparks et al. (1993). The mul- ture zone. In most cases, the upwelling limb of the
tigrid buoyant flow solver is described in Sotin roll is centered beneath the nearer spreading seg-
et al. (1992). ment, so that rolls of opposing vorticity are present
on either side of the spreading center (Fig. 6).
These rolls are thermally driven and advect com-
Structure ofThree-Dimensional Flow positionally less dense, depleted mantle down into
Buoyant flow has two effects on the overall the asthenosphere. There is also thermally driven
flow pattern: the enhancement and localization of axis-parallel circulation (Fig. 5). This circulation is
upwelling beneath the centers of spreading seg- not continuous along-axis, but is most prominent
ments, and the creation of thermal boundary layer in the downwelling limbs of axis-perpendicular
instabilities that grow into axis-perpendicular con- rolls.
vective rolls (Figs. 4-6), similar to the flow struc- The distance from the spreading axis at which
tures that develop beneath a non-offset spread- axis-perpendicular rolls (Fig. 6) are well devel-
ing center (Sparks and Parmentier, 1993). The . oped is a function of spreading rate and mantle
localized upwelling is driven by temperature gra- viscosity (Sparks and Parmentier, 1993). Thermal
dients across the transform. At. fast spreading rolls form when the thermal boundary layer has
rates, upwelling beneath the axis is relatively uni- penetrated through the layer of compositionally
form along spreading segments, except near trans- less dense depleted mantle. Therefore, for a given
forms, where the upwelling is weaker and conduc- viscosity, thermal rolls form closer to the spread-
tive cooling penetrates deeper. At slow spreading ing axis at slow spreading rates than at fast spread-
rates, buoyant flow is relatively stronger than the ing rates. At lower mantle viscosities, the strength
passive flow, and causes larger along-axis varia- of buoyant flow is increased relative to passive
tions in upwelling rate beneath the entire length of flow, and the rolls are more vigorous and form
spreading segments. closer to the axis.
 % depletion

Figure 4 Horizontal cross sections of a numerical convection experiment at depths of (A) 25 km, (B) 43 km, and (C) 100 km.
The shading represents the degree of depletion of the mantle, in 3% increments, with darker shades more depleted. Temperature
isotherms are plotted at 200°C intervals. The position of the plate boundary is shown in each plane by double lines for ridge
segments and a single line for the transform. The spreading half-rate is 2.7 cm/yr, and the mantle viscosity is 5 X 1019Pa' s. In
the two shallower cross sections (A-B), the isotherms reflect cooling with increasing distance from the nearest spreading seg-
ment. In the deepest section (C), linear downwelling regions carry cooler, more depleted material to depth.

% depletion

Figure 5 Axis-perpendicular vertical cross sections of the numerical experiment shown in Fig. 4, along the (A) southern and
(B) northern edges of the region. The shading and contour interval is as in Fig. 4. The position of the spreading axis is shown by
an arrow along the top of each figure. Each section cuts through two convective rolls, with an upwelling on one side of the axis
and a downwelling on the other side. The downwellings are evident from the downwarping of the isotherms and depletion
contours.
 4. Partial Melt Beneath Oceanic Spreading Centers 65

% depletion --~~ north

Figure 6 Axis-parallel vertical cross sections of the experiment shown in Figs. 4 and 5: (A) beneath the southern spreading
segment, (B) beneath northern spreading segment, (C) 250 km to the west of (A), and (D) 250 km to the east of (B). The shading
and contour intervals are as in Fig. 4. The cross sections beneath spreading axes (A-B) show no sign of the convective rolls
which form off-axis (C-D).

Melt Migration beneath Spreading Centers the very different length scales of melt genera-
tion and decompacting layer thickness (Sparks
Since decompacting layers form where freezing and Parmentier, 1991). This approach was used to
occurs, the position and slope of the layers are .estimate the efficiency of melt extraction to the
controlled by the temperature distribution. The spreading axis in two dimensions, but has not yet
upwelling and temperature distributions in the nu- been extended to three dimensions. However, we
merical experiments described above determine can make simple predictions about the direction
the distribution of melting. The freezing boundary and strength of lateral melt migration and the re-
layer, where decompaction occurs, lies along the sulting distribution and composition of basaltic
top of the melting region, so the position of the crust that may be formed at the spreading axis.
layer and the direction of melt migration can be The melting region in these experiments is es-
estimated from the shape of the melting region. sentially triangular in cross section; i.e., melting
A description of the shape of the melting region begins at a roughly uniform depth and extends to
and an approximate model for porous flow within shallower levels beneath the axis than it does off-
the layer can be combined to take advantage of axis (Fig. 7). The progressive penetration of con-
66 David W. Sparks and E. M. Parmentier

km

km 0

km

krn 0
Figure 7 The three-dimensional shape of the top of the melting region in two typical numerical experiments. The orientation
of the region is as in Fig. 3. Only part. of the 1200-km-long region is shown here, to focus on the region beneath the spreading
axis. Where there is no melting the surface drops to a plane at a depth of 120 km. The half-spreading rate is 2.7 cm/yr, and
mantle viscosity is (A) 2 X \020Pa· s and (B) 5 X \o'·Pa· s. The formation of axis-perpendicular convective rolls in (B) causes
significant along-axis variation in the width of the melting region. Melting occurs to a great distance off-axis within the linear
upwelling limbs of the convective rolls, while off-axis melting vanishes where downwellings occur.

ductive cooling with increasing distance from the some extent, but do not change its fundamental
axis causes melting to stop at deeper levels off- shape (cf. Sotin and Parmentier, 1989). If a sig-
axis. Thermal and compositional buoyancies may nificant melt fraction (>5%) is retained in the ma-
reduce the across-axis width of the melt region to trix, then the upwelling is focused, and the melt-
 4. Partial Melt Beneath Oceanic Spreading Centers 67
ing region is not very wide, even at its base (Scott eral melt migration in a decompacting layer at the
and Stevenson, 1989; Buck and Su, 1989; Jha top of the melting region, we examine more closely
et al., 1992). the expected shape of the layer. Plots of the depth
In the case of purely passive spreading to the top of the melting region show that the pres-
(Fig. 7A), at an intermediate spreading rate of ence of a transform offset can have a large effect
2.7 cm/yr, the melting region extends to nearly on the shape of the layer (Fig. 8). The arrows on
200 km from the axis. This distance decreases Fig. 8A show the direction and magnitude of the
slightly near a transform, but the cross section local slope of a decompacting layer in a purely pas-
of the melting region remains relatively constant sive upwelling. The predicted direction is roughly
along-axis. When there is strong buoyant flow toward the nearest spreading axis, but there is a
(Fig. 7B), so that convective rolls form near the saddle point in the layer beneath the center of the
melting region, there are large along-axis varia- transform. A line that passes through that saddle
tions in the width of the melting region. Near the point divides the melting region into two parts
downwelling limb of a roll, the melting region which provide melt to each spreading segment. If
abruptly pinches out relatively close to the axis. In migration in the melting region is vertical, and lat-
the upwelling limb, the melting region is extended eral migration occurs only along the top of the
in the spreading direction, with a narrow region melting region, then melt formed on one side of
of deep melting extending to over 300 km from this melting divide cannot be extracted to the seg-
the axis. ment on the other side. When strong convective
To determine the strength and direction of lat- rolls form, melting occurs along the upwelling

A
km
140
120
100
80
60
40 -,
-,
,

20
a
400 500 600 700 800 km

B
km
140
120
100
80
60
40 northL
east
20
a
400 500 600 700 800 km
Figure 8 The depth to the top of the melting region, contoured in 5-km intervals. Arrows show the local slope of the surface.
The thick gray line divides the area into two melt extraction regions. Melt formed in one region can be extracted only at the
spreading axis in the same region. The two cases are for the same experiments shown in Fig. 7, with mantle viscosities of (A) 2 X
lO,oPa· s and (B) 5 X IO'9Pa· s.
68 David W. Sparks and E. M. Parmentier

limb of the roll (Fig. 8B). but the position of the culated by summing the melt generation in axis-
melting divide is relatively unchanged. perpendicular planes along each segment.
Along-axis variations in melt production and
extraction may produce observable variations in
the distribution and composition of basaltic crust. Implications for Crustal Thickness and Gravity
To make predictions about the crustal distribu-
Thickness and Composition
tion from these numerical experiments, simple as-
of the Oceanic Crust
sumptions about melt migration are made. Melt
must migrate through the subsolidus lithosphere The effect of a transform offset on melt pro-
by fracture. Since the neovolcanic zone is very duction is shown in Fig. 9, where melt production
narrow, we assume that fractures are vertical and is expressed as thickness of basaltic crust (melt
confined to the plane of the spreading axis. There- production rate divided by the spreading rate).
fore melt migrates vertically within the melting The overlaps in the curves are the result of extract-
region into the decompacting layer. Within the ing melt as it reaches the plane of a spreading cen-
layer, melt migrates laterally toward the plane of a ter. The position of the melting divide is chosen
spreading axis, where it is extracted from the by inspection of plots like Fig. 8. The resulting
mantle to form crust. The position of a melting small discretization inaccuracy in the region of
column relative to the melting divide determines overlap, where the amount of melt generated is
at which segment the melt will be extracted. The small, does not affect the results.
along-axis distribution of melt production is cal- At high mantle viscosities, when the flow is

9
8
E viscosity (Pa-s)
~ 7 --2x10"20

--
c - -5x10"19 ..-
.E 6 ,
U 5 ---2.5x10"19 " ..........
:l
'tl -----10 .. 19 /
,./ ---,
...
0
a.
4 e:
3
sVI 2
...
:l
1.1
~_?'-
~ ......
•
-
0
0 25 50 75 100 125 150
A
distance along-axis (km)

7
spreading rate(cm/yr)
6
E --6.0 - -2.7 - - -1.0
~
5
---...:::
c <, ..........-
.S!
U 4
:l
<, -:
'tl <, /
...a.0 3 <,
<, -, / ......
......
<,
I'll
u; 2 <,
.....
-, /'
/

...
:l
>z
/1"- /

1.1
.:>
<:
'--
0
0 25 50 75 100 125 150
B
distance along-axis (km)

Figure 9 (A) Crustal production as a function of distance along axis for a range of mantle viscosities and a spreading rate of
2.7 cm/yr. The bold arrow shows the position of the transform. (B) Crustal production for a range of spreading rates and a
viscosity of 5 X 10 19 Pa . s.
 4. Partial Melt Beneath Oceanic Spreading Centers 69
primarily passive, melt production is relatively dicular to the axis, using the same division of the
constant along the center of the spreading seg- melt extraction region as in the previous section,
ments and decreases by about a factor of 2 near to yield the average depth of melting, D, and av-
the transform (Fig. 9A). As viscosity is decreased, erage degree of melting, F. At high viscosity, i5
and buoyant flow becomes important, the along- and F are relatively constant along a spread-
axis variation becomes larger in amplitude and is ing segment, with greater depths of melting and
distributed across the entire segment. Decreases in smaller degrees of melting near a transform (see
crustal thickness of up to 50% toward transform Fig. 10). As the viscosity is decreased, the gradi-
faults have been measured seismically (Cormier ent in Dand F near the transform increases. Also,
et al., 1984; Purdy and Detrick, 1986). Gravity high-D, low-F features develop at the centers of
measurements near several fracture zones (Kuo the segments, similar to the features that form near
and Forsyth, 1988; Lin et al., 1990; Blackman and the transform, only broader. This feature is due to
Forsyth, 1991; Morris and Detrick, 1991) are also the small extents of deep melting in the upwelling
consistent with about 50% crustal thinning. limbs of off-axis rolls.
For a given viscosity, the along-axis varia- A positive correlation between Na.O and FeO
tion is more subdued at higher spreading rates contents of basalts has been recognized in data
(Fig. 9B). This is in agreement with the observa- collected at slow-spreading centers (Brodholt and
tion that the amplitudes of along-axis variations in Batiza, 1989). This correlation has been suggested
topography and the mantle Bouguer anomaly in- to result from imperfect mixing of melts formed
crease with decreasing spreading rate (Lin and at different depths, with deeper melts containing
Phipps Morgan, 1992). Crustal thickness is also higher Na.O and FeO contents (Klein and Lang-
much more variable at slow spreading rates (Reid muir, 1989). Our numerical experiments indicate
and Jackson, 1981; Chen, 1992; White et al., that melts formed beneath a transform fault should
1992). have a signature of deeper, smaller extents of melt-
The shape of the melting region also affects the ing, particularly at slow spreading centers. While
composition of the basalts that form the crust. The this signature has been found near some transform
partition coefficients of elements between perido- faults (Langmuir and Bender, 1984; Batiza et al..
tite and basaltic melt can be sensitive to the pres- 1988), smooth along-axis variations in melt com-
sure of melting and the composition of the solid. position are not common. If such a variation in
The composition of an increment of partial melt basalt chemistry does not exist, it is probably due
depends on the instantaneous depth of melting to some process during crustal emplacement, such
and the integrated melting history of the parcel of as the episodic emplacement of dikes that overlap
mantle that is melting. Therefore, a unique melt in space, but have slightly different compositions.
composition is generated at each point in the melt- A greater number of spreading centers need to be
ing region. This melt composition could be deter- systematically sampled with good spatial resolu-
mined with exact knowledge of the initial compo- tion to resolve this question.
sition of the primitive mantle, and the distribution
coefficients for each element, as functions of depth Melt Extraction Efficiency and Melting
and degree of melting. Experimental petrology in Off-Axis Convective Rolls
provides information on the behavior of some ele- A melt-rich decompacting layer will be least
ments, which can be used to predict the composi- effective at extracting melt formed in the upwell-
tion of melt formed at a given depth and degree of ing limbs of the convective rolls because of the
melting (cf. Hanson and Langmuir, 1978; Niu and small degrees of melting and the relatively small
Batiza, 1991; Kinzler and Grove, 1992a). slope of the layer in this region. This melt may
Having calculated the distribution of oceanic not be extracted at the axis, but may refreeze into
crust, we can also calculate the average degree of the mantle or pool and then erupt off-axis. In some
melting and average depth of melting at which the experiments, melting in the upwelling limbs oc-
basalt was formed. For each numerical experi- curs far enough off-axis 'that it is separated from
ment, the degree of melting and depth at each grid the melting region beneath the axis by a subsoli-
node is weighted by the melt production rate at dus impermeable mantle. Melt that is not ex-
that node. The melt is pooled in planes perpen- tracted at the axis may account for the production
70 David W. Sparks and E. M. Parmentier

viscosities (Pa-s)
--2x10"20 ---5x10"19 -----10"19
0.085

..
CI
s:
Qj
0.08

-
E
0
CIl 0.075
...CI
CIl

CIl
'C 0.07
CIl
CI
...
III
CIl 0.065
>
III

0.06
0 25 50 75 100 125 150
A
...... 46
E
:.
CI 48

-
s:
Qj 50

-
E
0 52
-
s:
Co
~ 54
CIl
~... 56
CIl
>
III 58
0 25 50 75 100 125 150
B distance along-axis (km)

Figure 10 (A) Average degree of melting and (B) average depth of melting for a range of viscosities and a spreading rate of
2.7 cm/yr, as a function of distance along axis. The position of the transform is marked by the bold, vertical arrow. The curves
are truncated just beyond the transform for clarity.

of off-axis seamount chains (Sparks and Parmen- V = k(p - Pr)g sin w, (18)
tier, 1993). fLrcPo
The formation and extraction efficiency of
decompacting layers will depend on a balance where w is the dip angle of the layer, cPo is the melt
among the supply of melt to the layer, the flow fraction in the layer, and k is given by Eq. (4). The
along the layer, and the freezing within the layer. melt fraction at any point in the layer is bounded
Therefore, the magnitude of the melt velocity is by the maximum degree of melting within the up-
not determined solely by the magnitude of the welling mantle directly below. If we substitute the
layer slope shown in Fig. 8. We can estimate the maximum degree of melting into Eq. (18), the re-
spatial variations in the magnitude of the flow sulting melt velocities reflect the maximum spa-
within a layer that result from the variations in tial variation in the local slope and melt supply
melt supply and layer slope. (Fig. 11).
If forces due to gradients in the rate of compac- The estimated melt velocities in the layer de-
tion are ignored, then melt flow in the layer can be crease toward the edges of the melting region, and
treated as one-dimensional Darcy flow. The melt within the convective rolls are one-third or less
velocity in the upslope direction, V, is than the maximum velocities in the layer. The
 4. Partial Melt Beneath Oceanic Spreading Centers 71
A
km

1401~U.IDJffin
120
100
80
60
40
20
a L--,--~:..:....:J.d~EUJ!flElill£L::t::::zb:±~------,--_~
500 600 700 800 km

B
km
140
120
'00
80 I
60 I nOrlhL
40
east
20
0L..:.-L..:..-.:...J..:...-.:....:-.:....:-~.:....:-.:....;L~==I.Q...;;..c;....:~~~~:;....l..-:;L.l.----_--L_----'

500 600 700 800 km

Figure 11 Contours of the decompaction length scale. as given by Eq. (6) for the two experiments shown in Figs. 7 and 8. The
contours denote the thickness of the decamp acting layer in meters. The arrowheads represent the approximate flux of melt within
the layer. The arrows point in the direction of the steepest local slope, as in Fig. 8, but the magnitudes are given by the product
of the melt velocity along the layer (Eq. (18)), the porosity in the layer, and the layer thickness. The thick gray line divides the
two extraction regions, as in Fig. 8.

thickness of the layer, as estimated from Eq. (6), Predicted Topography and Gravity Signal
also decreases toward the edges of the melting re-
gion (shown by the contours in Fig. 1I) since it When the effects of the two major density in-
is a function of both upwelling velocity and the terfaces, crust-water and mantle-crust, are sub-
amount of melt production. Therefore, the melt tracted from the free-air gravity anomaly, the re-
formed in convective rolls may not contribute to sulting mantle Bouguer gravity anomaly (MBA)
crustal production at the axis. We can attempt to indicates either mantle density variations or devia-
account for this by fixing the width of the region tions from an assumed constant crustal thickness
from which melt is extracted. (see Kuo and Forsyth, 1988). Negative ellipsoidal
Figure 12 shows the amount of crustal produc- "bulls-eye"-shaped anomalies indicate that the
tion in a single numerical experiment for a vary- mantle is less dense and/or the crust is thicker be-
ing width of melt extraction. Melting occurs more neath the center of ridge segments. Observed mag-
than 50 km from the axis along most of the spread- nitudes typically range from 20 mgal (Kane Frac-
ing center, as shown by the difference between the ture Zone: Morris and Detrick, 1991) to 50 mgal
curves for 50 and 100 km. Beyond 100 km from (Atlantis Fracture Zone: Liri et 01., 1990).
the spreading axis, melting is limited to the up- Mantle density variations are insufficient to
welling limb of axis-perpendicular rolls. account for the observed magnitudes of the
72 David W. Sparks and E. M. Parmentier

7
E half-width of melt extraction (km)
~ 6
c --200 ---100 -----50
.2 5
"..
:::J
'0
0
4
Co 3
iii
..
Vi
:::J
()
2

0
0 25 50 75 100 125 150
distance along-axis (km)

Figure 12 Crustal production in an experiment with a spreading rate of 2.7 cm/yr and viscosity of 2.5 X 10'9 Pa . s, for
different widths of the region of melt extraction. Melt generated beyond the half-width of melt extraction does not contribute to
crust emplaced at the axis.

MBA, producing maximum along-axis variations mantle at density p (given by Eq. (17) for a tem-
of about 10 mgal (Sparks et al., 1993). Even the . perature of 14100C and no melt depletion). For
small variations in crustal thickness produced in a column with crustal thickness, h., and mantle
the high-viscosity experiments double these mag- density (given by Eq. (17)) averaged over the
nitudes, and at low viscosities, the magnitudes in- mantle column, p, the topography, h, is given by
crease tenfold. The magnitude of the along-axis
temperature variations is small, since temperature 200 km (p - p)
h = ----'-'----"-'-
within the melting region is buffered by the ab- (p - Pw) (19)
sorption of latent heat. These small, deep density (he - 6 km)(p p,.)
variations produce only a small effect on the sur- +-'-'------'--'----'---'--
(p - Pw)
face gravity field. However, variations in crustal
production may be large, even though the tem- where p; is the density of oceanic crust and p;
perature field is relatively uniform, due to along- the density of seawater, taken to be 2800 and
axis variations in upwelling velocity. Therefore 1000 kg' m -3, respectively.
crustal thickness variations dominate the MBA We calculate the MBA by treating horizontal
signal. density variations at a each depth as a sheet of
When buoyant flow is not important, there is anomalous mass. A two-dimensional fast Fourier
not much along-axis variation in topography and transform (FFT) transforms the mass variations
MBA, except near the transform. As the viscos- into wavenumber space. The magnitude of the
ity is decreased, patterns similar to the observed surface gravity anomaly, t:J.g. that is created by a
"bulls-eyes" emerge (Fig. 13). At intermediate sheet of anomalous mass is given by
viscosities the mantle Bouguer anomaly increases
by 40-50 mgal from the center of the spreading t:J.g = 21TG exp( -lTZ )M, (20)
segment to the transform, and the depth of the axis
increases by over 500 m. At low viscosities, con- where M is the magnitude, IT is the wavenumber,
vective rolls form near the axis causing variations and Z is the depth of the mass sheet. G is the gravi-
of over 100 mgal and 1.5 km. tational constant (6.67 X 10 -II m' . kg -1 . S -2).
Isostatic topography for the numerical experi- The contributions from each wavenumber of the
ments is calculated column by column, assuming mass sheets at each depth in the grid are summed,
compensation at the bottom of the asthenosphere. and an inverse FFT yields the total gravity anom-
The topography is calculated with reference to a aly due to mantle density variations. The crust-
column consisting of 6 km of crust and 200 km of mantle interface is treated as a mass sheet at the
 4. Partial Melt Beneath Oceanic Spreading Centers 73

Figure 13 Mantle Bouguer anomaly (mgal, shaded) and bathymetry (krn, contoured) for a spreading rate of 2.7 cm/yr, mantle
viscosity of 5 X 10 19 Pa . s, and ridge length ratio of 3 to I. (A) Crustal thickness assuming all melt is extracted in axis-
perpendicular planes. (B) Crustal thickness assuming two melt extraction regions and no extraction of melt formed beyond the
end of the transform. (C) Crustal thickness as a function of distance along axis. used in calculating the gravity and topography
in (A) and (B). In each case "bulls-eye"-shaped gravity anomalies form over each segment, but in (B) the magnitude is greater.

average depth of the interface from the sea sur- beneath the transform is considerably thinner, the
face, assuming the water depth to be 3 km. Crustal variations in gravity and bathymetry are also more
thickness is a function of distance along-axis and pronounced.
is taken to be the total of the crust generated in Figure 13 also shows the effect of variations in
planes perpendicular to the axis. the length of spreading segments. When the melt-
To demonstrate the sensitivity of the gravity ing region is divided, the longer segment is seen
field to crustal thickness, two different methods of to extract more than its proportional share of melt.
estimating crustal thickness were used for a single The resulting thicker crust produces a larger grav-
numerical experiment (Fig. l3C). First, the crustal ity anomaly beneath the longer segment, as is ob-
production was simply integrated in planes per- served south of the Atlantis Fracture Zone (Lin
pendicular to the axis (Fig. l3A); in the second et al.. 1990). A greater number of detailed along-
case, the region was divided into two extraction axis seismic refraction studies is needed to con-
regions to produce crustal thickness curves like firm the relation between gravity and crustal thick-
those in Fig. 9, and then melt formed beyond the ness. These studies coupled with modeling of
transform fault was neglected. Because the crust the mantle flow and melt generation can provide
74 David W. Sparks and E. M. Parmentier

important clues to the style and paths of melt temperature-dependent viscosity (Shen and For-
migration. syth, 1992) do not qualitatively change this result.
Therefore, except at East Pacific Rise-type spread-
ing rates, melt must migrate through a significant
Discussion thickness of subsolidus lithosphere by fracture.
Although flow in the decompacting layers may be
Further work is needed to assess the implications unsteady, it is likely to be a less time-dependent
of the three-dimensional numerical experiments process than transport through fractures. The
for observable characteristics of spreading cen- greater distance melt must travel in fractures may
ters. The relationship between plate boundary seg- contribute to the more episodic character of slow
mentation and buoyant flow is still unclear. Time spreading centers, e.g., the absence of steady-state
dependence develops in some numerical experi- crustal magma chambers (Phipps Morgan, 1991;
ments at low viscosities (:510 19 Pa . s, Sparks Solomon and Toomey, 1992) and the greater am-
et al., 1993). The effect of this time dependence plitude and spacing of abyssal hill topography
in the mantle flow on observable temporal variabil- (Malinvemo and Pockalny, 1990).
ity at the ridge crest also needs to be explored. The Decompacting layers may also have an observ-
amount of along-axis variability and the depths of able effect on basalt chemistry. The "local trends"
decompacting layers are two unexplored aspects in basalt composition have also been interpreted
of these experiments that may have important im- as the result of fractional crystallization at a vari-
plications for spreading center morphology and ety of depths during the ascent of the melt (Kinz-
episodicity. ler and Grove, 1992b). Since at least a small
Even in numerical experiments at moderate amount of freezing is occurring within decom-
mantle viscosities, the along-axis variations in pacting layers, if the melt in the layers does not
crustal production are as large as or larger than any fully reequilibrate during its ascent, it should re-
observed variations in crustal thickness. Buoyant tain a chemical signature of deep fractionation.
flow due to retained melt will further increase The analysis and calculations presented here are
along-axis variations (Jha et al., 1992). In some ex- a first attempt at describing the three-dimensional
periments crustal production vanishes along parts character of spreading centers by including pro-
of the spreading center. This indicates that there cesses that occur on a variety of scales: passive
must be along-axis melt migration toward trans- mantle flow on the scale of oceanic plates, convec-
forms that limit the variation of crustal thickness. tion on the scale of individual spreading seg-
The predicted migration paths in a decompacting ments, and porous flow on the scale of decom-
layer (Fig. 8) indicate along-axis flow away from pacting layers. A better understanding of the
transforms. Therefore, to limit crustal thickness interaction of processes at all these scales is cru-
variations, significant migration toward transforms cial to describing the nature of mantle flow, melt-
must occur during transport through the litho- ing, and melt migration in the mantle beneath
sphere or emplacement in the crust. spreading centers.
Figure 13 illustrates the degree to which observ-
able quantities are controlled by the distribution
of crust. The dynamics of crustal formation .are Acknowledgments
just starting to be addressed in a quantitative way
This work was supported by NSF Grants aCE 92-02599 and
(cf. Phipps Morgan, 1991)., A three-dimensional
aCE 93-96097, and by a grant from the University Collabo-
model for the emplacement of the various com- rative Research Program of the Institute for Geophysics and
ponents of the oceanic crust is the next step nec- Planetary Physics at Los Alamos National Laboratory. Sparks
essary to more closely link models of mantle dy- was supported during part of the preparation of the manuscript
namics and melt formation with observations. by a Postdoctoral Fellowship at Lamont-Doherty Earth Ob-
The minimum depth to the decompacting layer servatory of Columbia University. We thank Neil Ribe, Mi-
chael Ryan, and Norm Sleep for thoughtful reviews of the
varies from about 10 km at fast spreading rates manuscript, and Jason phipps Morgan and Marc Spiegelman
to about 25 km at very slow spreading rates. for useful discussions on mantle flow and melt migration. We
The slightly shallower depths that result from also thank Jason for providing a program for calculating pas-
 4. Partial Melt Beneath Oceanic Spreading Centers 75
sive flow and Greg Neumann for a program for generating 3D Frank, E C. (1968). Two-component flow model for convec-
surface plots. tion in the Earth's upper mantle, Nature 220, 350-352.
Hanson, G. N.. and Langmuir, C. H. (1978). Modelling ofma-
JOT elements in mantle-melt systems using trace element
References approaches, Geochim. Cosmochim. Acta 42, 725-742.
Hess, P. C. (1989). "Origins of Igneous Rocks," Harvard
Ahem, J. L., and Turcotte, D. L. (1979). Magma migration be- Univ. Press, Cambridge.
neath an ocean ridge, Earth Planet. Sci. Lett. 45, 115- 122. Jha, K, Parmentier, E. M., and Phipps Morgan, J. (1992).
Barcilon, v., and Richter, E M. (1986). Non-linear waves in Mantle flow beneath spreading centers due to mantle de-
compacting media, J. Fluid Mech. 164,429-448. pletion and melt retention buoyancy, EOS Trans. Am. Geo-
Batiza, R., Melson, W. G., and O'Hearn, T. (1988). Simple phys. Unioll73, 291.
magma supply geometry inferred beneath a segment of the Johnson, K T., Dick, H. J. B., and Shimizu, N. (1990). Melting
Mid-Atlantic Ridge, Nature 335, 428-431. in the oceanic upper mantle: An ion microprobe study of
Blackman, D. K, and Forsyth, D. W. (1991). Isostatic com- diopsides in abyssal peridotites,l. Geophys. Res. 95,2661-
pensation of tectonic features of the Mid-Atlantic Ridge: 2678.
25-27°30'S, J. Geophys. Res. 96, I 1741- 11758. Kinzler, R. J., and Grove, T. L. (1992a). Primary magmas of
Brodholt, J. P., and Batiza, R. (1989). Global systematics of mid-ocean ridge basalts. I. Experiments and methods, J.
unaveraged mid-ocean ridge basalt compositions: Com- Geophys. Res. 97, 6885-6906.
ment on "Global correlations of ocean ridge basalt chem- Kinzler, R. J., and Grove, T. L. (l992b). Primary magmas of
istry with axial depth and crustal thickness" by E. M Klein mid-ocean ridge basalts. 2. Applications, J. Geophys. Res.
and C. H. Langmuir, J. Geophys. Res. 94, 4231-4239. 97,6907-6926.
Buck, W. R., and Su, W. (1989). Focussed mantle upwelling Klein, E. M., and Langmuir, C. H. (1989). Local versus global
below mid-ocean ridges due to feedback between viscosity variations in ocean ridge basalt compostion: A reply, J.
and melting. Geophys. Res. Lett. 16,641-644. . Geophys. Res. 94, 424 1-4252.
Ceuleneer, G. (1992). Distribution of melt-migration struc- Kohlstedt, D. L., and Hirth, G. (1992). The effect of melt frac-
tures in the mantle peridotites of Oman: Implications for tion on the strength of olivine aggregates deformed in the
magma supply processes at mid-ocean ridges, EOS Trans. diffusion creep regime, EOS Trans. Am. Geophys. Union
Am. Geophys. Union 73, 537. 73,517.
Cheadle, M. (1989). "Properties of Texturally Equilibrated Kuo, B.-Y., and Forsyth, D. W. (1988). Gravity anomalies of
Two-Phase Aggregates," Ph.D. Thesis, Cambridge Univ. the ridge-transform system in the South Atlantic between
Chen, Y. (1992). Oceanic crustal thickness versus spreading 31 and 34SS: Upwelling centers and variations in crustal
rate, Geophys. Res. Lett. 19,753-756. thickness, Marine Geophys. Res. 10,205-232.
Cooper, R. E, and Kohlstedt, D. L. (1986). Rheology and Langmuir, C. H., and Bender, J. E (1984). Petrological and
structure of olivine-basalt partial melts, J. Geophys. Res. tectonic segmentation of the East Pacific Rise, 5°30'-
91,9315-9323. 14°30', Earth Planet. Sci. Lett. 69,107-127.
Cordery, M. J., and Phipps Morgan, J. (1992). Melting and Lin, J., Purdy, G. M., Schouten, H., Sempere, J.-c., and Zer-
mantle flow beneath a mid-ocean spreading center, Earth vas, C. (1990). Evidence from gravity data for focussed
Planet. Sci. Lett. 111,493-516. magmatic accretion along the Mid-Atlantic Ridge, Nature
Cormier, M. H., Detrick, R. S., and Purdy, G. M. (1984). 344, 627 -632.
Anomalously thin crust in oceanic fracture zones: New Lin, J., and Phipps Morgan, J. (1992). The spreading rate de-
seismic constraints from the Kane fracture zone, J. Geo- pendence of three-dimensional mid-ocean ridge gravity
phys. Res. 89,10,249-10,266. structure, Geophys. Res. Lett. 19, 13-16.
Daines, M. J., and Richter, E M. (1989). An experimental Macdonald, K. C. (1984). Mid-ocean ridges: Fine scale tec-
method for directly determining the interconnectivity of tonic, volcanic and hydrothermal processes within the
melt in a partially molten system, Geophys. Res. Lett. 15, plate boundary zone, Annu. Rev. Earth Planet. Phys. 10,
1459-1462. 155-190.
Forsyth, D. w., and Wilson, B. (1984). Three-dimensional Malinverno, A., and Pockalny, R. A. (1990). Abyssal hill to-
temperature structure of a ridge-transform-ridge system. pography as an indicator of episodicity in crustal accretion,
Earth Planet. Sci. Lett. 70, 355-362. Earth Planet. Sci. Lett. 99, 154-169.
Fowler, A. C. (1985). A mathematical model of magma trans- McKenzie, D. (1984). The generation and compaction of par-
port in the asthenosphere, Geophys. Astrophys. Fluid Dyn. tially molten rock, J. Petrol. 25, 713-765.
33,63-69. Morris, E., and Detrick. R. S. (1991). Three-dimensional anal-
Fowler, A. C. (l990a). A compaction model for melt transport ysis of gravity anomalies in the MARK area, Mid-Atlantic
in the Earth's asthenosphere. Part I. The basic model, in Ridge 23°N, J. Geophys. Res. 96,4355-4366.
"Magma Transport and Storage" (M. P. Ryan, ed.), pp. 3- Neumann, G. A., and Forsyth, D. W. (1993). The paradox of
14, Wiley, Chichester/Sussex, England. the axial profile: Isostatic compensation along the axis of
Fowler, A. C. (l990b). A compaction model for melt trans- the Mid-Atlantic Ridge? J. Geophys. Res. 98, 17891-
port in the Earth's asthenosphere. Part II: Applications, 1791 I.
in "Magma Transport and Storage," (M. P. Ryan, ed.), Niu, Y., and Batiza, R. (1991). An empirical method for cal-
pp. 15- 32, Wiley, Chichester/Sussex, England. culating melt compositions produced beneath mid-ocean
76 David W. Sparks and E. M. Parmentier

ridges: Application for axis and off-axis (seamounts) melt- Sleep, N. H. (1974). Segregation of magma from a mostly
ing, J. Geophys. Res. 96,21753-21777. crystalline mush, Geol. Soc. Am. Bull. 85, 1225-1232.
O'Hara, M. 1. (1975). Is there an Icelandic mantle plume?, Na- Sleep, N. H. (1984). Tapping of magmas from ubiquitous
ture Phys. Sci. 253, 708-710. mantle heterogeneities: An alternative to mantle plumes?,
Oxburgh, E. R., and Parmentier, E. M. (1977). Compositional J. Geophys. Res. 89, 10029-10041.
and density stratification in oceanic lithosphere-causes Sleep, N. H. (1988). Tapping of melt by veins and dikes, J.
and consequences, J. Geol. Soc. London 133, 343-355. Geophys. Res. 93, 10255-10272.
Parmentier. E. M., and Phipps Morgan, 1. (1990). The spread- Solomon, S. c., and Toomey, D. R. (1992). The structure
ing rate dependence of three-dimensional structure in oce- of mid-ocean ridges, Annu. Rev. Earth Planet. Sci. 20,
anic spreading centers, Nature 348, 325-328. 329-364.
Phipps Morgan, J. (1987). Melt migration beneath mid-ocean Sotin, C, and Parmentier, E. M. (1989). Dynamical conse-
spreading centers, Geophys. Res. Lett. 14, 1238-1241. quences of compositional and thermal density stratifica-
Phipps Morgan, 1. (l99\). Mid-ocean ridge dynamics: Obser- tion beneath spreading centers, Geophys. Res. Lett. 16,
vations and theory, Rev. Geophys. Suppl. 14, 807-822. 835-838.
Phipps Morgan, 1., and Forsyth, D. W. (1988). Three- Sotin, C; Parmentier, E. M., Carey-Gailhardis, E., and Stoclet,
dimensional flow and temperature perturbations due to P. (1992). A 3D multigrid Poisson solver on the connection
a transform offset: Effects on oceanic crustal and upper machine: Application to convection within planetary inte-
mantle structure. J. Geophys. Res. 93, 2955-2966. riors, submitted, 1. Comput, Phys.
Phipps Morgan, J., and Chen. Y. 1. (1992). Magma injection, Sparks, D. w., and Parmentier, E. M. (1991). Melt extraction
hydrothermal circulation. crustal flow and the genesis of from the mantle beneath spreading centers, Earth Planet.
oceanic crust. EOS Trans. Am. Geophys. Union 73, 290. Sci. Lett. 105, 368-377.
Purdy. G. M., and Detrick. R. S. (1986). Crustal structure of Sparks, D. W., and Parmentier, E. M. (1993). The structure of
the Mid-Atlantic Ridge at 23°N from seismic refraction three-dimensional convection beneath oceanic spreading
studies, J. Geophys. Res. 91,3739-3762. centres, Geophys. J./nt. 112,81-91.
Rabinowicz, M., Nicolas, A., and Vigneresse, 1. L. (1984). A Sparks, D. w., Parmentier, E. M., and Phipps Morgan, 1.
rolling mill effect in asthenosphere beneath oceanic spread- (i993). Three-dimensional convection beneath a seg-
ing centers, Earth Planet. Sci. Lett. 67, 97 -I 08. mented spreading center: Implications for along-axis vari-
Raitt, R. W. (1963). The crustal rocks, in "The Sea" (M. N. ations in crustal thickness and gravity, J. Geophys. Res. 98,
Hill, ed.), pp. 85-102, Wiley-Interscience, New York. 21,977-21,995.
Reid, I., and Jackson, H. R. (1981). Oceanic spreading rate and Spence, D. A., and Turcotte, D. L. (1985). Magma-driven
crustal thickness. Marine Geophys. Res. 5,165-172. propagation of cracks, J. Geophys. Res. 90, 575-580.
Ribe, N. M. (1985). The deformation and compaction of Spiegelman, M., and McKenzie, D. (1987). Simple 2-D mod-
partially molten zones, Geophys. J. R. Astron. Soc. 83, els for melt extraction at mid-ocean ridges and island arcs,
487-501. Earth Planet. Sci. Lett. 83,137-152.
Richter, F. M., and McKenzie, D. (1984). Dynamical models Spiegelman, M., and Kenyon, P. (1992). The requirements for
for melt segregation from a deformable matrix, J. Geol. chemical disequilibrium during magma migration, Earth
92,729-740. Planet. Sci. Lett. 109,611-620.
Ryan, M. P. (1988). The mechanics and three-dimensional in- Spiegelman, M. (l993a). Flow in deformable porous media.
ternal structure of active magmatic systems: Kilauea Vol- I. Simple analysis, J. Fluid Mech. 247, 17-38.
cano, Hawaii, 1. Geophys. Res. 93, 3213-4248. Spiegelman, M. (l993b). Flow in deformable porous media.
Scott, D. R., and Stevenson, D. 1. (1984). Magma solitons, II. Numerical analysis-The relationship between shock
Geophys. Res. Lett. 11, 1161-1164. waves and solitary waves, J. Fluid Mech. 247,39-63.
Scott, D. R., and Stevenson, D. 1. (1986). Magma ascent by Spiegelman, M. (l993c). Physics of melt extraction: Theory,
porous flow, J. Geophys. Res. 91, 9283-9296. implications and applications, Philos. Trans. R. Soc. Lon-
Scott, D. R., and Stevenson, D. 1. (1989). A self-consistent don A 342, 23-41.
model of melting. magma migration, and buoyancy-driven Stevenson, D. J. (1989). Spontaneous small-scale melt segre-
circulation beneath a mid-ocean ridge, J. Geophys. Res. gation in partial melts undergoing deformation, Geophys.
94,2973-2988. Res. Lett. 9, 1064-1070.
Sempere, 1.-c., Rabinowicz, M., Rouzo, S., and Rosemberg, Turcotte, D. L., and Ahern, 1. L. (1978). A porous flow model
C. (1992). Three-dimensional mantle flow beneath spread- for magma migration in the asthenosphere, 1. Geophys.
ing centers: Implications of the models, EOS Trans. Am. Res. 83, 767-772.
Geophys. Union 73, 494. Waff, H. S., and Bulau, 1. R. (1979). Equilibrium fluid distri-
Shaw, H. R. (1980). The fracture mechanism of magma trans- bution in an ultramafic partial melt under hydrostatic stress
port from the mantle to the surface, in "Physics of Mag- conditions, J. Geophys. Res. 84, 6109-6114.
matic Processes" (R. B. Hargraves, ed.), pp. 201-244, Watson, E. B. (1982). Melt infiltration and magma evolution,
Princeton Univ. Press, Princeton, N1. Geology 10, 236-240.
Shen, Y., and Forsyth, D. W. (1992). The effects of temperature- White, R. S., MacKenzie, 9., and O'Nions, R. K. (1992). Oce-
and pressure-dependent viscosity on three-dimensional anic crustal thickness from seismic measurements and rare
passive flow of the mantle beneath a ridge-transform sys- earth element inversions, J. Geophys. Res. 97,
tem, J. Geophys. Res. 97,19717-19728. 19683-19716.
 Chapter 5 Dike Patterns in Diapirs beneath Oceanic Ridges:
The Oman Ophiolite

A. Nicolas, F. Boudier, and B. IIdefonse

Overview the active ridge axis and center of crustal accre-

tion. Evidence from ophiolites and from oceanic
Dike orientation was systematically measured in the observations (Auzende et al., 1989) or from ocean
vicinity of mantle diapirs exposed in the Oman ophio- crust drilling (Anderson et al., 1982) indicates
lite. Collectively,diking reflects a continuous history of that below the blanket of basaltic extrusives is a
melt injection, from the early gabbro and pyroxenite sheeted dike complex composed of vertical dia-
dikes emplaced in a concurrently melting peridotite, to base dikes that are parallel to the oceanic ridge
the dominantly gabbro dikes emplaced in peridotites at
trend. In addition, the existence of melt lenses at
subsolidus temperatures, and, finally, to microgabbro
Moho level, 5-20 km away from the East Pacific
and diabase dikes emplaced in peridotites that have
cooled below SOO°e. Preferred orientations of dikes are Rise, has been proposed by Garmany (1989) and
generally poor, but become progressively better with Barth et al. (1991) on the basis of seismic experi-
decreasing wall-rock temperatures in the host perido- ments. If these results are confirmed, they show
tite. An orientation that corresponds to steep dikes par- that both sills and extensional dikes feed the ac-
allel to the inferred ridge trend (itself determined by the creting oceanic crust. Vertical fracturing related to
trend of the diabase sheeted dike complex) is first rec- extension and underplating by sills are also the
ognized and is ascribed to control of the tensional lith- two principal modes of basaltic injection envis-
ospheric stress field. A second diffuse intrusion orien- aged for the deep continental crust, as illustrated
tation corresponds to sills or moderately dipping dikes. by structural studies of diabase dike injection at
It is mainly detected among the early, high-temperature
the mantle-crust interface exposed on Zabargad
injections. Sill formation is ascribed to the stress field
Island in the Red Sea (Nicolas et al., 1987; Bou-
created by the mantle diapir's divergent flow in the
sub-ridge asthenosphere just below the newly accreted dier et al., 1988). There, a mantle diapir pene-
crust. Advecting melt can circulate either within "lith- trated the continental crust during an early stage
ospheric dikes" or within" asthenospheric sills." Over- of Red Sea opening. The diapiric peridotites and
all, these relations can be explained by a sudden change the surrounding deep crustal gneisses were cut by
from a lithospheric to an asthenospheric stress field. vertical dikes oriented parallel to the Red Sea
The change could result from the relaxation of litho- trend and thought to represent extensional frac-
spheric stresses in response to tension fracturing of the tures; emplacement of thick diabase sills followed,
lithospheric lid, triggered by a sudden melt surge. with intrusion taking place selectively at the inter-
face between peridotites and overlying rocks.
In this chapter, we explore the problem of
Introduction magma transport structures and diking at oceanic
ridges by reporting a structural study of dike ori-
Direct knowledge of the diking system in the up- entations in the peridotite section of the Oman
permost mantle below the currently active oceanic ophiolite. This study is similar to previous work
ridges is not, unfortunately, possible. This funda- on Zabargad Island. Oman ophiolite is particu-
mental question, however, is intimately related to larly well suited to such a study because, during
the overall problems of melt extraction from the the course of a long-term project on structural
mantle and of the subsequent melt moving toward mapping conducted since 1980, mantle diapirs

Magmatic Systems Copyright © 1994 by Academic Press, Inc.

Edited by M. P. Ryan 77 All rights of reproduction in any form reserved.
78 A. Nicolas et al.

through which melt was channeled to the overly-

ing accreting ridge were discovered (Nicolas et
al., 1988; Ceuleneer et al., 1988; Ceuleneer, 1991;
Nicolas and Boudier, submitted 1993). Diapiric
structures in the mantle peridotite are associated
with a dense network of gabbro and pyroxenite
dikes, melt impregnation veins and clots, mainly
within the several-hundred-meters-thick transition
zone located just below the Moho and above the
relatively homogeneous harzburgitic mantle. It has
been suggested that massive melt impregnation
within this zone could have reduced the mantle
viscosity and thus facilitated the sharp rotation of
flow lines within the diapir from a vertical to a
horizontal orientation (Rabinowicz et al., 1987).
Another favorable circumstance is that these dia-
pirs must have been detached, as part of a future
ophiolite, from the ridge of origin while they were
still active or in a waning stage of activity; other-
wise the high-temperature structures, characteris-
tic of their uprise, would not have been preserved.
Figure 1 Map of the Oman ophiolite with the locations of
Thus, a ridge structure and a geologic "moment the massifs considered in this study. Black pattern, mantle sec-
in time" were frozen and the resulting measured tions; gray pattern, crustal sections.
dike system relates directly to the active diapiric
structures of the sub-ridge mantle. It should be
noted that in most other studies of dike orien- pristine oceanic spreading centers subsequently
tations within peridotites, the continuing plastic dissected by erosion. The near absence of any late
flow in peridotites after dike emplacement par- deformation in these massifs is also better under-
tially blurs the intrusion picture (Fig. 4c, Nicolas stood by considering the results of a new gravi-
and Jackson, 1982). metry study, suggesting a keel of relatively strong
Finally, although data on dikes from the entire peridotite that is about 6 km in thickness (Ravaut,
ophiolite belt of Oman are available, we present in 1992).
this chapter only the results obtained in two mas- Within this well-identified ridge frame and in
sifs from the southeastern part of the belt (Fig. I) contrast with the other clearly organized struc-
where a detailed structural study was conducted tures, the orientation of dikes has been found to be
(Nicolas and Boudier, submitted 1993). A similar irregular on every scale, as shown by the diffuse
study, with similar results, is now being conducted patterns of preferred orientation presented in this
in a third massif (Wadi Tayin) of Oman (Ildefonse chapter (see also Nicolas and Boudier, submitted
et al., 1993). These massifs were selected be- 1993). Because of this dispersion in dike orienta-
cause of the flat-lying and regular attitude of the tions, it is necessary to further analyze these data
Moho, which (1) tends to limit the errors related by individual rotations in the reference frames of
to the rotation necessary to restore the structures both the ridge and the local mantle plastic flow.
in the paleo-ridge reference frame (Moho hori- Consideration of the structure of individual dikes
zontal) and (2) suggests that these massifs have and mutual relationships is also important.
experienced only moderate deformation during
and after their obduction onto the Arabian shield
margin. One main conclusion of the study cited Regions of Diapirism and Sub-ridge Flow
above is that the two massifs under consideration
(Nakhl-Rustaq and Maqsad) have escaped any Paleo-ridge structures were reconstructed in the
visible deformation subsequent to ridge accretion Maqsad (Fig. 2) and Nakhl-Rustaq (Fig. 3) mas-
and, thus, can be structurally considered relatively sifs from detailed structural mappings of the high-
 5. Dike Patterns in the Oman Ophiolite 79

temperature foliations and lineations induced by sense) of the local mantle flow (thick arrows in
solid-state flow in the mantle section and from Figs. 2A and 3A show the shear flow away from
magmatic foliations and lineations in the layered the diapir). Additional information on the spread-
gabbro unit. The ridge trend itself is defined by ing axis location comes from the convergence of
the azimuth of the diabase sheeted dike complex. upper gabbro foliations toward the inferred ridge
The peridotite foliations and lineations-normally axis, thus pointing to the along-strike closure of
flat-lying and parallel to the Moho-plunge lo- the magma chamber. Finally, the clustering of dia-
cally and thus define the contours of mantle dia- base dikes in the lower gabbros and in peridotites
pirs. The vertical lineations within the diapir tend suggests close proximity to the spreading axis.
to diverge outside this area as seen in Fig. 3. These These intrusions have been interpreted as prod-
diapirs, delineated in a map view by isodip lines ucts of the activity of the ridge in its dying phase,
of the foliations and lineations (30° in Fig. 2A, 45° with late-stage injections of basaltic melt in a gen-
in Fig. 3A), permit one to locate paleo-spreading erally cooling spreading center.
centers. As discussed further by Nicolas and Bou- In the Maqsad massif, the diapiric area is largely
dier (submitted 1993), this location is indepen- exposed, disappearing only in the southeast below
dently confirmed by a kinematic analysis (shear the lower crustal lithologic units (Fig. 2). It is thus

Figure 2 Maqsad area. (a) Paleo-ridge model (location in Fig. I; detailed map in Fig. 2b). Darker shaded area, mapped contour
of the mantle diapir (see text); lighter shaded area, presumed limit of the magma chamber. The ridge axis is independently
located as follows: (I) crossing through the center of the diapir, parallel to the sheeted dike trend; (2) separating opposed shear
senses of plastic flow in the mantle just below the Moho (thinner dashed line); (3) corresponding to the convergence of magmatic
foliations in upper gabbro units (dotted line); and (4) being preferentially intruded by the last melts (diabase dikes) issued from
the dying and cooling ridge (modified after Nicolas and Boudier, submitted 1993). (b) Geological map and dike trajectories.
(c) Map of individual dike measurements. Note the convergence of dike trajectories toward the ridge axis NW of the diapir
center (see Fig. 2a). The data on the sheeted dike attitude come mostly from the NE corner of the map. Continuous black lines
represent faults and, dominantly, hydrothermal fracture zones parallel to the sheeted dike complex (Nicolas and Boudier, sub-
mitted 1993).
continues
80
 58·
"\

J )c·:-;",
..• - <. ".~

v- '\

.c-: .
: : ""
;:. -':~.\\
:" , _:;

...... ., \~)"': "J .

~;'i
:t-<--
\~
~ ~ ./
I;/! :,~% ""
'. . : ..
','
•.. ..
~:
". ~

~ v.. ~ ..,
~~.r;'~· () ~\\~
.':"'>..
.". i'\

:;:!.....
"-

-c
r7: i; ~~!,3~~;. '
-
»,
~
':::~ '-, ('" :.\\.:
\1 ~<>
~ ."

'"
.'
....
...:;..
23' 23'

,l.
'i. dikes in Maqsad massif
.:' %
J [If d" diabasedikes
'"
" '"11iI"'..::'--
i/
gabbroand pyroxenite dikes
"-'¥Iy:."'>"
"'::::--"1.1 impregnation veins

.
.......)

"'\,
-::-:t,,::.•.__....!......\
............., (.:, /'j.":: .s-: )1. .
o 10Km
$'
c
58'

Figure 2 Continued.
82 A. Nicolas et al.

possible to compare the dike orientations on the Dike Typology in Relation to Diapiric Areas
NE and SW sides of the ridge axis. The center of
the diapir may also be exposed in the mantle sec- Several generations of basaltic dikes may be dis-
tion, making it possible to compare dike orienta- tinguished and mapped separately in peridotite
tions along the strike of a ridge segment. Unfor- massifs (Nicolas and Jackson, 1982; Nicolas,
tunately, mantle outcrops are limited in the SE 1989, pp. 21 and 65-67). Mapping the vast Oman
part of the segment compared to the northwestern ophiolite and delineating diapiric areas in mantle
portion, restricting possible comparisons. sections made it possible to clarify the relation be-
In the Nakhl-Rustaq massif (Fig. 3), two dia- tween diapirs and successive generations of dikes.
piric areas have been identified and are separated Dike density in the mantle sections of this ophio-
by a high-temperature shear zone. In the west- lite is very irregular, but it is always high in dia-
ern portion, mantle lithologies are exposed only piric areas. As noted previously, this feature is
around the southern termination of a diapir whose explained by the channeling of melting astheno-
detailed relationship with the overall ridge seg- sphere and of melt itself by diapirs. In the frozen
mentation is unclear. In the eastern area, mantle diapirs considered here, it is still possible to ob-
lithologies are exposed along the southwestern serve this original dike organization. In contrast,
side of the ridge axis, making it possible to com- in mantle areas far from diapirs such as at the tip
pare dike attitudes along the (25-km) half-length of a presumed propagating ridge (Fig. 3), the vig-
of a ridge segment. orous mantle flow radiating from the diapir has

23° 2Q'
c t~ds of lineations···..
in the mantleand . . .
flow sense o 5km
5]040' a
Figure 3 Nakhl-Rustaq massif. (a) Paleo-ridge model (location in Fig. I; detailed map in Fig. 3b). Two diapiric areas have
been identified (dark shading) and treated separately. The dikes in the western diapir (Fig. 9) were measured in the mantle rocks
around the southern tip of the diapir; those in the eastern diapir (Fig. 10) along the SW flank of the ridge axis. Same symbols as
in Fig. 2a (modified from Nicolas and Boudier, submitted 1993). (b) Geological map and dike trajectories. (c) Map of individual
dike measurements. The northerly tilt of the massif exposes the mantle section to the south and the sheeted dike complex to the
north. Local northward curvature in the sheeted dike trend is visible in the eastern part of the map and has been related to tectonic
activity at the ridge (Nicolas and Boudier, submitted 1993).
continues
 57°30' .57°40'
23°40' 23°40'
I

"C ~~ .••• '~r ' \

", ". '" \ '\ \
. . ;:~:~:,:- -c, ~" J }~': 1
, ~. ..
. ~-''-. i 1I
J \1\ ,P'
~
............:/.'(

...... ),
....> '-'~.'
" ..\...

.;;.:
~ "'" If I
"'/L
.. ,·'·'~T
I.···
L
\
~.
\

Rustak
.................
\.,.. -,--'- . ......... ~~

II J -
J-

'>-.. dikes in Nakhl-Rustaq massif

""= diabases
and hydrothermal dikes
23°20' .? -», gabbroand pyroxenne 23°20'
dikes

~ ..: o 5Km

57°40'
c

Figure 3 Continued.
 5. Dike Patterns in the Oman Ophiolite 85
deformed and transposed parallel to the flat-lying 450°C (Nehlig, 1989). Their orientation in the
foliation, all dikes and melt segregations formed field is generally regular and their thickness is
inside the diapir; only late dikes are discordant about 1 m.
with respect to this foliation. This classification of dikes as a function of
The first generation of dikes comprise early decreasing host rock temperature and, generally,
"indigenous" gabbro and pyroxenite dikes (Figs. the timing of intrusion should not be viewed too
4a, 4c, and 4d), crossing melt-impregnated peri- strictly, since there is continuity between the in-
dotites and having equilibrated at temperatures digenous and the intrusive dikes, suggesting that
above the solidus (l200°C). The pyroxenite dikes magma-induced fracture is a continuous process
are dominantly olivine-bearing websterites or eli- and one that spans the evolving structure of a dy-
nopyroxenites and are less commonly orthopy- namic diapir and its surroundings. In addition,
roxenites. Both gabbro and websterite dikes may gabbros or pyroxenites can intergrade into each
grade into impregnation clots (Fig. 4c) or into other and are encountered equally as often within
chromite schlierens and pods. Dunite veins are the different categories of high-temperature dikes.
related to these dikes and are formed by wall-
rock reaction of the circulating melt in the dike
(Figs. 4a, 4b, 4c; see recent reviews by Kelemen, Dike Patterns in the
1990, and Kelemen et al., 1990). Within diapirs, Southeastern Oman Diapirs
this first generation of dikes and sills is often
poorly defined in orientation. In contrast, away In their study of the diapirs of southern Oman,
from diapirs in areas where post-intrusion plastic Nicolas and Boudier (submitted 1993) concluded
flow has been intense, these early dikes are de- that in the peridotite section, two orientation
formed and transposed parallel to the foliation and groups of dikes and veins are defined. This con-
have thus been incorporated into the ubiquitous clusion is based on consideration of all the dikes
peridotite banding or layering (Fig. 4d). The next and veins collectively, regardless of their nature
generational type of gabbro and pyroxenite dikes and emplacement timing. These two groups are
has been called "intrusive" because such dikes apparent in maps of the dike trajectories (Figs. 2b
have sharp and straight walls without reactions and 3b). One well-defined group of dikes that
with surrounding peridotites and because they dis- are steep and parallel to the sheeted dike complex
play a comb structure that has been induced by is primarily composed of low-temperature dikes:
melt crystallizing from the cooler walls toward the diabase and fine-grained gabbros. It also includes
dike center (Fig. 5a). Thus, in contrast to the in- some high-temperature dikes that are indigenous
digenous dikes, these dike types were emplaced in as well as intrusive. Indigenous dikes, however, as
peridotites below their solidus. They commonly well as the dunite veins, have more dispersed ori-
constitute swarms of parallel dikes and range in entations than intrusive dikes and, geographically,
thickness from a few centimeters to a few tens of they tend to be restricted to the diapiric areas. Dia-
centimeters (Fig. 5b). They are generally unde- base dikes are not common in the peridotite or in
formed and discordant with respect to the folia- the lower gabbro sections. As mentioned above,
tion both within the vicinity of diapirs and far these dikes cluster near the inferred ridge axis.
from them. Through microgabbro dikes, showing The orientation of the second group of dikes is
evidence of brittle fracturing, they grade into dia- more diffuse than that of the first group. They usu-
base dikes, first without chilled. margins and fi- ally form moderate angles with respect to folia-
nally with chilled margins. Diabase dikes corre- tion in the enclosing peridotite, and they grade
spond to temperatures in the host rocks lower than into sills, thus plotting close to the center of ste-

Figure 4 Photographs of indigenous dikes. (a) Irregular indigenous clinopyroxenite dike (hammer) s~rrounded by a dark dunite
wall in contact with the lighter harzburgite. A coeval dunite sill is present as an off-shoot (parallel to the horizontal foliation and
banding of the harzburgite). Another thin and more regular indigenous clinopyroxenite dikelet (left) is cut by an intrusive gabbro
dikelet. (b) Dunite vein network in darker harzburgite; the sill and dike orientations are visible to the right. (c) Irregular indige-
nous gabbro dike grading into melt impregnations (to the left). (d) Melt impregnations, marked by lighter plagioclase, which
have been tectonically transposed parallel to the peridotite foliation.
 5. Dike Patterns in the Oman Ophiolite 87

Figure 5 Photographs of intrusive gabbro dikes in harzburgites. (a) Comb structure perpendicular to the sharp-margin dike
wall. (b) Network of parallel dikes emplaced into a harzburgite with a flat-lying foliation.
 88 A. Nicolas et at.

Figure 6 Relations between sills and dikes. (a) Apparent feeding of gabbro sills in dunites by a vertical dike (to the right);
other dikelets are visible near the lower left comer (Buri, Sumail Massif). (b) Network of gabbro sills and dikes with melt that
circulated into one another. This photograph is from northern Oman (Wadi Fayd); the dikes are EW and the sills are NNW-SSE,
parallel to the foliation and to a locally steeply dipping Moho. Note an olivine segregation in the center of intrusions.
 5. Dike Patterns in the Oman Ophiolite 89
reograms (Figs. 8, 9, 10). As a consequence, in lineation (oriented EW in these stereograms). If
map views, they tend to wrap around the diapir. the orientation of the dikes were controlled by
This second group of dikes is composed of dun- high-temperature plastic flow in the country rocks,
ite veins, indigenous dikes, and intrusive coarse- one would expect the dikes to display a geomet-
grained gabbros. Diabase dikes and fine-grained rical relationship with the structural elements. We
gabbros intruded into cooling peridotites are pres- show in Fig. 7 the results for the (a) intrusive and
ent but not common in this group. (b) indigenous dikes. The preferred orientation of
An observation, mentioned several times in the intrusive dikes (Fig. 7a) is weak and does not ex-
field, is that the same melt can be injected in both hibit a pattern consistent with control related to
a dike and a sill configuration. In a few places, high-temperature foliation. On the other hand, in-
dikes oriented parallel to the inferred ridge trend digenous dikes (Fig. 7b) plot as a weak girdle,
have been observed feeding sills, which are them- with a maximum normal to the foliation that cor-
selves parallel to the Moho within the transition responds to fractures parallel to the foliation and
zone (Fig. 6a). More commonly, in areas of heavy may suggest some structural control.
diking, dikes and sills show no cross-cutting rela-
tionships, and flow structures are continuous and Dikes in the Ridge Reference Frame
locally wrap around the corner of the intersections
(Fig. 6). To study the dike orientations in the paleo-ridge
A closer analysis of these data is now presented reference frame, we rotated the Moho in the two
to examine the relationships between dike orien- massifs considered into a horizontal position. The
tation in the uppermost mantle and either the plas- rotation is 15° down to the west about a north-
tic flow framework or the paleo-ridge framework. south axis (symbolized as OOW15°) in Maqsad and
0SW32°
from 110 to 105°SW30° in Nakhl-Rustaq
Dikes in the Plastic Flow Reference Frame from west to east. These rotations-deduced from
structural maps-have been indirectly checked by
These relationships have been studied in the Maq- the fact that they result in nearly vertical preferred
sad area. All dikes, whatever their mineral con- orientations of the diabase dikes from the sheeted
tent, grain size, or structure, have been individu- dike complex (continuous line in the stereograms
ally rotated into the structural reference frame of of Figs. 8-10). All dikes were rotated following
their enclosing peridotites: the high-temperature these axes and refer to these NW-SE paleo-ridge
foliation (oriented perpendicular to the projection directions.
plane of the stereograms in Fig. 7) and the mineral Figure 8 compares the dike orientations from

Figure 7 Stereograms (lower hemisphere) showing the pole distribution of (a) intrusive and (b) indigenous dikes from the
Maqsad massif. plotted in the plastic flow reference frame: foliation vertical EW (straight line), lineation EW (dots at ends of the
line). Contours 1.2,4% net area between 100 and 200 measurements and 0.5,1,2% net area above 200 measurements. Numbers
indicate the hemisphere subtotal of individual measurements.
90 A. Nicolas et at.

Figure 8 Stereograms (lower hemisphere) of the various dikes in Maqsad massif, plotted into the paleo-ridge reference frame:
Moho horizontal, ridge trend defined by that of the sheeted dike complex (continuous line). The computed best plane is projected
as dashed lines. Contours: 2, 4, 8% between 30 and 100 measurements; I, 2, 4, 8% between 100 and 200 measurements; and
0.5, 1,2,4% above 200 measurements. Numbers indicate the hemisphere subtotal of individual measurements. Wand E symbols
refer to data from southwest (W) and northeast (E) of the inferred Maqsad ridge axis.

Figure 9 Stereograms (lower hemisphere) of (a) pyroxenite and (b) gabbro dikes around the western diapir of Nakhl-Rustaq
massif, plotted into the paleo-ridge reference frame. Same conventions as in Fig. 8.
 5. Dike Patterns in the Oman Ophiolite 91

Figure 10 Stereograms (lower hemisphere) of the various dikes along the SW flank of eastern diapir of Nakhl-Rustaq massif,
plotted in the paleo-ridge reference frame. Same conventions as in Fig. 8.

the SW side of the Maqsad ridge segment (Figs. Nakhl-Rustaq massif. The intrusive dikes were
8a-8c) with the orientations from the NE side not distinguished here from indigenous ones. Py-
(Figs. 8d-8f). The diabase dikes emplaced in the roxenite and gabbro dikes have diffuse but com-
peridotites and lower gabbro sequence (Figs. Sa 'parable orientations in the ophiolite. This agree-
and Sd) plot nearly parallel to the similar diabase ment is not surprising since they grade into one
dikes of the sheeted dike complex, thus clearly be- another, with a tendency for the pyroxenite dikes
ing related to regional lithospheric tensional frac- to postdate the gabbro dikes. Both categories of
turing. Incidentally, the sheeted dikes used as ref- dike plot as diluted girdles composed of dikes
erence are exposed 25 km from the diapiric area, parallel to the sheeted dike complex and others
as seen in Fig. 2, illustrating our previous con- approaching the attitude of sills parallel to the
clusion that the massif was not significantly de- paleo-Moho, with, however, a large number of in-
formed after the original ridge accretion events. A termediate orientations.
first inspection of the intrusive dikes (Figs. 8b and In the mantle section SW of the eastern seg-
8e) leads to the conclusion that they are also pref- ment, the diabase dikes exposed in the mantle and
erentially emplaced according to a lithospheric lower crust close to the paleo-ridge axis (Fig. lOa)
tensional orientation. However, their overall ori- plot in a dominant orientation that is parallel to
entation is more scattered, with the occurrence of the sheeted dikes, with, however, a significant
sills plotting near the center of the nets. The best number of admixed sills. Measurements of pyrox-
computed plane (dashed lines) of these two stereo- enite dike orientations are simply too scarce to be
grams shows that the dike azimuths tend to conver- representative (Fig. lOb). The plot of gabbro dikes
gence toward the ridge axis, meaning that dikes (Fig. 1Oc) produces a weak girdle, again reflecting
from the SW flank plot with a preferred NNW the dichotomy of dikes oriented approximately
trend, and, conversely, dikes from the NE flank, parallel to the sheeted dikes and of sills oriented
with a WNW trend; this feature was already vis-' horizontally or with a moderate dip. The girdle
ible on the trajectory map (Fig. 2b). Finally, the obliquity with respect to the sheeted dike orienta-
indigenous dikes (Figs. 8c and 8f) produce diffuse tion shows the same sense as that recorded in the
stereograms with, possibly, a similar tendency to Maqsad massif, consistent with the inferred struc-
split into a group of steep NW-SE dikes and tural setting on the SW flank of the ridge axis.
another group of sills. The gropp of indigenous
dikes of NW-SE lithospheric orientation suggests
the same convergence toward the ridge axis. Discussion
Figure 9 compares the orientations of c1inopy-
roxenite and gabbro dikes in the mantle around In considering dike orientations, one must be
the southern termination of the western diapir of aware that, except for diabase dikes, there is a
92 A. Nicolas et al.

spectrum of orientations, even at the scale of a massifs considered, where the presence of mantle
single outcrop. Dikes as remarkably parallel as diapirs permitted the location of the ridge axis in
those in Fig. 5b are uncommon. This spectrum ex- the field (Figs. 2a and 3a), it is observed that dikes
plains why the most representative stereograms, from the SW flank trend more northerly than the
which include a few hundred measurements made ridge axis and dikes from the NE flank trend more
over areas of several tens of square kilometers, southerly (Fig. 8).
show diffuse preferred orientations. Therefore,
their interpretation, in an effort to understand When projected in the structural reference
regional and local contributions to their mechan- frame of the enclosing peridotites (high-tempera-
ics, is not straightforward. It is also necessary ture foliation and lineation, Fig. 7), dikes do not
to observe a large number of individual situa- exhibit clear preferred orientations. This fact in-
tions before reaching conclusions. On the basis of dicates that dike orientation is not controlled by
the present study and measurements and observa- local flow structure or by a mechanical weakness
tions in other areas of the Oman ophiolite (Ilde- plane, which would relate to these structures.
fonse et al., 1993), the following points seem well There is, however, a weak tendency for indige-
established. nous dikes to plot parallel to the foliation. This
tendency may reflect the tectonic transposition of
• Whatever the point of view, there is a con- dikes due to high-temperature flow, as described
tinuum between the different dike intrusions. Pet- in previous studies (Nicolas and Jackson, 1982).
rologically, pyroxenites and gabbros grade into . Many dikes were observed to be wholely plasti-
one another, coarse gabbros grade into microgab- cally deformed and close in orientation to the fo-
bros, and these, in tum, grade into diabases. In liation plane. This trait was also found to be true
terms of the time sequence and thermal structure for a large number of chromite deposits (concor-
of wall-rocks, indigenous dikes, intruded into con- dant deposits; Cassard et al., 1981). In both situ-
currently melting peridotites, subsequently grade ations, these concordant features are explained by
into intrusive dikes with evidence of progressively the tectonic rotation of the dikes (tectonic trans-
cooler wall-rocks, the coolest corresponding to position), which is induced by a large plastic flow
diabase dikes with chilled margins. This is also occurring subsequent to the initial dike or chro-
true of their preferred orientations. mite pod intrusion. Because our study was con-
• When plotted in the inferred paleo-ridge ref- ducted mainly in diapiric areas where this plastic
erence frame (Moho rotated to be horizontal, flow (related to spreading away from the ridge) is
ridge trend defined by the diabase sheeted dike reduced, this problem is partly avoided. However,
trend), the intrusions cluster in two main orienta- the deepest indigenous dikes, dunite veins, and
tions, being either vertical and parallel to the chromite pods emplaced in the harzburgite below
sheeted dike trend or flat-lying, comprising sills the transition zone have been tectonically trans-
that are approximately parallel to the Moho. These posed parallel to the peridotite foliation (Ceule-
two preferred orientations grade into one another neer and Nicolas, 1985). They contribute to the
and thus define girdles in their resulting stereo- weak preferred orientation in Fig. 7b, which is
grams. They are observed in all dike types, al- parallel to the plastic foliation plane.
though sills are more strongly represented in the Following earlier conclusions (Nicolas and
indigenous group of intrusions (and are rare in the Jackson, 1982), dike orientations in peridotites are
diabase group). In contrast, vertical dikes domi- believed to be controlled by the ambient stress
nate in the diabase group, where the preferred ori- field, being parallel to the (0"1' O"z) principal stress
entations are also the strongest. directions and, thus, normal to the 0"3 least princi-
• Individual observations confirm that a given pal stress. This conclusion is based on the results
parcel of melt can be injected either as a sill or as of detailed kinematic analyses and on the fact that
a dike and can flow from one orientation to the peridotites are comparatively massive rocks in
other (Fig. 6). which the foliation cannot be considered a me-
• Dikes tend to deviate slightly from the paleo- chanical weakness plane. Our study of dike ori-
ridge trend as defined by the sheeted dikes. In the entations in relation to the structural reference
 5. Dike Patterns in the Oman Ophiolite 93
frame of high-temperature enclosing peridotites is brought to the spreading center by hydrofrac-
tends to support this conclusion by failing to show tures produced within and issuing from the con-
any clear relationship between intrusion-preferred currently rising and melting mantle diapir. Hy-
orientations and this reference frame. drofracturing events are episodic and relatively
This conclusion is readily applied to all vertical violent. They bring to the accreting crust a quan-
dikes parallel to the diabase sheeted dike com- tity of melt large enough to create 1 m of crust, the
plex, which, being parallel to the presumed paleo- width of individual diabase dikes in the sheeted
ridge plane, would be in the orientation of pri- dike complex. Independent pieces of evidence
mary lithospheric fracturing. The occurrence of (melt budget, cooling time of dikes) show that a
sills and moderately inclined dikes suggests the single event lasts only a few weeks. Taking this
existence of another stress field, such that CT3' in- fact into account, the following scenario is pro-
stead of being horizontal and normal to the ridge, posed. In the oceanic lithosphere at the ridge,
is vertical or nearly vertical. Our two-dimensional which is already under tensile stress, the sudden
modeling of asthenospheric mantle diapirism be- release of melt triggers the opening of a fracture,
low ridges (Rabinowicz et al., 1984, 1987) shows which widens to I m during a time lapse of a few
that at the outskirts of diapirs, the forced mantle weeks. This opening relaxes completely the litho-
flow generates (just below the Moho) a stress spheric stress. It is speculated that at the onset of
field such that CT3 is steeply plunging and the (CT" the magmatic event, even at the Moho depth be-
CTz ) plane is flat-lying or moderately inclined to- neath the ridge axis, the regional lithospheric
ward the axis. It is tempting to conclude that the stress field predominates over the more local
sills are controlled by the asthenospheric stress stress field that is related to the relatively small-
field related to mantle diapirism. The predomi- scale diapir activity. Thus, the first intrusion of
nance of sills in the general environment of melt propagates along regional fractures parallel
high-temperature dikes (as contrasted with "low- to the ridge plane (lithospheric fracturing). Once
temperature" intrusions) suggests this tendency, the primary magma batch is injected and litho-
as the former are intruded into asthenospheric spheric tensional stresses are thus relaxed, the
mantle and the latter into lithospheric mantle. magma tends to be subsequently channeled into
A difficulty with the model of dikes and sills sills controlled by the more local asthenospheric
that are controlled by two independent stress stress field that is associated with the divergence
fields-lithospheric and asthenospheric, respec- of diapiric flow. This scenario explains that, dur-
tively-is that the same melt injection can form a ing the same melt surge, "lithospheric" and "as-
sill and a dike, melt moving from one into the thenospheric" controls can be successively ex-
other. The necessarily rapid change in the orien- erted on the magma-induced fracture orientations.
tation of the stress field can be attributed to spatial If the orientation of the sill group is controlled
or temporal changes. Local stress field deviations by the stress field related to diapiric activity, one
produced by shear motions are well documented could expect to observe different orientations de-
in earth science and have been invoked to explain pending on the location with respect to the dia-
the synchronous injection of a fluid phase in both piric center. Our efforts in Maqsad have been un-
shear and tension orientations. This situation is lo- successful in providing a definitive answer here
cally observed in our dikes on the scale of a few (see Fig. 8), probably because there is too much
meters, but we do not believe that it applies gen- scatter in the sill orientations. However, around
erally because within the dikes studied, oblique the Batin diapiric center (Wadi Tayin massif),
foliations symptomatic of shear motion are un- which is still under study, (I1defonse et al., 1993)
common (Figs. 4-6). These intrusions are also we have detected a tendency for the sills to wrap
oriented at 90° angles to each other (Fig. 6) and around the diapir center with a moderate dip away
not at the moderate angle that one should expect from it. The only consistent figure emerging from
in a shearing environment. the present study in terms or' local orientations
The temporal change in stress field is an inter- with respect to diapir location is that the azimuth
esting alternative. In previous publications (Nico- of vertical dikes tends to converge toward the
las, 1986, 1990), it was proposed that basaltic melt ridge axis, which means that to the SW of our
94 A. Nicolas et al.

ridge axes, the dikes have a more northerly azi- gram of the Institut National des Sciences de I'Univers-Centre
muth than the ridge itself (Figs. 8a-8c and 10) National de la Recherche Scientifique.

and, conversely, for those dikes located NE of the

ridge (Figs. 8e and 8f), the trend is southerly. This
tendency may be ascribed to local deviations of References
the "lithospheric" stress, but their detailed origin
Anderson, R. N., et 01. (1982). DSDP Hole 504B, the first ref-
is not well understood.
erence section over I km through layer 3 of the oceanic
crust, Nature 300, 589-594.
Auzende, J. M.. Bideau, D., Bonatti, E., Cannat, M., Hon-
Conclusions norez, J.. Lagabrielle, Y" Mallavieille, J.. Mamaloukas-
Frangoulis, V., and Mevel, e. (1989). Direct observation
of a section through slow-spreading oceanic crust, Nature
The diapiric areas of Oman represent exception-
337,726-729.
ally favorable sites to study the plumbing of a Barth, G.A., Mutter, J.e., and Madsen, J.A. (l99\). Upper-
ridge at the Moho level and immediately below. mantle seismic reflections beneath the East Pacific Rise,
The numerous dikes observed and measured are Geology 19, 994-996.
poorly constrained in their orientations as shown Boudier, E, Ceuleneer, G., and Nicolas, A. (1988). Shear
zones, thrusts and related magmatism in the Oman ophio-
by the stereograms in this chapter. This tendency
lite: Initiation of thrusting on an oceanic ridge, Tectono-
is mainly true for the earliest or more central dikes physics 151, 275-296.
emplaced into the melting peridotites of diapirs. Boudier, E, and Nicolas, A. (1972). Fusion partielle gabbroi-
In contrast, the diabase dikes-emplaced when que dans la lherzolite de Lanzo (Alpes piemontaises), Bul/.
the ridge was cooling-have excellent preferred Suisse Mineral. Petrol. 52, 39-56.
Boudier, E, and Nicolas, A. (1977). Structural controls on par-
orientations that are mainly parallel to the sheeted
tial melting in the Lanzo peridotites, in "Magma Genesis"
dike trend. Whatever the timing and location of (H. J. B. Dick, ed.), pp. 63-77, State Oregon Dpt. Geo!.
the intrusions, a group of dikes is oriented parallel Min. Ind.
to the sheeted dike complex and ascribed to a Cassard, D., Moutte, J., Nicolas, A., Leblanc, M., Rabinowicz,
relatively regionally organized lithospheric stress M., Prinzhofer, A., and Routher, P. (1981). Structural clas-
sification of chromite pods from New Caledonia. Econ.
field. Another group is composed of flat-lying sills
Geol. 76, 805-831.
and moderately inclined dikes, thought to be con- Ceuleneer, G. (l99\). Evidences for a paleo-spreading center
trolled by the locally organized asthenospheric in the Oman ophiolite: Mantle structures in the Maqsad
stress field created by the diapir flow divergence area, in "Ophiolite Genesis and Evolution of Oceanic
just below the ridge. Melt is injected in either the Lithosphere" (T. Peters, A. Nicolas, and R. G. Coleman,
eds.), pp. 147-173, Kluwer, Dordrecht.
sills or the dikes. This feature suggests a relatively
Ceuleneer, G., and Nicolas, A. (1985). Structures in podi-
rapid shift in the orientation of the local stress form chromite from the Maqsad district (Sumail ophiolite,
field, which is attributed to the effects of a large- Oman), Mineral. Deposita 20, 177-185.
volume and relatively violent melt surge from the Ceuleneer, G., Nicolas, A., and Boudier, F. (1988). Mantle
rising diapir into the newly accreting crust. flow patterns at an oceanic spreading centre: The Oman
peridotite record, Tectonophysics 151, 1-26.
Because orientations in the high-temperature
Garmany, J. (1989). Accumulations of melt at the base of
dikes are, at present, so poorly clustered, more young oceanic crust, Nature 340,628-632.
work is necessary before solid conclusions on this Ildefonse, B., Nicolas, A., and Boudier, E (1993). Evidence
interesting class of intrusions may be drawn. from the Oman ophiolite for sudden stress changes during
melt injection, Nature 366,673-675.
Kelemen, P.B. (1990). Reaction between ultramafic rock and
fractionating basaltic magma. 1. Phase relations, the origin
Acknowledgments of calc-alkaline magma series, and the formation of dis-
cordant dunite, J. Petrol. 31, 51-98.
This chapter has benefited from reviews and constructive re- Kelemen, P.B., Joyce, D.B., Webster, J.D., and Holloway, J.R.
marks by John S. Pallister, Michael P. Ryan, and an anony- (1990). Reaction between ultramafic rock and fractionat-
mous reviewer. A number of colleagues and students have ing basaltic magma. II. Experimental investigation of re-
contributed to field measurements: M. Misseri, S. Crambert, action between olivine tholeiite and harzburgite at 1150-
J. L. Bouchez, G. Ceuleneer, 1. Reuber, e. Sotin, V, Thomas, 1050°C and 5 kb. J. Petrol. 31, 99-134.
e. MacLeod, K. Berm, F. Quatrevaux, and S. Tait. Data were Nehlig, P. (1989). "Etude d'un systeme hydrothermal ocea-
computed by A. Saintenoy and A. Replumaz. This is contri- nique fossile: l'Ophiolite de Semail (Oman)," These de
bution No. 535 from the Dynamique et Bilans de la Terre pro- Doctorat, Universite Brest.
 S, Dike Patterns in the Oman Ophiolite 95
Nicolas, A. (1986). A melt extraction model based on struc- Structural mapping in the Oman Ophiolites: Mantle diapir-
tural studies in mantle peridotites, J. Petrol. 27, 999-1022. ism along an oceanic ridge, Tectonophysics lSI, 27-56.
Nicolas, A. (1989), "Structures of Ophiolites and Dynamics Nicolas, A., and Jackson, M. (1982). High-temperature dikes
of Oceanic Lithosphere," Kluwer, Dordrecht, in peridotites: Origin by hydraulic fracturing, J. Petrol. 24,
Nicolas, A. (1990). Melt extraction from mantle peridotites: 188-206.
Hydrofracturing or porous flow consequence on oce- Rabinowicz, M., Ceuleneer, M., and Nicolas, A. (1987). Melt
anic ridge activity, in "Magma Transport and Storage" segregation and flow in mantle diapirs below spreading
(M. P. Ryan, ed.), pp. 160-174. Wiley, Chichester/Sussex, centers: Evidence from the Oman ophiolites, 1. Geophys.
England. Res. 92, 3475-3486.
Nicolas, A., and Boudier, E (1993). Mapping mantle diapirs Rabinowicz, M., Nicolas, A., and Vigneresse, J. L. (1984).
and oceanic crust segments in Oman ophiolites, submitted A rolling mill effect in asthenospheric beneath oceanic
for publication. spreading centers, Earth Planet. Sci. Lett. 92,3474-3486.
Nicolas, A., Boudier, E, and Montigny, R. (1987). Structure Ravaut, P. (1992). "Le leve gravirnetrique de la chaine om a-
of Zabargad Island: An early rifting of the Red Sea, J. naise: Contribution a l'etude des mecanismes de com-
Geophys. Res. 92, 461-474. pensation," Dipl6me d' Etudes Approfondies, Universite
Nicolas, A., Ceuleneer, G., Boudier, E, and Misseri, M. (1988). Montpellier 2.
This page intentionally left blank
Chapter 6 Neutral-Buoyancy Controlled Magma Transport and
Storage in Mid-ocean Ridge Magma Reservoirs and
Their Sheeted-Dike Complex: A Summary of Basic
Relationships
Michael P. Ryan

Overview Pb is the bulk (melt ± crystal) density. The generalized

zonation is:
This chapter reviews the neutral-buoyancy phenome- (I) Negative-buoyancy region. This region extends
non as it applies to the injection and high-level storage from the volcanic surface to "'" I km depth. The den-
of magma in the mid-ocean ridge environment. Petro- sity of nonvesiculated magma is greater than that of
physical data, seismic surveys, and analytic and nu- the country rock such that magma may descend under
merical inversions of geodetic data indicate that the pri- neg~tive-buoyancy forces. This low-density region is
mary mode of along-axis magma transport and storage produced by high amounts of fracture and grain-scale
within volcanic rift systems is regulated by the neutral porosity that also locally lower the elastic moduli and
buoyancy of picritic and tholeiitic melt, and melt and in situ elastic wave velocities. It is a gravitational "no-
crystal mixtures. The depths of the horizon of neutral man's-Iand" for magma. and to erupt, nonvesiculated
buoyancy (HNB) that controls the dynamics of magma magma must traverse the region aided by deeper re-
storage and lateral injection along the East Pacific Rise, plenishment. Theoretical dike shapes for descending
the Valu Fa Ridge-Lau Basin, Kilauea and Mauna Loa magma batches have bulbous (lower) noses and slender
volcanoes, Hawaii, and the Krafla Central volcano, (upper) tails. Lithologically, this region corresponds to
Iceland, are remarkably similar, although depths vary pillow basalts ± local sediments.
somewhat from system to system. (II) Neutral-buoyancy region. This region extends
The horizon of neutral buoyancy for MORB may be from ""'I to ""'3 km depth. The overall density of
subdivided into a picritic horizon of neutral buoyancy, magma is equal to that of the country rocks, and magma
HNBp, and a tholeiitic counterpart, HNB T' For the East is in local mechanical (gravitational) equilibrium with
Pacific Rise at gON, elastic crack stability relations sug- its surroundings. The differentiation of pieritic melt,
gest that ""'600 :s HNB T :s ""'1400 m, whereas mix- and the separation of tholeiitic melt from suspensions
tures of picritic melt and olivine crystals are in gravita- of melt + crystal mixtures will render magma parcels
tional equilibrium over > 1400 m :s HNBp:S ""'3000 m within this region positively buoyant. These tholeiitic
depth beneath the ridge axis. Differentiation processes parcels will ascend to form a tholeiitic layer at the top
therefore transform melt in equilibrium at HNBp into a of the magma chamber. Rupture of the chamber roof
new equilibrium setting at HNB T • Thus elastic crack will nucleate a dike-forming event within the dike com-
stability relations are consistent with: (i) the paucity of plex. Thus parcels of magma within the region of neu-
picritic eruptive products on the sea floor; and (ii) the tral buoyancy may flicker between states of neutral and
dominantly tholeiitic nature of the sheeted-dike com- positive buoyancy, depending on their extent of differ-
plex and ridge-crest eruptives. For the aggregate den- entiation and their load of suspended crystals. Within
sity range 2.6 :s Pb :s 2.82 g . em -3, and the composi- the upper portions of this region, the dynamics of lat-
tional range pieri tic melt + olivine crystals through eral magma injections that produce the mid-ocean ridge
tholeiitic basalts to ferrobasalts, the transition region sheeted-dike complexes mimic their counterparts in ac-
between negative buoyancy and neutral buoyancy has tive Icelandic and Hawaiian rift zones. Theoretical dike
an average depth of "'"I000 m beneath the East Pacific shapes along the HNB have slender tails and enlarged
Rise surface. midsections (across-axis profile): whereas the along-
The generalized buoyancy zonation of the East Pa- axis cross sections show advancing dikes with para-
cific Rise magma reservoir and its surroundings may be bolic noses. Local retardations of along-axis intrusions
considered in light of the complete compositional and force dike keels to descend while the dike tops rise to-
magma density range: 2.6 :s Pb :s 2.82 g . em -3, where ward the axial valley floor, driven by flow rates from

Magmatic Systems
Edited by M. P. Ryan 97
98 Michael P. Ryan

the ruptured chamberon the order of = 100-500 m '> s - I. Units

High flow rates and continued crack-front arrest pro-
Vs Shear (secondary) elastic wave kIn· S-I
mote the ascent of the dike top and subsequent eruption.
velocity
Breakthroughs along the path of lateral advance propel
the intrusion forward, bringing its top and bottom back w Dimensionless fluid-filled fracture dimensionless
half-width
to = 1-=3 km depth and gravitational equilibrium po-
sitions. The overall lateral intrusion process is, there- X Molar fraction of minerals being dimensionless
fore, the integrated contribution from each incremental removed from the melt
crack advance. An entire sheeted-dike complex is thus X Cartesian coordinate m
produced by successive neutrally buoyant lateral intru- y Cartesian coordinate m
sions. The lithology of the HNB is sheeted basaltic dikes
Z Cartesian coordinate m
and the upper levels of the isotropic gabbro complex.
(III) Positive-buoyancy region. This region extends
C Constant; index for flow regime dimensionless

from =3 to =75 km depth where melt densities are al- CI Constant in the melt + material dimensionless
ways less than the country rock. Melt migration begins property coefficient D
along a hierarchy of grain-scale microporous networks C, Constant in the dike height dimensionless
within the disaggregating parent lherzolite, supplying expression
vein swarms and deep dikes. Theoretical dike shapes c Constant in magma-fracture height dimensionless
for ascending melt batches have bulbous (upper) noses evolution
and slender (lower) tails. Lithologically, this region d Constant in magma-fracture height dimensionless
corresponds to garnet and spinel Iherzolites, grading evolution
further upward into dunite-, websterite-, and gabbro- e Constant in magma-fracture height dimensionless
impregnated harzburgite. evolution
g Gravitational acceleration constant m· S-2

h Height of a fluid-filled fracture m

Notation m Elastic modulus MPa. GPa
Units
Il Exponent dimensionless
D Melt + matrix material property dimensionless r, Mole fraction ratio dimensionless
coefficient Adjustable Cartesian coordinate m
HNB Horizon of neutral buoyancy m Time elapsed since fluid injection
HNB p Pieritic horizon of neutral buoyancy m
i, Time of residence of olivine crystals s
HNBT Tholeiitic horizon of neutral m in a MORB and olivine suspension
buoyancy
to Time, initial
K lc Critical (mode I) stress intensity MPay!iTI
V, Summed molar volume of compo- em'
factor
nents removed from melt
M Gram formula weight of melt g
Vj Molar volume of the jth component ern'
M, Summed gram formula weights of g
W Fluid-filled fracture half-width m
components removed from melt
MJ Gram formula weight of the jth g x, Mole fraction of the jth component dimensionless
component 0 porosity dimensionless
N Number of components dimensionless 00 Initial reference porosity dimensionless
P Fluid pressure MPa ~ The summation convention
Po Initial fluid pressure MPa e Density contrast parameter for the dimensionless
Q Flow rate of injected magma m's I horizon of neutral buoyancy
V Volume of injected magma m' e Conjugate density contrast dimensionless
parameter
V Partial molar volume of melt em'
Q' Time exponent dimensionless
V, Partial molar volume of components em'
removed from melt <; Dimensionless fluid-filled fracture dimensionless
Vj Molar volume of the jth component height
em'
Vp Compressional (primary) elastic krn- S-I TJ Melt shear viscosity Pa- s
wave velocity JL Shear modulus MPa, GPa
 6. Neutral-Buoyancy Controlled Magma Dynamics 99
Units logic variations and compressional wave velocity
/I Poisson's ratio dimensionless
relationships for ophiolite complexes are com-
g -cm ?
bined to provide a physical basis for the forma-
P Density
tion and sustained existence of mid-ocean ridge
P, Density of the (summed) compo- g. cm- 3
nents removed from melt
magma chambers.
g. cm- 3
This chapter reviews the buoyancy-zonation
Pr Density of the final evolved melt
structure for mid-ocean ridge magma reservoirs
Pg Density of the matrix grains g. em:"
and considers the role of fractional crystallization
P, Density, initial, of the melt g. cm:"
in altering melt buoyancy, followed by an over-
P, Density of the region below the g. cm- 3
view on magma dynamics in the sheeted-dike
HNB
complex, with comparisons of the analogous dy-
Po Density of the region above the HNB g. cm- 3
namics of Icelandic and Hawaiian rift zones.
Pw Density of sea water g. cm "!
Pis In situ country rock density g. cm- 3
Pm Melt density g. cm ? In Situ Density-Depth Relationships
!:J.p Density contrast g -crn ?
U Stress MPa In situ density-depth relations for the East Pacific
U3 Minimum principal compressive MPa Rise were estimated from an evaluation of seis-
stress mic profiles (Orcutt et al. 1976) and consideration
V' Del, nabla, the gradient operator m- I of the tabulated velocity-density relations for ba-
V'P. Vertical gradient in picritic magma MPa·km- ' saltic, diabasic, and gabbroic rocks (Christensen,
pressure (negative-buoyancy region) 1982). Figure 1 presents in situ density vs depth
V'P~ Vertical gradient in picritic magma MPa· km- ' relations based on the velocity-depth profiles of
pressure (neutral-buoyancy region) Orcutt et al. (1976) for refraction surveys on
V'Ft Vertical gradient in picritic magma MPa· km- I 2.9 X 106 and 5.0 X 106 yr crust. Density values
pressure (positive-buoyancy region) were assigned on the basis of the density spreads
V'P T Vertical gradient in tholeiitic MPa·km- ' shown by the acoustic results for H 20-saturated
magma pressure (negative- rock samples to 200 MPa confining pressure
buoyancy region)
(Christensen, 1982)-appropriate conditions for
V'P~ Vertical gradient in tholeiitic MPa·km- 1
the crustal reservoir depth. Both profiles show ap-
magma pressure (neutral-buoyancy
region) proximately the same type of density-depth be-
havior: a nonlinear increase in density over the 0-
V'FT Vertical gradient in tholeiitic MPa· km '
magma pressure (positive-buoyancy to 8-km-depth range. Ranges of inferred in situ
region) density at specific depth intervals correspond to
V'uH Vertical gradient in horizontal com- MPa. km I the variable scatter encountered in Vp-P plots of
ponent of confining pressure the tabulated data of Christensen (1982): the den-
sity range reflecting the total data range at a spe-
cific velocity value corresponding to each seismic
profile. Superposed in the figure is the band of den-
Introduction sity values for tholeiitic melt experimentally deter-
mined by Fujii and Kushiro (1977) with extension
Why do mid-ocean ridge magma chambers exist? to the total range inferred for picritic melt based
Why do relatively flat-topped sheeted-dike com- on the work of Stolper and Walker (1980), Sparks
plexes exist? What is it about newly created oce- et al. (1980), and Sparks and Huppert (1984). The
anic crust that enables it to capture and retain density bandwidth from 2.6 to 2.8 g . em? thus
magma at shallow depths-and thus makes pos- spans the composition rang~ tholeiite to picrite.
sible the long-term storage and differentiation of The in situ country rock trend and the melt band
basaltic melt? In this chapter, petrophysical prop- show a cross-cutting relationship with melt den-
erties of basaltic rock and melt, results from seis- sities less than rock densities beneath =3 km
mic surveys over mid-ocean ridges, and the litho- depth, melt densities greater than rock densities
 100 Michael P. Ryan

OrrT.,..,..,,..,...,r---,---.----.,r---,---,--,--;-;,....-;r-;-.,..-,-.,--,,,---,---.....,---.-----,-----,

1
2 HNB

T
4
·········~mfi~~=~
E : : : : :~: : : :
6 .:.:.:.:.~:.:.:.:.
I ........ ·u "t. tJ;~~#;~;;¢;:;<;;:;:;:;;:;t;J
~
0-
W
:.:.:.:.:E:':':':':'
0 :::::::::~:::::::::
6 ....... ·.0·.·.·. '.' 1LLLL..::.L..4~~~~~~~~~
: : : : :~: : : : :
EAST PACIFIC RISE
........
.........·C·········;:: .
9°N .:.:.:.:.<:C:.:.:.:.:
: : : : :~: : : : :
02.9 x 10 6 Yr CRUST
8 :::::::::~:::::::::
822l5.0 x 10 6 Yr CRUST :.:.:.:.:0:.:.:.:.:.
......... .
.... . ... .....
.........
....
.........
....
.... ...
....
.........
•••••

... .... 0, • • • •

.........
10 L-_--'--_ _. l - _ - - - - l_ _- ' -_ _ 1-._--'-~~-"-'-~-'-" _ _-'--_ __'___--'-_ _...L-_--'
2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3
IN SITU DENSITY (g·cm- 3)

Figure 1 In situ density-depth relations for the East Pacific Rise and for tholeiitic and picritic melt. Shaded and hatchured
boxes correspond to density ranges based on the seismic profiles of Orcutt et al. (1976) and density-velocity distributions for
mafic and ultramafic rock types. Between = I and =3 km depth, a crossover occurs between crustal and melt densities. This
crossover defines the HNB. The density spreads correspond to the ranges in density for laboratory-derived acoustic measure-
ments as compiled by Christensen (1982). The depth ranges for each box correspond to the depth steps for specific inferred
seismic velocities in the refraction profiling of Orcutt et al. (1976).

above = 1 km depth, and a rough equity in melt Hawaii and Iceland are based on combinations of
and country rock densities in the depth interval seismic and gravity surveys, and the source refer-
=1 to =3 km. Ryan (1993) has discussed the ences are given in the figure caption and are dis-
probable roles of H 20 and suspended olivine phe- cussed in Ryan (1987b). All country rock densi-
nocrysts in altering magma densities. ties were compared with the pressure-corrected
In Fig. 1, the horizon of neutral buoyancy tholeiitic melt and picritic melt bands as above.
(HNB)-by definition-coincides with the =1- All three sections show: (i) a characteristic non-
to =3-km-depth interval where in situ melt densi- linear increase in in situ host rock density with
ties are just balanced by the country rock density. depth and (ii) a similar buoyancy zonation with
Under conditions of local density balances be- depth.
tween the host rock and magma, the HNB thus rep-
resents a mechanical equilibirium position: the net
(resultant) integrated forces for inducing magma Definitions
ascent and descent are in balance, and long-term
stability is attained. Magma reservoir. For the sub-ridge oceanic crust,
Figure 2 compares the in situ melt-rock den- a magma reservoir as used in this chapter refers to
sity crossover regions in the Hawaiian, mid-ocean that domain of fluid-filled matrix that is capable
ridge, and Icelandic sections. Density ranges for of storing and transmitting magma. This includes
 6, Neutral-Buoyancy Controlled Magma Dynamics 101

IN SITU DENSITY (gocm- 3)

02022242628303234 2.2 2.4 2,6 2.8 3,0 3.2 3.4 2.0 2.2 24 26 28 30 3 2 34
I I I I I I I -: I: IIITT
o
f~~ II It : ~:

2
f-

f-
l~
1\'
"1j\
-

-
-\'N
f- .:
-

- 2

3 f- ;..,: - f- - 3

4 f- - f- - 4

E
C. 5 f- - f- - 5
w
o
<l:
U.
a: 6 f- I.: - f- - 6
::>
en
Q
z
<l:
',"'"
, '-'\'
,'lU '
o
...J
7 f- - -:~, f- - 7
o 'u
> :'t:':
I -:~,
I-
<l:
8 f- - , -.I, f- - 8
w '0-
Z
w : ~:
CD ,'Q',
I - ,'<:' , - 9
I-
0..
9 f-
-x
',u'
f-
"'"-
W
o ..i=:..
>i?:
10 - - - 10
" '~' f-

11 f- - f- - 11
' \ -: -:
HAWAII EAST PACIFIC RISE
12 f - MID-SECTIONS - 9°N l-
ICELAND
I - 12
... KILAUEA-I 02.9 xl0 6 Yr CRUST
• KILAUEA-II
[ZJ 5.0 xl~6,Yr CRUST • MID-SECTIONI
13 f- • MAUNA LOA
\
- f- - 13

14
I I I I I -:1:- I I I I I I I I I :,' I I I I
14
Figure 2 In situ density-depth profiles for the East Pacific Rise, in relation to analogous profiles for active Hawaiian and
Icelandic rift zones. Collectively, the crossover between the country rock density and that of picritic and olivine tholeiitic melt
occurs within the = 1- to =7-km-depth interval. Within an individual system, the crossover defines the horizon of neutral buoy-
ancy and shows some variance between these three systems. In situ density values for Hawaii are from the gravity inversions
and seismic surveys of Zucca et af. (1982) consistent with the seismic refraction surveys of Hill (1969). In situ density values
for Iceland are from Palmason (1971) and Palmason and Saemundsson (1974).
102 Michael P. Ryan

single- and rnuitiple-connected microscopic and with depth in the Earth's crust. Concomitant in-
macroscopic magma-filled cavities. The reservoir creases in density, seismic velocities (V p , V s) , the
may be logically subdivided into a shallow magma bulk modulus (K), and decreases in fluid perme-
chamber and a deeper more extensive crystal-liq- ability (Kr) and compressibility ([3) accompany
uid mush region (Sinton and Detrick, 1992). See the reduction in porosity with increases in depth
Fig. 8. and confining pressure. States of anelastic volume
Magma chamber. The magma chamber is the decrease under applied loading. This process may
top of the magma reservoir. It is relatively fluid be contrasted with dilatancy-an anelastic vol-
rich and has a sill-like overall geometry with a ume increase under differential stress. In regions
ceiling at the base of the sheeted-dike complex. of active basaltic magmatism, rock contractancy
The floor is relatively diffuse and becomes more and matrix compaction combine with zeolite and
crystal rich with depth, grading downward into greenschist facies hydrothermal mineralization to
the top of the crystal-melt mush that makes up produce the crossover between melt and country
the volumetrically more significant portion of the rock density.
reservoir. In the chapter by Phipps Morgan et al.
(this volume) it is termed the "magma lens." See
Fig. 8. The Magma Reservoir Environment
Neutral buoyancy. Pm = Pis, where Pm is the and the Region of Neutral Buoyancy
melt density and Pis is the in situ country rock
density. Local contributions to driving forces that. What is the correspondence between regions of
tend to induce ascending or descending motion in low elastic wave velocity in mid-ocean ridges and
the melt have been removed, and a state of local the neutral buoyancy region for tholeiitic melt?
mechanical equilibrium exists between the molten Seismic surveys conducted across and along the
region and the subsolidus surroundings. ridges provide a means of answering this question.
Horizon of neutral buoyancy (=Ievel of neu- The locations of the seismically defined magma
tral buoyancy (LNB)). The HNB is a layer ofnar- reservoir and the region of neutral buoyancy can
row vertical extent and wide lateral extent within be compared on the basis of the region of the
which melt has achieved mechanical equilibrium. crossover in the in situ density-depth profiles for
The HNB corresponds to the depth extent of shal- rock of the oceanic crust and that of picritic-to-
low subcaldera magma stabilization and long- tholeiitic melt. Summary results from several rep-
term shallow magma accumulation, and the level resentative surveys are discussed as follows in
of lateral dike emplacement in active volcanic light of their relationships to the principal buoy-
systems. ancy zonation of the ridge.
Negative buoyancy. Pm > Pi" and gravitational In "Webster's Third New International Dictio-
potential energy is released through the descent nary of the English Language Unabridged" (Gove,
of magma. The region of negative buoyancy lies 1964), the term "region" is defined as "one of the
above the horizon of neutral buoyancy in the major subdivisions into which the body or one of
Earth's crust, and melt parcels may therefore de- its parts may logically be divided." In this spirit,
scend to the HNB throughout this region. The this paper defines three regions of buoyancy for
Earth's free surface is the top of the region of magma: the region of negative buoyancy (sea floor
negative buoyancy. to = 1 km depth); the region of neutral buoyancy
Positive buoyancy. Pm < Pis, and the gravita- (= I to =3 km depth); and the region of positive
tional potential energy of the system is reduced by buoyancy (=3 km to the depth of melt separation
the ascent of magma. The region of positive buoy- in the parent lherzolite). As used here, the term
ancy is bounded below by the site of magma gen- "region" is finite or reasonably constricted in its
eration and bounded above by the horizon of neu- vertical or lateral extent. Thus, the region of neu-
tral buoyancy. tral buoyancy coincides with the upper levels of
Contractancy. The progressive reduction in the magma storage reservoir.
macroscopic and microscopic pore space produced The term "horizon" is defined as (Gove, 1964)
by a progressive increase in confining pressure "a stratigraphic level or position in the geologic
 6. Neutral-Buoyancy Controlled Magma Dynamics 103
column; a natural ... layer. Any of the reasonably region of neutral buoyancy that correspond to the
distinct layers . . . in a vertical section or pro- interval inferred in Fig. 1 have been overlayed.
file and gradually developed as a result of Multichannel seismic reflection profiles across
natural processes." In this chapter, the term the East Pacific Rise north of the Siqueiros Frac-
"horizon of neutral buoyancy" thus refers to an ture Zone have revealed a compressional wave
interval whose vertical extent is finite but very low-velocity region consistent with magma stor-
small compared to its great lateral extent. The age at a depth of =2 km beneath the sea floor
HNB's great lateral extent guides the relatively (Herron et al., 1978). Compressional wave ve-
rapid lateral magma intrusions that are individual locities rise to values in excess of 6.5 km . s- 1
dike-forming events. Thus the horizon of neutral just above the reflector and then abruptly drop to
buoyancy is important in the dynamics of the lat- =4.5 km . s-) within the 2 to 3-km-depth interval
eral intrusion process and contains within it the inferred to be magma rich.
active region of neutral buoyancy that coincides Hale et al. (1982) have reexamined Lamont-
with sub-ridge magma storage. This usage is con- Doherty Earth Observatory multichannel reflec-
sistent with terminology used previously by Ryan tion data taken across the East Pacific Rise at
(1985, 1987a, 1987b). lat. 9°N in light of elastic wave velocity values
The correspondence between the region (hori- from ophiolite samples. Laboratory-based com-
zon) of neutral buoyancy and the preferred depth pressional wave velocity data from samples of
of magma residence and lateral injection is shown the Sumail ophiolite (Christensen and Smewing,
in Figs. 3A and 3B. Depth ranges for sub-ridge 1981) have been combined with the temperature
magma storage from the East Pacific Rise, and the distribution expected for a double spreading rate
Valu Fa Ridge, Lau Basin, have been deliberately of 12.2 cm· yr " ' (Sleep, 1975) to help constrain
pooled with ranges for Kilauea and Mauna Loa the velocity profile and the cross-sectional ve-
volcanoes, Hawaii, and the Krafla Central vol- locity structure of the rise axis. Figure 6 provides
cano, Iceland, to provide one perspective on the the Vp-depth profile of Hale et al. (1982) and
overall similarities of these depth ranges. When compares it with the profile determined by Orcutt
considered in light of the density crossover rela- et al. (1976). Both profiles show a pronounced
tionships above, the data illustrate that the pre- low-velocity region over the 1- to 3-km-depth
ferred depth for magma storage and lateral in- core of the region of neutral buoyancy. The up-
trusion is in a virtual J : J correspondence with per portion of the high-velocity-gradient region of
the horizon of neutral buoyancy. Figure 3B sub- the magma reservoir roof (0-2 km depth) corre-
divides the ranges on a regional basis, illustrat- sponds in part to heavily fractured and porous pil-
ing the differences in magma depth between and low basalts and brecciated dikes, where the in situ
within these centers. density of these porous rocks is substantially less
Seismic refraction profiles have been conducted than-and subsequently approaches with increas-
along the East Pacific Rise crest near lat. 9°N by ing depth-that of nonvesiculated melt. Accord-
Orcutt et al. (1976) (Fig. 4). Along the crest, com- ingly, magma within this upper veneer (0-= 1 km)
pressional wave velocities increase from low val- may descend under negative buoyancy forces. Be-
ues (=2.5 - 3.5 km . s - I) near the sea floor, to about low =3 km depth, Pm < Pis and melt parcels rise
6.7 km . s - I at the top of a low-velocity region at within the magma reservoir driven by positive
2 km depth beneath the ridge surface (Fig. 5). Be- buoyancy forces. Figure 7 illustrates the across-
neath this compressional wave low-velocity zone, axis velocity model of Hale et al. (1982) and its
Vp increases again to values of about 7.5 km s- I at relation to the principal buoyancy zonations for
7 km depth. Profiles that have resulted from sur- tholeiitic melt.
veys parallel to the rise crest but on 2.9 X 106 and Seismic refraction profiling on the crest of the
5.0 X 106 year old crust do not show evidence of East Pacific Rise at lat. 21°N by Reid et al. (1977)
a low-velocity zone, but instead show a continu- also has revealed a region of high shear wave at-
ous and progressive increase in Vpthrough the base tenuation at a depth of 2.5 km beneath the sea
of the oceanic crust (Fig. 5). In Fig. 5, the re- floor. This region is in essential coincidence with
gions of negative and positive buoyancy and the the low compressional wave velocity region deter-
 104 Michael P. Ryan
o,-----------------------:;:-----~-_____,,:__---~ 0
T,
T
I,'T
I
I
I,
o I,
,I
T
~
I
I
I,
I

f
2 2

!
on
lEw

3 3
E , CD
:!!.
I-
a: , I
2
<J)
I
I I I
I

ii
z
~
I-
4 I
I
I I'
I
ii
I ,
1
4
..J
« I I'I II II
a: I I I I
w
o I Ii II I
S :;: I

i1
I
, I
I
o
z
«
w
5 N
:I:
1 1~
CD:;: i~
1

I
5
<.:J :;: I
« I
a:
~
. I, CD
lE
II

« "'
:I:
... 1 w I,
:::!: !o,
.L 6
~
:::!:
6 I
I
N
N
:I:
u,
o I
I
:J:
Ii:
w
I
1
o
7 I, 7

o
N
I
:I:

8 o EAST PACIFIC RISE 8

• VALU FA RIDGE
o ICELAND
• HAWAII

9 9

10L....- A -J 10

Figure 3A Pooled mean depth ranges for magma storage and transport beneath the rift-zone surface for the East Pacific Rise,
the Valu Fa Ridge, the Krafla central volcano, Iceland, and Kilauea volcano, Hawaii. The pooling process illustrates the substan-
tial similarities in the nature of the depth distributions. The collective core of the region of neutral buoyancy, as expressed by the
mean depths, lies predominantly within the 1- to 4-km-depth interval. High-level, mid-ocean ridge magma storage and intrusion
within the sheeted dike complex is in a I : I correspondence with the region of neutral buoyancy. (Data sources are provided in
the appendix.)
 6. Neutral-Buoyancy Controlled Magma Dynamics 105
HAWAII MID·OCEAN RIFT SYSTEM ICELAND

•
10
E
~
f-
a:
o
g, 2 2
Z
<t
a:
f-
-'
<t
ffi 3 3
~
o
Z
<t
~ 4 4
<t
a:
~
(/) 10
<t
~ 5 2 5
-c
::!;
u, .L
o
:I:
17A
h:w 6

I
6
o I
I

I KILAUEA I I• EAST PACIFIC KRAFLA :

l VOLCANO. I RISE , CENTRAL
7 2'I0 Io VALU FA RIDGE 12
VOLCANO: 7
LAU BASIN
I
B
Figure 3B Regional patterns in depths of magma reservoirs and lateral dike-forming magma injections. Note that the total
range in depths of mid-ocean ridge reservoirs is less than the vertical variation for Kilauea volcano alone. Symbols are the same
as in Figure 3A. "E" signifies the initial and final gravitational equilibrium position of magma after a series of wave-like
oscillations about the horizon of neutral buoyancy, induced by the periodic halting of the advancing magma fracture front. The
vertical dashed lines above the "E" position are thus transient nonequilibrium excursions during lateral intrusion and are damped
down upon the completion of intrusion. Note that the final resting place is always at the HNB (equilibrium) position after an
intrusion transient. For Kilauea, the HNB is centered at 3 km depth.

mined by Orcutt et al. (1976) at 9°N and also cor- Structure of the Ridge Magma Reservoir in
responds to the region of neutral buoyancy in Relation to the Region of Neutral Buoyancy
Fig. I.
Multichannel seismic reflection surveys across Detrick et al. (1987) have conducted multichan-
the Valu Fa Ridge (Lau Basin) by Morton and nel seismic surveys along the East Pacific Rise be-
Sleep (1985) have revealed a reflector at a depth tween lat. 8°50'N and lat. 13°30'N. The study was
of 3.5 km beneath the ridge. This reflector has designed to resolve questions related to along-axis
been interpreted as the relatively flat-lying roof of magma continuity as well as to constrain the
a magma chamber. The roof width is about 2 to depth-width variations of the axial magma res-
3 km. The Valu Fa Ridge is a back-arc spreading ervoir. The reservoir width (the chamber por-
center with a spreading rate of about 70 mm . yr -1 tion) was resolved at its top as a relatively high-
(Weissel, 1977). Like the East Pacific Rise, the amplitude reflection, with a maximum of 2 to
Valu Fa Ridge magma chamber coincides with the 3 km. The reflector is relatively flat-lying, further
(slightly deeper) position of neutral buoyancy for suggesting a roof structure that dips gently out-
picritic-to-tholeiitic melt (Fig. 3A, VF1 at left- ward from the ridge axis. Reflections from the
center of plot). Moho, however, extend to 2 to 3 km on either side
2l' _ ---=~-'----' ...L

Figure 4 Marine bathymetry of the East Pacific Rise. The horizon of neutral buoyancy for the EPR lies at the = 1- to =3-km-
depth interval beneath the ridge crest: it controls the dynamics of lateral dike-forming injections in the sheeted-dike complex
and modulates magma storage in the uppermost portion of the reservoir. The EPR surveys discussed in this chapter were not
plotted to avoid crowding the map and obscuring bathymetric contours. They are: (i) 9°:'-/ (Orcutt et al., 1976; Hale et al., 1982);
oN
(ii) 9°_W (Toomey et al .. 1990); (iii) south of the Clipperton Fracture Zone (Vera et al .. 1990); (iv) north of the Siqueiros
Fracture Zone (Herron et al., 1978): (v) 8°50'N-13°30'N (Detrick et al.. 1987); and (vi) 21ON (Reid et al .. 1977). Based on the
compilation of Mammerickx and Smith (1980).
 6. Neutral-Buoyancy Controlled Magma Dynamics 107

°1 •[ -..= --T

BUOYANCY
L--------

HORIZ N BUOYANCY

I
f-
a.
LU
o EAST PACIFIC RISE
9' N
I

l_- ---, I
L...<e------,.
i
- - - Ridge axis
-.-..-..- 2.9 x 106 Ma crust
I
I
- - - 5.0 x 106 Ma crust
I
I
'-I

REGION OF POSITIVE BUOYANCY

I
I
I

I
COMPRESSIONAL WAVE VELOCITY (krn-sJ)

Figure 5 Compressional wave velocity profiles for the East Pacific Rise atlat. 9'N in relation to the region of neutral buoyancy
for tholeiitic melt. For the zero-age rise axis profile, most of the low compressional wave velocity region is contained within the
region of neutral buoyancy. The dashed lines at I and 3 km depth denote the approximate limits of the region of neutral buoyancy.
Modified after Orcutt et al. (1976).

of the Rise axis, suggesting a magma reservoir devals appear to be superficial features with re-
base within a range 4 to 6 km wide. The depth to spect to the reservoir, and the reservoir roof passes
the top of the magma chamber is constrained by continuously beneath some 70% of them.
the two-way travel time between the ridge surface The East Pacific Rise just south of the Clipper-
and the chamber roof and is generally 1.2- 2.4 km ton Fracture Zone (Fig. 4) contains a bright upper
beneath the sea floor (see also Figs. 5 and 6). crustal reflector suggestive of the roof of a vol-
Minimum estimates of roof thickness tend to cor- ume with high magma-to-rock ratios (Fig. 8; Vera
relate with the shallowest positions of the ridge et al., 1990) and lies beneath a well-developed
axis beneath sea level. These positions tend to be axial graben (not illustrated). Three-dimensional
near the segment center and attest to the relatively seismic tomography by Toomey et al. (1990) has
high magma budgets (Macdonald et al., 1984) related the low compressional wave velocity core
that have constructed these topographic highs. of the along-axis magmatic region to the surficial
The lat. 8°50'N to lat. 13°30'N study area con- structures and to the surrounding distribution of
tains prominent overlapping spreading centers as relatively high- and low-velocity material within
well as several along-axis deviations from axial the newly created oceanic crust. Inversion of com-
linearity (devalsy. Apparent along-axis continuity pressional wave travel time residuals has pro-
of the magma chamber was observed for an aggre- duced a succession of sections through the com-
gate length of =350 km (or =61 %) of the 500 km pressional wave velocity structure in horizontal
of ridge crest surveyed, with laterally continuous depth slices from the sea floor to 3 km depth, as
reflections on 40- to 50-km-long traverses. Most well as in cross sections across and along the rise
 108 Michael P. Ryan

REGION OF NEGATIVE BUOYANCY

HORIZON OF

E
:OS
I
f0-
n,
UJ
o

REGION OF POSITIVE BUOYANCY

I EAST PACIFIC
go N
RIS~
------

COMPRESSIONAL WAVE VELOCITY (krn-s")

Figure 6 Compressional wave velocity profiles for the East Pacific Rise at lat. 9°N in relation to the region of neutral buoyancy
for tholeiitic melt. For both profiles, the low compressional wave velocity region approximates a I: I correspondence with the
region of neutral buoyancy. The square-stepped (heavy line) profile has been derived by Orcutt et al. (1976) whereas the
piecewise-linear (fine line) profile is the product of the multichannel reflection survey reexamined by Hale et al. (1982).

-------- --- ---

Figure 8 Compressional wave velocity contours for the East Pacific Rise at lat. 9°N in cross section, and their relation to the
region of neutral buoyancy, and the regions of negative and positive buoyancy. The lowest velocity core is totally contained
within the region of neutral buoyancy. The magma chamber corresponds to the lowest velocity upper portion of the region, where
the fluid/rock ratios are highest and the configuration is sill-like (Sinton and Detrick, 1992). The magma reservoir includes the
chamber, but extends downward into the high-velocity (Vp = 5.5-6.0 km . s -') cumulates that may be sufficiently porous and
permeable to permit the accumulation (and transmission) of melts. Magma in the lower half of the reservoir is always positively
buoyant, whereas magma in the upper half may be either neutrally buoyant or positively buoyant depending on the suspended
crystal load and the degree of differentiation. Diverging asthenospheric flow patterns (diverging arrows beneath 7 km) corre-
spond to mineral lineations in the harzbugite of the sub-Moho sections of ophiolite complexes. Vertical long and short rulings
denote the sheeted-dike complex. Based in part on contour distributions provided in Vera et al. (1990), and modified after Ryan
(1993).
 6. Neutral-Buoyancy Controlled Magma Dynamics 109
EAST

REGION
OF
POSITIVE
BUOYANCY

I !
4 2 0 4 10
DISTANCE FROM RIDGE AXIS (km)

Figure 7 Cross-sectional compressional wave velocity structure of the East Pacific Rise at lat. 9°N based on the multichannel
seismic survey of Hale et al. (1982) and employing the thermal model of Sleep (1975) with laboratory-determined \'p measure-
ments and the temperature derivatives of compressional wave velocities.

WEST EAST
o

COMPRESSIONAL WAVE
CONTOUR INTERVAL
_ _ _ (05~M_S-,-) _ _-,

J, REGION
-:.- OF-_. - _ - -
------
-POSlTllif'
--
r
6

- 7.5-
7I BUOYANCY
- - - ----
----~r(:-- -8.0----
• __ . 1 --'---
2 1 0 1 2 3 4 6 8 10

DISTANCE FROM RIDGE AXIS (KMI

Figure 8
110 Michael P. Ryan

:I: -
1-:2:
«:.<:
w-
ZC/.l
~~
:I:w
1-(:1
o.e
w_
co:

:I: -
1-:2:
«:.<:
w-
ZC/.l
2
~~
:I:w
1-(:1
3
o.e
w_
co:
4

50 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
DISTANCE ALONG RIDGE AXIS (KM)

Figure 9 East Pacific Rise compressional wave velocity structure in relation to the region of neutral buoyancy from lat. 9°N to
lOoN (darkest shading). Vp contour intervals are in units of 0.2 km . s -I, and contours begin at the 3 km . S·I minimum centered
in the region of neutral buoyancy. (A, Top) Cross-sectional structure showing the core region (darkest shading) of lowest velocity
and presumably melt enrichment centered along the axis; regions of reduced velocity beneath and bilaterally flanking the reser-
voir above, and the relatively high-velocity newly created oceanic crust. The vertical dashed line indicates the along-axis, cross-
section of Figure 9B. (B, Bottom) Longitudinal along-axis cross section. For both the across-axis and along-axis sections, the
low-velocity core region is in approximate correspondence with the region of neutral buoyancy. Vertical dashed line indicates
the across-axis slice location of Figure 9A. Figures modified after Toomey et al. (1990). (From Ryan, 1993; reproduced with
permission of the American Geophysical Union.)

axis. Figures 9A and 9B relate the low-velocity Me-Down Massif, Bay of Islands Complex, New-
core region of the ridge to the regions of neutral, foundland, as well as Troodos, Cyprus.
negative, and positive buoyancy,
Region of Negative Buoyancy

Lithologic Associations in Pillow basalts and sheeted basalt flows compose

Ophiolite Complexes the dominant lithology. Low in situ densities are
produced by pervasive macroscopic fractures and
The regions (horizons) of negative buoyancy and microscopic porosity. The resulting high fluid per-
neutral buoyancy and the region of positive buoy- meabilities (Nelig and Juteau, 1988) promote the
ancy each have a distinctive igneous association. hydrothermal penetration of this layer, with fluids
Studies of ophiolite complexes, combined with sporadically reaching into the horizon of neutral
the density zonations observed for active mid- buoyancy, below. Locally, brecciated dikes make
ocean ridges, reveal the relationships between rock up a portion of the lower part of the region of
type and buoyancy zonation. These relationships negative buoyancy and attest to sporadic hydro-
are illustrated in Figs. 10 and II for the Blow- thermal explosions. The contact with the top of
 6. Neutral-Buoyancy Controlled Magma Dynamics 111
SEDIMENTS
0
~"l
PILLOW

- - ---I
BASALT
------------
ANDESITIC
PILLOW
BASALTS
HORIZON OF E
E
c
w
INEUTRAL) I C
w
0
2
111111111!1111111111111111:111111
2

1~--
0 BUOYANCY «

-1
« IX
LL

~1I~r
LL 3
IX W
w HORNBLENOE I- 3 -- 3
I- GABBRO ~
~ I-
I- 4 Z NORITE
z W
41- - 4
w 2
2 0 TROCTOLITE
0 --- w

«
w
(/J
r:.
....J
5 GABBRO
NORITE ~ (/J
r:.
....J
«
(/J
51-
DUNITE
- 5
(/J
Vs -c
«
I
CD

I-
6~TRocrOLl~ Vp
CD
I
I-
«
6-
Vp
6
w HARZBURGITE
«
w 7 DUNITE Z
Z W
w CD' 7 r----I----,,-.l 7
CD I TROODOS
I I-
I- B a. CYPRUS
a. W
w HARZBURGITE 0
8 8
0
9
DUNITE
9 9
LHERZOLITE 23456789
10
ELASTIC WAVE
VELOCITY (krn-s")

11 Figure 11 The relationship between the lithologic succes-

2 3 4 5 6 7 B 9 sion for the Troodos, Cyprus, ophiolite and the reconstructed
ELASTIC WAVE depth range for the paleo-horizon of neutral buoyancy. (Com-
VELOCITY (krn-s") pressional and shear wave velocity profiles are based on
laboratory-derived acoustic measurements as summarized in
Figure 10 The relationship between the lithologic succes-
Christensen, 1978.)
sion for the Bay of Islands ophiolite complex and the recon-
structed depth ranges for the paleo-horizon of neutral buoy-
ancy. The paleo-horizon of negative buoyancy lies above the
sheeted-dike complex, within the pillow basalts. (Profiles are
zone. This would correspond to the transition in-
based on laboratory-derived acoustic velocities and have been terval between the region of negative buoyancy
summarized in Christensen, 1978.) (porous, fractured basalts) and the region of neu-
tral buoyancy (sheeted diabase dikes).

the sheeted dike swarm may be relatively sharp. Region of Neutral Buoyancy
Beneath the "geotimes" volcanics of the Oman
ophiolite (Lippard et al., 1986;' Nicolas, 1989), The sheeted diabase dike swarms and the upper-
for example, the transition region from the ba- most portions of the gabbro complex make up
salts to the top of the sheeted dikes is tens of the region of neutral buoyancy. It is at this level
meters to = 100 m thick. The upper portions of that rising melt begins to spread laterally, and the
both the vertically and laterally propagating dikes isotropic gabbros thus must have experienced sig-
thus would encounter the influences of the region nificant along-axis melt migration in response to
of negative buoyancy-inhibiting upward magma replenishment from below. Periodic reservoir rup-
movement-within a fairly restricted transition ture produces the sheeted-dike complex, and, by
 112 Michael P. Ryan

analogy with well-documented centers such as Table I

Krafla, Iceland, the dikes were formed dominantly Neutral Buoyancy Magma Transport Correlations: Intrusion
during lateral along-axis emplacement. The roots Dynamics, the III Situ Environment, and the Development of
Igneous Structure
of = l-m-thick fine-grained dolerite dikes have
been mapped in the Maydan syncline area of
• World-wide correlations of the depth of magma storage and
Oman. They may be traced downward and show intrusion in basaltic rift zone environments-indepen-
crosscutting relations with high-level gabbros, dent of local tectonic settings
where their chil1ed margins flare outward, attain- • Physical existence of basaltic magma chambers
ing widths of 15- 20 m. Progressing downward, • Long rift zone development
they grade into doleritic-textured microgabbro • Long (bladed) dike generation
and then more coarsely crystal1ine massive gab-
• Upward evolution of active magma reservoirs
bro (Rothery, 1983). They thus appear to represent
• Starting and stopping position (depths) of dike-forming
the magma chamber connection where subvertical magma intrusion
flow has drained the overpressured chamber com-
• Noneruptive nature of most shallow intrusions
partment beneath. In other areas, members of the
• Depth location of the sheeted dike swarm
base of the dike swarm do not root in the plutonic
• Episodic reequilibration to the HNB dike emplacement
complex, but crosscut the subsolidus gabbros be-
depth during incremental intrusion
neath and show true dike keels. This relationship
• Preferred horizon of magma mixing (intra- and inter-
should be .a common one and would be a natural reservoir)
consequence of neutral1y buoyant (and negatively
'. Location of subcaldera magma reservoirs
buoyant) lateral injection episodes from reser-
• Location of rift zone dike-induced microseismicity
voirs that are somewhat offset from each other in
• Geodetic signatures of dike intrusion and subcaldera
plan view. Intermittent and low-volume reservoirs
magma storage
maintained at low spreading rates would be ex-
• First-order fracture porosity and fracture networks in rift
pected to rupture at their up-rift and down-rift zones
margins (Ryan, Fig. 17, 1987b). At high spreading
• Eruption-drainback phenomenology: dynamics and
rate ridges such as the East Pacific Rise, the volumes
sheeted-dike complex is underlain nearly continu- • III situ density-depth profiles
ously by stored magma, and the rupture is expected
• III situ Vp , Yo-depth profiles
to occur in the chamber roof itself and directly over
• Experimental crack closure data at high pressures
the central1y located mantle replenishment source
• Experimental Vp V,data as a function of confining pressure
(Whitehead et al., 1984; Ryan, 1987b). In such
• Depth location of fluid pressure center (maximum magma
cases, the dike attempts to grow upward into the re-
driving pressure) in geodetic inversions
gion of negative buoyancy. but negative buoyancy
forces require it to "roll over," and-in cocks-
comb fashion-to propagate laterally along the
magma chamber roof axis. Examples of the tran- of crystal-melt phase equilibria (O'Hara, 1968a,
sition from vertical injection through the chamber 1968b), and of the crystal1ization products of ophi-
roof with subsequent rol1 over to a dominantly lat- olite complexes, especial1y the ultramafic rocks
eral injection mode have been wel1 documented at at and beneath the crust-mantle interface (e.g.,
Krafla, Iceland, during the September 8, 1977, Nicolas, 1989). How does the crystallization of a
multiple intrusion-eruption episode (Brandsdottir picritic melt influence the density of the evolving
and Einarsson, 1979; Ryan, 19S7b). Table 1 lists a liquid, and how would changes in melt density po-
number of magma transport and physical property tentially interact with the ambient density struc-
correlations within a neutral buoyancy context. ture of an active magma chamber?
Figure 12 il1ustrates the changes in melt den-
sity produced in picritic and tholeiitic melts as a
Differentiation and Melt Density function of olivine, plagioclase, clinopyroxene,
and ilmenite crystal1ization. Picritic liquids (tri-
Mid-ocean ridge tholeiites are the differentiation angles) are based on both rock sample analyses
products of picritic liquids, as suggested by studies (Clarke, 1970; Elthon, 1979) and modeled com-
 6. Neutral-Buoyancy Controlled Magma Dynamics 113
Fe/(Fe + Mg)mol
0.2 0.4 0.6 0.8
I

IMID-OCEANI
RIDGE!

PLAGIOCLASE1 ,CLINOPYROXENE N

OLiVINE
l
ILMENITE L
j-I
XjMj
M
2.8 Pi N
(1)
V
L Xj~
j j=1

2.7 where X, is the mole fraction, M, is the gram for-

mula weight, and Vj is the molar volume of the jth
component, and if and V are the gram formula
weight and the molar volume of the melt, respec-
EXPLANATION tively. Fractional crystallization processes may
@Walkeret al. (1979)
remove minerals composed of components in a
• Bender et al. (1978) mole fraction ratio If. where 2: 'i = I. Sparks and
• Residual glasses, Huppert (1984) defined the density of the com-
Shibata et al. (1979) ponents in the fluid that are selectively removed
... Picritic basaIts
through the fractional crystallization process as

"'V rM j _ M
__
LJ_._,_" ~
0.4 0.6 0.8 1.0 Pc - -, (2)
Fe/(Fe + Mg)mol "'V rV
L.J / 1 Vc

Figure 12 (Lower) Melt-density changes produced by frac- where M; and V, are the summed gram formula
tional crystallization of picritic and tholeiitic liquids. as illus- weights and molar volumes, respectively.
trated by changes in Fe-Mg content. Points of appearance of
For a molar fraction X of minerals that are
liquidus mineral phases are indicated by arrows. (Upper) The
preponderance of mid-ocean ridge basalts is associated with crystallizing and being removed from the original
the density minimum produced by olivine fractionation. Based melt, the density of the final evolved melt, p., is
on a presentation of Stolper and Walker (1980).
L (X j - rjX)M j
Pr = (3)
L (X j - -xiv.'
positions (Irvine, 1977). Liquidus temperatures
which may be reformatted as
were used to infer density values following the ap-
proach of Roeder (1974) and the empirical partial Pi[ I - «Pc V,J/(Pi Vi))Xj
molar volume-density relations of Bottinga and Pr = [I - (VjV;)Xj
(4)
Weill (1970). For the fractional crystallization-
density relations plotted in Fig. 12, the changes in Figure 13A illustrates the variations in indi-
(Fe/(Fe + Mg))mo, was used as an index of differ- vidual fractionation densitie~ for the components
entiation. That portion of the liquid line of descent forming the olivine, clinopyroxene, orthopyrox-
spanning the interval Fe/[Fe + Mg] (molar) from ene, and plagioclase solid-solution series. Note
0.3 to 0.8 is from the l-atm melting experiments that p; is a fictive parameter and relates to
of Bender et al. (1978) and Walker et al. (1979). the chemical components of an individual soIid-
At the most fractionated portion of the curve, solution series and not to the actual minerals. For
 I 14 Michael P. Ryan
FRACTIONATION AND MELT DENSITY (g'cm'3)
2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4
A 100

90
--i
:r:
0
r-
~
<t:
=1
0"- 80 ?i
0 OJ
E :t>
(f)
:t>
r
0 1

70>-

1 1
B 6·U8'-=== -- ~ ~

l.ower Upper
Reservoir Magma reservoir j

crust mantle
~ roOf~IO~
~~ ~ICELAND

~-----el-- • f-1MID-OCEAN RIDGE

~
I
. f--o-1HAWAII
I
I
.1 _~ 1 ~- -~-- -:--=--:c--=-
2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4
COUNTRY ROCK DENSITY (g'Crl' 3)

Figure 13 (A) Fractionation densities as a function of depletion of mol% A. where A represents the high-temperature
component in each mineral solid-solution series. Also shown is the reference range for tholeiitic melt density. The vertical
dashed line represents the density limit for picritic melts. (B) III situ density ranges for magma reservoir roof blocks, lateral
magma reservoir country rocks, and the subreservoir oceanic crust and upper mantle for Iceland, the mid-ocean ridges, and
Hawaii. As in (A), the density range that includes picritic melts extends to the dashed line. Modified after Sparks and Huppert
(1984).

those components forming olivine, clinopyrox- bulk densities at least to P = 2.82 g . em - 3 are
ene, and orthopyroxene, fractional removal re- expectations for the flowing suspension.
duces the remaining melt density, whereas the re- It is important to observe that the 2.723 g. cm- 3
moval of plagioclase components increases melt density for the melt phase, Pm' reference point
density. These results are in accord with Fig. 12 should be regarded as a minimum value, which has
and the results of Stolper and Walker (1980) and the effect of making the 2.82 g . em -3 value for the
Sparks et al. (1980). (See also related work in melt and 20% olivine crystal suspension a conser-
Koyaguchi, 1990, and in Huppert and Sparks, vative density estimate for the mixture. The picritic
1980. melt densities of Fig, 12approach 2.8 g- em 3,sug-
Forced and free convection within the upper gesting that the bulk density for a new estimate of
fluid-rich chamber will entrain olivine phenocrysts the melt and olivine suspension at 10 vol% would
within a flowing suspension. Thus the bulk (melt + approach o, == 2.847 g. cm " whereas at 20% crys-
crystal) density, Pb' of a picritic melt and olivine tals, the revised estimate is Pb = 2.897 g . ern 3.
mixture has a natural place in neutral buoyancy as- The long-term maintenance of a suspension of
sessments. Incremental additions of olivine (F0 9o ) picritic melt and olivine crystals depends on the
to a picritic melt may be expected to raise the strength of the buoyant lift provided by rising
crystal-free density (Pm=2.723 g' cm ":') through fluid in natural and/or forced convection. This lift
the range Ph = 2.746,2.770,2.795, and 2.820 g . occurs through a balance between the positive
cm-3, for additions of 5, 10, IS, and 20% olivine buoyancy of the melt and the Stokes settling of
(Huppert and Sparks, 1980; Ryan, 1993). Thus olivine (Stommel, 1949; Marsh and Maxey, 1985).
 6. Neutral-Buoyancy Controlled Magma Dynamics 115
The residence times (t r ) expected for olivine sus- 1200 m depth. Hooft and Detrick (1993) next as-
pensions in a MORB melt matrix cover roughly sumed that the contents of the EPR magma reser-
the range = 10 2 :s t, :s = 10 yr (Martin and voir may be characterized by a density of 2.70 ±
Nokes, 1988), and increases in crystal growth and 0.02 g . em -" and in a comparison with their pro-
diminished temperature contrasts within the cham- files outlined previously, then concluded that the
ber (that drive convection) will thus enhance the neutral buoyancy level lies at a depth 100-400 m
likelihood of olivine sedimentation (e.g., Sparks below the rise axis. This unfortunate assumption
et al., 1993). Within the chamber. the states of has implicitly confined attention to only minimum
magma buoyancy are expected to alternate be- density tholeiitic melts. The complete composi-
tween neutral buoyancy and positive buoyancy as tional range for MORB includes ferrobasalts as
the mixture(s) dynamically evolve. Thus rounds well as picrites, however. In addition, it is impor-
of sedimentation and differentiation may yield pro- tant to consider potential contribtions from sus-
gressive changes from neutral to positive buoy- pended crystal phases (e.g., olivine), which have
ancy, and those magma parcels affected are ex- been shown (Huppert and Sparks, 1980; Ryan,
pected to rise progressively toward the chamber 1993) to impact significantly the bulk or aggre-
roof in response. Continued differentiation to tho- gate density of the flowing melt and crystal mix-
leiitic "minimum-density" melts may be expected ture. Considering now the complete composi-
to correlate with magma chamber roof fracture tional and density range appropriate for EPR
events and injection in the sheeted-dike complex. magmas, we can return to the curves of Hoof! and
Density reductions due to H7.0 additions must Detrick (1993) and see that their density estimate
be understood within the perspective of the rela- of the Vera et al. (1990) profile (their Fig. 4a,
tively low H 20 contents found for MORE. The p. 425) is compatible with that of a picritic melt
infrared spectroscopy of Dixon et al. (1988) sug- and olivine crystal mixture through the 1000- to
gest dissolved H 20 contents over the range 0.07- 1200-m-depth interval. Indeed, it is also compat-
0.48 wt% (Endeavor segment of the Juan de Fuca ible with a tholeiitic melt and olivine crystal mix-
Ridge), whereas the Cobb offset lavas yielded ture at depths in excess of 1000 m.
0.15 -0.36 wt%. Density reductions induced by The inferred interactions between evolving
adding 1.08 wt% H 20 to anhydrous basalt (I. Ku- melt densities with fractional crystallization and
shiro, personal communication, 1986) produce a in situ rock density must be consistent with the
t::.p = 0.05 g . cm '. Thus the shaded melt density observed lithologic succession in ophiolite com-
bands of Figs. I and 2 contain all conceivable per- plexes as well as the seismic velocity structure of
turbations due to water in mid-ocean ridge basalts. the active mid-ocean ridges. Figure 14 schemati-
Hoof! and Detrick (1993) estimated the rock cally illustrates the crosscutting relationship be-
density-depth profile for the 0- to 1400-m-depth tween the density band for picritic-to-tholeiitic
interval of the EPR on the basis of three ap- melts and the country rock density profile(s).
proaches. Their initial approach made use of pub- Three rock profiles were drawn and range from
lished porosity-depth profiles (Berge et al.. 1992) high to low values of dp/dz. Note that while the
based on Christeson et al. (1992) and Vera et al. range in melt densities is constrained to be be-
(1990) seismic velocity profiles. This initial ap- tween 2.6 and 2.8 g. em -3, the surroundings of the
proach used the relation p(Z) = Pg - 0(Z)(p g - magma reservoir may have a high or low gradient
Pw), with an assumed grain density of Pg = 2.95 g in in situ density depending on factors such as the
. em -3. The resulting high-gradient profiles attain maturity of the crust (thickness of the low-density
inferred densities of 2.75 g . cm 3 over 200- basalt units), spreading rate(s), and associated ther-
500 m depth. Next, employing an exponential re- mal structures. For "high-gradient" crust (i.e., dp/
lationship between depth and porosity, 0(Z) = dz large), the density profile cuts the melt band at
0 0 e- A'. yielded the estimate P = 2.75 g . em - 3 by relatively shallow depth (Z,) and has a correspond-
about 500 m depth. Finally, application of the em- ingly thin horizon of neutral buoyancy HNB,
pirical velocity-density relation of Christensen (Fig. 14). Thus the equilibrium magma chamber
and Shaw (1970), P = 1.85 + 0.165 Vp • to the pro- would be expected to be relatively shallow and
file of Vera et al. (1990) yields P == 2.80 g . em 3 of small floor-to-ceiling height. Conversely, for
by 1000 m depth, and p == 2.85 g . cm by r ' "low-gradient" crust (i.e., dp/dz small), the pro-
116 Michael P. Ryan

DENSITY(P)

~
Z, sz,

T ~
Z2 OZ2

T
::I:
l-
c, T
oZ3 HNB 3
w Z3
0

THOLEIITIC
AND
PICRITIC
MELTS

Figure 14 Density-depth relations between mafic and ultramafic melts (shaded) and country rock (solid curves) for three
schematic host rock profiles. For each profile, cross-cutting relationships between melt (p-Z) and host rock (p-Z) define an HNB
whose thickness and depth depend sensitively on the gradient of in situ density with depth. For example, 8Z" and HNB"
correspond to the thickness and the position of the horizon of-neutral buoyancy for profile I. Arrows along a profile connect the
picritic values (right dot) with the tholeiitic values (left dot) in schematic fractional crystallization paths. While these relations
are general, applications to the oceanic crust suggest that the magma density band lies over the 2.6-2.8 g . ern , interval.

file cuts the melt band at greater depth (2 3 ) and density of the remaining liquid; thus interactions
thus has a thicker horizon of neutral buoyancy should be expected between melt and the subsoli-
HNB 3 (Fig. 14), tending to promote a more verti- dus surroundings. Figure 15 is a synoptic density-
cally extensive magma chamber. difference map for the depth evolution of equilib-
As discussed earlier, the fractional crystalliza- rium melt parcels during fractional crystallization.
tion of a picritic melt continuously changes the The equilibrium position of the picritic melt is
 6. Neutral-Buoyancy Controlled Magma Dynamics 117

(PENVIRONMENT - PMELT)

(-) o (+)

Z - - - - - - - - - HNB
T

FRACTIONAL
CRYSTALLIZATION
PATH

I
I-
2

I
Q..

W
o

Z+/iZ - - - - - - - - - HNB p

(1)

(2)

(4)

(-) (+)
Figure IS Density difference map showing schematic densitycontrasts between the environment and newly fractionated melt
batches during fractional crystallization from picritic (P) to tholeiitic (T) compositions, Radiating lines of potential density
contrast results from a spectrum of potential ill situ host rock density contrasts above and below the picritic and tholeiitic horizons
of neutral buoyancy. The map centerline is defined by (PENVIRO",IE" - PMELT) ~ 0 and is the line of neutral buoyancy. Melts on
the right half are positively buoyant, whereas those on the left are negatively buoyant.

point P and rests at the picritic horizon of neutral tion at P. These lines are, in tum, generated by the
buoyancy, A family of potential density contrast potentially high (line I) or low (line 4) values of
lines passes through point P and describes the dpidz, of the country rock as discussed above.
relative magnitudes of positive or negative buoy- As fractional crystallization proceeds, the equilib-
ancy for picritc melt parcels if suddenly displaced rium pathway from picrite to tholeiite lies directly
upward or downward from the equilibirum posi- along the line of neutral buoyancy, and a final
118 Michael P. Ryan

equilibrium position (at depth Z) is obtained at the mate pieritic horizon of neutral buoyancy (HNB p )
tholeiitic horizon of neutral buoyancy, HNB T • As for melt and crystal mixtures. Thus the summary
before, there is a family of potential density con- relationships are
trast lines (1-4) that passes through the tholeiitic
position, and mechanical equilibrium (and sudden
Horizon VPi > V(TH magma-
of nega- filled
departures from it) may be related to anyone line (5)
(but only one). Eruptions are considered nonequi- tive cracks
librium events, and volumetric displacements (in- buoyancy: descend
fluxes of deeper melt) are generally expected to be Horizon VP~ == V(TH magma-
required to lift tholeiitic liquids through the region of neutral filled
of negative buoyancy for extrusion at the surface. buoyancy: cracks are
[A second condition for "summit" eruption (di- (6)
gravita-
rectly above the replenishment site) is high values tionally
of crack-tip K 1C in the lateral sheeted dike com- stable
plexes and high values of (T3 normal to the lateral
intrusion pathway.] Thus the density minima (tho- Region of VPt < V(TH magma-
leiitic) melts of Fig. 12 are the recurrent and pre- positive filled
(7)
ferred products. buoyancy: cracks
Fractional crystallization progresses by the se- ascend.
quential appearance of liquidus phases, with each ln relations (5) thru (7), the superscripts ( -, N, +)
appearance inducing a change in melt density. indicate a magma pressure gradient within the
Figure 16 is a highly schematic illustration of one negative-, neutral-, and positive-buoyancy re-
mode in which fractionation can displace a melt gimes, respectively. Elastic fracture stability con-
batch from its former gravitational equilibrium ditions suggest that above the HNBI' VPi >
position. Removal of olivine, for example, from a V(T H, and negative-buoyancy forces promote the
picritic batch reduces the derivative melt density descent of pieri tic melt and crystal mixtures. This
(as shown in Figs. 12 (lower) and 13), producing conclusion is in accord with the relative rarity of
an increment of density reduction (op). Such re- ocean floor pieri tic eruptives.
duction renders the melt parcel unstable, and a For tholeiitic minimum-density melts (see
new mechanical equilibrium must be sought at a Fig. 12), VP T ~ 27 MPa . km- I , and gravitational
higher level (oZ). Continued episodes of crystal- equilibrium is approximated in the interval =600
lization and ascent are expected to progressively to = 1400 m depth below the volcanic surface.
elevate the parcel; however, the removal of rela- This position corresponds to the sheeted-dike com-
tively low-density components-through plagio- plex, in broad terms, and accords with the gener-
clase fractionation, for example-may tend to in- ally tholeiitic nature of these dikes. In the figures
crease the density of the remaining melt and lead of this chapter, the depth = 1000 m represents the
to negative buoyancy. mean transition, therefore, between the negative-
The overall horizon of neutral buoyancy may buoyancy region and the HNB for the combined
be subdivided into its ultrabasic and basic com- MORB density range 2.6 to 2.8 g . cm". As in
ponents. Gravitational equilibrium in the lower, relations (5) thru (7), an analogous set of rela-
portion requires matching the local gradient in tions describe the stability of tholeiitic minimum-
confining pressure, V(TH, with the local gradient in density melts: VPi > V(TH, V P~ == V(TH, and
picritic magmastatic pressure, VP p • For the East VPi- < V(TH, for the regions of negative, neutral,
Pacific Rise, the nonlinear Pi'(Z) gradient (Fig. 1) and positive buoyancy, respectively.
may be approximated by three piecewise-linear
segments: a near-surface 22.75 MPa· km- I seg-
ment, a central 27 MPa . km - 1 portion, and a lower Icelandic and Hawaiian Analogs
29 MPa . km - I gradient, respectively. Compari-
son with VPp == 28 MPa . km - 1 provides a best Episodes of rock fracture that accompany lateral
match within the interval = 1400 to =3000 m be- magma intrusions in active Icelandic rift systems
neath the volcanic surface. This is the approxi- produce microearthquakes that are routinely re-
 6. Neutral-Buoyancy Controlled Magma Dynamics 119
DENSITY(P)

T PLAGIOCLASE THOLEIITIC
CLINOPYROXENE _ _-.l__~ HORIZON OF '
I NEUTRAL
I
I BUOYANCY
I
I
I
I
I
I
I
I
I
I
I ILMENITE

I --4D
~ PICRITIC
0- HORIZON OF
NEUTRAL
W BUOYANCY
o

t-
Z
UJ
U
(fJ
«

op
/
CRYSTALLIZATION

Z L- ~-------------'

Figure 16 Density-depth relations for the depth evolution of the equilibrium position of a melt parcel during fractional crys-
tallization. "P" denotes picritic depth whereas "T" denotes tholeiitic depth. Increments of op correspond to liquidus phases
whose fractional removal induces increments of 02 (ascent). Picritic and tholeiitic horizons of neutral buoyancy are HNB p and
HNB T • respectively.

corded on portions of the Icelandic seismic net- on September 8, 1977 (Brandsdottir and Einars-
work. If the seismicity associated with the lateral son, 1979) as shown in Ryan (l987b». The se-
advance of the magma fracture front is systemati- quence began with the vertical ascent of magma
cally tracked during a dike-formation episode, the from the subcaldera inflation center that fed a
geometric relationships of dike evolution about small fissure eruption in the Gjastykki section
the horizon of neutral buoyancy then can be eval- of Krafla's fissure swarm. Within 19 h. the intru-
uated. Such a plot is shown in Fig. 17 and is based sion had "rolled over" the magma reservoir roof,
on the micro seismicity produced during and fol- and, driven by the continued overpressure in the
lowing the rupture of the Krafla magma reservoir ruptured reservoir and modulated by negative-
120 Michael P. Ryan

Figure 17 Depth-time-distance evolution of the magmatic fracture front at the Krafla central volcano, Iceland. The parabolic
fracture front continually adjusts its position to conform to the equilibrium position at the local horizon of neutral buoyancy.
Original data from the September 8, 1977, episode as reported in Brandsdottir and Einarsson (1979). (From Ryan, 1987b.)
Numerals on the progressive locations of the front positions are times (in hours) after the initial magma reservoir rupture. The
shaded arrow is the parabola peak pathway as the equilibrium density-depth position is approached. "F" denotes the final
position of the newly formed dike snout. (After Ryan (I 987b).)

buoyancy forces in the melt in the upper 2 km of frame (A), typically begin at the HNB and ad-
the rift, began a southward-directed lateral intru- vance until the fracture front is gradually arrested
sion into the Namafjall section. By to + 20 h, the and magma begins to be impounded behind the
intrusion had developed a characteristic parabolic crack tip. The arrested front forces melt above and
magma fracture front and by to + 22 h, it had below the HNB in ways that induce the dike keel
reached a position centered at the horizon of neu- to descend while the dike top grows toward the
tral buoyancy. The final resting position CF" in volcanic surface. Eventual lateral breakthrough
Fig. 17) was at the equilibrium neutral-buoyancy along the horizontal fracture path induces a simul-
level. It is coincident with the in situ melt country taneous lowering and rising of the magmatic col-
rock density crossover position in the Icelandic umn above and below the HNB, respectively. At
crust. Therefore, flowing melt tends to attempt to the conclusion of the intrusion episode, the para-
achieve a local mechanical equilibrium with its bolic fracture front comes to rest at the equi-
surroundings, which in effect makes the lateral in- librium position. Frame (B), on the other hand,
trusion process a natural moving in situ "density shows that an intrusion has experienced an early
meter." fracture front arrest and thus has a keel and top
Analogous behavior is commonly shown dur- that begin their motion with a rapid descent and
ing lateral magma intrusion episodes in Kilauea's ascent, respectively. Like intrusion (A), however,
rift zones. Figure 18 is from Ryan (1987b) and intrusion (B) also begins and ends at the horizon
illustrates depth-time portraits for dike-formation of neutral buoyancy and displays a crude top and
episodes. They are based on the depth-time bottom symmetry about the HNB.
swarms of microseismicity as depicted by the Multiple-increment intrusion cycles show a
shaded envelopes. These earthquakes were gen- more complex depth-time portrait that, however,
erated by rock fracture events induced by the hy- has several features in common with the single-
draulic pressure transients during intrusion. Two increment mode. For example, in (F), the initial
broad classes of intrusion are illustrated: (A) and intrusion event beneath the PUll 06 eruption site
(B) are single-increment intrusions and (C)-(F) contained three discrete path blockage events that
are multiple-increment events. correspondingly induced three depth-time pul-
Single-increment intrusions, as illustrated in sations in the height of the within-dike magma
 6. Neutral-Buoyancy Controlled Magma Dynamics 121
0 Or-------- =~---::::--- ....
1 1
2 2 t;.:::r::::7!:rr::::;;~2Mil I
3 3 c.:~=="'=_:_="'~ .
4 4
5 5
6
I 12 HRS. I
6
7 7
8 12 HRS. I KILAUEA 8 SOUTHWEST RIFT
9 EAST RIFT ZONE: JUNE, 1976 9 AUGUST, 1981
E A
------::~------~...::;
~E
10 10.. ...
~ 21 22 23 10 11
LU LU
o 0 o
-e
IL
1 ;f
II: 2 II:
::J ::J
Ul 3 Ul
o 4 o
Z 5 Z
-e -e 12 HRS.
c
...J 6 c...J
0 7 o
> 8
12 HRS. > MIDDLE EAST RIFT ZONE
J: J: DECEMBER, 1982
I- 9 EAST RIFT ZONE: MAY, 1979
-e
LU 10 B !CLU E
Z 30 Z 10
LU LU
III III
J: 0 J:
I- .1-
a. 1 a.
LU
LU 2
0 o
3
4
5
12 HRS.
6 H
7 12 HRS. I PU'U 0'0 SUPPLY
8 MIDDLE EAST RIFT ZONE
EAST RIFT ZONE: MARCH, 1980 JANUARY, 1983
9
10
c F
11 12 1 9
DATE WITHIN MONTH

Figure 18 Depth-time portraits of neutrally buoyant dike-formation episodes at Kilauea volcano, Hawaii. In all cases. the
intrusion stops and starts at the horizon of neutral buoyancy and has a mean height that is regulated by the HNB equilibrium
position. All horizontal time bars correspond to a 12-h duration. Frames (A) and (B) illustrate single-increment intrusions.
whereas frames (C)-(F) illustrate multiple-increment intrusion cycles. The timing of the three wave-like oscillations in frame
(F) corresponds to changes in the summit subsidence rate as recorded by the Uwekahuna tiltmeter (1. Dvorak. written commu-
nication). Envelopes are approximate. and are influenced by the standard errors in hypocenter locations. Only frame (F) was
associated with surface outbreaks of eruptive activity January 1983, corresponding to the maximum vertical surges of magma.
as illustrated. In general. multiple wave-like fluctuations in the dike height are responses to multiple incidents of crack-front
arrest during intrusion. each arrest producing oscillations above and below the equilibrium HNB. as magma is impounded behind
the crack tip. (From Ryan 1987b.)

column (i.e., the dike height). Upon breakthrough sociated with abrupt increases in dike length, due
at each blockage point, the vertical extent of the to conservation of mass. When time-integrated
dike contracted symmetrically about the hori- through an entire set of intrusion cycles, a wave-
zon of neutral buoyancy. Similarly, the multiple- like rise and fall of the vertical magma column in
increment cycles depicted in frames (C)-(E) show the growing dike is a characteristic feature. These
the pressure buildup-dike inflation-patterns be- oscillations may have substantial variations in the
hind the arrest site (increases in dike height) and amplitude (intrusion height) and frequency (time
the local pressure reduction-dike deflation- variations in dike height); 'however. they always
pattern that accompanies breakthrough events and exhibit approximate symmetry about the horizon
subsequent dike-height reductions along the intru- of neutral buoyancy. Thus wave-like patterns (in
sion pathway. Decreases in dike height are also as- depth-time sections) as illustrated in Fig. 18 are
122 Michael P. Ryan

Figure 19 Mid-ocean ridge environments at 9°_14°N (I) and at 21"N (2) along the East Pacific Rise where the highest levels
of magma storage and lateral injection in the sheeted dike complex is controlled by neutral buoyancy. Additional regions with

the result of magma encountering several obstruc- Neutral-Buoyancy Control in

tions along the rift zone path, with each obstruction Lateral Intrusion Dynamics:
forcing an increase in dike height as impounded The Sheeted-Dike Complex
magma rose above and descended below the hori-
zon of neutral buoyancy. Figure 19 summarizes the Figure 20 illustrates the structural and kinematic
well-documented locations where magma storage context within which the lateral high-level intru-
and intrusion dynamics are controlled by neutral sion of magma takes place. Integrated through
buoyancy. time, this process builds the sheeted-dike com-
 6. Neutral-Buoyancy Controlled Magma Dynamics 123

demonstrated neutral buoyancy control are Iceland (3) and Hawaii (4). The back-arc spreading center of the Valu Fa Ridge (5)
in the Lau Basin is also inferred to have similar control. (Modified after the Mercator projection of Heezen and Tharp, 1977.)

plex. The environment has two important attri- high amounts of fracture porosity produce a re-
butes, both essential for lateral intrusion: non- gion of negative buoyancy, such that for nonvesic-
linear increases in the in situ density structure ulated tholeiitic melt, Pm > Pis' This negative
with depth and available magma driving pressures buoyancy condition will be, therefore, accentu-
that reach maximum values at the HNB and can ated if the melt is picritic. Increases in confining
thus counteract the (horizontal) (T3 stress compo- pressure with depth progressively eliminate this
nent normal to the rift zone axis. Near the surface fracture porosity, such that a crossover is pro-
and beneath the axial valley floor, volumetrically duced between the density of the melt and that of
 124 Michael P. Ryan

BASALTIC
EXTRUSIVES

GABBROIC
INTRUSIVES

HARZBURGITIC
UPPER MANTLE
SECTION

Figure 20 Schematic isometric relationships between the evolving lateral magma injection pathway (short arrows) and the
structure of the axial valley and ridge axis. The orientation of the injection pathway is orthogonal to the sub-Moho flow of
harzburgite. Sea level corresponds to the top surface of this reference volume.

the surrounding rock, generating a horizon of neu- HNB counterbalance positively buoyant contri-
tral buoyancy. As discussed earlier, it is along this butions to dike-height establishment from fluid
horizon that magma is injected laterally. Beneath beneath the HNB. When combined with the posi-
the HNB, continued increases in confining pres- tive pressure differential from the ruptured reser-
sure, combined with progressive changes to ultra- voir that drives the fracture front forward, these
mafic lithologies, produce the upper portions of negative and positive contributions to melt buoy-
the region of positive buoyancy. Thus in Fig. 20, ancy continuously modulate the elevation of the
the basaltic extrusives blanket the sheeted-dike dike centerline during lateral growth. Profiles 1and
complexes, whereas the layered gabbros and the 2 (Fig. 21) thus correspond to an unobstructed-
harzburgites of the uppermost mantle form an en- and unarrested-mode of forward growth. By
vironment that produces positive melt buoyancy. profile 3, the intrusion front has been arrested by
Coordinates for lateral intrusion are shown in locally high values of if], or high K,c values along
Fig. 21 and illustrate the range of melt dynamics the crack-advance path. Magma then impounds
that produce the characteristic parabolic profile of behind the crack tip. Continued melt outflow from
the magma fracture front as well as the important the ruptured reservoir now begins to inflate the
melt-height perturbations induced by crack-front dike, and the walls move outward, swelling the
arrest. Symmetric intrusion modes are expected, dike and increasing its width. Conservation of
with the HNB serving as the plane of symmetry. mass now also forces magma up above the previ-
Thus negatively buoyant fluid elements above the ous dike top, while the keel begins a simultane-
 6. Neutral-Buoyancy Controlled Magma Dynamics 125

ERUPTIVE VENT
FRACTURED AND
POROUS BASALTIC
EXTRUSIVES
z

,,
:
,
SHEETED
i 4~
3
2
,,

DIKE
COMPLEX
1
8~ 9\ ,i
,,
10

"':.·d:::::::::::::::::::::::.::.::..:::::.\ ::.::.::.::. :::::::::::::::::::::::::::::::::::::::: :::::::::::::::::::"

-:_-_"'0 :::::~:=~--

y I
I
,
I

,
I

,,
I
I
I
,}--------

MAFIC
CUMULATES

Figure 21 Symmetry relationships and temporal evolution of the magma fracture front during lateral dike emplacement along
the horizon of neutral buoyancy. Numbers along profiles correspond to sequential positions of the front. At position 3, high
resistance to continued lateral advance has arrested the crack and impounded magma behind the crack tip. Continued out-flow
from the ruptured reservoir inflates (widens) the dike and forces magma up above and down below the equilibrium depth level
(curves 4, 5, and 6). Curve 7 breaks the surface, producing an eruption. Eventual rupture of the path-advance barrier tends to
return parcels of relatively shallow and deep magma to the neutrally buoyant equilibrium position, and curves 8, 9, and 10
therefore track subsequent magma crack tip positions; the eruption stops with intrusion being the only mode of magma move-
ment. Repeated episodes of the lateral injection process build the sheeted-dike complex. (The horizon of neutral buoyancy was
schematically compressed into a plane of symmetry for this illustration. In nature, it has a finite vertical extent much larger than
that implied here.)

ous descent, thus providing profiles 4, 5, and 6. to sill formation. Each accommodates the rapidly
Should the fracture front continue to be arrested, varying spatial changes in the resolved least com-
melt is forced further up along profile 7, which pressive stress orientation (0'3) and/or heteroge-
may subsequently intersect the Earth's surface, neously distributed elastic moduli.
producing an eruption. Breakthrough at the arrest Consider a fluid-filled crack in an isotropic and
region once again advances the dike, draws down linear elastic solid where the flow regime is lami-
melt from the strongly negatively buoyant region nar and undergoes Poiseulle flow between plane
high above the HNB-while simultaneously shut- parallel walls. The flow is driven by the primary
ting off the eruption-and simultaneously draws pressure differential VP within the fluid, and the
up relatively positively buoyant melt from the temporal evolution of the fracture width is given
dike keel. This process reestablishes the parabolic by (Lister, 1990)
fracture front, and profiles 8, 9, and 10 then track
dw I
the continued down-rift progression of the intru- - V . (w 3V P ), (8)
dt 3TJ •
sion. Deviatoric stress states induce important
modulations in the crack-advance pathway. These where TJ is the viscosity of the magma and w is the
modulations may include inflections in the crack fluid-filled crack half-width. The relations of Lis-
plane, fracture-front breakups, bifurcations and ter (1990) and Lister and Kerr (1990, 1991) are
dike-splitting events, and the transition from dike summarized in the following outline.
 126 Michael P. Ryan

The elastic pressure in the solid is given by static pressure differences between the fluid and
the solid are
p = -m ex (~;), (9) p = Po(x) e (PI - pJ gZ (11)
(above the HNB)
where m = f.L/(1 - v), f.L is the shear modulus of
and
the country rock, v is Poisson's ratio, and ex is the
Hilbert transform (Muskhelishvili, 1963; Erdelyi p = Po(x) + e (PI - pJ gZ (12)
et al., 1954). The total pressure is the buoyancy (below the HNB),
pressure resulting from density contrasts between
the melt and the country rock, as transmitted from
e
where is the density contrast parameter (given in
(10)) and the conjugate-density contrast parame-
the magma reservoir, in addition to the elastic
pressure of the crack walls. Taking the vertical co-
e
ter is
ordinate as Z. the pressure is Pm - Pu
e= 1 - e (13)
PI - Pu
p = - tipgZ - m ex (ddzW). (10)
The notations PI and p; refer to country rock den-
sities below (lower) and above (upper) the hori-
For the laterally directed intrusion along the zon of neutral buoyancy, respectively; they are
HNB, under density step conditions, the hydro- . thus a further subdivision of the (PiJ in situ

Figure 22 Relations between dike height and dike width for melt-country rock density contrasts (tip), when the elastic wall
rock stress is just balanced by fluid pressure in the flowing magma. Average dike/vein widths for Hawaii and Iceland as well as
the maximum range for dike heights are given by the shaded bands. At Kilauea. the two major rift zones are the East rift zone
(ERZ) and the Southwest rift zone (SWRZ).
 6. Neutral-Buoyancy Controlled Magma Dynamics 127
density. Ahove the HNB, the melt is more dense and Kerr (1991), modified for the applications of
than the country rock, and the condition Pm - p; this chapter.
> 0 helps to drive the flow laterally along the Relationships between dike height and dike and
HNB. Similarly, below the HNB, the condition vein widths for conditions approaching neutral-
PI - p.; < 0 drives fluid upward, and then out- buoyancy flow are illustrated in Fig. 22. The rela-
ward along the horizon of neutral buoyancy. Both tionship is given for density-contrast values (l1p)
conditions in (II) and (12) combine to modulate spanning the range 1.0-0.1 g. ern 3. These curves
the flow and to provide an overall VP to drive lat- move progressively from positively buoyant con-
eral intrusion of magma. In addition to the pri- ditions (l1p = 1.0-0.3 g . em -3), appropriate for
mary fluid pressure accumulation in the magma ascent through the basal region of a mid-ocean
reservoir, long rift zone intrusions produced by ridge magma reservoir, to the virtually neutrally
high-batch volume injections will also include a buoyant conditions (l1p == 0.1 g. em - 3), appropri-
pressure contribution produced by the topographi- ate for the uppermost isotropic gabbro and the
cally derived hydraulic head. sheeted-dike regimes of the reservoir and rift sys-
Lister and Kerr (1990, 1991) and Lister (1990) tems. Superposed are the fields of neutrally
have provided similarity solutions for the first- buoyant dike widths and dike heights for Hawaii
order fracture parameters under conditions that in- and Iceland, since such dike dimensions for active
clude neutral-buoyancy controlled crack advance. mid-ocean ridges are not available.
This section makes use of those solutions, with The relationship between the dimensionless
application to intrusion in mid-ocean ridge, Ice- dike height (c) and dimensionless dike width (w)
landic, Hawaiian, and ophiolite rift systems. In is illustrated in Fig. 23, for end-member classes of
particular, Figs. 22-26 were adapted from Lister the buoyancy parameter e. Collectively, they re-

NEGATIVE
BUOYANCY 1.0
1.0

Q 0.5 0.5
l-
I
o
iii
I
CIl
-HNB- CIl
W 0 0
...J
Z
0
Vi
z
ui
~
15 -0.5 -0.5

-1.0
POSITIVE
BUOYANCY
~
6 P<P
t m -1.0

Pm <Pb
.9
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4
DIMENSIONLESS WIDTH (W)

Figure 23 Cross-sectional aspect ratios of dikes as a function of dimensionless height (0 and dimensionless width (w), for
select values of the density contrast parameter e.
 128 Michael P. Ryan

spectively portray the cross-sectional aspect ratios cient combining material-property parameters for
of inflated dikes in the strongly positively buoy- the fluid and matrix:
ant, the fully developed negatively buoyant, and
e
the neutrally buoyant intrusion modes. reflects D = CI[(PI - pJgr. (17)
the density contrast between rocks that lie imme- 7][ILI(I - v)]2
diately below and above the horizon of neutral For conditions of laminar flow, c = 1, while
buoyancy, as well as the contrast between the melt the problem posed for a density step condition at
density and the surrounding country rocks during the HNB requires e = 3, in (IS), and d = 5.
initial ascent. It is defined (Lister, 1990) as For a lateral injection along the horizon of neu-
e = (PI - Pm). tral buoyancy, Lister and Kerr ( 1991) have related
(14) the crack height to fluid and rock properties as a
(PI - pJ
specialization of expression (IS),
e values that approach 1 denote strongly posi-
tively buoyant ascent modes associated with dikes r"l I - V)]4 Q2t2a-I]I/l1
hex. t) = <g~/5 [ 7]11-" (0 meg),
that tend to develop bulbous tops as they approach cIC;;[(PI - Pu)g]5
the HNB and have tails that are progressively (18)
squeezed shut by local components of the confin-
ing pressure. Similarly, as e approaches zero, where
dikes above the HNB may descend with bulbous
(lower) noses and tails squeezed closed by the' <g= (19)
local values of the effective stress. One could
further refine the mechanical equilibrium-non- and
equilibrium arguments by defining a picritic e and 1
a tholeiitic e.
e
Along the horizon of neutral buoyancy, = 0.5
<gn = [L H 3 d<g rS/I (20)

and is associated with dikes that have a maximum

In (20), H satisfies the ordinary differential
dilation (width) centered on the HNB, and both a
equation
keel and a top that are progressively (and sym-
metrically) squeezed shut by the horizontal com-
aH3 -
Sa +3 [<gH3]'
ponent of the effective stress. Each of the three 11
major intrusion modes with respect to e is illus-
trated in Fig. 23 and schematically sketched in the and boundary condition
figure inset box. The dimensionless nature of the
H(I) = O. (22)
plot, combined with the scaling chosen, results in
cross-sectional aspect ratios that, while preserving Physical-property values recommended for the
an accurate and faithful set of ratios (t, 11'), also modeling were derived from a combination of lab-
exaggerate the inflated widths so that one may il- oratory measurements and geophysical surveys.
lustrate the relevant trends. Values for the rigidity (IL = 20 GPa) reflect the in
From a general similarity theory for buoyancy- situ high temperatures at depth under conditions
driven flows (Lister. 1990), the height evolution of of partial fracture closure and were determined
a vertically oriented fluid-filled fracture is given by by torsional resonance spectroscopy (see Ryan,

dhdte Ddx~[hd+e~,
1987a). The Poisson's ratio was taken as 0.25. The
= (dh)']
dt
(IS) shear viscosity of tholeiitic melt (7] = 102 Pa . s)
was determined by Couette viscometry (Ryan and
with a relation between the flow rate and the Blevins, 1987, pp. 455-457, 466). The ill situ
fracture height integrated along the length and density (p; = 2.3 g . c;m -') of the near-surface
given by rocks (0.0-2 km depth) was inferred from the
combined seismic refraction and gravity studies of
(16) David Hill and co-workers (Zucca et al., 1982)
and was summarized in (Ryan, 1987b, Fig. 7b). It
where c. d. and e are constants, and D is a coeffi- is an average value for the horizon of negative
 6. Neutral-Buoyancy Controlled Magma Dynamics 129
buoyancy. Similarly, the density (p I) for the deeper inferred during observatory-based real-time intru-
region of positive buoyancy (ibid.) is 2.9 g . cm -3. sion monitoring in Iceland and Hawaii. Analogous
The density of the magma was inferred from to the dike-width relationships, the constant flow
the high-pressure falling-sphere melt density mea- rate curves show a gradually increasing height-
surements of Fujii and Kushiro (1977) on olivine but at a progressively diminished rate-as the
tholeiite and is 2.6 g . cm- 3 • These values have flow progresses. Similarly, the constant batch vol-
also been adopted by Lister (1990) and by Lister ume conditions suggest a progressive decrease in
and Kerr (1990, 1991). Importantly, note that melt height that is consistent with the conservation of
densities for the reservoir and sheeted-dike com- mass requirement: the continued lateral spread
plex will be in the range 2.6-2.8 g . ern:", reflect- along the HNB (dike lengthing) will correspond-
ing a compositional range that includes picrites as ingly diminish the dike height. The measured flow
well as tholeiites. durations at Krafla and Kilauea are representative,
Neutrally buoyant dike heights as a function of but by no means exhaustive. Thus, both the con-
flow duration are compared in Fig. 24 with heights stant flow rate and the constant batch predictions

FLOW DURATION, t (hours)

2.7 27. 277.
10,000

1 5,000 >-
I
KILAUEA
SW RIFT
8-1981

f-
I
o
Lij
I
w
~
o
f-
Z
~
o KILAUEA KILAUEA
::J
tIl KILAUEA ERZ PU'U 00
~ TROODOS ERZ 3-1980 1-1983
..J OMAN 12-1982
«
a:
f- KILAUEA
::J ERZ
w
Z 6-1972

FLOW DURATION, t (seconds)

Figure 24 Neutrally buoyant dike-height variations as a function of flow duration. Constant flow rates (10' m' . S-I :5 Q :5
102 .5 m' . s-') appropriate for Icelandic magma reservoir rupture as well as the rupture of mid-ocean ridge reservoirs illustrate
the declining rates of dike height evolution. Constant magma batches (10' m' :5 V:5 10' rn') produce initially high dikes that
progressively shrink in height as melt more closely approaches the horizon of neutral buoyancy and becomes increasingly spread
out laterally along this equilibrium depth level. The Kilauea and Krafla data have a dike height bar that corresponds to the
maximum depth range (heights) observed, while the shaded circles lie at the average height. The Oman, Bay of Islands, and
Troodos dikes are constrained in height only; however. their dimensions are compatible with the short-to-moderate flow dura-
tions shown here. These ophiolite points have thus been tentatively plotted to be compatible with the predicted dike heights.
 130 Michael P. Ryan

correspond with at least some part of virtually all has discussed the kinematics of rift zone intru-
observed ranges of dike height. For Kilauea, yet sion along the HNB in Hawaii and in Iceland,
lower values of both the flow rates and the magma respectively.
batch volumes appear required to pick up the entire Neutrally buoyant dike widths as a function
range of observed heights. Ophiolite dike dimen- of flow duration are compared with field data in
sions and geometric flow parameters for Oman Fig. 25. The widths were estimated by assuming
(Juteau et al., 1988; Christensen and Smewing, two types of "end member" flow regimes: cases
1981), the Bay of Islands complex, Newfound- of constant flow rate (Q = 102 to 102.5 m 3 • S-I)
land (Salisbury and Christensen, 1978), and the and constant magma batch volume (107 ::5 V ::s
Troodos, Cyprus, complex (Christensen and Sal- 108 m"), Experience in monitoring Kilauea and
isbury, 1975) are constrained in height only, and Krafla has demonstrated that episodes of reservoir
not, of course, in flow duration. Their modest rupture have outflow rates that are of order 102 to
heights, however, are roughly compatible with 5 X 102 m>- S-I; thus the plotted flow rates have
both the Icelandic and Hawaiian flow rates during a scaling appropriate for dike-forming reservoir
early 'flow periods, immediately after magma res- rupture in Hawaii, Iceland, and, by analogy, the
ervoir rupture. They have, therefore, been tenta- mid-ocean ridge system. Plots of water-tube tilt-
tively plotted as illustrated. Ryan (1988, 1990) meter and electronic tiltmeter signals during the

FLOW DURATION, t (hours)

2.7 27.
'0

1---------------,..,..-;-ICElAND ---------------1

O.',LO·:--- -'-_ _...L_--L_...L----ll.-...L...L...L-'-:- -"- -'-_---L_...l--L_L--.L-l-J

FLOW DURATION, t (seconds)

Figure 25 Neutrally buoyant dike widths as a function of flow duration. End-member cases of constant flow rate (10 2 m 3 • S-I
S Q S 10 2.5 m 3 • S-I) and constant volume (10 7 m 3 S V S 108 rn") span ranges appropriate for the mid-ocean ridges, Hawaii
and Iceland. The gray band brackets average Icelandic dike widths (circle + dot) and average Sumail ophiolite, Oman dike
widths. Inferred widths for Hawaiian dikes are plotted (circles) for their measured flow durations at Kilauea and are compatible
with the predictions based on both constant flow rate and constant magma batch (volume) conditions.
 6. Neutral-Buoyancy Controlled Magma Dynamics 131
rupture event, however, clearly demonstrate that scales with the segment half-length has been as-
magma reservoir-exiting flow rates are exponen- sumed, consistent with a dike-forming intrusion
tially decreasing functions of time (e.g., Dvorak that has a centrally located origin within a spread-
and Okamura, 1987), and thus the constant rates ing center segment magma reservoir (Whitehead
illustrated in Fig. 25 must be understood to be and Helfrich, 1990; Ryan, 1987b). Thus points 8
highly idealized. Constant batch volumes have a (Oman; Nicolas, 1989),9 (Juan de Fuca; Nicolas,
physical basis that is solidly grounded in obser- 1989), and IO (East Pacific Rise average; Bonatti,
vation: the abrupt rupture of the fluid-swollen res- 1985) are roughly constrained in length only, while
ervoir releases, in certain cases, a discrete volume the flow durations are, of course, unknown. Nev-
of magmatic fluid into the rift system. After re- ertheless, the confidence gained in the Icelandic
lease, the reservoir deflation process may reseal and Hawaiian comparisons, which are well con-
fluid passageways, and thus a finite volume of strained, suggests that the upper right-hand side of
magma is introduced into the sheeted-dike com- the graph is broadly consistent with their lengths.
plex. Superposed on Fig. 25 are the flow durations If this agreement is true, then one has the physical
for closely monitored Hawaiian and Icelandic in- basis for making some rough flow duration esti-
trusions, respectively representing Kilauea and mates for observationally inaccessible or ancient
Krafla. Average dike widths for Icelandic and Ha- magma intrusions.
waiian rift systems as well as the Sumail ophiolite,
Oman, define a broad field that is believed to
include the active mid-ocean ridge sheeted-dike Notes on Hydrothermal Interactions
swarms. A consideration of that portion of the plot
that is well-populated by data points suggests that The interrelationships between spreading rate,
there is a broad correspondence between the ophi- neutral-buoyancy zonation depths, and mid-ocean
olite and Hawaiian widths and the constant vol- ridge thermal structure are as yet unknown, but
ume/flow rate predictions. The larger Icelandic such relationships are now required to complete
dikes require yet higher flow rates and volumes, our understanding of the detailed balances that
however. The constant flow-rate curves reflect an promote the long-term stability of shallow magma
increase in width with flow duration, but at a chambers. That the position of the shallow sub-
greatly reduced rate of growth after the initial ridge magma chambers and their sheeted-dike
27-h period. Conversely, the release of constant complexes approximate a 1: I correspondence
batch volumes into a growing dike produces a with the horizon of neutral buoyancy is evident
diminution in widths with flow duration, reflect- from this chapter and from Ryan (1985, 1987a,
ing the conservation of mass consequences of in- 1987b, 1993). What is not clear, however, is the
creased lateral spreading (dike lengthening) along additional detailed range of nonequilibrium effects
the HNB.. associated with hydrothermal heat withdrawal and
Neutrally buoyant dike lengths as a function of the details of the solidification process.
flow duration are compared with real-time flow Magma positive buoyancy and heat losses are
durations and dike length data for Iceland and Ha- competing influences in modulating the ascent
waii in Fig. 26. Again, the observed values are process. Understanding this phenomenon within
believed to be representative of each location, the context of long-term magma storage requires,
but are not exhaustive. As expected, the constant among other things, a three-dimensional model of
flow-rate conditions predict ever-increasing dike a spreading center that properly incorporates the
lengths through time. Constant batch volumes also fracture mechanics of deep dilatant cracking in
suggest increases in dike length, as the finite vol- the magma reservoir roof. Fracture processes are
ume becomes increasingly spread out and attenu- an important function of spreading rate and ther-
ated along the finite thickness horizon of neutral mal structure, bearing in mind that much of the
buoyancy. For both the constant batch and con- density-reducing porosity in the roof is fracture
stant flow-rate conditions, fairly good agreement porosity. It is fracture porosity, that, for example,
is found with the overall spread of observed intru- dramatically lowers the in situ Vp , Vs values in the
sions at Krafla and at Kilauea. For ophiolite and upper 2 km of the oceanic crust. Moreover, frac-
mid-ocean ridge comparisons. a dike length that ture porosity and fracture permeability dominate
 132 Michael P. Ryan
FLOW DURATION, t (hours)

2.7 27. 277.

100~ I I I I --.---.---,-----,-.-.r-~

~
I-

:§
I
I-
(9-
~i:
-1(9
wZ
~~
01- 5
~ ~
«~
>-(9
Ow
::::JU"l
10
•
aJu.
~~
-II
«-
a:
I-
::::J

I
w
Z
10 KILAUEA

L----_J
1
0 KRAFLA

FLOW DURATION, t (seconds)

Figure 26 Neutrally buoyant dike length variations as a function of flow duration. For constant flow rates, mid-ocean ridge
dike lengths are expected to increase at rates compatible with the documented growth rates at Kilauea, Hawaii, and Krafla,
Iceland. Constant magma batch volumes (l O" m' oS V -s 10' m') are similarly compatible with both the Hawaiian and the
Icelandic rift zone injection environments. Ranges (and averages) at positions 8, 9. and 10 are for Oman. the Juan de Fuca Ridge,
and the East Pacific Rise, respectively, and are constrained in segment length only.

the efficacy of the hydrothermal system and its gressive shallowing of the fracture front as the
depth extent. spreading rate increases. Because much of the
Phipps Morgan et al. (1994, this volume) and density reductions that produce a negatively buoy-
Phipps Morgan and Chen (1993) have found broad ant environment (that overlies and induces the
agreement between the apparent depth dependence neutral-buoyancy horizon) are associated with
of the tholeiite solidus (predicted from numerical fracture porosity, the Huang and Solomon (1988)
experiments with Nusselt numbers (Nu) over the data suggest that the horizon of negative buoy-
range 8 :=:; Nu :=:; 12) and half-spreading rates, as- ancy may have a greater depth extent over slow
suming hydrothermal penetration temperatures of spreading centers-thus tending to deepen some-
600°C (see their Fig. 9 (Phipps Morgan et al., what the horizon of neutral buoyancy and thus the
1994) and Fig. 7 (Phipps Morgan and Chen, 1993), sheeted-dike complex and the uppermost magma
respectively). They have also plotted the depth reservoir. This implication should, in turn, be
extents of axial earthquake centroids, on the basis consistent with the progressive elevation of the
of the teleseismic study of Huang and Solomon brittle-ductile transition as expected as spread-
(1988) in Fig. 9 of their chapter in this volume. ing rates increase. Such a relationship is borne
These results indicate deep fracturing (to depths out by the results of their numerical experiments.
of 8-12 km) at slow spreading rates and a pro- An overlay of the =1- to =3-km-depth range
 6. Neutral-Buoyancy Controlled Magma Dynamics 133
inferred for the horizon of neutral buoyancy of nounced where the spreading rates are low and
this chapter, with the modeled magma lens depth the deep fracture network promotes deep hydro-
of Phipps Morgan and Chen (1993) or Phipps thermal penetration. The Mid-Atlantic Ridge fits
Morgan et al. (1994, this volume), shows com- into this category. Ryan (I987b) has demon-
plete agreement throughout the half-spreading rate strated, however, that the magma chamber of the
range 30-80 mm . yr -I. This agreement included Krafla central volcano is coincident with the
all EPR points plotted. Comparisons with other HNB, and this region has a double spreading rate
areas (e.g., Juan de Fuca and the Lau Basin) of 20 mm . yr - 1. Thus the purely hydrothermal-
should utilize seismic-velocity profiles specific to dominated portion of the spreading rate range
the region for detailed study. I conclude here that must be confined to low rates indeed.
the gross density zonation of mid-ocean ridges
over the full spreading rate range 20-160 mm .
yr - I, and the position of the horizon of neutral Summary Relationships and Evolution
buoyancy in particular, provides fundamental con- of the Oceanic Crust
trol of the location of the uppermost portions of
the magma reservoir as well as the location and The region of negative buoyancy in the newly
the intrusion dynamics within the sheeted-dike created lithosphere begins at the ocean floor and
complex. Within this overall picture, thermal ef- extends to = I km depth. It comprises the pillow-
fects that include both solidification and hydro- basalt flow units (± sediments) and the upper-
thermal penetration may provide significant lo- most brecciated dikes and corresponds to oceanic
cal variations on what is dominantly a magma- crustal layers I and 2A (Fig. 27). In situ densities
buoyancy and gravitational equilibrium theme. are lower than non vesiculated tholeiitic melt and
Hydrothermal interactions that locally retard the are kept low by deeply penetrating fractures and
achievement of equilibrium neutral-buoyancy po- large amounts of grain-scale porosity. For tholei-
sitions (depths) are expected to be most pro- itic melt, this region represents a gravitational "no

o~--==-== _
1111n-,-,.~ EGAT/VE BUOYANCY 2A

l NEU;R~~/Z6N6F ! I I ' , T I 28
BUOYA. NCY JlillilllLlllLlLillLllllllllWlillllilllllWllilill
:2
~
-~ ~
5~
I~
'I '"
-

r:;:-
-...."..~
:----
- - :--=--
-=.....--=::::::--·---------------+-M-
LAYERED
GABBROIC
COMPLEXES

~ ,I
c: >-
::c 10~1 ~
~ u, ~ REGION OF
LJ.J 0 0
Z Z:::l
~ 0 co COMPRESSIONAL
::c (3 w WAVE RETARDATION
I- W >
t

~ l5"11t1
10 20 30
AGE OF OCEANIC CRUST (Mal
~- 40 50 60

Figure 27 Evolution of the horizons of neutral buoyancy and negative buoyancy and their lithologic products as sea floor
spreading progresses astride the ridge. "M" denotes the Mohorovicic discontinuity. Seismic layer numbers (I, 2, and 3) are
given in the right-most column (modified after Christensen and Salisbury, 1975).
134 Michael P. Ryan

man's land," and to erupt, magma must traverse Table 2

the region of negative buoyancy aided by contin- A World Without Neutral Buoyancy-Controlled Magma
ued volumetric displacements and reservoir re- Storage and Along-Axis Transport in the Mid-Ocean Ridge

plenishment from positively buoyant melt, pro-

• Roughly steady-state artesian-like eruption of basaltic melt
viding vertically integrated lift from depth. directly above regions of melt production and sub-
The generalized HNB begins at = I km and ex- Moho melt focusing
tends to =3 km depth. Lithologically, it produces • Extrusion of basaltic melt only above the site of produc-
the sheeted-dike swarms, as well as the upper- tion, and thus an excessive concentration at the seg-
most isotropic gabbros. It corresponds to layer 2B ment center
and to the uppermost portion of oceanic layer 3 • No lateral injections of magma from crustal reservoirs
(Fig. 27). From data derived from seismic reflec- • No well-developed sheeted-dike complexes in ophiolites
tion and refraction surveys of the East Pacific and oceanic crust
Rise, for example, the HNB approximates a I: I • No long-lived magma reservoirs beneath mid-ocean ridges
correspondence with the low-velocity region ly- with high melt/crystal ratios
ing just beneath the rise axis. In tomographic reso- • No significant development of isotropic gabbro
lutions of the compressional wave velocity struc- • Marked increases in the ratio of extrusives to intrusives in
ture, the horizon of neutral buoyancy corresponds ridge and ophiolite environments
to the lowest velocity core region and suggests the
highest fluid-to-rock ratios. Late-stage differenti-
ates such as trondjhemite intrusions, are contained. DSDP hole 418A. [Their Figs. 5 and 6 (pp. 9161-
within the horizon of neutral buoyancy at the base 9162) contain an order of magnitude error in the
of the sheeted-dike root region. Proper discussion depth axis labeling.] Table I summarizes several
of buoyancy zonation must incorporate the com- correlations of the neutral-buoyancy hypothesis.
plete expected compositional range: picritic to
tholeiitic to ferrobasaltic melts, with due consid-
Comments at Closure:
eration of suspended phenocryst phases. A World without Neutral Buoyancy
As the oceanic crust cools and is rafted aside
on a flowing asthenosphere, the region of negative It is interesting to contemplate a world without
buoyancy and the horizon of neutral buoyancy neutral buoyancy (Table 2) and its consequences.
ride with it (Fig. 27). Hydrothermal alteration due In such a world, crustal rocks are everywhere of
to continued zeolite and greenschist facies meta- greater density than the melts that pass through
morphism will gradually increase the density of them and, accordingly, only a universal regime
permeable basaltic flow units (Carlson and Ras- of positive buoyancy exists-from the Earth's
kin, 1984; Carlson and Herrick, 1990; Nelig and surface to sites of grain-scale melt mobilization.
Juteau, 1988); however, in situ densities that are Table 2 lists several consequences of such a world
lower than tholeiitic melt densities should persist and, taken collectively, they suggest a very differ-
to at least 25-30 Myr and become progressively ent picture than that observed for both the rock
elevated (Carlson and Herrick, 1990, p. 9168). record in ophiolite complexes, and the eruptive,
Therefore, upward migrating off-axis melts may intrusive, and gross geomorphic character of mid-
continue to be trapped in oceanic crust as old as ocean ridges.
25-30 Myr, and high-level sills may be expected
at the most shallow depths. Thus sea mounts de-
veloping in off-ridge locations, for example, may Acknowledgments
be expected to be underlain at relatively shallow
depths by rather small magma storage reservoirs. Discussions with Charles Carrigan (Lawrence Livermore Na-
These relationships are consistent with the down- tional Laboratory), Robert Coleman (Stanford University), and
Robert Crosson (University of Washington) have contributed
hole sonic logging inferences of Carlson and Her-
to this work. John Lister (University of Cambridge) kindly
rick (1990), which indicate that in situ densities provided a preprint of his JGR paper with Ross Kerr. Helpful
match those of tholeiitic melts at depths of 200- review comments by Paul Delaney, John Dvorak, Terrence
450 m within the llO-Ma Atlantic ocean crust of Edgar, John Lister, Janet Morton. Wayne Shanks, and Rob-
 6. Neutral-Buoyancy Controlled Magma Dynamics 135
ert Tilling improved progressive revisions of the manuscript. 116. Tryggvason (1980)
Shirley Brown, Lendell Keaton, and Nancy Polend of the U.S. 117. Pollard et al. (1983)
Geological Survey assisted in the preparation of several of the 118. Ewart et al. (1990)
figures. Michelle Urie and Debby Pasquale prepared the type- 119. Ewart et al. (1990)
script. John Sinton and Jason Phipps Morgan are thanked for 120. Ewart et al. (1990)
preprints. Parts of this work were presented at the Prince- 121. Ewart et al. (1990)
ton-Conoco Symposium in Earth Sciences, Princeton Univer- 122. Ewart et al. (1990)
sity; the V. M. Goldschmidt Conference. Reston, Virginia; the 123. Ewart et al. (1990)
Symposium on the Generation, Segregation, Ascent and Stor- IV. Kilauea volcano, Hawaii
age of Magma at the 29th International Geological Congress, HI. Dieterich and Decker (1975)
Kyoto, Japan; the Geophysical Institute of Kyoto University; H2. Duffield et al. (1982)
the Earthquake Research Institute of the University of Tokyo; H3. Walsh and Decker (1971)
and the Department of Earth and Planetary Sciences of Har- H4. Mogi (1958)
vard University. H5. Jackson et al. (1975)
H6. Fiske and Kinoshita (1967)
H8. Davis etal. (1974)
Appendix H9. Ryan et al. (1981)
HIO. Dvorak et al. (1983)
Data sources for depth distributions of neutrally buoyant melt HII. Ryan etal. (1983)
H12. Eaton (1962)
beneath the East Pacific Rise, the Valu Fa Ridge, Lau Basin.
H13. Koyanagi etal. (1976)
the Krafla Central volcano, NE Iceland, and Kilauea volcano.
Hawaii. Symbols correspond to those in Figs. 3A and 3B. H14. Ryan etal. (1983)
H15. Ryan etal. (1983)
I. East Pacific Rise H16. Ryan et al. (1983)
EPR 2. (9°N) Seismic refraction profile. Orcutt et al. Hl7. Moore and Fiske (1969)
(1976) H18. Swanson etal. (1976)
H19. Jackson et al. (1975)
EPR 3. (9°N) Multichannel seismic reflection. Herron
et al. (1978) H20. Wright and Weiblen (1968)
H21. Swanson et al. (1976)
EPR 4. (9°N) Multichannel seismic reflection, Hale
et al. (1982)
EPR 5. (9°_14°N) Multichannel seismic reflection, De-
trick et al. (1987) References
EPR 6. (9°_IO ON) Multichannel seismic reflection.
Vera et al. (1990)
Bender, J. E, Hodges, EN., and Bence, A. E, (1978). Petro-
EPR 7. (9°N) Seismic reflection, Sleep ct al. (1983)
genesis of basalts from the project FAMOUS area: Experi-
EPR 8. (21°N) Seismic refraction profile, Reid et al.
mental study from 0 to 15 kbars, Earth Planet. Sci. Left.
(1977)
41,277-302.
EPR 9. (9°N) Seismic refraction, Rosendahl et al.
Berge. P" Fryer, G., and Wilkens, R. (1992). Velocity-
(1976)
porosity relationships in the upper oceanic crust: Theoreti-
II. Valli Fa Ridge, Lau Basin cal considerations, J. Geophys. Res. 97, 15239-15252,
VF I. Multichannel seismic reflection. Morton and Bjornsson, A" and Sigurdsson, O. (1978), Hraungos iir bor-
Sleep (1985) holu i Bjarnarflagi, Ndtturu Fraedingurinn, 19-23.
III. Krafla central volcano. Iceland Bjornsson, A., Johnsen, G., Sigurdsson, S., Thorbergsson, G.,
I I. Einarsson (1978) and Tryggvason, E. (1979). Rifting of the plate boundary
12. Einarsson (1978) in north Iceland: 1975-1978. J. Geophys. Res. 84, 3029-
13. Bjornsson et al. (1979) 3038.
14. Tryggvason (1981, 1986) Bonatti, E. (1985). Punctiform initiation of seafloor spreading
15. Ryan, unpublished numerical results (1983) in the Red Sea during transition from a continental to an
16. Johnsen et al. (1980) oceanic rift, Nature 316,33-37,
17. Marquart and Jacoby (l985) Bottinga, Y, and Weill, D.E (1970). Densities of liquid sili-
18. Marquart and Jacoby (1985) cate systems calculated from partial molar volumes of ox-
19. Marquart and Jacoby (1985) ide components, Am, 1. Sci. 269, 169-182,
110. Larsen et al. (1979); Bjornsson and Sigudsson Brandsdottir, B., and Einarsson, P, (1979). Seismic activity as-
(1978) sociated with the September 1977 deflation of the Krafla
Ill. Marquart and Jacoby (1985) central volcano in northeastern Iceland, J. Volcanol. Geo-
112. Marquart and Jacoby (1985) therm. Res, 6,197-212.
113. Marquart and Jacoby (1985) Carlson, R. L.. and Raskin, G. S. (1984), Density of the ocean
114. Brandsdottir and Einarsson (1979) crust. Nature 311, 555-558.
115. Brandsdottir and Einarsson (1979) Carlson, R. L.. and Herrick, C. N, (1990). Densities and po-
 136 Michael P. Ryan

rosities in the oceanic crust and their variations with depth Transport and Storage" (M. P. Ryan, ed.), Wiley, Chi-
and age, J. Geophys. Res. 95, 9153-9170. chester/Sussex, England.
Christensen, N. I. (1978). Ophiolites, seismic velocities and Fiske, R. S., and Kinoshita, W. T. (1967). Inflation of Kilauea
oceanic crustal structure. Tectonophysics 47. 131-157. volcano prior to its 1967-1968 eruption, Science 165,
Christensen. N. I.. and Salisbury, M. H. (1975). Structure and 341-349.
constitution of the lower oceanic crust. Rev. Geophvs. Fujii, T., and Kushiro, I. (1977). Density, viscosity, and com-
Space Phys. 13.57-86. pressibility of basaltic liquid at high pressures, Carnegie
Christensen, N. I., and Shaw, G. H. (1970). Elasticity of mafic lnst, Washington Yearbook 76, 419-424.
rocks from the mid-Atlantic ridge, Geophys. J. Roy. As- Gove, P. B. (1964). "Webster's Third New International Dic-
tron. Soc. 20, 271-284. tionary of the English Language Unabridged," G.&C
Christensen. N. I., and Smewing, 1. D. (1981). Geology and Merriam Co., Springfield, MA.
seismic structure of the northern section of the Oman Hale. L. D., Morton, C J., and Sleep, N. H. (1982). Reinter-
ophiolite. J. Geophys. Res. 86. 2545-2555. pretation of seismic reflection data over the East Pacific
Christensen. N. I. (1982). Seismic velocities, in "Handbook of Rise, 1. Geophys. Res. 87,7707-7717.
Physical Properties of Rocks" (R. S. Carmichael, ed.), Heezen, B. C, and Tharp, M. (1977). (Map of) the world's
Vol. II, pp. 1-228, CRC Press, Boca Raton, FL. ocean floor. Mercator Projection, Scale: I: 23,230,300,
Christeson. G. L., Purdy, G. Moo and Fryer, G. 1. (1992). Struc- United States Navy, Office of Naval Research.
ture of young upper crust at the East Pacific Rise near Herron, T. 1., Ludwig, W.1., Stoffa, P. L., Kan, T. K" and Buhl,
9°30'N, Geophys. Res. Lett. 19, 1045-1048. P. (1978). Structure of the East Pacific Rise crest from
Clarke, D. B. (1970). Tertiary basalts of Baffin Bay: Possible multichannel seismic reflection data, 1. Geophys. Res. 83.
primary magma from the mantle, Contrib. Mineral. Petrol. 798-804.
25,203-224. Hill. D. P. (1969). Crustal structure of the Island of Hawaii
Davis. P. M., Hastie, L. M., and Stacey, F. D. (1974). Stresses from seismic-refraction measurements, Bull. Seis. Soc.
within an active volcano-with particular reference to Ki- Am. 59, 101-130.
lauea, Tectonophysics 22, 355-362. Hill, D. P" and Zucca. 1. 1. (1987). Geophysical constraints on
Detrick. R. S., Buhl, P., Vera. E, Mutter. 1., Orcutt, 1., Madsen, the structure of Kilauea and Mauna Loa volcanoes and
J., and Brocher, T. (1987). Multi-channel seismic imaging some implications for seismomagrnatic processes, in "Vol-
of a crustal magma chamber along the East Pacific Rise. canism in Hawaii" (R. W. Decker. T. L. Wright, and P. H.
Nature 326,35-41. Stauffer, eds.), U.S. Geological Survey Prof. Paper 1350.
Dieterich, J. H., and Decker. R. W. (1975). Finite element 917.
modeling of surface deformation associated with volca- Hooft, E. E., and Detrick, R. S. (1993). The role of density in
nism, 1. Geophys. Res. 80,4094-4102. the accumulation of basaltic melts at mid-ocean ridges,
Dixon. 1. E., Stolper, s., and Delaney. J. R. (1988). Infrared Geophys. Res. Lett. 20,423-426.
spectroscopic measurements of CO, and H,O in Juan de Huang, P. Yoo and Solomon. S. C (1988). Centroid depths of
Fuca Ridge basaltic glasses, Earth Planet. Sci. Lett. 90. mid-ocean ridge earthquake: Dependence on spreading
87 -104. rate, J. Geophys. Res. 93.13445-13477.
Duffield, W. A.. Christensen. R. L., Koyanagi, R. Y.. and Pe- Huppert, H. E.. and Sparks, R. S. J. (1980). Restrictions on the
terson. D. W. (1982). Storage. migration and eruption of compositions of mid-ocean ridge basalts: A fluid dynami-
magma at Kilauea volcano, Hawaii, 1971-1972. J. Vol- cal investigation, Nature 286, 46-48.
canol. Geother. Res. 13,273-307. Irvine, T. N. (1977). Definition of primitive liquid composi-
Dvorak. 1., Okamura. A.. and Dieterich, J. H. (1983). Analysis tions for basic magmas, Carnegie lnst, Washington Year-
of surface deformation data, Kilauea volcano, Hawaii, hook 76. 454-461.
October 1966 to September 1970, J. Geophys. Res. 88. Jackson, D. B" Swanson, D. A., Koyanagi, R. Y., and Wright.
9295-9304. T. L. (975). The August and October 1968 East Rift Erup-
Dvorak, 1., and Okamura, A. T. (1987). "A Hydraulic Model tions of Kilauea Volcano, Hawaii," U.S. Geological Sur-
to Explain the Variations in Summit Tilt Rate at Kilauea vey Prof. Paper 890.
and Mauna Loa Volcanoes. U.S. Geological Survey Prof. Johnsen, G. v., Bjornsson, A.. and Sigurdson, S. (1980). Grav-
Paper 1350. pp. 1281-1296. ity and elevation changes caused by magma movement be-
Eaton, 1. (1962). Crustal structure and volcanism in Hawaii. in neath the Krafla caldera, northeast Iceland, J. Geophys.
"Crust of the Pacific Basin," pp. 13-29. Am. Geophys. Res. 47. 132-140.
Union Monograph. Juteau, T., Beurrier, M" Dahl, R., and Nehlig. P. (1988). Seg-
Einarsson. P. (1978). S-wave shadows in the Krafla caldera in mentation at a fossil spreading axis: The plutonic sequence
NE-Iceland, evidence for a magma chamber in the crust, of the Wadi Haymiliyab area (Haylayn Block, Sumail
Bull. Volcanol. 41-3. 1-9. Nappe, Oman), Tectonophysics 151.167-197.
Elthon, D. (1979). High magnesia liquids as the parental Koyaguchi, T. (1990). Graphical estimations of magma densi-
magma for ocean floor basalts, Nature 278. 514-518. ties and compositional expansion coefficients, Contrib.
Erdelyi, A., Mangus, w.. Oberhettinger, E, and Tricomi. E G. Mineral. Petrol. 105. 173-176.
(eds). (1954). "Tahles of Integral Transforms," McGraw- Koyanagi, R. Y., Unger, J. D., Endo. E. T.. and Okamura, A.
Hill, New York. T. (1976). Shallow earthquakes associated with inflation
Ewart. J. A., Voight, B., and Bjornsson, A. (1990). Dynamics episodes at the summit of Kilauea volcano, Hawaii. Bull.
of Krafla caldera, north Iceland: 1975-1985, in "Magma Volcano I. 39. 621-631.
 6. Neutral-Buoyancy Controlled Magma Dynamics 137
Larsen, G., Gronvold, K., and Thorarinsson, S. (1979). Vol- tion of basic and ultrabasic rocks. Earth Sci. ReI'. 4,
canic eruption through a geothermal borehole at Nama- 69-1330.
fjall, Iceland. Nature 278, 707 - 710. Orcutt, J. A., Kennett, B. L. N., and Dorman, L. M. (1976).
Lippard, S. 1., Shelton, A. W, and Gass, I. G. (1986). The Structure of the East Pacific Rise from an ocean bottom
ophiolite of Northern Oman, Geol. Soc. London Memoir, seismometer survey, Geophys. J. Roy. Astron. Soc. 45,
11. 305-320.
Lister, J. R. (1990). Buoyancy-driven fluid fracture: Similarity Palmason, G. (l97\). "Crustal Structure of Iceland from Ex-
solutions for the horizontal and vertical propagation of plosion Seismology," Soc. Sci. Islandica, R.l.T. XL.
fluid-filled cracks. 1. Fluid Mech. 217. 213-239. Palmason, G., and Saemundsson, K. (1974). Iceland in rela-
Lister, 1. R., and Kerr, R. (1990). Fluid-mechanical models of tion to the mid-Atlantic ridge, Annu. ReI'. Earth Planet.
dyke propagation and magma transport, in "Mafic Dyke Sci. 2.
and Emplacement Mechanisms" (A. 1. Parker, P. C. Rick- Phipps Morgan, J., and Chen, Y. 1. (1993). The genesis of oce-
wood. and D. H. Tucker. eds.), pp. 69-80. A. A. Balkema anic crust: Magma injection, hydrothermal circulation, and
Publishers, Rotterdam. crustal flow, J. Geophys. Res. 98, 6283-6297.
Lister, 1. R., and Kerr, R. (1991). Fluid-mechanical models of Phipps Morgan, 1., Harding, A., Orcutt, J., Kent, G., and Chen,
crack propogation and their application to magma trans- Y. 1. (1994). An observational and theoretical synthesis of
port in dykes, J. Geophvs. Res. 96, 10049-10077. the magma chamber geometry and of crustal genesis along
Macdonald. K. c, Sempere, 1. c, and Fox. P. 1. (1984). East a mid-ocean spreading center, in "Magmatic Systems"
Pacific Rise from Siqueiros to Orozco fracture zones: (M. P. Ryan, ed.), Academic Press, San Diego.
Along-axis-strike continuity of axial neovolcanic zone and Pollard, D. D., Delaney, P. T., Duffield. W A., Endo, E. T., and
structure and evolution of overlapping spreading centers. Okamura, A. T. (1983). Surface deformation in volcanic
J. Geophys. Res. 89, 6049-6069. rift zones. Tectonophysics 94, 541-584.
Mammerickx, J., and Smith, S. M. (1980). General bathymet- Reid, I., Orcutt, 1. A., and Prothero, W A. (1977). Seismic
ric chart of the oceans (GEBCO), (Mercator Projection. evidence for a narrow zone of partial melting underlying
Scale: I: 10,000,000 at equator). Canadian Hydrographic the East Pacific Rise at 21°N, Geol. Soc. Am. Bull. 88,
Service, Ottawa, Canada. 678-682.
Marquart. G., and Jacoby, W (1985). On the mechanism of Roeder, P. L. (1974). Activity of iron and olivine solubility in
magma injection and plate divergency during the Krafla basaltic liquids, Earth Planet. Sci. Lett. 23, 397-410.
rifting episode in NE-Iceland. 1. Geophys. Res. 90, 10178- Rosendahl, B. R., Raitt, R. W, Dorman, L. M., Bibee, L. D.,
10192. Hussong, D. M., and Sutton, G. H. (1976). Evolution of
Marsh, B. D., and Maxey, M. R. (1985). On the distribution oceanic crust. Part I. Physical model of the East Pacific
and separation of crystals in convecting magma, J. Vol- Rise crest derived from seismic refraction data. J. Geo-
canol. Geother. Res. 24, 95-105. phys. Res. 81, 5294-5304.
Martin, D., and Nokes, R. (1988). Crystal settling in a vigor- Rothery, D. A. (1983). The base of a sheeted dyke complex,
ously convecting magma chamber. Nature 332, 534-536. Oman ophiolite: Implications for magma chambers at
Mogi, K. (1958). Relations between the eruptions of various oceanic spreading centers, J. Geol. Soc. London 140,
volcanoes and the deformation of the ground surfaces 287-296.
around them. Bull. Earthquake Res. Inst. Tokyo Univ, 36, Ryan, M. P. (1985). The contractancy mechanics of magma
99-134. reservoir and rift system evolution: EOS Trans. American
Moore, 1. G., and Fiske, R. S. (1969). Volcanic substructure Geophysical Union, vol. 66, no. 46, p. 854.
inferred from dredge samples and ocean bottom photo- Ryan, M. P. (l987a). The elasticity and contractancy of Ha-
graphs, Hawaii, Geol. Soc. Am. Bull. 80,1191-1202. waiian olivine tholeiite, and its role in the stability and
Morton. J. L., and Sleep, N. H. (1985). Seismic reflections structural evolution of sub-caldera magma reservoirs and
from a Lau Basin magma chamber, in "Geology and Off- rift systems, in "Volcanism in Hawaii" (R. W Decker,
shore Resources of Pacific Island Arcs-Tonga Region T. L. Wright. and P. H. Stauffer, eds.), Vol. 2. U.S. Geo-
(D. W Scholl and T. L. Vallier, eds.), Vol. 2, pp. 441- logical Survey Prof. Paper 1350.
453, Circum-Pacific Council for Energy and Mineral Re- Ryan, M. P. (l987b). Neutral buoyancy and the mechanical
sources, Earth Sciences Series, Houston, TX. evolution of magmatic systems, in "Magmatic Processes:
Muskhelishvili, N. I. (1963). "Some Basic Problems of the Physicochemical Principles" (B. O. Mysen, ed.), Geo-
Mathematical Theory of Elasticity." Noordhoff, Leiden, chemical Society Special Publication No. I, pp. 259-
The Netherlands. 287, Univ, Park, PA.
Nelig, P., and Juteau, T. (1988). Flow porosities. permeabili- Ryan, M. P. (1988). The mechanics and three-dimensional in-
ties and preliminary data on fluid inclusions and fossil ternal structure of active magmatic systems: Kilauea vol-
thermal gradients in the crustal sequence of the Sumail cano, Hawaii, J. Geophys. Res. 93,4213-4248.
ophiolite (Oman), Tectonophysics 151,199-221. Ryan. M. P. (1990). The physical nature of the Icelandic magma
Nicolas, A. (1989). Structures of ophiolites and dynamics of transport system, in "Magma Transport and Storage"
oceanic lithosphere, Kluwer, New York. (M. P. Ryan, ed.). Wiley. Chichester/Sussex, England.
O'Hara, M. J. (I 968a). Are any ocean floor basalts primary Ryan, M. P. (1993). Neutral buoyancy and the structure of
magma? Nature 220, 683-686. mid-ocean ridge magma reservoirs. 1. Geophys. Res., 98,
O'Hara, M. 1. (I 968b). The bearing of phase equilibria studies 22.321-22,338.
in synthetic and natural systems on the origin and evolu- Ryan, M. P., and Blevins, 1. Y. K. (1987). "The Viscosity of
138 Michael P. Ryan

Synthetic and Natural Silicate Melts and Glasses at High T L. (1976). "The February 1969 East Rift Eruption of
Temperatures and One Bar (10' Pascals) Pressure and at Kilauea Volcano, Hawaii," U.S. Geological Survey Prof.
Higher Pressures," U.S. Geological Survey Bulletin 1764. Paper 891.
Ryan, M. P.. Blevins, 1. Y. K., Okamura, A. T, and Koyanagi, Toomey, D. R., Purdy, G. M., Solomon, S., and Wilcox, W.
R. Y. (1983). Magma reservoir subsidence mechanics: (1990). The three-dimensional seismic velocity structure
Theoretical summary and application to Kilauea volcano, of the East Pacific Rise near latitude 9°30'N Nature 347,
Hawaii, J. Geophys. Res. 88,4147-4181. 639-644.
Ryan, M. P., Koyanagi, R. Y., and Fiske, R. S. (1981). Model- Tryggvason, E. (1980). "Observed Ground Deformation dur-
ing the three-dimensional structure of magma transport ing the Krafla Eruption of March 16, 1980," Nordic Vol-
systems: Application to Kilauea volcano, Hawaii, J. Geo- cano!. Inst. Rept. 80-05, Univ. of Iceland.
phys. Res. 86, 7111-7129. Tryggvason, E. (1981). "Pressure Variations and Volume of
Salisbury, M. H., and Christensen, N. 1. (1978). The seismic the Krafla Magma Reservoir," Nordic Volcano!. Inst. Rept.
velocity structure of a traverse through the Bay of Islands 81-05, University of Iceland.
ophiolite complex, Newfoundland, and exposure of oce- Tryggvason, E. (1986). Multiple magma reservoirs in a rift
anic crust and upper mantle, 1. Geophys. Res. 83, 805- zone volcano: Ground deformation and magma transport
817. during the September 1984 eruption of Krafla, Iceland,
Shibata, T, DeLong, S. E., and Walker, D. (1979). Abyssal J. VolcanoI. Geotherm. Res. 28, 1-44.
tholeiites from the oceanographer fracture zone. 1. Petrol- Vera, E. E., Mutter, J. c., Buhl, P., Orcutt, J. A., Harding, A. 1.,
ogy and fractionation, Contrib. Mineral. Petrol. 70, 89- Kappus, M. E., Detrick, R. S., and Brocher, T. M. (1990).
102. The structure of 0- to 0.2-m.y.-old oceanic crust at 9°N
Sinton, J. M., and Detrick, R. S. (1992). Mid-Ocean ridge on the East Pacific Rise from expanded spread profiles,
magma chambers, J. Geophys. Res. 97,197-216. J. Geophys. Res. 95, 15529-15556.
Sleep, N. H., Morton, 1. L., Bums, L. E., and Wolery, T. 1. Walker, D., Shibata, T., and DeLong, S. E. (1979). Abyssal tho-
(1983). Geological constraints on the volume of hydro- leiites from the oceanographer fracture zone. II. Phase equi-
thermal flow at ridge axes, in "Hydrothermal Processes libria and mixing, Contrib. Mineral. Petrol. 70, 111-125.
at Sea Floor Spreading Centers" (P. Rona, K. Bostrom, Walsh, 1. B.. and Decker, R. W. (1971). Surface deformation
L. Lambier, and K. Smith, eds.), NATO Conference Se- associated with volcanism, J. Geophys. Res. 76, 3291-
ries II, Marine Sciences 12, Plenum Press. New York. 3302.
Sleep, N. (1975). Formation of oceanic crust: Some thermal Weissel,1. K. (1977). Evolution of the Lau Basin by the growth
constraints, J. Geophys. Res. 80, 4037 -4042. of small plates, ill "Island Arcs, Deep Sea Trenches and
Sparks, R. S. 1., and Huppert, H. E. (1984). Density changes Back-Arc Basins" (M. Talwani and W. C. Pitman, III,
during the fractional crystallization of basaltic magmas: eds.), American Geophysical Union, Maurice Ewing Se-
Fluid dynamic implications, Contrib. Mineral. Petrol. 85, ries I, pp. 429-436.
300-309. Whitehead, 1. A., Dick, H. 1. B.. and Schouten, H. (1984). A
Sparks, R. S. 1., Huppert, H. E., Koyaguchi, T, and Hallworth, mechanism for magmatic accretion under spreading cen-
M. A. (1993). Origin of modal and rhythmic igneous lay- ters, Nature 312, 146-147.
ering by sedimentation in a convecting magma chamber. Whitehead, 1. A., and Helfrich, K. R. (1990). Magma waves
Nature. 361, 246-249. and diapiric dynamics, ill "Magma Transport and Stor-
Sparks, R. S. J., Meyer. P., and Sigurdsson. H. (1980). Density age" (M. P. Ryan. ed.), pp. 53-76, Wiley, Chichester/Sus-
variation amongst mid-ocean ridge basalts: Implications sex. England.
for magma mixing and the scarcity of primitive lavas, Wright, T L., and Weiblen, P. W. (1968). "Mineral Composi-
Earth Planet. Sci. Lett. 46, 419-430. tion and Paragenesis in Thoeliitic Basalt from Makaopuhi
Stolper, E., and Walker, D. (1980). Melt density and the aver- Lava Lake, Hawaii [abs]," Geol, Soc. Amer. Special Paper
age composition of basalt, Contrib. Mineral. Petrol. 74, 115. pp. 242-243.
7-12. Zucca, 1. 1., Hill, D. P., and Kovach, R. L. (1982). Crustal
Stommel, H. (1949). Trajectories of small bodies sinking structure of Mauna Loa Volcano, Hawaii, from seismic
slowly through convection cells, 1. Marine Res. 8, 24-29. refraction and gravity data, Bull. Seismol. Soc. Am. 72,
Swanson, D. A., Jackson, D. B., Koyanagi, R. Y., and Wright, 1535-1550.
Chapter 7 An Observational and Theoretical Synthesis of
Magma Chamber Geometry and Crustal Genesis
along a Mid-ocean Ridge Spreading Center

J. Phipps Morgan, A. Harding, J. Orcutt, G. Kent, and Y. J. Chen

Overview Notation
Units
In this chapter we review seismological evidence and
D, Lagrangianderivative
other geophysical evidence that the axial magma cham-
ber beneath a fast spreading ridge is a narrow (-l-km- K permeability m'
wide), thin (-50- to 200-m-thick), magma lens that lies L latent heat J. kg-I
at the sheeted dike/gabbro cumulate transition region L" velocity gradienttensor
roughly 1.2- 1.5 km beneath the seafloor and overlies Nu Nusselt number dimensionless
a broader region of "hot rock" with at most -3-5%
Q attenuation dimensionless
partial melt fraction. This axial magma chamber ap-
pears to contradict earlier ophiolite-based studies that T temperature K,OC
used the dip and dip relations within the "cumulate" Tm mantle temperature K,oC
gabbro layer to argue for a broad, gabbro-layer- T'",off temperaturelimit for hydrothermal K,oC
thickness magma body that deposited cumulates along circulation
its base and sides. However, it is compatible with an T~cawaler seawater temperature K
emerging theoretical paradigm that crustal accretion oc-
b crack spacing m
curs by magma emplacement and solidification within
this magma lens, with cumulates subsiding and flowing p pressure Pa
to form the lower crust as initially proposed by N. H. qhydm axial hydrothermal heat flux W·m- 2
Sleep (J. Geophys. Res. 80, 4037-4042, 1975). Here time
we present a theoretical thermal and mechanical model
u, u, velocity vector
for crustal genesis that incorporates this paradigm and
appears to explain successfully the observed depth de- x, X; position vector m

pendence of the axial magma lens with spreading rate y, y; particle displacementvector m
(and the fact that no axial magma lens has been seen :'"<off depth limit for hydrothermal m
in a ridge with a median valley morphology), as well circulation
as observed relationships among axial morphology, :hyd," axial depth of hydrothermal m
spreading rate, and magma supply. We suggest that a penetration
fairly delicate balance between magmatic heat input K 1, °C-I
a thermal expansivity
and hydrothermal heat removal determines the thick-
ness (i.e., yield strength) of the axial lithosphere, which 1> porosity dimensionless
in turn controls the axial morphology associated with K thermal diffusivity
plate boundary extension. Thus the depth (and exis- A hulk viscosity Pa s
tence) of an axial magma lens and the axial morphology IJ- shear viscosity Pa· s
along a spreading center share a common thermal ori-
7;; deviatoric stress tensor Pa
gin that is a function of spreading rate and magma
supply. "Local" trend geochemical systematics may ifJ magma emplacementrate
also share this common thermal origin. ,I, ,I;; finite strain tensor dimensionless

Magmatic Systems Copyright © 1994 by Academic Press, tnc.

Edited by M. P. Ryall 139 AH rights of reproduction in any form reserved.
140 J. Phipps Morgan et al,

Notation for Oceanic Crust Expanded Abbreviation Definition

Seismic Analogous
velocity ophiolite Typical Mg# Magnesium number [MgO/(MgO + FeO)
Seismic (km' crustal Ophiolite depth range moles]
layer s") layer structure (m) MORB Mid-ocean ridge basalt Most common basalt
extruded at a mid-ocean
2a 2.5-5 ia Pillow 0-500 ridge, less Si-rich than
basalt ocean island or island
2b 5-6.5 ib Sheeted 500-1500 arc basalts.
dike 'Na "Sodium 8" A geochemical observ-
3 6.5-7 Cumu- able is corrected for
t"
late the effects of shallow
gabbro 1500-6000 crustal fractionation
along a "proper" liquid
line of chemical evolu-
Note. We use a separatenotation for seismicand ophiolitecrustal tion to its appropriate
layering to keep these very distinctobservations conceptualIy distinct. value when the MgO
Western numerals are used for seismiclayering (the standardconven- content of the liquid
tion)and Arabicnumeralsare used for ophiolitelayering. was 8%. [cf. Klein and
Langmuir, 1987]
Acronym Glossary
OBS Ocean bottom Self-contained seafloor
Expanded Abbreviation Definition
seismograph seismograph with in-
strument and recording
AMC Axial magma chamber The melt lens + under-
package.
lying mush that makes
up an EPR magma OSC Overlapping spreading A (small) offset in the
chamber. center ridge axis where two
ridge segments overlap
CDP Common depth point Standard sorting tech-
instead of being linked
gather nique where shot-
by a lransfonn fault
receiver pairs with a
offset.
shared "midpoint" are
gathered. P-wave Pressure wave Compressional mode of
seismic energy
DEVAL Deviation from axial A kink in the ridge axis.
propagation.
linearity
PmP P-moho-P P-wave arrival that has
EPR East Pacific Rise Type-example of a fast-
reflected from the
spreading ridge.
Moho.
ESP Expanding spread Seismic reflection tech-
S-wave Shear wave Shear mode of seismic
Profile nique where source and
energy propagation.
receiver pairs are sited
equal distances from a
central point to image Note. Marine geologists. seismologists. and geochemists share a
the structure beneath fondness for creatingacronyms that mayconfusea newcomer. Wehope
the central point at mul- this glossarymakes amends for our introduction of a separatenotation
tiple "look" angles. for seismicand ophioliteoceaniccrustallayeringby describingthe ac-
This is essentially a ronymsthat are liberalIy used in this chapter.
CDP gather extending
to tens of kilometers,
but involves two ships
steaming tens of kilo- Introduction
meters away from this
common point. The mid-ocean ridge system is the most produc-
",1AR Mid-Atlantic Ridge Type-example of a tive volcanic system on Earth, annually adding an
slow-spreading ridge. average of 20 km- of new material to the oceanic
\tiCS Multichannel seismic Seismic technique that crust. Almost all the material added to the oceanic
reflection technique uses a multiple-source/
crust is thought to experience some degree of
multiple-receiver ge-
ometry to enhance ob-
low-pressure fractionation within a crustal magma
servation of seismic chamber. However, the nature of the magma cham-
reflectors. ber and of crustal accretion at fast- and slow-
 7. Magma Chamber Geometry and Crustal Genesis 141
spreading ridges is fundamentally different. Early ophiolite observations (see Fig. lA), which were
thermal models of ridge crest structure (Sleep, used to support the crustal-scale magma chamber
1975; Kusznir and Bott, 1976) predicted that the hypothesis, with magma deposition at the base of
heat supplied by magma rising from the mantle this large molten magma chamber leading to the
would lead to a large (- 5-km-thick, -15-km- formation of cumulate layered gabbro structures
wide at its base) magma chamber beneath the axis (Cann, 1974; Smewing, 1981; Pallister and Hop-
of a fast-spreading ridge. Sleep (1975, 1978) and son, 1981). The "fly in the ointment" came from
Dewey and Kidd (1977) proposed much smaller early seismic measurements that showed no large
molten magma bodies, but these ideas were ap- molten zone present beneath the axis of pres-
parently contradicted by early interpretations of ent day ridges. Subsequent seismic studies have

A Gabbro Layering & Genesis -- Oman Ophiolite

Lavas
diking
Dyke swarm
\I Isotropic gabbro
and Plagiogranite
magma chamber
Layered
sequence

--....=---
--- - -
_ _ _•
-:, ~~'il
:-....::. ... \1
J:i'(~
If .. ~-:
--- -
_
Harzburgite
tectonite
. --- - - -- ---,',,)1 1(:,'.---
.- - - - l\I~'\I'1 I {I~:( L.-.J
Ikm
I 1'1,
feeder zone

B Seismic Image of Axial Crustal Structure

Pillow Basalts
Dikes

Gabbro

Peridotite

Figure 1 (A) Schematic model of a mid-ocean ridge spreading center derived from Oman ophiolite studies (after Smewing,
1981). The stratigraphic sequence below the pillow and sheeted dike complexes is shown on the right side of the figure. Isotropic
gabbros just beneath the sheeted dike complex grade into gabbros with a (weakly developed) near vertical dip that becomes both
more developed (Nicolas et 01., 1988; Nicolas, 1989) and more shallowly dipping as you move deeper into the gabbro section,
(The layering is best developed and parallel to the Moho directly above the gabbro-peridotite "petrologic Moho" contact.) This
dip structure was used by Smewing (1981) to infer that gabbro layering reflects cumulate deposition on the floor of the large
magma chamber sketched here. (B) Model of magma chamber structure for the East Pacific Rise derived from seismic studies.
Molten magma is concentrated in a lens approximately I km wide that resides at the base of the sheeted dike complex. Beneath
the magma lens and extending to mid-crustal depths is a broader region of rock at elevated temperatures that contains a few
percent partial melt. Details of the structure immediately above the Moho are unclear but it is known that the seismic Moho
forms within a few kilometers of the axis. The surficial layer of pillow basalts and sheet flows increases in thickness away from
the axis as a result of successive volcanic eruptions.
continues
142 J. Phipps Morgan et al.

C Magma Lens Model for Crustal Flow and Layering

crustal "isochrons";
predicted layering due to lower crustal flow flowlines & accumulated strain

Figure 1 continued. (C) Schematic theoretical model of crustal extension and flow showing flowlines. accumulated strain, and
the layering predicted by lower crustal flow away from a shallow injection site at the sheeted dike/gabbro interface. The left side
of the figure shows predicted layering "isochrons" generated from crustal flow away from a magma lens at the base of the
sheeted dike complex. The right-hand side shows several typical crustal flowlines and the accumulated strain along each flowline.
Strain is most intense in the lowermost part of the gabbro section. Note the similarity in form between the crustal-flow-generated
layering and the Oman ophiolite layering shown in (A). .

found that a very small (50 to hundreds of meters Henstock et al., 1993) and thus significantly re-
thick, l-km-wide) magma chamber structure (see duce the size of any magma chamber or cause it
Fig. IB) is ubiquitously present beneath the axis to freeze (Lister, 1983). Ophiolite layering can
of a fast-spreading ridge (Detrick et al., 1987; also be reconciled with a small shallow magma
Harding et al., 1989; Vera et al., 1990), but a lens if most lower crustai material is initially in-
magma body of even this size has yet to be de- jected into this shallow lens-the subsequent flow
tected along the axis of a slow-spreading ridge to make the lower crust will impart a strain-
(Detrick et al., 1990). Sleep's (1975) original induced fabric that is similar in form to the fabric
"magma chamber" and crustal flow scenario has seen in the ophiolite record (Phipps Morgan and
been revived and its consequences have been fur- Chen, 1993; Quick and Denlinger, 1993; Hen-
ther explored (cf. Fig. lC). Magma injection into stock et al., 1993). In this chapter we first review
a small magma lens at the base of the sheeted dike the current body of geophysical constraints on
complex appears to have the potential to explain ocean ridge magma chamber structure. Then we
both the flow patterns seen in the ophiolite record present the theoretical foundations and implica-
and the small magma lens that seismic methods tions of this emerging physical paradigm for oce-
image at active fast-spreading ridges (Phipps anic crustal genesis.
Morgan and Chen, 1993; Quick and Denlinger,
1993; Henstock et al., 1993). Thermal predictions
and geophysical observations are reconciled by The Seismic Record: Seismic Constraints on
including the effects of hydrothermal circulation Magma Emplacement and Crustal Accretion
in the heat budget. Early thermal models (Sleep, at a Fast-Spreading Ridge Axis
1975; Kusznir and Bott, 1976) cooled the crust
purely by a combination of conductive cooling Geophysical experiments, particularly seismic ex-
through the seafloor and advection of heat away periments, conducted in the last decade support a
from the ridge crest by the diverging plates. Hy- basic division between a temporally variable ac-
drothermal circulation may enhance the cooling cretionary pattern at slow-spreading ridges with
through the roof of a magma chamber by a factor cycles of magmatic injection and amagmatic ex-
of about 8-12 compared to conduction (Morton tension, and an essentially steady-state magma
and Sleep, 1985; Phipps Morgan and Chen. 1993; chamber sustaining crustal accretion at fast-
 7. Magma Chamber Geometry and Crustal Genesis 143

spreading ridges. So far, experiments at slow- periments. Although these experiments demon-
spreading ridges have failed to detect anything strated that a magma chamber could exist beneath
that could be considered to be the seismic signa- the East Pacific Rise they simultaneously demon-
ture of a molten magma body (Detrick et al., strated that an extensive body of magma could not
1990), although seismic velocity anomalies that exist, thereby confounding expectations based on
can be considered the thermal afterglow of the thermal and ophiolite models.
most recent site of magmatic injection have been Individually the seismic results from the north-
found in isolated locations beneath the median ern East Pacific Rise are not unique. Magma
valley of the Mid-Atlantic Ridge (Purdy and De- chamber reflections have been recorded within
trick, 1986; Kong, 1990; Kong et al., 1992). In the Lau Basin (Morton and Sleep, 1985; Collier
contrast, seismic experiments conducted along and Sinha, 1990) and at the Juan de Fuca Ridge
the fast-spreading East Pacific Rise have found (Morton et al., 1987; Rohr et al.. 1988). Zones of
a largely continuous shallow reflector (Detrick reduced seismic velocity, a sign of at least ele-
et al., 1987, 1993) that is interpreted as the top of vated temperatures, have been reported beneath
a similarly continuous magma body. The magma the Juan de Fuca Ridge (White and Clowes, 1990)
body is underlain by a broader region of reduced and also the median valley of the Mid-Atlantic
seismic velocity that is even more continuous Ridge. Rather, the significance of the northern
along-axis and is probably disrupted only by East Pacific Rise results lies in the fact that they
larger ridge discontinuities such as transforms. can be united into a coherent and consistent pic-
In the following sections we review the observa- ture, albeit still somewhat blurred, of the axial
tions that constrain the current seismic picture magma chamber structure. In particular, reflec-
of magma chamber structure at fast-spreading tion, refraction, and tomography experiments
ridges. were collocated within the 9°N segment so that re-
sults of one experiment could be compared di-
Northern East Pacific Rise rectly to another.
The 9°N segment of the East Pacific Rise is
The geometric constraints on magma chamber bounded at its northern end by the Clipperton
structure at fast-spreading ridges are in large part transform at lOolO'N, where the axis is offset by
the result of a diverse series of seismic experi- 85 km, and at its southern end by the 9°Q3'N over-
ments conducted along the northern East Pacific lapping spreading center (OSC), where the rift
Rise during the 1980s. These experiments include zones are separated by 8 km. A finer-scale seg-
the MAGMA ocean bottom seismograph (OBS) mentation into 10 or more segments has been pro-
tomography experiment at 12°N (McClain et al., posed on the basis of ridge crest morphology and
1985), a 1985 two-ship multichannel seismic ex- detailed near-bottom mapping of the axial summit
periment that included a regional reflection sur- graben or caldera (Haymon et al., 199Ia). In ad-
vey between 9° and 13°N, plus detailed refraction dition to the geophysical experiments, the seg-
and reflection experiments at two sites at 9° and ment has been the site of systematic dredging, an
I3°N (Harding et al., 1989; Vera et al., 1990), extensive submersible dive program, and a recent
and an ocean bottom hydrophone tomography ex- attempt at bare rock drilling by the Ocean Drilling
periment at 9°N (Toomey et al., 1990). During Program. The segment was also the site of a vol-
the previous decade, refraction experiments con- canic eruption during the early part of 1991 (Hay-
ducted as part of project RISE at 21°N (Reid et al.. mon et al.. 1991b).
1977) and project ROSE at 12°N (Lewis and Gar-
many, 1982; Ewing and Meyer, 1982; Bratt and Correlation between Seismic and Ophiolite
Solomon, 1984) as well as at 9°N (Orcutt et al., Observations of Crustal Structure
1975; Rosendahl et al.. 1976) had all found evi-
dence for an axial magma chamber, as had a pair A detailed interpretation of seismic velocity re-
of reflection lines shot across the rise axis at 9°N sults in terms of the structure and petrology of the
(Herron et al., 1978; Hale et al., 1982). However, oceanic crust is the subject of a constantly evolv-
no generally accepted model for the structure of ing debate. The basis for interpretation is the as-
the axial magma chamber emerged from these ex- sociation of the seismically determined layering
144 J. Phipps Morgan et al.
EPR velocity model & Oman ophiolite interpretation
o r--------.::------------,==~
Layer2A

Layer 2B iB
Dike swarm
-j-Ll--'-L-'-.L.L-Lj Isotropic gabbro
f.-.----.-...,...-j and Plagiogranite
2

,,-...
S 3
,..
C Layer]
Layered
oS sequence
fr4
'"Cl

_ _ -----1 Harzburgite
6 Moho Transition Region tectonite

2 3 4 5 6 9
velocity (km . sol)
Figure 2 A common means of inferring oceanic crustal structure from seismic measurements is to relate them to ophiolite
observations. This figure shows a possible correlation between an East Pacific Rise velocity model and a stratigraphic column
from the Oman ophiolite (cf. Figure lA). The correlation of layer 2 with the upper volcanics, and layer 3 with the plutonic section
is generally accepted. However, the relationship of the subdivisions of seismic layer 2 to structure is.less certain and is almost
certainly not uniform across all spreading rates.

of the oceanic crust with the structural sequence that shown in Fig. 2 (Christensen and Smewing,
found in ophiolites as shown in Fig. 2. For a fast- 1981; Kempner and Gettrust, 1982; Karson et al.,
spreading ridge, the following "simple" layering 1984). The association is complicated by, among
pattern appears to be appropriate. Seismic layer 2 other factors, the strong dependence of seismic
is associated with the volcanic basalt section velocity on porosity (Spudich and Orcutt, 1980),
(layer r) of the ophiolite crust; 2A is associated the relatively weak dependence of seismic ve-
with pillow lavas, and layer 2B with the sheeted locity on mineralogy, and uncertainty as to how
dike complex of the upper part of the ophiolite representative ophiolites are of oceanic crust (e.g.,
sequence. (Seismic layer 1 is associated with deep Moores, 1982). Support for the ophiolite model of
marine sediments that cap the pillow basalts in the oceanic crust comes from drilling at DSDP/ODP
ophiolite section. However, sediment cover is at hole 504B, which has penetrated the pillow lavas
most only a few meters thick in local sediment and sheeted dikes that form the upper part of the
ponds at the extremely young seafloor near a fast- ophiolite sequence (Becker et al., 1989), and from
spreading ridge axis.) Seismic layer 3 is associ- recent submersible observations at Hess Deep
ated with the plutonic gabbro section (layer r-) of (Francheteau et al., 1992). The observations at
ophiolite crust (e.g., Kempner and Gettrust, Hess Deep, which is a rifted portion of normal
1982). Several reasonably successful attempts East Pacific crust, give structural thicknesses for
were made to correlate seismic velocities of the upper crust that are' comparable to seismic
ophiolite rocks and structural associations with layer thicknesses derived from rise-axis experi-
the marine crustal velocity structure leading to ments. These recent comparisons make it reason-
ophiolite-derived velocity structures much like ably certain that the base of layer 2 does indeed
 7. Magma Chamber Geometry and Crustal Genesis 145
lie at the base of the sheeted dike complex and that the magma lens, even though this may be the clos-
this is the level of the top of the seismically im- est approximation to the one in the conventional
aged axial magma chamber at a fast-spreading physical sense, as the interaction between the
ridge. magma lens and the underlying mush zone with
the exchange of melt and crystals may be impor-
Current Seismic Picture of an Oceanic Magma tant in the chemical differentiation of the oceanic
Chamber at a Fast-Spreading Ridge crust (Langmuir, 1989; Sinton and Detrick, 1992).
From a seismological viewpoint it is convenient
In this section, the current geometrical model of to divide the magma chamber into two distinct
the structure of a fast-spreading magma chamber components, a magma lens and an underlying low
is presented as fact without many of the necessary velocity region. The low velocity zone contains
qualifications and caveats that are inevitable for a hot rock with a possible core of a crystal mush
model based on indirect measurements. Subse- zone that surrounds the magma lens-both ele-
quent sections more thoroughly discuss the geo- vated temperatures and partial melt cause a reduc-
physical evidence for the model. tion in seismic velocities. The seismological divi-
In cross section, the magma chamber at a fast- sion reflects the fact that different elements of
spreading ridge is a narrow (-I-km-wide), thin the magma chamber structure are detected by dif-
(- 50 to hundreds of meters thick) melt lens that ferent seismological methods. Evidence for the
overlies a broader (r-fi-km-wide), thicker (2- to magma lens comes principally from reflection
4-km) region of hot rock that may include a small data while evidence for the low velocity zone
«3-5%) melt fraction (Fig. lB). The lens accu- comes from refraction and tomography data. The
mulates at the base of the sheeted dike complex next sections review the primary seismic con-
through the trapping of buoyant magma that is straints on the structure of this type of magma
injected into the crust from the mantle and is chamber.
rereleased from the (initially cumulate) crystal
mush as the mush subsides and shears to form the There Is an Axial Magma Lens at the Base of
lower part of the oceanic crust. Vigorous cooling the Sheeted Dike Complex that Contains a
of the lens from above by hydrothermal circula- Large (>25%) Fraction ofMelt
tion causes freezing or plating of gabbro from the The best evidence for a magma lens or sill be-
lens onto the base of the sheeted dike complex and neath the axis of the East Pacific Rise comes from
cumulates freezing and settling to a subsiding multichannel seismic reflection data. When prop-
floor of the lens. Only the top of the magma cham- erly migrated, the mid-crustal reflection inter-
ber, where a lens or sill of predominantly molten preted as the magma lens typically appears on
magma accumulates, is well defined. While the cross-axis reflection profiles as a bright, narrow
top edges of the magma lens are quite distinct, event centrally located about 0.6 s beneath the
both the sides and the base of the magma chamber rise-axis (Fig. 3). The same event, termed the
are indistinct with melt percentages within a crys- axial magma chamber reflection or AMC reflec-
tal mush zone falling toward the edges of the tion, also appears at the same location on inter-
chamber and grading into plutonic rocks with secting along-axis reflection profiles and is con-
residual elevated temperatures. The maximum tinuous for tens of kilometers along strike (Fig. 3).
width of the magma chamber with a mean melt When correlated with velocity studies, it can be
fraction less than ~ 3% is typically on the order of shown that this reflection marks the top of a
6-8 km, whereas the base of the chamber lies at crustal low velocity zone and that the reflection is
mid-crustal depths, 3-4 km below the seafloor, caused by a negative acoustic impedance contrast.
within the gabbroic section. Further information on conditions within the
It is evident that this idea of a magma "cham- magma lens can be deduced QY detailed examina-
ber," in the conventional sense of a well-defined tion of the amplitude and phase characteristics of
cavity within the host rock, is something of a mis- the AMC reflection, which depends on the con-
nomer at fast-spreading ridges. Petrologists have trast of material properties across the top of the
been loath to restrict the label magma chamber to magma lens. Data on the variation of the seismic
 146 J. Phipps Morgan et al,

\1;.,\ rJo(l~ 6
P-wave
5
-;"
v:
4
E
C
Q 3 S-wave
I1J 'u
E 0
~
E::: ~ 2

OL--_.l.--_~_",--_-'-_...l-_-'-----'

600 800 1000 1200

Temperature (0C)
Figure 4 The variation of seismic velocity with temperature
Figure 3 An intersecting pair of migrated seismic sections of a basalt (modified from Murase and McBimey, 1973 l. Al-
from 9°40'!,/ on the East Pacific Rise. On the cross-axis profile. though the general behavior is correct. the detailed variation
COP 29, the magma chamber reflection appears as a narrow, of velocity between the solidus and liquidus as a function of
< I-km-wide, event at about 4.0 s two-way traveltime directly melt fraction is not well resolved.
beneath the axial summit caldera and the thinnest portion of
layer 2A. The base of layer 2A is interpreted as marking the
transition between the extrusive unit and the sheeted dike peratl;lreshows a similar knee to the velocity func-
complex. On the intersecting along-axis profile, COP 41, the tion (Marsh, 1981; Sinton and Detrick, 1992) and
magma chamber reflection is continuous along-axis for tens of
there may be a more nearly linear relationship be-
kilometers although reflection strength is variable due in part
to the wandering of the profile with respect to the narrow tween velocity and melt fraction over a larger su-
reflection. persolidus temperature range (Sato et al .. 1989;
Sato and Sacks, 1990). Murase and McBirney's
results show shear wave velocities decrease only
velocities of basalts with temperature and degree slightly as P-wave velocities are reduced and then
of partial melting is limited, particularly for pres- fall discontinuously to zero, suggesting that it is
sures approximate to mid-ocean ridges. Those not possible to have large changes in S-wave ve-
studies that do exist (Murase and McBirney, 1973; locity without also having large changes in P-
Christensen, 1979; Khiratov et al., 1983; Mangh- wave velocity. The combined results indicate that
nani et al., 1986) plus appropriately scaled a velocity anomaly on the order of 0.5 km . S' 1, a
(Henstock et al .. 1993) analogous data on the -10% reduction, requires only a small percentage
high-pressure melting of peridotite (Sato et al., of melt and that a fully molten magma chamber
1989) can be used to infer the temperature and should produce a large amplitude reflection. How-
melt fraction distribution within and around the ever, it is not possible, at present, to quantify pre-
magma lens. cisely such phrases as "nearly completely mol-
Over the temperature range from O°C to the ten" or "partially solidified" in terms of melt
solidus at around 1000°C, P-wave velocities are fraction and it may be possible for a magma to
reduced on the order of 0.5 km . s 1 (Fig. 4). have a crystal fraction of 25% or greater, not be
The velocity variation is probably approximately able to convect (Marsh, 1989), and still produce a
linear at subsolidus and immediately supersolidus large-amplitude reflection indicative of a nearly
temperatures and there is no discontinuous change completely molten magma chamber.
in velocity at the solidus. At greater temperatures If P-wave velocities are halved at an interface,
there is a knee in the velocity-temperature curve then the normal incidence reflection is phase-
with P-wave velocities decreasing rapidly above reversed and has an amplitude of 0.33 relative to
the knee, and at the liquidus P-wave velocities the incident wave. Estimates of normal incidence
have been approximately halved to less than 3 km . reflection coefficients from the northern EPR data
s- '. The variation of crystal fraction with tem- range up to 0.3-0.4 (Barth et al., 1987), while
 7. Magma Chamber Geometry and Crustal Genesis 147
the AMC reflection often appears to be phase- perton transform. Even at these "gaps," it is pos-
reversed relative to the seafloor reflection (e.g., sible that the ship wandered off the narrow magma
Fig. 3). Both observations are thus consistent with lens rather than the lens itself being absent. Given
the existence of a large negative velocity contrast such an observation, it is reasonable to think in
at an AMC. Estimates of normal incidence reflec- terms of a typical magma chamber structure for a
tion coefficients tend to be highly variable due to, fast-spreading ridge and to interpret seismic re-
among other factors, the sensitivity of recorded sults within such a framework rather than viewing
amplitudes to small bathymetric features that can each seismic measurement as potentially unique.
focus or defocus seismic energy. A more reliable Similar continuity of the magma lens is seen in
estimate of material properties at the top of the recent seismic experiments along the southern
magma chamber may be made from the amplitude East Pacific Rise. To a first approximation, crustal
versus offset behavior of the AMC reflection. accretion along the northern East Pacific Rise
Normal incidence reflection coefficients are sen- appears remarkably uniform both spatially and
sitive primarily to P-wave velocities. To a lesser temporally, particularly when compared to slow-
extent reflection coefficients are sensitive to den- spreading environments. The majority of well-
sity (the density decrease associated with melting constrained crustal thickness determinations for
is on the order 5-10%). However, the amplitude the East Pacific Rise cluster around 6 km with a
versus offset behavior of the reflection is strongly spread of ± I km (Chen, 1992), and along-strike
dependent on the difference in shear modulus variations in mantle Bouguer anomalies between
across an interface (Harding et al., 1989). A large 9° and l3°N, an indicator of crustal variations, are
decrease in shear modulus, as would be the case small, on the order of 10 mGal (Madsen et al.,
at the top of the magma chamber, causes a rapid 1990).
decrease in the reflection amplitude with offset
and the reflection can switch signs becoming a The Magma Lens is Thin: -50-200 m Thick
low-amplitude normal polarity arrival at larger At present no reflection has been identified in
offsets. On the other hand, a small decrease in the CDP data that can unambiguously be associ-
shear modulus, even if accompanied by a small ated with the bottom of a magma chamber, nor for
drop in P-wave velocities, results in a reflection that matter has a side-wall reflection. The absence
that has reversed polarity at all offsets and can in- of a bottom reflection can be explained by a
crease in amplitude at larger offsets. Amplitude gradual increase in the crystal content with depth
versus offset behavior of the AMC reflection in within the magma chamber and a concomitant
rise-axis ESPs and CDP data at 9°N falls into the gradual increase in seismic velocity, which would
former pattern (Vera et al., 1990) and has a very not produce a distinct reflection. Alternatively, if
rapid decrease with increasing offset while similar the magma chamber is a thin lens, the top and bot-
data from l3°N follows the latter pattern, leading tom reflection will coalesce into a single arrival
to the inference that the top of the magma cham- effectively masking the bottom of the chamber.
ber at 9°N is nearly completely molten while the Figure 5 plots the primary reflection response for
magma chamber at 13°N is partially solidified. a thin layer with equal but opposite polarity reflec-
tions at either boundary as the layer thickness
The Magma Lens Is Usually a Continuous varies using an appropriate source wavelet. For
Feature along the Axis of a Fast-Spreading small thicknesses, the reflection response has re-
Ridge duced amplitude and appears to have reversed
The 1985 regional reflection survey also pro- polarity. At larger separations, there is a construc-
vided the first convincing geophysical evidence tive interference or tuning of the two reflections,
that an axial magma chamber was a reasonably which increases the peak amplitude by about
continuous and steady-state feature beneath a fast- 50%, and even a simple classification between ei-
spreading ridge. A magma chamber reflection was ther normal or reversed polarity is difficult. Fi-
found along more than 60% of the rise-axis be- nally the top and bottom reflections separate into
tween 9° and l3°N with the major gaps occurring two distinct arrivals. Morton and Sleep (1985) re-
near major ridge discontinuities such as the Clip- ported evidence for closely spaced negative and
148 J. Phipps Morgan et al.

presence of diffractions is consistent with the

10
A abrupt cross-axis termination of a magma lens or
68m B chamber. Time migration of the stacked section
ESP 11 ITS with the appropriate velocities derived from re-
0.2
8
fraction data results in a collapsed reflection with
a width of around 0.7 km. Migration using stack-
-0.2 68m
ing velocities results in only a partial collapse of
46m the diffraction tails and a larger apparent magma
6 0 0.06 0.12 0.18 chamber. Stacking velocities over the rise-axis are
biased low by the axial bathymetry and the thick-

,~ ~v- :::
ening of layer 2A while stacking velocities in-
crease along the diffraction tails in a predictable
manner because of the offset reflection of the re-

2 ~\~
-l : \.. 15m
-0.2

o 0.06 0.12 0.18

flection midpoint from the edge of the magma
chamber (Kent et al.. 1993a). When the AMC re-
flection is narrow and centrally located beneath
\.J;~V the rise-axis, widths estimated from migrated sec-
o .A\'~{f\'..- 5.m _ tions are comparable to more careful estimates
V!.: V - made by forward modeling the diffraction hyper-
"
, bolas (Kent et al.• 1993a). It should be noted that
the width of the top of the low velocity zone as
0.06 0.12 0.18 o 0.06 0.12 0.18 determined by refraction and tomography results
Time (sec) Time (sec) exceeds that of the AMC, which may be explained
Figure 5 (A) Response of a thin magma layer with equal as truncation of the base of the magma lens
amplitude but opposite polarity top and bottom reflections as against the dipping roof of the chamber or as a
a function of separation between the top and bottom of layer
rapid lateral gradient in melt fraction. Subsidence
(solid line) for an airgun source wavelet. Dashed line is the
response of the top of the layer alone. Separations are based
of the sheeted dike complex combined with the
on a nominal velocity of 2.6 km . s I. At separations of about freezing or plating of gabbro on the roof of the
15 m the amplitude of the response is enhanced relative to a chamber should ensure a roof with dips on the or-
single interface by tuning of the layer, and the secondary posi- der of tens of meters, rather than hundreds of me-
tive pulse exceeds the primary negative pulse in amplitude. At
ters, over widths typical of the AMC reflection.
around 30 m positive and negative pulses have approximately
equal amplitudes and at greater separations the wavelets sepa-
Sometimes the Magma Lens Is Offset from the
rate in distinct arrivals. (B) Comparison of magma reflections
at different localities (solid line), with the predicted response
Ridge Axis-Then It Is Wider (2-4 km Wide)
of a thin-layer model (dashed line). Source wavelets in each and Slopes up toward the Ridge
case were estimated from seafloor reflection response, Re-
A strikingly different AMC reflection is pres-
sponse based upon a ID model that includes attenuation within
the upper crust.
ent in Fig. 7, which shows a stacked cross-axis
profile, CDP 33, shot about 35 km to the south of
the previously discussed line at 9°l9'N. The im-
positive polarity reflections in data from the Lau age of layer 2A is not as clear or continuous as in
Basin. Kent et al. (1990) demonstrate that a the previous example and the AMC reflection is
magma lens with a thickness between 10 and 50 m considerably more complex and is displaced to
will mask the bottom reflection and produce a cal- the west of the axis. Layer 2A deepens asymmet-
culated response that is comparable to the data at rically away from the rise-axis, being thicker to
9°N (cf. Fig. 5). the west of the axis than to the east. With such a
flat rise-axis profile the location of the layer 2A
The Magma Lens Is (Usually) Narrow:
-I km Wide
a
minimum thickness is more reliable indicator
than bathymetry of the location of the rise-axis.
When carefully stacked, cross-axis reflection The large variation in the thickness of layer 2A
profiles characteristically show diffraction tails above the AMC reflection is responsible for its
emanating from the AMC reflection (Fig. 6). The complex appearance in the stacked section. The
 7. Magma Chamber Geometry and Crustal Genesis 149

Figure 6 (A) Unmigrated version of COP 29 profile, cf. Figure 3. The magma chamber reflection has extended diffraction tails,
suggesting an abrupt cross-axis termination to the magma chamber reflector. (B) A finite difference simulation of the profile
including a narrow elliptical magma lens. 750 m wide and 50 m thick, reproduces the general character of the reflection response
including the extended diffraction tails. The increased thickness of the magma lens in the simulation compared to that estimated
for the profile extends the reflection response in the simulation.

low velocities within layer 2A cause the AMC re- sequence of the adjustment of magma supply
flection to be pulled up beneath the rise-axis and across the 9°03'N OSc. Away from the OSC the
pushed down west of the axis (Harding et al., magma supply from the mantle is concentrated di-
1993; Kent et al.,1993b). In this case, estimates rectly beneath the rise-axis. This configuration of
of AMC width and depth have been derived mantle supply and rift results in the prototypical
through the use of image rays after time migration cross-axis reflection profile with an approximately
(Hubral, 1977) and checked with forward model- I-km-wide AMC reflection centered beneath the
ing (Kent et al., 1993b). The resultant AMC re- rise-axis and Moho reflections that can be traced
flector lies at a nearly constant depth below the to within a few kilometers of the edge of the AMC
base of layer 2A and has a total width of 4.2 km. reflection.
The average dip on the reflector is estimated as
10° with one edge located directly beneath the Synthetic Experiments Can Reproduce These
rise-axis. The asymmetry and dip of the reflector Basic Magma Lens Features Quite Well
may be possibly explained as a melt migration The justification for describing the molten part
path; buoyant melt collects off-axis at the base of of the axial magma chamber as typically a narrow
the sheeted dike complex and then flows upward lens rests in part on its satisfying the constraints
and across-axis to erupt along the axial rift zone. imposed by the reflection data and in part on such
This interpretation of the AMC reflector suggests a geometry being physically simple to achieve, re-
a displacement of the underlying magma source quiring only that the magma is buoyantly, and
region relative to the rise-axis. The AMC reflec- usually stably, trapped against an impermeable
tion in CDP 33 is rare and is thought to be a con- lid. The consistency of the magma lens model
150 J. Phipps Morgan et al.
. - - - - - - - --- - -
Rise-axis
West

----
u
V
~
v

'E 4.0
Q)
>
o:l
b
»
o:l
~
0
S!:
f-;

5.0

A
,-.,
o<l)
C/J West Sea Floor Refl~ec::t:.;;io:!.n:..==:::& East
~:::,
'-'
<l)

....E
a3
>
~
b
>-.
~
S!:
0
I Forward Modeling Hyperbolae
~
E--- -5 -4 -3 -2 -1 o 2 3

2.0
Sea Floor--------...
2.5
,-., 3.0 Base of Layer 2A ....
E 3.5
C
...c: 4.0
0.
V
4.5
c 5.0
5.5
6.0
-5 -4 -3 -2 -I 0 1 2 3
C Along-track distance (km)
Figure 7 (A) Cross-axis profile, CDP 33, at 9°19'N. Profile is unmigrated to aid identification ofreflection arrivals. The magma
chamber reflection is displaced to the west of the axis and is considerably wider than the CDP 29 profile. The time separation of
the magma chamber reflection from the base of layer 2A is approximately constant, indicating that the complexity of the reflec-
tion is a consequence of propagation through a thickening extrusive layer. (8) Line drawing of the reflections in (A) plus the
travel time hyperbola resulting from forward modeling the magma chamber reflector as a series of point diffractors. (C) Magma
chamber reflector determined from ray theoretical depth migration. The reflector is 4 km wide with the eastern edge located
beneath the rise-axis. The reflector is located approximately 1.5 km beneath the rise-axis and has a dip of about 10°.

with the data can be demonstrated through a finite 0.75 km and a maximum thickness of 50 m. The
difference simulation of the cross-axis profile, response of the lens in the simulated section is
CDP 29, Fig. 6B. The magma chamber in the comparable to that of the AMC in the stacked
simulation was an elliptical lens with a width of data, with diffractions originating from the edge
 7. Magma Chamber Geometry and Crustal Genesis 151
of the lens. There is no obvious bottom response pace above the magma chamber (Haymon et al.,
from the lens although the finite thickness has ex- 1991a). Thus, the layer 2A reflection is interpreted
tended the reflection signature. as marking the top of the sheeted dike complex;
i.e., the dikes that feed eruptive activity lie within
Other Seismic Constraints on Crustal Accretion 100-200 m of the seafloor at the rise-axis. The
at a Fast-Spreading Ridge symmetric deepening of the layer 2A reflection
away from the rise-axis is interpreted as the result
Crustal Accretion and Moho Formation of the progressive burial of the sheeted dike com-
Occur in a Narrow Zone about the Ridge Axis plex by successive lava flows. Near the rise-axis,
Several lines of evidence support the hypothe- seismic velocities within the upper part of layer
sis that the magma lens is a fairly stable feature 2A lie between 2.5 and 3.0 km . s - I , velocities that
and that crustal accretion is occurring within a can reasonably be explained only by large porosi-
narrow region indeed. A distinct Moho reflection ties within the upper crust. The transition to ve-
can be traced to within -2 km of the ridge axis locities in excess of 5.5 km . s- 1 below the base of
(Detrick et al., 1987); i.e., a distinct seismic re- layer 2A, values more nearly comparable to those
flection Moho is present immediately after one of unfractured basalt, can be accounted for solely
moves away from the near-axis ranges where a by a downward decrease in porosity, and it is con-
noticeable shallow AMC diffraction tail obscures ceivable that the base of layer 2A is simply a po-
any deeper, weaker reflections. As well as being rosity horizon that deepens as the result of tec-
evidence for a region of (crustal) magma supply tonic fracturing and is not structurally controlled.
directly beneath the axis, the proximity of Moho However, recent submersible dives at Hess Deep,
reflections to the axis suggests that the basic struc- on tectonically exposed sections through normal
ture of the lower crust is formed close to the rise- East Pacific crust, found a total extrusive section
axis. Shallow crustal extrusion also appears to on the order of 400-600 m comparable to the off-
occur within a narrow zone directly above the axis thickness of layer 2A. In addition, the off-axis
magma lens. Seismic studies along the northern thickening of layer 2A observed in sections such
East Pacific Rise have found evidence of a sys- as CDP 33 (Fig. 7) does not occur at the expense
tematic variation of upper crustal structure within of the rest of layer 2 above the axial magma cham-
1-2 km of the axis. The tomography results from ber, as would be the case for a deepening of a po-
9°N found an axial positive velocity anomaly rosity horizon within a constant thickness crustal
within the upper 1 km of crust. By correlating layer, but rather indicates an overall thickening of
ESP-derived velocity models for the upper crust layer 2. The 2- to 4-km width of the neovolcanic
with CDP reflection images, it can be demon- zone inferred from layer 2A thickening is compa-
strated (Harding et al., 1993) that the magnitude rable in magnitude to earlier estimates based on
of the velocity anomaly can be accounted for by the transition width of seafloor magnetic linea-
the deepening of the surficial low velocity layer, tions (Sempere et al., 1988).
layer 2A, seen in the CDP profiles (Fig. 6). It has
been proposed that the base of layer 2A marks the The Magma Lens Is Underlain by a Broader
top of the sheeted dike complex and thus that the Low Velocity Zone ofHot Rock and Possible
thickening of layer 2A is a consequence of thick- Small Amounts ofMelt «3 -5%)
ening of the extrusive section to a full thickness of A largely solidified cumulate mush has only a
-400-600 m within 1-2 km of the axis (Harding small impedance contrast relative to the host rocks
et al., 1993; Christensen et al., 1992). The layer and is essentially invisible to reflection methods.
2A reflection reaches its minimum depth at the However, such a chamber can be detected by re-
rise-axis directly above the AMC reflection and fraction and tomography methods that are sensi-
directly below the axial summit graben or axial tive to seismic velocity variations. The divisions
summit caldera. The axial summit caldera marks between refraction and tomography are somewhat
the loci of the most recent volcanic eruptions and arbitrary but in the context of studies of the East
is believed to a syn- or post-eruptive feature Pacific Rise most seismic refraction lines have
resulting from the collapse of the volcanic cara- been shot parallel to the rise-axis to a single re-
 152 J. Phipps Morgan et at.

ceiver and have been used to determine a one- of the axial low velocity zone is wider than the
dimensional depth-dependent profile. In this con- AMC reflection at 9°N and also that a low velocity
text, a low-velocity zone is a decrease in seismic zone can exist where there is no distinct AMC re-
velocity with depth, which creates a shadow zone flection visible in the CDP data (Harding et al.,
in the arrivals. Seismic tomography experiments 1989).
use the travel times from multiple shots to multiple Although the crude dimensions are known,
receivers to create two- and three-dimensional the detailed structure of the low velocity zone is
velocity models of the rise-axis. Tomographic poorly constrained by the seismic data. The hori-
results are usually expressed in terms of veloc- zontal and vertical resolution of the tomography
ity anomalies and a low velocity zone is a nega- data is at its best on the order of I km for resolving
tive velocity anomaly relative to a lD reference shallow crustal structure and much poorer than
model. Vertical velocity gradients within layer 3 this in the lower crust (Caress et al., 1992).
of the oceanic crust are sufficiently small, on the Sharply delineated structures will be smeared out
order of 0.1 (km . s -I) . km - 1, that the refraction- in the tomographic images. For example, a
derived and tomographic axial low-velocity zones magma lens is not resolved by the tomographic
roughly coincide. images although the largest low velocity anomaly
The axial low velocity zone determined by is centered at the location of the magma lens. The
tomographic results from 9°N (Toomey et al., ESP data have good vertical resolution down to
1990) and l2°N (McClain et al., 1985; Burnett the top of the low velocity zone lid and can re-
et al., 1989; Caress et al., 1992) as well as the. solve the depth of the lid to within a few hundred
expanding spread profile (ESP) refraction results meters. However, the velocities and depths of
from 9° and l3°N (Harding et al., 1989; Vera structures within and beneath the low velocity
et al., 1990) is 4-5 km wide at its top, broadening zone are poorly constrained. No rays pass through
to 8-10 km at its base. Maximum velocity pertur- the lowest velocity region in tomographic experi-
bations are on the order of 0.5 -I km . s 1, corre- ments, which is essentially why the lowest veloci-
sponding to absolute velocities of around 6 km . ties are not imaged. (Were it a high velocity region
s- 1, and are confined principally to the upper part like a salt dome, then the tomographic image of
of the crust above a depth of 4 km. In the ESP the high velocity structure would be much better
data, wide-angle AMC reflections are asymptotic because seismic ray paths preferentially sample a
to upper crustal refractions at the ranges just be- high velocity anomaly.) Bounds may be placed on
fore the start of the shadow zone (Fig. 8), showing the vertically integrated velocity anomaly within
clearly that the AMC reflection lies at the top of the low velocity zone using the arrival times and
the axial low velocity zone and corroborating the position of wide-angle arrivals from the Moho,
interpretation of the AMC reflection as being from PmP, in the ESP data (Harding et al., 1989). These
the top of a molten magma body. The ESP results bounds are consistent with the tomographic re-
also demonstrate that the relatively flat-lying top sults, indicating that a large 0.5- to l-km . s- 1 ve-

. 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0

Range (km)
Figure 8 An expanding spread profile at I TS on the East Pacific Rise. The solid line marks the reflection from the axial magma
lens, which can be traced from a vertically incident reflection out to a distance of 7.5 km from the axis where it asymptotically
approaches the upper crustal refraction arrival that emerges from the seafloor reflection at 3.5 km. There is a shadow zone at
ranges greater than 7.5 km that indicates that a low velocity zone underlies the axial magma lens.
 7. Magma Chamber Geometry and Crustal Genesis 153
locity anomaly can extend only to mid-crustal EPR, the axial high is dominated by a narrow,
depths and that the vertically integrated strength triangular ridge that is 1-2 km wide and 100-
of the anomaly decreases with distance from the 200 m high (Fig. lOb). One explanation for such
axis. An additional difficulty with proposing a de- ridges at intermediate spreading rates is that they
tailed structure for the low velocity zone is that, are constructional features that are built during
as discussed earlier, the relationship between seis- periods of increased magmatic activity (Kappel
mic velocity and partial melt fraction is relatively and Ryan, 1986). However, a constructional ori-
poorly known even for small hand samples at gin for the ridge at \3°N is precluded by the
the appropriate pressures. Extrapolation of hand uniform thinness of layer 2A along-axis and its
sample measurements to crustal scales suggests increased thickness off-axis (Kappus, 199 I). In-
that on average the low velocity zone is largely stead, the seismic results support the conjecture
solidified with only a small percentage of partial that the axial high is buoyantly supported and the
melt although at elevated temperatures. However, relatively modest thickness increase of layer 2A,
isolated conduits, dikes, and sills that are of pet- by 60%, is consistent with the notion that the tri-
rologic importance could be contained within the angular cross section of the \3°N segment is in-
average structure but not be resolvable from the dicative of relative magmatic starvation (Macdon-
seismic data. A I-km . s - I velocity anomaly at ald et 0/., 1984; Macdonald and Fox, 1988).
mid-crustal depths implies that melt, if uniformly Assuming relatively low densities for layer 2A
distributed over this depth range, must fill less of 2.35 Mg . m:" (Stevenson, 1992), and "erupt-
than 3% of the rock volume at these depths (Ca- able" densities of -2.6 Mg· m 3 for the magma,
ress et 0/.. 1992). the triangular neovolcanic zone can be Airy com-
The tomography data set at 9°30'N has also pensated near mid-crustal depths by a narrowly
been used to obtain models of crustal P-wave at- confined magma chamber. As an order of magni-
tenuation (Wilcock et 0/., 1992). Spectral esti- tude estimate, this value is consistent with the
mates of the attenuation of P-waveforms are in- tomographic images that show the main low ve-
verted for the reciprocal of the quality factor Q (a locity anomaly extending to mid-crustal depths
measure of anelasticity or elastic wave attenu- and indicates that the axial topography is sup-
ation), assuming the velocity structure and ray ported by a thicker buoyant and low viscosity re-
paths obtained by delay time tomography (Toomey gion than just the thin magma lens. Isostatic sup-
et 0/., 1990). Q is not imaged within the magma port is consistent with the axial high that exists
lens, since no rays pass through this region. Below within 1-2 km of the axis (Kappus et al., 1992).
the magma lens, Q values are 20-50 compared Crustal density variations may also be a much
with values of ~500 at similar depths well off- larger source of compensation for the broader re-
axis. The low on-axis values are very similar to gion of anomalous elevation within - 15 km of
Qp values inferred from torsional oscillation ex- the ridge axis than inferred by recent studies by
periments on gabbros at solidus temperatures Wilson (I992) and Wang and Cochran (1993);
and seismic frequencies (Kampfmann and Berek- these results are more consistent with the large
herner, 1985). Thus, the attenuation models also component of crustal compensation of axial relief
suggest that the melt fraction at mid- and lower found in Madsen et 01. (1990). A simple estimate
crustal depths beneath the melt lens is no more of these crustal effects considers the isostatic ef-
than a few percent. fects of two crustal columns, one on-axis and the
other located - I5 km off-axis. The isostatic effect
of crustal cooling due to hydrothermal circulation
A "Hot Zone" beneath the Magma Lens Is can be estimated from the thermal model for a
Needed to Support the Axia/ Topography at fast-spreading ridge shown in Fig. l6A. This ther-
Fast-Spreading Ridges mal structure would predict the axial crust to be
If one accepts that the base of layer 2A marks 4.5 km of -1200°C crust capped by - 1.5 km of
the top of the sheeted dike complex, then the be- - 300°C. Fifteen kilometers off-axis, the litho-
havior of this horizon may be used to infer struc- sphere is a -7-km-high hydrothermally cooled
tural information about the nature of the axial column of crust and mantle with an average tem-
ridge along the East Pacific Rise. At \3°N on the perature of -300°C. This results, assuming a ther-
154 J. Phipps Morgan et al.
mal expansivity a = 3 X 1O~ K I, in a thermal 9°35'N and 9°l7'N DEVALs. North of 9°35'N,
subsidence of -150 m for this 5.5-km-thick litho- widths are less than 0.7 km, from 9°35'N to
sphere column of crust and mantle that cools by 9° l7'N widths vary between 1.0 and 1.2 km,
900°C within 10 km of the ridge axis, a number whereas south of 9°17'N to the 9°03'N OSC
that is -120 m larger than that used in the Wilson widths exceed 3 km. South of the 9°03'N OSC the
(1992) and Wang and Cochran (1993) studies. width of the reflector is reduced once again to be-
Similarly, a 6-km-high axial crust with an average tween 0.7 and 1.3 km. The increased widths co-
of 3% retained melt will produce an isostatic up- incide with increasing displacement of the AMC
lift of -50 m, and a 100-m-thick magma lens will reflection to the west and, as discussed previously,
lead to an axial isostatic uplift of -25 m. These are believed to be due to accommodation of the
isostatic effects account for roughly 225 of the mantle supply to the 9°03'N OSc. South of the
-350-400 m of relief associated with the near- 9°35'N DEVAL, the westward displacement of
axis bathymetric high, leaving at most -175m the low velocity zone is comparable to that of
of relief to be produced by density variation (due AMC reflection. Thus at 9°35'N there is evidence
to the presence of melt?) beneath the Moho, for a segment boundary in the low velocity zone,
i.e., leaving only about one-half the anomalous in the AMC reflection, and in the axial mor-
topographic high assumed in Wang and Cochran phology. However, the 9°N section will, in detail,
(1993) and Wilson (1992) to infer that a deep low- be more segmented than the MCS results indicate
density root is needed beneath a fast-spreading as the MCS data do not, for example, resolve a
ridge. This important question clearly merits fur- 'segment boundary at 9°28'N.
ther investigation. The seismic results do not support the idea of a
single focus of magma supply for the 9°N seg-
Delivery of Melt from the Mantle ment. Instead they provide evidence for segmen-
to Supply the Magma Lens tation of the axial magma chamber on length
scales of 10- 30 km, and, by inference, of the un-
Geophysical information on the nature of the derlying magma supply. Additionally, there is no
mantle delivery of magma to the crust is at present evidence for a decrease in magma supply toward
extremely limited. There is some evidence from the ends of the segment as would be predicted by
time delay and Q tomography measurements of a single injection model; to the contrary, gravity
anomalous structure at the base of the crust be- measurements, seismic estimates of crustal and
neath the rise-axis (Solomon and Toomey, 1992; extrusive thicknesses, and the AMC width es-
Wilcock et al.. 1992). Madsen et al. (1984) used timates indicate a relative increase in magma
compensation depths of gravity anomalies to ar- supply near the 9°03'N OSc. The extent and im-
gue for a magma chamber at Moho depths, while portance of along-axis magma migration at fast-
Garmany (1989) has suggested that certain con- spreading ridges in smoothing out fluctuations in
verted shear phases are due to isolated pockets of crustal structure are still unresolved. However, the
melt residing at Moho depths. However, none of petrologic boundaries at certain DEVALs as well
these measurements can shed light on the geome- as the discrete jumps in AMC width indicate that
try of magma delivery to the crust. along-axis mixing of magma is incomplete.
One means of inferring information about
magma delivery is from the segmentation of the Spreading Rate Dependence of the Magma Lens:
crustal magma chamber along-axis. At 9°N, infor- Seismic and Gravity Constraints
mation on segmentation comes from the tomog-
raphy results (Toomey et al., 1990) and the com- A similar set of seismic experiments have been
paratively densely spaced set of MCS profiles performed along the southern East Pacific Rise
(Kent et al.. 1993a). The low velocity anomaly where spreading rates are close to the upper end
imaged by the tomography experiment narrows of the spectrum. Preliminary results from these
and thins toward a pair of ridge axis kinks or experiments (Detrick et al., 1993; Kent et al.,
DEVALs at 9°28'N and 9°35'N where the local 1993c) support the magma chamber model devel-
ridge trend changes abruptly. Magma lens width oped from the northern East Pacific Rise experi-
estimates show distinct jumps in widths at the ment and demonstrate that the size of the magma
 7. Magma Chamber Geometry and Crustal Genesis 155
chamber is not strongly dependent on spreading which is spreading at a slower rate. The interme-
rate. Once the transition to an essentially two- diate-rate Juan de Fuca and Lau Basin ridges are
dimensional accretion pattern has been made at spreading at lower rates and have deeper magma
spreading rates between 25 and 35 mm . yr: 1 lenses than the northern East Pacific Rise. A
(half-rate), the basic magma chamber structure magma lens has yet to be observed on any slow-
appears to be set. spreading ridge. There is a similar spreading rate
There is a small spreading rate dependence to dependence of the thickness of layer 2 supporting
the depth of the magma lens where it has so far the observed on-axis spreading rate dependence
been observed as shown in Fig. 9. The magma (Purdy et al., 1992). However, at slow spreading
lens is slightly shallower at the southern East Pa- rates, the thickness of layer 2 appears to plateau at
cific Rise than the northern East Pacific Rise, a maximum thickness of roughly 3 km or ~50%
of the crustal thickness (Purdy et al., 1992). The
thickness of the axial seismogenic zone, or brittle
lithosphere, is also a sensitive function of spread-
Magma Lens Depth and Axial Lithosphere Thickness
ing rate as shown in Fig. 9.
0
....EPR As noted above, seismic experiments to date
have found no direct evidence for an axial magma
2
III :; EPR
chamber beneath a slow-spreading ridge. This
4 1 ! 1+• •
+ nonobservation suggests that magmatic emplace-
E 6 ment occurs in a much more episodic fashion
c.
..c
0.
r • Seismic Reflector
6km thick crust
at slow-spreading ridges, in agreement with pet-
rologic observations that argue for small, inter-
<> 8
Q mittent magma bodies at these spreading centers
T cutoff = 600°C
10 (e.g., Natland, 1980; Sinton and Detrick, 1992).
-----... 750°C Nu = 8
I
Slow-spreading ridges typically differ signifi-
1200°C Nu= 8
12
1200°C Nu =12 cantly from fast-spreading ridges in their axial
no lens - - morphology-they have a 15- to 30-km-wide, I-
14
0 10 20 30 40 50 60 70 80 to 2-km-deep median valley as their axial expres-
Half-spreading rate (rnm- yr') sion instead of a 1- to 2-km-broad, several hun-
Figure 9 Depth to the top of the magma lens as a function of dred meter high axial high as shown by character-
spreading rate, all other parameters being held constant. Solid istic axial bathymetry in Fig. 10. In addition, the
squares (and associated uncertainties from Purdy et 01., 1992) axial relief along the axis of a slow-spreading
show multichannel seismic observations of the depth of the ridge is much more variable than at a fast-
magma lens along intermediate and fast-spreading ridges.
Curves show results from a suite of numerical experiments
spreading ridge. Figure 11 shows profiles of axial
with Nu = 8 and Nu = 12 and T,",off = 600°C. For these hy- bathymetry and axial Bouguer gravity as a func-
drothermal cooling parameters a steady-state magma lens (ex- tion of spreading rate. The axial relief along a fast-
isting when a 1200°C isotherm lies within the crust) can exist spreading ridge is extremely uniform in contrast
only within the crust at half-spreading rates greater than about to the ~ l-km variation in median valley depths
20 (or 30) mm . yr- t • A well-developed shallow magma lens
exists only for spreading rates greater than 30 mm . yr- t •
along a single segment of a slow-spreading ridge.
There is good agreement between model predictions of the The axial mantle Bouguer anomaly can be used to
depth dependence of the magma lens with spreading rate and infer along-axis variations in crustal thickness.
multichannel seismic observations. Depths to the 750°C iso- The along-axis mantle Bouguer gravity profiles in
therm are also shown for a suite of numerical experiments with Fig. 10 suggest ~ 3 km of along-axis variation in
Nu = 8 and T,",off= 600°C. The model 750°C isotherm corre-
lates well with the spreading rate'dependence inferred from
crustal thickness at a typical segment of a slow-
axial earthquake centroid depths (shown by small dots and spreading ridge in contrast to at most ! km of
associated bars representing inferred total rupture depth) of along-axis crustal thickness variation where an
Huang and Solomon's (1988) teleseismic study. Triangles are axial magma chamber is present (e.g., Lin and
microearthquake focal depths beneath the axis of the northern Phipps Morgan, 1992). This dramatic change in
Mid-Atlantic Ridge (Toomey et 01., 1988; Kong et 01.• 1992).
There is good agreement between model isotherms and the
along-axis crustal variation could reflect a funda-
spreading rate dependence of the depth of the seismically ob- mental change in upwelling between fast- and
served brittle-ductile transition. slow-spreading ridges, with fast-spreading ridges
 156 J. Phipps Morgan et al.

Figure 10 (A) Regional axial high bathymetry of the relatively fast spreading East Pacific Rise. This section is near 9°N.
Selected profiles are shown by solid lines. Note the changes in axial segmentation between this typical fast- (EPR) and slow-
(MAR, Figure JOe) spreading ridge.
continues
 7. Magma Chamber Geometry and Crustal Genesis 157
Profile at 9° 40' N Profile at 9° 19' N
B
2.6 W 2.6 W E
E E
c- 2.8 c- 2.8
.c .c
E.. E..
3.0 3.0
'"
Ci c'"
3.2 3.2
-15 -10 -5 0 5 10 -10 -5 0 5 10 IS
Distance from rise-axis (km) Distance from rise-axis (km)

Figure 10 continues (B) Axial bathymetry across the EPR profiles noted in Figure lOA. Note the changes in axial relief
between these typical fast- (EPR) and slow- (MAR, Figure IOC) spreading ridges.

Figure 10 continues (C) Regional axial bathymetry and across-axis bathymetric profiles for the relatively slow-spreading Mid-
Atlantic Ridge, at the 29°N segments. The median valley bathymetry is illustrated. Selected profile is shown by a solid line.
 158 J. Phipps Morgan et al.
Profile at 29 0 1 2 ' N spreading ridges may be fundamentally colder ac-
D LOr-::-:---------''-----------~
w E cretion sites than fast-spreading ridges.
E 2.0
Co
..c: 3.0
Closure on Observational Constraints on the
15..
<U
Cl 40 Structure of an Axial Magma Chamber

-30 -25 -20 -15 -10 -5 0 5 10 15 20 25 Although many of the details of magma chamber
Distance from rise-axis (krn) structure have not been elucidated by current seis-
Figure 10 continued. (D) Axial bathymetry across the pro- mic experiments. it is important to recognize that
file noted in Figure IOC. these models are fundamentally different from
earlier models that proposed a large, well-mixed,
molten magma chamber that was 4-5 km deep
and extended throughout the lower crust. Even a
having a 2D form of upwelling and melting while narrow molten magma chamber with a width on
slow-spreading ridges are underlain by diapiric the order of the AMC reflection is not supported
3D upwelling points along-axis (Parmentier and by the seismic data. Similarly, the low amplitude
Phipps Morgan, 1990; Lin and Phipps Morgan, of the mantle Bouguer anomal y, -10 mGal, at the
1992). However, these observations are also com- axis of the EPR is incompatible with a large mol-
patible with three-dimensional upwelling and . ten magma body, although a narrow magma body
melting at both slow- and fast-spreading ridges would be permitted by the gravity data (Madsen
with a magma lens acting to smooth along-axis et al.; 1990; Stevenson, 1992).The basic 3D inter-
crustal thickness variations where it is present pretation of the previous colocated seismic studies
(Phipps Morgan. 1991; Lin and Phipps Morgan, is shown in Fig. 12 (after Kent et al .. 1993b). The
1992). seismic experiments in the 1980s have proposed
One consequence of the type of fast-spreading this picture as a challenge to the theoretical and
axial magma chamber inferred from seismic stud- ophiolite-based models, which presumed a large
ies is that the low-viscosity melt lens and hot crustal magma body. In the next sections we pres-
crustal rocks beneath the lens will be an extremely ent an emerging theoretical paradigm that may be
efficient zone for along-axis crustal flow. (This able to explain successfully this apparent discrep-
will certainly be true if the magma lens is truly ancy and lead to a picture of crustal accretion in
"molten." Marsh (1989) suggests that a -55% striking concordance with seismic constraints.
crystal fraction is the rheological equivalent of a
"viscous solid." This type of strong Marsh mush
may still be a good seismic reflector. It is still al- A Theoretical Model for Crustal Genesis
most certain to be the lowest viscosity region in
the axial crustal section.) Thus plate extension due Introduction
to plate spreading is likely to force significant
along-axis flow if crustal emplacement from the In this section we construct a thermal and me-
mantle is focused at several sites along a spread- chanical model for the genesis of oceanic crust
ing segment. This flow will preferentially occur in that we feel is a good candidate to integrate the
the magma lens if a molten zone is a persistent above observations into a coherent synthesis. Its
feature. If not, this along-axis flow is likely to be most important conceptual ingredient is that it is
concentrated in the '.'hot mush" zone beneath the the interplay between magmatic crustal injection
magma lens. In contrast. where no magma lens and hydrothermal cooling that is responsible for
and associated basal mush zone is present, crustal the presence of a quasi-steady-state magma lens
rocks will be much colder and stronger so that they beneath a fast-spreading -ridge and the dramatic
can support large lateral variations in axial crustal change in axial morphology between a fast- and
structure. This inference suggests that the lack slow-spreading ridge axis (because axial mor-
of an observed axial magma chamber at slow- phology directly reflects the lithospheric strength
spreading ridges is not mere bad luck-slow- or thickness across a ridge axis that is quite sen-
 7. Magma Chamber Geometry and Crustal Genesis 159

East Pacific Rise

-I~r ~.
,~-,- _n on' ~ r ~~ A .~~ j2200
___ . L I j f- I L I 3000
9° N 9° 30' 12° 30' 13° N 13° 30'

Cocos-Nazca Ridge
or -
... ~.,.:-----.----,--,,-j
~
· .."::-1.." < >
,,-,(~
2800

I
-20 r .'
] 3600
r-=----- ~ I I 1
96° W 95° W

Mid-Atlantic Ridge
17--·,------ .. ,Ie'r-. ... \ ", -
,/\
--1

~ 3000
"';"'
, ,
'

,--,si

"
... \ I \ I I

, \'
:'~ 3800 -;::
, \'-\,'\ \/ \," -l
-20 I'
\,,'
~ ~
3~S . ~01 ~
c-,
• 'l3000 ~
-t

-'--'\ 3800

j
'-::-'::-::-:-::-:---::':-:-:-:---=2--=3~0 30' N

I
-20~
r-_ .1 _ _ .
30° N

Figure 11 Along-axis profiles of seafloor depth (dotted lines) and mantle Bouguer anomalies (solid lines) for ridge segments
spreading at fast (EPR). intermediate (Cocos-Nazca), and slow (MAR) rates. Arrows at the top show locations of transform,
nontransform. and overlapping spreading center ridge axis offsets. Zero level for mantle Bouguer anomalies is arbitrary. Mantle
Bouguer anomalies were generated by subtracting from the free-air anomaly the attraction of seafloor topography and the attrac-
tion of relief on the crust-mantle interface. The crust was assumed to be 6 km thick. Densities of 1.03, 2.7, and 3.3 Mg . m'
were assumed for sea water, crust, and mantle, respectively. The resulting mantle Bouguer anomaly predominantly reflects along-
axis variations in crustal thickness and density but also reflects along-axis variations in mantle density. All calculations use
gravity reduction techniques developed at Brown University (cf. Kuo and Forsyth, 1988).

sitive to this thermal balance). The basic model velocity structure and potential episodicity, and
geometry is shown in Figs. lC and 13. The result- Quick and Denlinger (1993) explore the petro-
ing magma injection and crustal flow structure logic implications of this form of crustal genesis.
that we envision has been the subject of at least This emerging paradigm is extremely similar to
three recent efforts, all apparently inspired by field the qualitative crustal accretion scenario sketched
trips and discussion at the 1989 Oman Ophiolite by Sleep (1975, 1978) and also similar to the sub-
Conference. We follow in this chapter the treat- siding magma chamber floor model of Dewey and
ment of Phipps Morgan and Chen (1993). Hen- Kidd (1977) and Browning (1984). The common
stock et al. (1993) explore the effects on seismic thread in these studies is the realization that crustal
 160 J. Phipps Morgan et al.
-100 m
... ~

Layer 28, -1.4 km

I
L

Figure 12 A schematic cross section of the proposed melt-lens model for crustal accretion at a fast-spreading ridge axis. The
upper crust is partially stripped away to show segmentation of the magma lens near a small DEVAL "kink" in the spreading
axis. Off-axis thickening of the extrusive section occurs within the first few kilometers of the ridge axis. The low velocity zone,
corresponding to a region of elevated temperatures with at most a few percent partial melt, originates at the base of the magma
lens and extends through the upper half of the gabbro section. Crustal strain of cumulate gabbro formed in the magma lens leads
to the development of gabbro layering and fabric, which becomes progressively more intense near the Moho. Schematic crustal
layering is shown by "isochron" lines of similar crustal age. and the crustal flow pattern is shown by the flow lines with embodied
arrowheads.

subsolidus flow away from a small, shallow level cyclicity without an apparent depth dependence
magma lens is potentially able to generate the gab- (Browning, 1984). There should be no depth de-
bro "layering" seen in the Oman ophiolite- pendence if all the gabbro section is emplaced
cumulate freezing at the sloping sides of a crustal- through the same magma lens (Browning, 1984).)
thickness-scale magma body is not necessary to Phipps Morgan and Chen (1993) propose that a
produce this general layering pattern. (This flow fairly delicate balance between magma input and
pattern also provides an appealing explanation for hydrothermal cooling at the ridge axis leads to the
the observation that cryptic geochemical variation observed spreading rate dependence of the depth
within the gabbro section has a maximal 200-m of the magma lens. To consider the time-averaged
 7. Magma Chamber Geometry and Crustal Genesis 161
Crustal Flow at a Fast Spreading Ridge

u
D Hydrothermal cooling --. T=O
seafloor

crust

, mantle
.

stress free at base of computational region

Figure 13 Geometry of the theoretical model.

thermal structure at a ridge it is useful to parame- Henstock et al. (1993) suggest that neutral
terize these processes in a reasonably simple man- buoyancy is responsible for the fact that the ob-
ner, which we next describe. served magma lens is at roughly a constant depth
where magmas fractionate until they are light
Magma Ascent and Emplacement enough to reach the surface. However, a recent as-
sessment of this hypothesis by Hooft and Detrick
Magma ascent within the oceanic crust is prob- (1993) suggests that the neutral buoyancy level is
ably controlled by two complementary processes: at most 500 m deep at the East Pacific Rise, in
(a) Magma will ascend only if buoyancy forces contradiction with the observed ~ 1500-m depth
(or, if large enough, viscous pressure gradients) of the axial magma lens. In contrast, Phipps Mor-
cause it to; thus magmas may rest in their ascent gan and Chen (1993) were stuck by the apparent
for a while at zones of neutral buoyancy until they spreading rate dependence of the depth of the
have fractionated and reacted to the point where magma lens, which suggested to them that magma
the magma's density is less than that of surround- freezing may be the process that more strongly
ing host rock (this idea has been explored in limits the depth of magma ascent at a fast-
particular by Ryan (1987, and this volume). . spreading ridge. Here we initially consider only
(b) Magma can also be halted in its ascent when it the effects of magma freezing in limiting melt as-
reaches a freezing horizon, where the dilational cent and ignore density effects on lower crustal
volume change associated with magma freezing melt transport, reaction, and segregation. Later we
leads to viscous stresses that favor magma pond- lightly touch on potential observational tests of
ing within roughly one viscous "compaction whether magma freezing or neutral buoyancy play
length" of the freezing horizon (term coined by the more important role in limiting magma ascent.
McKenzie (1984». This idea has been recently Since magma freezing must 'occur (by definition)
proposed in a particularly simple form by Sparks within the sheeted dike and extrusive sections of
and Parmentier (1991, and this volume) to explain oceanic crust, these accretion processes, unlike a
melt focusing to a narrow neovolcanic zone at a magma lens, must be fundamentally transient in
spreading center. nature. To model the steady-state thermal impact
162 J. Phipps Morgan et al.

of crustal accretion via dike injection and pillow mechanically as a strong, viscous fluid-this ex-
flows we treat all crustal accretion in rocks below treme physical simplification must be improved if
a magma "solidus" to occur in a narrow 250-m- we wish to model the chemical evolution of the
wide dike-like region centered about the ridge crust associated with magma fractionation and cu-
axis. Within this region magma emplacement mulate segregation. For the same reason we do not
rates are taken to be equal to the spreading rate address convection and magma fractionation pro-
divided by the width of the diking region-i.e., all cesses within the magma lens. See Quick and
shallow crust is modeled to be emplaced within Denlinger (1993) for an analysis of some of the
this region. Although correctly treating heat injec- petrologic consequences of crustal genesis by pre-
tion within the sheeted dike section, this approxi- dominant melt injection within a small magma
mation does somewhat overemphasize the impor- lens. The thermal structure at a spreading center is
tance of magma heat injection within the pillow predominantly influenced by two factors in this
section. Since pillow flows are extruded on the model: (a) the depth and injection rate within a
seafloor, burying previous flows, they rapidly potential steady-state magma lens; (b) the effi-
cool to the ambient seawater temperature and ciency of hydrothermal circulation in removing
essentially advect this cold boundary layer down- heat through rocks that are cool enough to permit
ward with subsequent pillow burial and subsi- cracking and hydrothermal heat transport.
dence as discussed previously. In addition, seis-
mic observations discussed previously suggest Hydrothermal Cooling
that the thickening of the pillow layer occurs
within roughly -1- 2 km of the axis, instead of The primary remaining uncertainty in these nu-
within the central 250 m. The neglect of these ef- merical experiments is how to parameterize ap-
fects implies that this model will tend to underes- propriately the form and magnitude of the effects
timate the depth of the magma lens isotherm, by of hydrothermal circulation on shaping heat trans-
an amount that is no more than the thickness of port within the crust. We choose the formulation
the pillow section above the magma lens, i.e., developed in Phipps Morgan et al. (1987) which
-200 m (e.g., Fig. 6). uses the results of Combamous and Bories (1975)
The rest of the oceanic crust is modeled to be and Combarnous (1978) to treat hydrothermal
emplaced as a steady-state magma lens directly heat transport as an enhanced thermal conductiv-
beneath the solidus freezing horizon. We take the ity within the temperature and depth range where
-l-km-wide, -250-m-thick prismatic shape in- hydrothermal activity occurs. The enhanced con-
ferred from recent EPR seismic studies discussed ductivity is parameterized by the Nusselt number
previously to be a kinematic constraint on the Nu, which is defined as the ratio of hydrothermal
shape of this magma lens. Once we have this heat transport within a permeable layer to heat
shape, the steady-state emplacement rate is de- transport by heat conduction alone. Rock that is
termined by the constraint that this lens supply either at temperatures greater than 600°C or at a
all crust not emplaced through diking/extrusion depth greater than 6 km is assumed to be imper-
above the magma lens. The depth of this lens is meable. Since water as hot as 400°C discharges
controlled by the depth of the magma solidus from vents on the seafloor, this is a minimum
(here taken to be 1200°C) determined from a self- value of the maximum temperature through which
consistent thermal structure for a spreading cen- water must circulate. Phipps Morgan et al. (1987)
ter. Thus the lens will cease to exist in this model and Sleep (1991) use a value of 400-450°C for
if the steady-state thermal structure places this this cutoff temperature, while Morton and Sleep
solidus isotherm beneath the crust. In addition, the (1985) and Wilson et al. (1988) prefer a hotter
injection rate within the lens will diminish as the 600°C cutoff for hydrothermal circulation. Mor-
lens moves deeper into the crust, since magma in- ton and Sleep (1985) suggest that the limit tem-
jection into the lens supplies magma only for perature should be at 'least that obtained by
crustal sections below the sheeted dike complex. extrapolating the surface hydrothermal venting
Note that this solidus temperature is more prop- temperature down to 45 MPa pressures at a shal-
erly viewed as the effective temperature at which low magma lens, which yields a 465°C cutoff
the magma is sufficiently crystallized to behave (Bischoff and Rosenbauer, 1984). In addition,
 7. Magma Chamber Geometry and Crustal Genesis 163
they suggest that hydrothermal cooling will rap- depth of water penetration. Nehlig and Juteau
idly cool the region close to a fluid-filled crack so (1988) find a similar depth limit to near-ridge hy-
that the average rock temperature can be -600°C drothermal circulation and also find that the high-
with fluid present in locally cooler cracks. For the est recorded hydrothermal fluid inclusion trapping
purposes of this study we can sidestep this ques- temperatures are -400-530°C. In these studies
tion to some degree because the hydrothermal we generally choose to model hydrothermal heat
heat loss will be governed by an approximate transport as Tcutoff = 600°C and Nu = 8. We
product of the hydrothermal heat transport en- choose these values because they lead to a solu-
hancement factor Nu times the hydrothermal cut- tion where a steady-state magma lens can exist at
off temperature Tculoff, e.g., 1.2- to 1.5-km depths beneath a fast-spreading
ridge and not exist beneath a slowly spreading
qhydro = Nu(Tculoff - Tseawaler)/Zhydro, (1)
ridge (cf. Fig. 9). Note, however, that the results
where Zhydro is the maximum depth of axial hydro- in Fig. 14 show that this parameterization of heat
thermal penetration. Thus for a lower Tcutoff we transport leads to the same hydrothermal heat
find that we will need a higher Nu value to pro- flow as TCU10ff = 530°C and Nu = 9, or any pair of
duce a magma lens at a given depth for a fast- Nu and Tculoffwith Nu(Tcutoff - Tseawater) = 4800°C.
spreading numerical experiment. Figure 14 shows Studies of the relationship between Rayleigh num-
the strong tradeoff between Tcutoff and Nu in cool- ber and Nusselt number of hydrothermal convec-
ing the axial upper crust, illustrating graphically tion (Combarnous and Bories, 1975; Combar-
that once we can determine TCU10ff from rock and nous, 1978) imply that a Nusselt number of 10
water chemistry observations, we can use this ap- corresponds to a Rayleigh number of -400, about
proach to determine the effective additional heat 10 times the critical Rayleigh number for the on-
transport by hydrothermal circulation and so esti- set of convection. The rock permeability needed
mate the effective permeability within the hydro- for this hydrothermal flow through an axial hydro-
thermal system. Gregory and Taylor (1981) report thermal layer that is 1200 m thick is K = 8 X
that sub solidus oxygen isotope exchange occurred 10- 15 , roughly -800 X larger than the measured
mainly within the upper 5-6 km of the Oman value of 10- 17 in DSDP hole 504B at depths
ophiolite, thus giving an estimate of the maximum greater than 600 m in 6.5-Myr-old crust (Becker
et al., 1989). The low measured value at DSDP
hole 504B would clearly not account for convec-
Or----------,-------, tive hydrothermal heat transfer with a Nusselt
50 mm- yr' number as large as 10. However, near the ridge
on
c:
------------- axis where active faulting and cracking associated
2
~
1.0 ----- with rapid cooling occur (Lister, 1974), it is not
E
~
unreasonable to expect significantly larger per-
E
.s 2.0
meabilities, which are rapidly reduced by cracks
...o -,
-,
, -,
and inactive faults being filled by hydrothermal
mineral deposition. Simple geometric models
§ -,
.9 3.0 T cutoff =(ore -,
,,
(e.g., Turcotte and Schubert, 1982) for the rock
..c
0..
- T cutoff e
= 600 0

,, permeability K suggest that

Tcutoff = 800 e
0
<l) - - -
Q
(2)
2 4 6 8 ro 12 14
Nusselt number (Nu) where b is the crack spacing and ¢ is the crack
Figure 14 Plot of the depth to the top of an axial magma lens
porosity of a cubic matrix of circular tubes (other
described in the text for a half-spreading rate of 50 mm . yr ' ' regular geometrical forms, e.g., a network of flat
and various hydrothermal heat transport enhancement factors cracks, will differ by geometrical factors of order
(Nu) and cutoff isotherms (Teo,"rr) above which hydrothermal 1 (Turcotte and Schubert, 1982). For a crack spac-
flow ceases. In this model, the effectiveness of hydrothermal
ing of 1 m (a typical sheeted dike thickness), this
cooling determines how deep a steady-state magma lens will
reside. Thus if we can determine T,,(off by geochemical means,
implies an active crack porosity of 10 4%. This
then the depth of a steady-state magma lens will directly con- porosity is equivalent to the volume change due to
strain Nu. a secular cooling of 0.3°C, or the porosity opened
164 J. Phipps Morgan et al.

by 1 day of amagmatic extension over an axial where P is the pressure, and J.L and A are the shear
width of 250 m for a typical East Pacific Rise and bulk viscosities, respectively (cf. Tritton,
spreading rate of 100 mm . yr: I. 1977). In the preceding equations we use a comma
As noted above, deep hydrothermal alteration to represent differentiation by the subsequent in-
within the Oman ophiolite is concentrated near dex (i = 1,2 correspond to x- and z-directions),
faults. This observation suggests that it may be and the summation convention applies to an im-
a reasonable hypothesis that crustal extension plicit sum of repeated indices. While the shear
through faulting opens channels for deep hydro- viscosity of mantle and crustal rocks is poorly
thermal flow-channels which lead to a higher known, the bulk viscosity is even less well known,
effective hydrothermal cooling enhancement in so for this suite of numerical experiments we
slow-spreading environments where large median choose A = J.L. In addition, in this suite of numeri-
valley bounding normal faults and 6- to lO-km- cal experiments we do not treat buoyant contri-
deep seismically active faults are present (Fig. 9). butions to mantle upwelling that are discussed in
In this study we only briefly explore a spreading Scott (1992) and Turcotte and Phipps Morgan
rate dependence on the efficiency of hydrother- (1992). Here we try to limit our model complexity
mal heat transport. If this dependence does exist to the oceanic crust. The viscous crust and mantle
(which we feel is likely), it is likely to be an en- rheology that we use is a simplification of that pre-
hancement from Nu = 8-10 at fast-spreading sented in Chen and Morgan (1990). We assume a
ridges to Nu = 12-15 at median valley ridges. Newtonian lithosphere, mantle, and lower crustal
We choose not to include this effect because it .viscosity structure with a lithosphere viscosity
would only enhance the already strong trends that that is 104 greater than the asthenosphere (mantle)
are seen in the following suite of numerical viscosity, which in turn is 103 times more viscous
experiments. than crust that is hotter than 750°C (both crust and
mantle that are colder than 750°C have a litho-
Model Formulation sphere rheology). This results in a total viscosity
contrast that can be as large as 107 , the maximum
To assess this conceptual model, we have imple- strength range that can be treated by the penalty
mented it as a finite-element code that models the finite-element solution algorithm that we employ
flow and heat transport associated with incom- to solve the flow problem (Hughes, 1987). The
pressible flow for a crust and mantle that have a dike injection region has a strength 10 times less
small magma lens and dike region of magma in- than that of the lithosphere. These strength con-
jection. The system of equations governing the re- trasts approximate the rheology in Chen and Mor-
sulting steady-state heat and mass transport is gan (1990) and were chosen to eliminate an addi-
tional nonlinear solution iteration needed to solve
Conservation of Energy
the non-Newtonian flow problem studied in Chen
Ltfr + V . (uT) = KVZT, (3) and Morgan (1990). The finite-element code that
we use has a flow-solver that is an extension of the
where tfr is the magma injection rate, L is the mag-
standard penalty formulation for incompressible
ma's latent heat of cooling (334 kJ . kg :") con-
flow that uses a penalized Lagrange multiplier
verted into a "superheat" of 320°C, the thermal
term to satisfy the incompressibility constraint.
diffusivity K = 10- 6 m2 • s- I, U is the velocity, and
Reddy (1984) and Hughes (1987) provide good
T is the temperature.
summaries of this technique. The difference be-
Conservation ofMass tween the code in this study and standard penalty
formulations is that this code prescribes the dila-
V·u l/J. (4) tion to have a nonzero value in regions where
magma injection is occurring. We solve for heat
Conservation ofMomentum transport using a standard streamline-upwind
aUi+aUj] Petrov -Galerkin technique that is well summa-
J.L [-
aX
-
ax,. rized in Brooks and Hughes (1982). The advan-
j
(5) tages of this formulation are that it has no cross-
P,i + r.: 0, wind artificial diffusion that would introduce a
 7. Magma Chamber Geometry and Crustal Genesis 165
"spurious" numerical diffusion in the direction move out of the sides of the box, and that the top
perpendicular to fluid flow and that it also pro- (seafloor) temperature is O°C. For flow, the verti-
vides a numerically consistent weighting for the cal velocity and horizontal shear stress are zero at
energy source L ~ in the numerical formulation of the seafloor; on the sides of the box we assume
the equation describing energy conservation. Ad- that passive plate spreading occurs beneath the
ditional details of our finite-element implementa- rigid part of the lithosphere, and the bottom of the
tion are discussed in the appendix to Phipps Mor- box is a shear- and normal-stress free surface. See
gan and Chen (1993). Chen and Morgan (1990) for further discussion of
Although we are primarily interested here in these boundary conditions. The boundary condi-
solving for crustal flow and thermal structure, we tion that is unique to the magma lens problem is
also need to solve for the mantle flow to properly that for the influx of magma in the zone of crustal
treat the thermal and mechanical effects of a accretion. We treat this magma, for the geometry
growing lithosphere. Thus we solve for heat and described above, as inflowing at a steady-state
mass transport within a 90 X l40-km region on rate, which exactly balances the rate that crust
one side of a symmetric ridge as shown in Fig. 15. leaves the box due to plate spreading. The effec-
All subsequent figures of the temperature and flow tive temperature of this inflowing material in the
structure near a spreading axis are extracted from axial dike region is set to be higher than the
this larger computational region. The temperature magma solidus by an amount equal to the energy
boundary conditions for this problem are that the re~eased as latent heat of cooling. See Phipps
mantle is flowing into this region at a constant Morgan et al. (1987) for further discussion of this
temperature Tm = l350°C, that heat is free to type of boundary condition. For magma lens ac-
cretion, all of the latent heat of cooling for the
crust that lies below the depth of the magma lens
Problem Solution Region and Boundary Conditions
is released at the top of the magma lens. The depth
° FLOW & TEMP (10 mm· yr')
To=O C w=crs=o of the lens is determined by the depth of the axial
o l200°C isotherm, which is determined after each
solution iteration. This procedure leads to a stable
solution after 0(20) iterations. A typical numeri-
cal experiment takes -I h on a Sun Sparcstation.

Numerical Determination
of Accumulated Crustal Strain

Once we have solved for a steady-state flow field,

crs=crn=o 60 TM=1350°C 120 the accumulated crustal strain associated with
Distance from ridge axis (km)
crustal flow is found using the formulation and
Figure 15 Problem geometry and boundary conditions for techniques summarized in McKenzie (1979). If
the numerical experiments performed in this study. The prob- we imagine the vector y'(t), which joins two
lem region that we consider is a 140-km-wide by 90-km-deep
nearby particles in a fluid element at a time t, then
region on one side of a symmetric ridge axis. Mantle flow is
driven solely by plate spreading, and crust is emplaced at the
this vector is related to a previous vector y joining
ridge axis according to thermal and geometrical criteria de- these two particles at t = 0 by
veloped in the text. We solve the problem.on a 63 (r-dir) X
80 (y-dir) variable spacing tensor-product grid with an .r- and
y' (t) = ~(t)y. (6)
y-nodal spacing shown by tick marks along the top and right- The matrix ~ is initially the unit matrix. At time t,
hand sides of the region. The problem boundary conditions are
shown on each side of the box. The lightly shaded box is the y' (t) = x~(t) x; (t) (7)
subregion from which the solution is extracted to make the
detail plots shown in Figure 16. This subregion contains 43 and at a later time t + tit,
vertical by 30 horizontal grid points. The sample temperature
and flow field shown here is for a half-spreading rate of 10 mm
y' (t + dt) = x~(t + dt) - x; (t + dt). (8)
. yr: I, where no steady-state magma lens exists within the Subtracting and going to the limit as tit ~ 0 gives
crust (the run from which Figure 16 was extracted). Solution
isotherms are contoured at 200°C intervals. D,y'(t) = v'(x], t) - v'(x], t), (9)
166 J. Phipps Morgan et at.

where v'is the velocity of the fluid at position x' relatively uniform with depth (like in Fig. 16B)
and time t, and D, is the Lagrangian derivative. If then there will be no major differences in accu-
we approximate v' (x~,t) with a first-order Taylor mulated strain with depth in a crustal or ophiolite
expansion about x;, then we arrive at section. In the accumulated strain hypothesis for
the development of layering, it is where flow
D ,Yi(t - LikYk(t) -
I ) _ , _ [av:] ,
- , Yk(t), (10) streamlines turn sharply that the straining is most
aXk intense; thus if injection occurred at the bottom of
where summation over repeated subscripts is im- the crust with crustal flow to shallower levels, then
plied. Substitution of Eq. (6) into Eq. (10) gives we would expect a sense of layer development op-
posite that seen in the Oman ophiolite. It is only a
(11)
shallow-level intrusive center that leads to iso-
We use this relation to determine :5(t) by tracing tropic gabbros at shallow stratigraphic levels un-
the flowline and accumulated strain from seed- derlain by progressively more deformed gabbro
points just outside the region of magma injection. sections. The layering development, intensity, and
Equation (6) is numerically integrated with a vari- gabbro dips in the Oman ophiolite support a sce-
able time step that is chosen to place -3 evalua- nario where crustal injection to form this crust oc-
tion points within each element. curred by predominant magma freezing within a
narrow sill near the sheeted dike-isotropic gabbro
Model Results contact-in agreement with previous assertions
·that this crust was created at an analog to a fast-
At a fast-spreading ridge with Nu = 8, Tcutoff = spreading ridge like the present East Pacific Rise
600°C, magma freezing at 1200°C, and a latent spreading center (Nicolas, 1989). (Note that Ni-
heat of solidification of 334 kJ . kg- I, a steady- colas (1989) favors an interpretation of the Oman
state magma chamber exists 1.35 km below the ophiolite record where the ophiolite section lies
seafloor O°C bounding isotherm (see Fig. 16A). on the side of the spreading axis opposite that
All magma that forms the lower crust rises to this determined by Pallister and Hopson (1981) and
level, solidifies, and then flows to deeper crustal Smewing (1981). In this case, mantle must flow
levels. faster than the lower crust for the accumulated
strain to have the observed pattern. If future work
Crustal Strain Can Lead to the Development can resolve on which side of the ridge the ophiol-
of Gahhro Layering ite was created, then the crustal and mantle strain
Figure 16A shows that the accumulated strain seen in the Oman ophiolite will give us a strong
during this flow process is most intense at deeper constraint on the relative importance of buoyant
crustal levels for a ridge with a shallow axial re- vs plate-spreading induced mantle upwelling be-
gion of crustal emplacement. Strain becomes pro- neath a ridge.)
gressively more intense and more flat-lying as the
Moho is approached-the strain within the low- Theoretical Spreading Rate Dependence of
ermost kilometer of the crust is too strong to ef- the Existence and Depth of a Magma Lens
fectively show with the "stretched ellipse" con- Figure 9 shows that theoretical depth to a
vention shown in Fig. 16. A comparison of magma injection lens plotted vs spreading rate
Fig. 16A (or Fig. lC) with Fig. lA clearly shows where Nu and the solidus temperature are fixed,
that crustal strain can produce. layering with the and only the spreading rate is varied. In this case
dip and layer development seen in the Oman there is a fairly abrupt transition with spreading
ophiolite, thus providing an appealing explanation rate from a shallow steady-state magma lens at a
for the orientation and strength of lower crustal 30 mm . yr : I half-spreading rate to no steady-state
layering. In contrast, for a slow-spreading ridge magma lens within the crust at a 20 mm . yr: 1
(Fig. 16B), no steady-state magma lens exists half-spreading rate. There'is good agreement with
within the crust for the same hydrothermal cool- seismic observations of the depth to a magma lens
ing enhancement. We can use Fig. 16 to assess the as a function of spreading rate, which are also
implications of a deeper magma injection lens be- plotted in Fig. 9. (Note, however, that a higher
neath a fast-spreading ridge. If magma injection is Nu = 12 is more consistent with the 30-35 mm·
 7. Magma Chamber Geometry and Crustal Genesis 167

Figure 16 (A) Temperature, flow, and crustal strain for a model spreading center with an opening half-rate of 50 mm . yr : I,
Nu = 8, and Teo'off = 600°C. Thermal structure is shown by levels of gray shading; crustal flow by arrows. The extent of the
magma intrusion zone is shown by the heavy contour-in this case intrusion is limited to the "dike" section and a shallow
magma lens directly beneath the diking zone. Crustal flow lines away from the intrusion zone are shown by light lines, and
accumulated strain is shown at 15,OOO-year time steps along the crustal flow lines. We suggest that the gabbro fabric will reflect
the accumulated strain pattern resulting in a fabric development like that seen in Figure IA that is a consequence of a crustal
flow away from a quasi-steady-state shallow-level intrusion zone. This figure was extracted from a calculation done in the larger
computational domain shown in Figure 15. (B) Temperature, flow, and crustal strain for a model spreading center with an opening
half-rate of 10 mm . yr-', Nu = 8, and Teo'off = 600°C. Same notation conventions as in (A). In this case intrusion extends
completely through the crust since none of the crust is hot enough to sustain a quasi-steady-state magma lens. Crustal flow lines
away from the intrusion zone are shown by light lines, and accumulated strain is shown at 160.000-year time steps along the
crustal flow lines. There is not nearly as strong a gradient in accumulated crustal strain in (B) as that for the fast-spreading case
in (A) because the intrusion rate is constant with depth. A small gradient exists because passive mantle flow near the Moho
beneath the ridge axis moves more slowly than the plate opening velocity. This flow pattern occurs because the weak lower crust
effectively acts as a near-ridge stress free boundary condition on mantle upwelling in response to plate spreading. This figure
was extracted from a calculation done in the larger computational domain shown in Figure 15.

yr : I spreading half-rate, which is roughly ob- at fast-spreading ridges-results which support a

served (Macdonald, 1986; Small and Sandwell, slight hydrothermal enhancement associated with
1989) as the transitional spreading rate between median valley extension.) We performed a suite of
median valley and axial high relief. However, it is numerical experiments in which the magma lens
less consistent with the depth of the magma lens was assumed to be a 2-km-wide, 500-m-thick
168 J. Phipps Morgan et al.

body, i.e., twice the width and more than four and (ii) where the crust is described by a weaker
times the volume of the lens in the preceding nu- diabase rheology while the mantle is described
merical experiments. We found that the depth of by an olivine rheology. (See Chen and Morgan
the lens is most strongly controlled by the balance (1990) for more discussion of an appropriate ridge
between the rate of magma injection within the axis rheological structure.) Again in all numerical
lens and hydrothermal cooling-to first order a experiments the only physical parameter that var-
bigger lens does not influence the net rate of ies is the spreading rate. Independent of detailed
magma injection and hence does not affect the crustal rheology, there is a strong increase in axial
depth of the lens. To second order, a wider lens is yield strength once the half-spreading rate drops
more efficiently cooled at the axis, resulting in a below =20 mm . yr : I. The large variation in in-
slightly (200-300 m) deeper 2-km-wide lens than tegrated axial yield strength with spreading rate
a l-km-wide lens for the same Nu, hydrothermal shown in Fig. 17 is a likely reason for the typi-
cut-off temperature, and spreading rate. cal presence of a strong lithosphere extension-
generated median valley at a slow-spreading ridge
Implications/or Axial Morphology and its absence at a fast-spreading ridge where a
Tapponier and Francheteau (1978) proposed shallow melt lens is commonly seen. Figure 9
that the extension of a strong ridge axis litho- shows that the depth to the 750°C isotherm found
sphere layer may be the origin of median valley in these experiments correlates well with the
topography. This hypothesis has been extended in spreading-rate-dependent maximum earthquake
more recent work by Phipps Morgan et al. (1987), slip depths inferred from teleseismic and micro-
Lin and Parmentier (1990), and Chen and Morgan seismic studies.
(1990). Phipps Morgan et al. (1987) showed that
moments due to lithospheric stresses within a Effects ofVariations in Crustal Thickness
brittle plate that is 8 km thick at the ridge axis at a Given Spreading Rate
and thickens by only a few kilometers within the These model results suggest that a fairly deli-
30-km half-width of the axial valley can produce cate balance exists between magmatic heat injec-
the typical axial topography of a slow-spreading tion during crustal accretion and hydrothermal
ridge. Lin and Parmentier (1990) developed an heat removal, which leads to a strongly differ-
elastic/plastic idealization of plate extension that ent crustal thermal structure at fast- and slow-
allows them to explore the transient development spreading ridge axes-a difference in thermal
of extensional rift valley topography. They found regime that is directly responsible for the ob-
that the form of the rift valley depends on the served differences in axial morphology. This
thickness and thickness variations in the stretch- suggests a strong observational test for this
ing lithosphere and that the lithosphere stress- model-can it explain known axial variability at
supported topography remains after extension a given spreading rate? Hydrothermal cooling is a
stops, successfully explaining the persistence of function of temperature difference and (fault-)
failed rift topography. This emerging paradigm permeability structure, which are not intrinsically
for axial accretion incorporates the idea that it a function of spreading rate. (The fault-extension
is the balance between magmatic heat input and induced component of the permeability structure
hydrothermal heat removal that determines the is quite likely a function of axial thermal struc-
thickness (= yield strength) of the axial litho- ture, which we predict to be strongly dependent
sphere, which in turn controls the axial morphol- on spreading rate. As we discuss later, this effect
ogy associated with plate boundary extension. will enhance the basic predictions of this model.)
The integrated axial strength of the lithosphere is In contrast, the rate of crustal injection is poten-
a qualitative measure of the magnitude of the hori- tially variable at a given spreading rate. Along-
zontal extensional stress that can be supported axis gravity variations discussed earlier suggest
during ridge axis extension. Figure 17 shows the that crustal accretion typically varies by ~ 3 km
axial yield strength as a function of spreading along the axis of a given slow-spreading ridge
rate for two crustal and mantle rheologies: segment (Kuo and Forsyth, 1988; Lin et al., 1990;
(i) where the crust and mantle rheologies are both Blackman and Forsyth, 1991; Lin and Phipps
described by an olivine brittle-ductile rheology, Morgan, 1992).
 7. Magma Chamber Geometry and Crustal Genesis 169
A
YIELD STRENGTH ENVELOPE
O~_~ III~--
~
..c:
r----- I

0..C,)
12
a I 30mm· yr" I 1
16

20
~lom~,<1 I I ! I I I • I
50mm' Yr"
I!IIJ~
1

0 60 120 180 0 60 120 180 0 60 120 180

(MPa) (MPa) (MPa)

B YIELD STRENGTH

IO~ Olivine Rheolo~;-l

Z I Olivine & Diaba~e I
S:l
0
,....,
~ 8 r- \

..c:
.......
\
\
Nu =8
00 \
t:: 6 \
C,)
....
.......
\
\
\
'"
""0
\
\
Q3 4 \

'>-' \
\
\
~ \

0 2 \
\
E-< \
,
o.;
0 10 20 30 40 50 60 70
Half-spreading rate (mm . yr'
Figure 17 Integrated axial lithosphere strength for a suite of numerical experiments plotted as a function of spreading rate.
(A) Mantle yield strength envelopes used in the calculations for 10, 30, and 50 mm . yr- '. Solid lines show the yield strength
envelope for a crust and mantle with an assumed brittle-ductile olivine rheology. Dashed lines show the yield strength envelope
where the crust has a diabase rheology and the mantle has an olivine rheology (see Chen and Morgan, 1990, for more details of
this rheological structure). (B) Depth-integrated total axial yield strength as a function of spreading rate. There is a strong
increase in IOtal axial yield strength at spreading half-rates below 20 mm . yr-', which suggests that extension of strong axial
lithosphere is a likely mechanism for the presence of a median valley at slow-spreading ridges and the absence of a median
valley relief along fast-spreading ridges.

To explore the effect of crustal thickness varia- tential to reach (but just reach) the Moho where
tion at a given spreading rate we have performed cracks become rapidly filled with precipitates.
a suite of experiments at half-spreading rates of Figures 18A and 18B show the depth of the
10, 30, and 50 mm . yr 1 for a range of potential 1200°C isotherm as a function of crustal thickness
crust thicknesses. Figure 18 shows these results at half-spreading rates of 10, 30, and 50 mm .
for Tcutor[ = 600°C, Nu = 8 and 12, and two as- yr : I. Figures 18C and 18D show the resulting in-
sumptions about the maximal depth of penetration tegrated axial yield strength as a function of
of hydrothermal circulation: (a) Hydrothermal crustal thickness at these spreading rates. The
penetration is limited by the depth or pressure at axial "strength" of the lithosphere is the depth-
which cracks close, i.e., is limited to -6 km in- integral of the axial yield-strength envelope. It
dependent of crustal thickness. (b) Hydrothermal is a measure of the magnitude of the horizontal
penetration is limited by the chemical contrast be- extensional stress that can be supported during
tween crust and mantle; i.e., it always has the po- ridge axis extension. According to the lithosphere
 170 J. Phipps Morgan et at.
A
Magma Lens Depth: Nu = 8
c
50 ---
Diabase Crust - Ohvme Mantle Rheology

30mm· yr"

10
~) "N'"

;;J
s
10
tJ
=~
'
~

--5
I
N

15
~
"i)
0;;"
'" r
4
I
,
~
,
9
..c
- - - - Zcutoff= He --' - c; 2
o \, /0 mrn yr- J
- - ZcutoJf= 6 kill E-o 30
oi~- '
~ 20 TcutoJf= 600°C
c .LS=="'==='3~===-. -- L..'_ - , - -

0 2 4 6 8 10 12 14 o 2 4 6 8 10 12 14
B D Diabase Crust - Olivine Mantle Rheology
Magma Lens Depth: Nu = 12 10 .------.--....------...=..:..-,
O ,
e-
C
50 z '.
,,
Nu = 12

~ 8-
...
E
5
I
,
\l)
..c ,,
I

0 ,,
.~
U 10- ,, ,
°0 ,, ,,
I
0 I
N
, I

\l) 15 - I
~ I 10 mm .yr-" .........
B - - - - Zcutojf= H e "
E. 20[-- Zcutoff= 6 km ........ .r :
"
... 1
Q ITcutoJf ~ 600o~ 4 6 8 10 12 14
o 2 4 6 8 10 -1~2----l14~ Crustal thickness (km)
Crustal thickness (km)
Figures 18C and 18D Axial yield strength plotted as a func-
Figures 18A and 18B Magma lens depth as a function of tion of crustal thickness at spreading rates of 10, 30, and
crustal thickness at spreading rates of 10, 30, and 50 mm . 50 mm . yr- '. Figures show the corresponding integrated axial
yr- '. Figures show the depth of the l200°C isotherm plotted yield strength plotted"as a function of crustal thickness. The
as a function of crustal thickness for Nu = 8 and 12, and for axial yield strength also changes dramatically near this "criti-
two different hydrothermal penetration limits Z'",off of 6 km cal" crustal thickness.
and the crustal thickness H,. The region above the lens/
no-lens transition is where the 1200°C isotherm will lie within
the crust. Note the sharp change in the depth of this isotherm
phology. Similar results for a 20 mm . yr - I half-
for small variations in crustal thickness around a "critical" rate typical for the southern Mid-Atlantic Ridge
crustal thickness. suggest that the along-axis crustal thickness varia-
tions of 4.5- 7.5 km inferred from axial seismic
studies (Tolstoy et al.• 1992) and axial gravity
stretching hypothesis for the origin of axial topog- studies (cf. Lin and Phipps Morgan, 1992; Neu-
raphy (Phipps Morgan et al.. 1987; Lin and Par- mann and Forsyth, 1992) are sufficiently large to
mentier, 1990; Chen and Morgan, 1990), the net change the ridge from an axial high to a median
horizontal axial force is proportional to the ampli- valley morphology. In general, for transitional
tude of the resulting median valley relief. At a spreading rates of 20-30 mm . yr- I a dramatic
half-spreading rate of 30 mm·· yr- I we find that a change in axial strength will be associated with
magma lens is present for a crustal thickness of these along-axis changes in crustal thickness and
6 km but disappears when the crust is slightly thermal structure as shown in Figs. 18C and 18D.
thinner (4.5 km, Fig. 18A). This result suggests At a half-spreading rate of 10 mm . yr : I a very
that the intermediate spreading rate Australian- large crustal thickness change from a "normal" 6-
Antarctic Discordance, which has a median val- km-thick crust is needed before a quasi-steady-
ley morphology, can be produced by crust that is state magma lens appears. At this crustal thickness
1-1.5 km thinner than the crust of ridge segments of 9-12 km (the exact value depends on which
to the east and west that have an axial high mor- assumption is used for Nu or the intrinsic depth
 7. Magma Chamber Geometry and Crustal Genesis 171
12,
• Axial High depth. This positive-feedback effect is why these
,, o Median Valley model thermal structures are more sensitive to
IOl spreading rate or crustal thickness variations than
E the thermal structures produced in previous ther-
C
'"
8~ mal models which treat crustal accretion always
"'"0:: as a vertical crustal-thickness dike (Sleep, 1975;
~
o
:§
6l- -- Phipps Morgan et al., 1987; Lin and Parmentier,
0
1989; Chen and Morgan, 1990).
~
2
u
4~ 1 ""0 The results in Figs. 18A-D show the existence
of a "threshold" crustal thickness at a given

2
oo
t Tcutojf = 600°C

'
- - Zcutojf= 6 km
- - - - Zcutoff= He

60
spreading rate about which small changes in
crustal thickness can produce a dramatic change
in axial thermal structure. Figure 18E shows a plot
10 20 30 40 50
Half-spreading rate (mm· yr') of this threshold crustal thickness as a function of
Figure 18E The minimal crustal thickness where a steady-
spreading rate for the several hydrothermal cool-
state magma lens will exist within the crust plotted against ing assumptions discussed above. (This threshold
spreading rate. Small variations about the "normal" crustal crustal thickness is defined as the point where the
thickness of 6 km (Chen, 1992) are most significant for the 1200° isotherm reaches the base of the crust.) This
lens/no-lens transition at intermediate half-spreading rates of figure supports the result found in Figs. 18A-
20-30 mm . yr-'.
18D-variations in axial thermal structure are
most sensitive at intermediate spreading rates of
limit for hydrothermal penetration) the transition 20- 30 mm . yr "! half-rate to small fluctuations in
is abrupt. This result roughly agrees with a seis- crustal input about the normal crustal thickness
mic refraction determination of 8- to l G-km-thick of 6 km. Greater hydrothermal heat removal in-
crust along a 9-Ma isochron at the slow-spreading creases the spreading rate of this transition and, at
axial high Reykjanes Ridge. (Bunch and Kennett, slow spreading rates, greater depths of hydrother-
1980). Finally, at half-spreading rates of 50 mm . mal penetration require a larger crustal thickness
yr: lor greater these results suggest that it is diffi- (magmatic heat input) for a quasi-steady-state
cult to ever generate median valley relief-the crustal magma lens to form. The observed Reyk-
crust must be less than - 3 km thick before the janes Ridge axial high for a crustal thickness of
magma lens structure disappears. This crustal -8-10 km (Bunch and Kennett, 1980) implies
thickness is lower than any oceanic crustal thick- that Tculoff = 600°C and Nu = 8 is a good descrip-
ness measurement to date (Chen, 1992). One tion of hydrothermal heat transport at a slow-
striking feature of these results is the sharp change spreading range. The observed variation in axial
in axial thermal structure with small changes in relief at intermediate spreading rates (Macdonald,
crustal thickness near the "threshold" crustal 1986) is also better fit in Figs. 18C and 18D by
thickness for the existence of a steady-state crustal models where the depth of hydrothermal circula-
magma lens. For the 10 mm . yr - I experiments, a tion is limited by the depth at which cracks close
change in crustal thickness from 9 to 9.5 km (at (pressure) instead of the depth of the crust-mantle
Nu = 8) leads to a shallowing of the 1200°C iso- transition. These results lead to a (weakly) pre-
therm from 16 to 8 km! This effect can be under- ferred parameterization of hydrothermal circula-
stood from the nature of the latent-heat release in tion by a cutoff depth Zeutoff of -6 km, thermal
a magma lens-it releases the latent heat of limit Teu/off = 600°C, and heat transport enhance-
the magma-cumulate phase change for all of the ment Nu = 8.
"cumulate gabbro" crust that resides below the
magma lens. Thus an upward displacement of Axial Variability along a Single Segment
the level of the magma lens because of the in- Currently, there is a debate about whether
crease of the total heat input for thicker crust, or magma emplacement along a fast-spreading ridge
the increase of spreading rate, would result in is fairly continuous along-axis or confined to a
more latent heat release (positive feedback), an ef- few discrete volcanic centers that are foci of axial
fect that would favor a lens at an even shallower accretion processes (e.g., Macdonald et al., 1990;
172 J. Phipps Morgan et al,
Phipps Morgan, 1991). Lin and Phipps Morgan leads to a spreading rate dependence of the
(1992) note that while gravity and topography depth of the magma lens, which is comparable to
data show that slow-spreading ridges have a clear that observed. At its simplest expression, neutral
along-axis variation in crustal thickness (i.e., in- buoyancy arguments would suggest that magma
tegrated magma supply varies along-axis), the intrusion depths are independent of spreading rate.
much smaller along-axis gravity and topography Hooft and Detrick (1993) point out that the ob-
variation at a fast-spreading ridge can be ex- served magma lens depths are invariably deeper
plained either by a more 2D pattern of upwelling than predicted neutral buoyancy depths of -100-
and melting beneath a fast-spreading ridge or by a 400 m below the seafloor. However, Ryan (this
well-connected, temporally persistent magma lens volume) suggests that Hooft and Detrick (1993)
(or low viscosity zone) that smooths the along- use a minimum magma density when inferring a
axis crustal structure at a fast-spreading ridge. too-shallow neutral buoyancy depth, whereas they
These results suggest that magmatic heat input as- should use an average magma density (i.e., in-
sociated with a mean 6-km crustal thickness is cluding ferrobasaltic magmas and entrained crys-
sufficient for a quasi-steady-state magma lens to tals) when inferring a magma's neutral buoyancy
form at fast-spreading ridges. At slower-spreading level, which would predict neutral buoyancy
rates with 6-km-thick crust, a continuous magma depths that are consistent with observed EPR
lens cannot be a quasi-steady-state feature along an magma lens depths. The neutral buoyancy hy-
entire segment. Again, along-axis crustal flow will . pothesis is also hard-pressed to explain the 4-km-
preferentially occur in the hottest, weakest regions wide, ridgeward sloping lens observed near the
since plate extension tends to confine flow of 9°03'N overlapping spreading center along the
stronger regions to the plate-spreading direction. East Pacific Rise (cf. Fig. 7), a feature that would
However, the crustal thickness at slow-spreading be a natural consequence of magma freezing-
ridges is (usually) not strongly smoothed by limited melt ascent if the primary magma source
along-axis flow, resulting in an observed ~3 km is offset from the ridge axis. Joann Stock (per-
along-axis crustal thickness variation (e.g., Black- sonal communication) has pointed out that if the
man and Forsyth, 1991; Lin and Phipps Morgan, depth of hydrothermal circulation is controlled by
1992; Neumann and Forsyth, 1993). The resulting a spreading-rate-dependent balance between the
along-axis variations in across-axis thermal struc- heat of magma injection and cooling by hydro-
ture lead to dramatic variations in axial yield thermal processes, then hydrothermal alteration
strength (Figs. 18C and 18D) and hence in the may change the shallow axial density structure so
along-axis relief of median valley topography. that magmas intrude to a deeper level at a slow-
This model predicts that along-axis variations in spreading ridge. Ryan (this volume) explores this
axial yield strength will be strongly correlated effect in more detail. (This mechanism for crustal
with crustal thickness-in agreement with obser- density reduction would imply pervasive crustal
vations of strongly correlated axial mantle Bou- serpentinization as opposed to fault-concentrated
guer anomaly (a poor man's proxy for crustal hydrothermal alteration.)
thickness) and axial depth profiles at slow- and These hypotheses may have different predic-
transitional-spreading ridges as seen in Fig. II tions for the relationship between basalt compo-
(Phipps Morgan, 1991; Lin and Phipps Morgan, sition and the depth of the magma lens. If the
1992; Neumann and Forsyth, 1993). depth is buoyancy controlled, then basalts erupted
from a ridge with a deeper magma lens should,
Neutral Buoyancy versus Magma Freezing as a in general, show more fractionation than basalts
Fundamental Limit to Magma Ascent erupted from shallower magma lenses. In contrast,
the magma freezing hypothesis would predict that
We favor the hypothesis that magma ascent at a basalts freeze a (slightlyj.thicker cumulate section
fast-spreading ridge is primarily limited by axial as the lens depth shallows, which would imply an
thermal structure-magma tends to pond beneath inverse correlation between the average concen-
a freezing horizon until sufficient melt buoyancy tration of incompatible elements in erupted ba-
builds up for the magma to fracture to sheeted salts and the depth to the magma lens. Petrologists
dikes and the surface. This mechanism naturally are just starting to explore whether there are criti-
 7. Magma Chamber Geometry and Crustal Genesis 173
cal geochemical tests for these two mechanisms of ments fit either a global or a local trend. Figure 19
limiting melt ascent. shows these data. In general, MORB variabil-
The effects of neutral buoyancy and axial ther- ity along fast-spreading ridge segments follows
mal structure may, in fact, be strongly linked in the "global" trend, whereas chemical variability
real mid-ocean ridge magmatic systems. If the along slow-spreading ridge segments follows the
floor of the subsiding magma lens consists pre- "local" trend (Niu and Batiza, 1993). We suggest
dominantly of dense cumulate phases that flow that the local trend may be due to magma crystal
away to form the lower crust, while the roof is fractionation (Grove et al., 1992) and/or reaction
compositionally lighter, then the magma lens it- within the uppermost, cooler mantle beneath a
self may be a strong mechanism to shape the den- slow-spreading ridge. In this case the local trend
sity structure of the crust, which in tum will limit should be strongly correlated with axial topogra-
magma ascent. A key ingredient to resolving phy-both the local trend and a median valley
some of these questions will be to understand bet- will exist only where the uppermost mantle be-
ter the evolution of the crystallate fraction of as- neath a ridge is "cold." The correlation of median
cending magma, as it plays a key role in shaping valley vs axial high morphology with global vs
the magma's density. In addition, if we can com- local trend is even better than the correlation with
pare basalts that are erupted from the flank with spreading rate. At fast-spreading ridges with an
those from the ridge crest of the sloping magma axial high the global trend dominates the chemical
lens at the 9°Q3 'N OSC, then these rocks may pro- systematics along a spreading segment, at slow-
vide a means to test if either neutral buoyancy or spreading ridges with a median valley the local
axial thermal structure is the primary limit to trend dominates, and at intermediate spreading
magma ascent. rates the global or local trend correlates extremely
well with both axial morphology and the presence
Geochemical Variability between of a magma lens seismic reflector as shown in
Ridge Segments Fig. 19. In addition, the three propagating ridge
segments in the Niu and Batiza. (1993) data set
Petrologists have long noted a difference in follow the local trend, in accord with previous
chemical systematics between fast- and slow- work that suggests that the growth of a propagat-
spreading ridges-fast-spreading ridge MORBs ing ridge segment is associated with a transient
(mid-ocean ridge basalts) have both a lower bulk "colder" accretion environment than that of a
Mg#, suggesting more cumulate fractionation similarly spreading stable ridge segment. (This
and magma evolution within the crust, and less idea was suggested by Sinton and Christie in the
magma diversity. Both observations support the early 1980s to explain the petrologic gradients
existence of a persistent magma reservoir at fast, seen approaching a propagating ridge tip and the
but not slow-, spreading ridges (cf. Sinton and De- diverse suite of basalts seen near the propagator
trick, 1992). MORB systematics appear to show tip proper (Christie and Sinton, 1981; Sinton
three distinct effects: (i) shallow crystal fraction- et al., 1983). We feelthat these results support the
ation, which workers usually "correct" for by re- hypothesis that the presence of a local trend and
ferring to an element concentration at a reference an axial median valley reflects the same system-
magma MgO content, which is found by extrapo- atic variations in the axial thermal structure with
lating a shallow fractionation trend through a suite spreading rate and with variations in crustal thick-
of surface samples (e.g., RNa in Klein and Lang- ness at a given spreading rate.
muir, 1987; (ii) local chemical variability, which
cannot be accounted for by low-pressure crystal-
lization, which Klein and Langmuir (1989) refer Closure
to as a local trend; and (iii) global chemical vari-
ability of average composition at a given MgO con- This chapter has attempted to summarize an
tent, which Klein and Langmuir (1987) refer to as emerging theoretical and observational paradigm
a global trend. Niu (1992) and Niu and Batiza for magma chamber structure at a fast- and slow-
(1993) have determined how well chemical syste- spreading ridge. This paradigm relates magma
matics along 32 different individual ridge seg- chamber structure and depth dependence to the
174 J. Phipps Morgan et al.

-2.0
- Median Valley
Axial High
-1.0
1I11l1l1ll11l1
No Distinct Segment Morphology
....... Propagating Ridge

e0 0
8 .
I, _~ L
• Magma Lens Observed with MCS

CI:l
.• " .
1.0 • • •
••
2.0
a 50 100 150
Spreading Rate (km . Myr- I )
Figure 19 Geochemical systematics along 32 ridge segments (Niu and Batiza, 1993) are plotted versus spreading rate. The
length of each line segment is an index to how well the variability along a segment parallels the Klein and Langmuir (1989)
"global" or "local" trends. Dark lines note where the axial segment morphology is a median valley, light lines where the segment
morphology is an axial high. Short-dashed lines denote ridge segments with no distinct segment morphology. Long-dashed lines
are segments with current/recent ridge propagation. Stars show axial segments that have an axial magma lens reflector that has
been observed in an axial seismic reflection study along at least part of the ridge segment (Detrick et al., 1987, 1993). This figure
clearly shows that chemical systematics along fast-spreading-ridges follow Klein and Langmuir's (1987) global trend while slow-
spreading ridges follow their local trend (Niu and Batiza, 1993). All axial high segments mapped so far with reflection seismic
methods have had a crustal magma lens as predicted by our model; the axial high segments without a star in this plot are segments
that have not yet been mapped with seismic reflection methods. Segment scale chemical systematics correlate even better with
axial morphology than with spreading rate-axial high segments follow the global trend while median valley segments follow
the local trend. This correlation suggests a common dependence of median valley relief and local trend chemical systematics on
a cold shallow mantle directly beneath the ridge axis. Propagating ridge segments also tend to follow a local trend, in agreement
with previous suggestions that propagating ridge tips have an anomalously cold initial accretion environment.

thermal structure of a mid-ocean ridge, which is persistent crustal magma reservoir. Finally it ap-
controlled by a delicate balance between heat in- pears to raise several further predictions on the re-
put by crustal injection and removal by hydrother- lationships among basalt chemistry, crustal thick-
mal circulation so that the presence or absence of ness, and spreading rate, questions that will let
axial topography directly reflects the axial thermal us actively explore and test the viability of this
structure along a ridge. This paradigm is consis- emerging paradigm.
tent with the melt lenses recently imaged along
fast-spreading ridges but that have not been found
where an axial valley morphology is present. It Acknowledgments
suggests that much of the variability in axial relief
at a given spreading rate is due to local variations We thank Mike Ryan and Bob Detrick for helpful reviews of
in crustal magmatic injection. It also suggests this chapter and for giving us preprints of their recent work on
that many of the differences in basalt composition the role of neutral buoyancy in magma ascent at a fast-
and chemical systematics between "slow-" and spreading ridge. We also thank Mike Ryan for his editorial
suggestions. Breck Betts created and drafted many of the fig-
"fast" -spreading ridges are due to differences in
ures. Maya Tolstoy and Bob Detrick generously allow us to
shallow mantle and crustal thermal structure be- quote their prepublication seismic and gravity observations of
neath these ridges-there is a fast-spreading "hot" crustal thickness variations shown in Fig. 18E, Yaoling Niu
environment that has little magma freezing and re- permitted us to use his and Batiza's prepublication measure-
acting with the mantle as it ascends beneath the ments of axial geochemistry used in Fig. 19, and Jian Lin and
l-C. Sempere provided the median valley bathymetric data
ridge and a temporally continuous magma lens
shown in Fig. 10. Their help is gratefully acknowledged, as
within the crust, and a slow-spreading "cold" en- are the years of collaboration with Don Forsyth, Jason Mor-
vironment with significant magma freezing/reac- gan, and Marc Parmentier during the formulation of the theo-
tion within the shallow mantle and no temporally retical synthesis that is presented here. The development of
 7. Magma Chamber Geometry and Crustal Genesis 175
these observations and ideas was supported by the National chamber beneath the Lau Basin back-arc spreading centre,
Science Foundation. Nature 348, 646-648.
Combarnous, M., and Bories, S. A. (1975). Hydrothermal con-
vection in saturated porous media, in "Advances in Hy-
droscience." (Y. T. Chow, ed.), pp. 231-307. Academic
References Press, New York.
Combarnous, M. (1978). Natural convection in porous media
Barth, G. A., Mutter, J. C, and Detrick, R. W (1987). Along- and geothermal systems, in "Proceedings, 6th Int. Heat
axis variability in East Pacific Rise seismic structure, EOS Transfer Conf.," pp. 45-59.
Trans. Am. Geophys. Union 44, 1491-1492. Detrick, R. S., Buhl, P., Vera, E., Mutter, 1., Orcutt, 1., Madsen,
Becker, K., Sakai, H., Adamson, A. C, Alexandrovich, 1., Alt, J., and Brocher, T. (1987). Multi-channel seismic imaging
1. C, Anderson, R. N., Bideau, D., Gable, R., Herzig, of a crustal magma chamber along the East Pacific Rise,
P. M., Houghton, S., Ishizuka, H., Kawahata, H., Kinosh- Nature 326,35-41.
ita, H., Langseth, M. G., Lovell, M. A., Malpas, 1., Ma- Detrick, R. S., Mutter, 1. C., Buhl, P., and Kim, 1. I. (1990). No
suda, H., Merrill, R. B., Morin, R. H., Mottl, M. J., Pariso, evidence from multichannel reflection data for a crustal
1. E., Pezard, P., Phillips, 1., Sparks, 1., and Uhlig, S. magma chamber in the MARK area on the Mid-Atlantic
(1989). Drilling deep into young oceanic crust, Hole 504B, Ridge, Nature 347, 61-64.
Costa Rica rift, Rev. Geophys. 27, 79-102. Detrick, R. S., Harding, A. J., Kent, G. M., Orcutt, J. A.,
Bischoff, 1. L., and Rosenbauer, R. 1. (1984). The critical point Mutter, 1. C; and Buhl, P. (1993). Seismic structure of the
and two-phase boundary of seawater, 200-500°C, Earth southern East Pacific Rise, Science 259, 499-503.
Planet. Sci. Lett. 68, 172-180. Dewey, 1. E, and Kidd, W S. E (1977). Geometry of plate
Blackman, D. K., and Forsyth, D. W (I 99\). Isostatic com- accretion, Geol. Soc. Am. Bull. 88, 960-968.
pensation of tectonic features of the Mid-Atlantic Ridge: Ewing, 1. I., and Meyer, R. P. (1982). Rivera ocean seis-
25-2rS,J. Geophys. Res. 96, 11741-11758. . mic experiment (ROSE) overview, J. Geophys. Res. 87,
Bratt, S. R., and Solomon, S. C (1984). Compressional and 8345-8358.
shear wave structure of the East Pacific Rise at II °20'N: Francheteau, 1., Armijo, R., Cheminee, 1. L., Hekinian, R.,
Constraints from three-component ocean-bottom seis- Lonsdale, P., and Blum, N. (1992). Dyke complex of the
mometer data, 1. Geophys. Res. 89, 6095-6110. East Pacific Rise exposed in the walls of Hess Deep and
Browning, P. (1984). Cryptic variation within the cumulate se- the structure of the upper oceanic crust, Earth Planet. Sci.
quence of the Oman ophiolite: Magma chamber depth and Lett. HI, 109-121.
petrologic implications, in "Ophiolites and Oceanic Litho- Garmany, 1. (1989). Accumulation of melt at the base of young
sphere." (1. G. Gass, S. 1. Lippard, and A. W. Shelton, oceanic crust, Nature 340, 628-632.
eds.), Geol. Soc. London, Spec. Publ., 71-82. Gregory, R., and Taylor, H. P. (1981). An oxygen isotope pro-
Bunch, A. W H., and Kennett, B. L. N. (1980). The crustal file in a section of Cretaceous oceanic crust, Samail
structure of the Reykjanes Ridge from seismic refraction, ophiolite, Oman: Evidence for dlKO buffering of the
Geophys. J. R. Astron Soc. 61, 141-166. oceans by deep (>5 km) seawater-hydrothermal circula-
Burnett, M. S., Caress, D. W, and Orcutt, J. A. (1989). Tom- tion at mid-ocean ridges, 1. Geophys. Res. 86,2737 -2755.
ographic image of the magma chamber at 12°50'N on the Grove, T. L., Kinzler, R. L., and Bryan, W. B. (1992). Frac-
East Pacific Rise, Nature 239, 206-208. tionation of mid-ocean ridge basalt (MORB): Evidence for
Cann, 1. R. (1974). A model for oceanic crustal structure de- fractionation in the uppermost oceanic mantle, EOS Trans.
veloped, Geophys. 1. R. Astron. Soc. 39, 169-187. Am. Geophys. Union 73, 615.
Caress,D. W, Burnett, M. S., and Orcutt, 1. A. (1992). Tom- Hale, L. D., Morton, C 1., and Sleep, N. H. (1982). Reinter-
ographic image of the axial low velocity zone at 12°50'N pretation of seismic reflection data over the East Pacific
on the East Pacific Rise, 1. Geophys. Res. 97,9243-9264. Rise, J. Geophys. Res. 87, 7707-7717.
Chen, Y., and Morgan, W 1. (1990). A nonlinear-rheology Harding, A. 1., Orcutt, 1. A., Kappus, M. E., Vera, E. E., Mut-
model for mid-ocean ridge axis topography, 1. Geophys. ter, 1. C., Buhl, P., Detrick, R. S., and Brocher, T. M.
Res. 95, 17583-17604. (1989). The structure of the young oceanic crust at 13°N
Chen, Y. 1. (1992). Oceanic crustal thickness versus spreading on the East Pacific Rise from expanding spread profiles,
rate, Geophys. Res. Lett. 19,753-756. J. Geophys. Res. 94,12163-12196.
Christensen. N. 1. (1979). Compressional wave velocities in Harding, A. 1., Kent, G. M., and Orcutt, 1. A. (1993). A multi-
rocks at high temperatures and pressures, critical thermal channel seismic investigation of upper crustal structure at
gradients, and crustal low velocity zones, J. Geophys. Res. 9°N on the East Pacific Rise: Implications for crustal ac-
84,6489-6857. cretion, J. Geophys. Res. 98, 13925-13944.
Christensen, N. I., and Smewing, 1. D. (1981). Geology and Haymon, R. M., Fornari, D. 1., Edwards, M. H., Carbotte, S.,
seismic structure of the northern section of the Oman Wright, D., and Macdonald, K. C (l99Ia). Hydrother-
ophiolite, J. Geophys. Res. 86, 2655-2555. mal vent distribution along the East Pacific Rise crest
Christenson, G. L., Purdy, G. M., and Fryer, G. 1. (1992). 9°.09'N -9°54'N and its relationship to magmatic and tec-
Structure of young oceanic crust at the East Pacific Rise tonic processes on fast-spreading mid-ocean ridges, Earth
near 9°30'N, Geophys. Res. Lett. 19, 1045-1048. Planet. Sci. Lett. 104,512-534.
Collier, 1., and Sinha, M. (1990). Seismic images of a magma Haymon, R., Fornari, D., Von Damm, K., Edmond, 1., Lilley,
 176 J. Phipps Morgan et al.
M., Perfit, M., Shanks, W. c., III, Grebmeier, J., Lutz, R.. Klein, E. M., and Langmuir, C. H. (1987). Global correlations
Carbotte, S., Wright, D.. and Smith, M., McLaughlin, E.. of ocean ridge basalt chemistry with axial depth and
Beedle, N., Seewald, J., Reudelhuber, D., Olson, E., John- crustal thickness, J. Geophys. Res. 92, 8089-8115.
son, F. (1991 b). Eruption of the EPR crest at 9°45' -54'N Klein, E. M.. and Langmuir, C. H. (1989). Local versus global
since late 1989 and its effects on the hydrothermal venting: variation in ocean ridge basaltic composition: A reply,
Results of the ADVENTURE program, an ODP site survey J. Geophys. Res. 94,4241-4252.
with Alvin, EOS Trails. Am. Geophys. Union 72, 480. Kong, L. S. L. (1990). "Variations in Structure and Tectonics
Henstock, T. J., Woods, A. w., and White, R. S. (1993). The along the Mid-Atlantic Ridge, 23°N and 26°N," Ph.D.
accretion of oceanic crust by episodic sill intrusion, J. Thesis, Mass. Inst. of Techno!./Woods Hole Oceanogr.
Geophys. Res. 98, 4143-4162. Inst., Woods Hole, MA.
Herron, T. J., Ludwig, W. J., Stoffa, P. L., Kan, T. K., and Buhl, Kong, L. S. L., Solomon, S. C., and Purdy. G. M. (1992).
P. (1978). Structure of the East Pacific rise crest from mul- Microearthquake characteristics of a mid-ocean ridge
tichannel seismic data, 1. Geophys. Res. 83, 798-804. along-axis high, J. Geophys. Res. 97,1659-1685.
Hooft, E. E., and Detrick, R. S. (1993). The role of density in Kuo, B.-Y., and Forsyth, D. W. (1988). Gravity anomalies of
the accumulation of basaltic melts at mid-ocean ridges, the ridge-transform system in the South Atlantic between
Geophys. Res. Lett. 20,423-426. 31 and 34.5°S: upwelling centers and variations in crustal
Huang, P. Y.. and Solomon, S. C. (1988). Centroid depths of thickness, Mar. Geophys. Res. 10,205-232.
mid-ocean ridge earthquakes: Dependence on spreading Kusznir, F. D., and Bott, M. H. P. (1976). A thermal study of
rate, J. Geophys. Res. 93, 13445-13477. oceanic crust, Geophys. J. R. Astron. Soc. 47, 83-95.
Hubral, P. (1977). Time migration-Some ray theoretical as- Langmuir. C. H. (1989). Geochemical consequence of ill situ
pects. Geophys. Prosp. 25, 738-745. crystallization, Nature 340, 199-205.
Kampfmann, w.. and Berckhemer, H. (1985). High tempera- Lewis, B. T. R., and Garmany, J. D. (1982). Constraints on the
ture experiments on the elastic and anelastic behaviour of structure of the East Pacific Rise from seismic refraction
magmatic rocks, Phys. Earth Planet. Int. 40, 223-247. data. J. Geophys. Res. 87, 8417-8425.
Kappel, E. S., and Ryan, W. B. F. (1986). Volcanic episodicity Lin, J., and Parmentier, E. M. (1990). A finite amplitude neck-
and a non-steady state rift valley along northeast Pacific ing model of rifting in brittle lithosphere, J. Geophys. Res.
spreading centers: Evidence from SeaMARC I, 1. Geo- 95,4909-4923.
phys. Res. 91, 13925-13940. Lin, J.. and Phipps Morgan, J. (1992). The spreading rate de-
Kappus, M. E. (1991)... A Baseline for Upper Crustal Velocity pendence of three-dimensional mid-ocean ridge gravity
Variations along the East Pacific Rise," Ph.D. Thesis, structure, Geophys. Res. Lett. 19, 13-16.
Univ. of California, San Diego. Lin, J. Purdy, G. M., Schouten, H., Sempere, J.-c., and Zervas.
Karson, J. A, Collins, J. A., and Casey, J. F. (1984). Geologic C. (1990). Evidence from gravity data for focussed mag-
and seismic velocity structure of the crust/mantle transi- matic accretion along the Mid-Atlantic Ridge, Nature 344,
tion in the Bay of Islands ophiolite complex, J. Geophys. 627-632.
Res. 89, 6126-6138. Lister, C. R. B. (1983). On the intermittency and crystalliza-
Kempner, W. C; and Gettrust, J. F. (1982). Ophiolites, syn- tion mechanisms of sub-seafloor magma chambers, Geo-
thetic seismograms, and ocean crustal structure. 1. Com- phys J, R. Astron. Soc. 73, 351-365.
parison of ocean bottom seismometer data and synthetic Macdonald, K. c, Sempere, r.c, and Fox, P. J. (1984). East
seismograms for the Bay of Islands ophiolite, J. Geophys. Pacific Rise from Siqueiros to Orozco Fracture Zones:
Res. 87, 8447-8462. Along-strike continuity of axial neovolcanic zone and
Kent. G. M., Harding, A. J., and Orcutt, J. A (1990). Evidence structure and evolution of overlapping spreading centers,
for a smaller magma chamber beneath the East Pacific Rise J. Geophys. Res. 89,6049-6069.
at 9°30'N, Nature 344, 650-653. Macdonald, K. C. (1986). The crest of the Mid-Atlantic Ridge:
Kent. G. M., Harding, A. J., and Orcutt, J. A. (I993a). Distri- Models for crustal generation processes and tectonics, ill
bution of magma beneath the East Pacific Rise between "The Geology of North America: The Western North At-
the Clipperton transform and the 9°17'N Deval from for- lantic Region, v. M." (P. Vogt. and B. Tucholke, eds.),
ward modeling of CDP data, 1. Geophys. Res. 98, pp. 51-68, Geol. Soc. Am., Boulder, CO.
13945-13970. Macdonald, K. c., and Fox, P. J. (1988). The axial summit
Kent, G. M., Harding, A J., and Orcutt, J. A. (l993b). Distri- graben and cross-sectional shape of the East Pacific Rise
bution of magma beneath the East Pacific Rise near the as indicators of axial magma chambers and recent volcanic
9°03 'N overlapping spreading center from forward mod- eruptions, Earth Planet, Sci. Lett. 88, 119-131.
eling of CDP data, J. Geophys, Res. 98, 13971-13998. Macdonald, K. c.
Fox, P. J., Perram, L. J., Eisen, M. F., Hay-
Kent, G. M., Harding, A. J., and Orcutt, J. A (l993c). The mon, R. M., Miller, S. P.. Carbotte, S. M., Cormier, M.-H.,
uniform accretion of oceanic crust south of the Garrett and Shor, A. N. (1990). A new view of the mid-ocean ridge
transform at 14°15'S on the East Pacific Rise, submitted from the behaviour of ridge-axis discontinuities, Nature
for publication, J. Geophys. Res. 335,217-225.
Khiratov, K I., Lebedev, E. B., Dorfman, A M" and Bagdass- Madsen, J. A., Forsyth, D. w., and Detrick, R. S. (1984). A
rov, N. S. (1983). Study of process of melting of the Kir- new isostatic model for the East Pacific rise crest, J. Geo-
gurich basalt by the wave method, Geochimica 9, p~vs.Res.89,9997-100015.
1239-1246. Madsen, J. A., Detrick, R. S., Mutter, J. C; and Orcutt, J. A
 7. Magma Chamber Geometry and Crustal Genesis 177
(1990). A two- and three-dimensional analysis of gravity Niu, Y. (1992). "Mid-Ocean Ridge Magmatism: Style of
anomalies associated with the East Pacific Rise at 9°N and Mantle Upwelling, Partial Melting, Crustal Level Pro-
13°N. J. Geophys. Res. 95,4967-4987. cesses, and Spreading Rate Dependence-A Petrologic
Manghnani, M. H., Sato, H., and Chandra, S. R. (1986). Ultra- Approach," Ph.D. Thesis, Univ. of Hawaii.
sonic velocity and attenuation measurements on basalt Niu, Y., and Batiza, R. (1993). Chemical variation trends at
melts to 1500°C: Role of composition and structure in fast- and slow-spreading mid-ocean ridges, J. Geophys.
the viscoelastic properties, J. Geophys. Res. 91, 9333- Res. 98, 7887-7902.
9342. Orcutt, 1. A.. Kennett, B. L. N., Dorman, L. M., and Prothero,
Marsh, B. D. (1981). On the crystallinity, probability of occur- W. A. (1975). Evidence for a low velocity zone underlying
rence and rheology of lava and magma, Contrib. Mineral. a fast spreading rise crest, Nature 256, 475-476.
Petrol. 78, 85-98. Pallister, 1. S., and Hopson, C. A. (1981). Samail ophiolite
Marsh, B. D. (1989). Magma chambers, Annu. Rev. Earth plutonic suite: Field variations. phase variations, cryptic
Planet. Sci. 17,439-474. variation and layering, and a model of a spreading ridge
McClain, 1. S., Orcutt, 1. A., and Burnett. M. (1985). The East magma chamber, J. Geophys. Res. 86, 2593-2644.
Pacific Rise in cross section: A seismic model, 1. Geophys. Parmentier, E. M., and Phipps Morgan, 1. (1990). The spread-
Res. 90, 8627-8639. ing rate dependence of three-dimensional structure in oce-
McKenzie, D. P. (1979). Finite deformation during fluid flow, anic spreading centres, Nature 348,325-328.
Geophys.1. R. Astron. Soc. 58, 689-715. Phipps Morgan. 1. (l99\). Mid-ocean ridge dynamics: Obser-
McKenzie, D. (1984). The generation and compaction of par- vations and theory, Rev. Geophys., U.S. National Report to
tially molten rock, 1. Petrol. 25, 713 - 765. IUGG. Suppl .. 807-822.
Moores, E. M. (1982). Origin and emplacement of ophiolites, Phipps Morgan, 1., Parmentier, E. M., and Lin, J. (1987).
Rev. Geophys. Space Phys. 20, 735-760. Mechanisms for the origin of mid-ocean ridge axialtopog-
Morton,1. L., and Sleep, N. H. (l985a). A mid-ocean ridge . raphy: Implications for the thermal and mechanical struc-
thermal model: Constraints on the volume of axial hydro- ture at accreting plate boundaries, 1. Geophys. Res. 92,
thermal flux, 1. Geophys. Res. 90, 11345-11353. 12823~ 12836.
Morton, 1. L., and Sleep, N. H. (l985b). Seismic reflections Phipps Morgan, J., and Chen, Y. J. (1993). The genesis of oce-
from the Lau Basin magma chamber, in "Geology and anic crust: Magma injection, hydrothermal circulation, and
Offshore Resources of Pacific Island Arcs-Tonga Re- crustal flow, J. Geophys. Res. 98, 6283-6298.
gion," (D. W. Scholl and T, L, Vallier, eds.), pp. 441-453, Purdy, G. M., and Detrick, R. S. (1986). Crustal structure of
Circum Pacific Council for Energy and Mineral Resources the Mid-Atlantic Ridge at 23°N from seismic refraction
Earth Science Series. studies. J. Geophys. Res. 9I, 3739-3762.
Morton, 1. L., Sleep. N. H., Normark. W. R., and Tompkins, Purdy, G. M., Kong. L. S. L., Christeson, G. L., and Solo-
D. H. (1987). Structure of the Southern Juan de Fuca o mon, S. (1992). Relationship between spreading rate and

Ridge from reflection records. J. Geophys. Res. 92, the seismic structure of mid-ocean ridges, Nature 355,
11315-11326. 815-817.
Murase, T., and McBirney, A. R. (1973). Properties of some Quick, 1. E., and Denlinger. R. P. (1993). Ductile deformation
common igneous rocks and their melts at high tempera- and the origin of layered gabbro in ophiolites, 1. Geophys.
tures, Geol. Soc. Am. Bull. 84,3563-3592. Res. 98, 14015-14028.
Mutter, J.c, Barth, G. A., Buhl, P., Detrick. R. S., Orcutt, 1., Reid. 1. D., Orcutt, 1. A., and Prothero. W. A. (1977). Seismic
and Harding, A. (1988). Magma distribution across ridge- evidence for a narrow zone of partial melting underlying
axis discontinuities on the East Pacific Rise from multi- the East Pacific Rise at 21°N, Geol. Soc. Am. Bull. 88,
channel seismic images, Nature 336, 156-158. 678-682.
Natland, 1. H. (1980). Effects of axial magma chambers be- Rohr, K. M., Milkereit, B., and Yorath, C. J. (1988). Asym-
neath spreading centers on the compositions of basaltic metric deep crustal structure across the Juan de Fuca ridge,
rocks, in "Initial Report of the Deep Sea Drilling Proj- Geology 16, 533-537.
ect," (B. R. Rosendahl and R. Hekinian, eds.), v. 54. Rosendahl. B. R., Raitt, R. w., Dorman, L. M., Bibee, L. D.,
pp. 833-850. Hussong, D. M., and Sutton, G. H. (1976). Evolution of
Nehlig, P., and Juteau, T. (1988). Flow porosities, permeabili- oceanic crust. 1. A physical model of the East Pacific Rise
ties, and preliminary data on fluid inclusions and fossil crest derived from seismic refraction data, 1. Geophys.
thermal gradients in the crustal sequence of the Sumail Res. 81, 5294-5304.
ophiolite (Oman), Tectonophysics i51, 199-221. Ryan, M. P. (1987). Neutral buoyancy and the mechanical evo-
Neumann, G. A., and Forsyth, D. W. (1993). The paradox of lution of magmatic systems, in "Magmatic Processes:
the axial profile: Isostatic compensation along the axis of Physiochemical principles," The Geochemical Society.
the Mid-Atlantic Ridge? 1. Geophys. Res. 98, 17891- Special Publication I, pp. 259-287.
17904. Ryan, M. P. (1994). Neutral-buoyancy controlled magma
Nicolas, A., Reuber. 1., and Benn, K. (1988). A new magma transport and storage in mid-ocean ridge magma reservoirs
chamber model based on structural studies in the Oman and their sheeted dike complex: A summary of basic rela-
ophiolite, Tectonophysics 151, 87 -lOS. tionships, in "Magmatic Systems" (M. P. Ryan, ed.), Aca-
Nicolas, A. (1989). "Structures of Ophiolites and Dynamics demic Press, San Diego.
of Oceanic Lithosphere," Kluwer, New York. Sato, H., Sacks, 1. S., and Murase, T. (1989). The use of labo-
 178 J. Phipps Morgan et al.
ratory velocity data for estimating temperature and partial ing of seismic refraction data, J. Geophys. Res. 85,
melt fraction in the low-velocity zone: Comparison with 1409-1433.
heat flow and electrical conductivity studies, J. Geophys. Stevenson, 1. M. (1992). "Applications of Marine Gravimetry
Res. 94, 5689-5704. to Mid-oceanic Spreading Centers and Volcanos," Ph.D.
Sato, H., and Sacks, I. S. (1990). Magma generation in the Thesis, Univ. of California, San Diego.
upper mantle inferred from seismic measurements in peri- Tapponier, R., and Francheteau, 1. (1978). Necking of the
dotite at high pressure and temperature, in "Magma Trans- lithosphere and the mechanics of slowly accreting plate
port and Storage" (M. P. Ryan, ed.), pp. 277-292. Wiley, boundaries, J. Geophys. Res. 83, 3955-3970.
Chichester, England. Tolstoy, M., Harding, A. 1., and Orcutt, J. A. (1992). An expla-
Scott, D. S. (1992). Small-scale convection and mantle melt- nation for "bull's eye" mantle bouguer anomalies on the
ing beneath mid-ocean ridges, in "Mantle Flow and Melt southern Mid-Atlantic Ridge, EOS Trans. Am. Geophys.
Generation at Mid-ocean Ridges," (1. Phipps Morgan, Union (Suppl.) 73, 495.
D. K. Blackman, and 1. Sinton, eds.), AGU Monograph 71, Toomey, D. R., Solomon, S. c., and Purdy, G. M. (1988). Mi-
327-352. croearthquakes beneath the median valley of the Mid-
Sempere, rc., Macdonald, K. c.. Miller, S. P., and Shure, L. Atlantic Ridge near 23°N: Tomography and tectonics,
(1988). Detailed study of the Brunhes/Matuyama 1. Geophys. Res. 93, 9093-9112.
boundary on the East Pacific Rise at 19°30'S: Implications Toomey, D. R., Purdy, G. M., Solomon, S. C., and Wilcock,
for crustal emplacement processes at an ultra fast spread- W S. D. (1990). The three-dimensional seismic velocity
ing center, Mar. Geophys. Res. 9,1-25. structure of the East Pacific Rise near latitude 9°30'N,
Sinton, 1. M., and Detrick, R. S. (1992). Mid-ocean ridge Nature 347, 639-645.
magma chambers, J. Geophys. Res. 97,197-216. Turcotte, D. L., and Schubert, G. (1982). "Geodynamics:'
Sleep. N. H. (1975). Formation of oceanic crust: Some thermal Wiley, New York.
constraints, J. Geophys. Res. 80,4037 -4042. Turcotte, D. L., and Phipps Morgan, 1. (1992). The physics of
Sleep, N. H. (1978). Thermal structure and kinematics of the magma migration and mantle flow beneath a mid-ocean
mid-ocean ridge axis, some implications to basaltic volca- ridge. in "Mantle Flow and Melt Generation at Mid-ocean
nism, Geophys. Res. Lett. 5, 426-428. Ridges" (J. Phipps Morgan, D. K. Blackman, and 1. Sin-
Sleep, N. H. (1991). Hydrothermal circulation, anhydrite pre- ton, eds.), AGU Monograph 71, pp. 155-182.
cipitation, and thermal structure at ridge axes, 1. Geophys. Vera, E. E., Mutter, 1. c., Buhl, P., Orcutt, 1. A., Harding, A. J.,
Res. 96, 2375-2387. Kappus, M. E., Detrick, R. S., and Brocher, T. (1990). The
Small, C; and Sandwell, D. T. (1989). An abrupt change in structure of 0- to 0.2-m.y.-old oceanic crust at 9°N on the
ridge axis gravity with spreading rate, J. Geophys. Res. 94, East Pacific Rise from expanded spread profiles, 1. Geo-
17383-17392. phys. Res. 95, 15529-15556.
Srnewing, J. D. (1981). Mixing characteristics and compo- Wang, X.. and Cochran. J. R. (1993). Gravity anomalies, isos-
sitional differences in mantle-derived melts beneath tasy, and mantle flow at the East Pacific Rise crest, submit-
spreading axes: Evidence from cyclically layered rocks in ted for publication, J. Geophys. Res.
the ophiolite of North Oman, J. Geophys. Res. 86, White, D. J., and Clowes, R. M. (1990). Shallow crustal struc-
2645-2659. ture beneath the Juan de Fuca ridge from 2-D seismic re-
Solomon, S. C.. and Toomey, D. R. (1992). The structure fraction tomography, Geophys. J. Int. 100,349-376.
of mid-ocean ridges. Annu. Rev, Earth Planet. Sci. 20, Wilcock, W S. D., Solomon, S. C; Purdy, G. M., and Toomey,
329-364. D. R. (1992). The seismic attenuation structure of a fast
Sparks. D. W, and Parmentier, E. M. (1991). Melt extraction spreading mid-ocean ridge, Science 258, 1470-1474.
from the mantle beneath spreading centers, Earth Planet. Wilson, D. S., Clague, D. A., Sleep, N. H.. and Morton, 1. L.
Sci. Lett. 105,368-377. (1988). Implications of magma convection for the size and
Sparks, D. W, and Parmentier. E. M. (1994). The generation temperature of magma chambers at fast spreading ridges,
and migration of partial melt beneath spreading centers, in J. Geophys. Res. 93, 11974-11984.
"Magmatic Systems," (M. P. Ryan, ed.), Academic Press, Wilson, D. S. (1992). Focussed upwelling beneath mid-ocean
San Diego. ridges: Evidence from seamount formation and isostatic
Spudich, P., and Orcutt. 1. A. (1980). Petrology and porosity compensation of topography, Earth Planet. Sci. Lett. 113,
of an oceanic crustal site: Results from wave form model- 41-55.
Chapter 8 Deep Structure of Island Arc Magmatic Regions
as Inferred from Seismic Observations

Akira Hasegawa and Dapeng Zhao

Overview Units

Ttib<. observed arrivaltimefor ith earth-

Recent seismic observations provide possible evidence quakeat jth station
for deep-seated magmatic activity in some of the sub- Teal
'1
calculated arrival timebasedon a ve-
duction zones of the world. Tomographic inversions for locitymodel
seismic wave velocity structures delineate low-velocity
V seismic wave velocity km. S-1
zones in the crust and in the mantle wedge beneath ac-
V, seismic wave velocity at /lthgrid km- S-1
tive volcanoes. Seismic attenuation tomography also
delineates similar zones of low-Q value in the crust and point
mantle wedge beneath active volcanoes, although they d column vectorconsisting of travel
are less clearly imaged due to lower spatial resolution. timeresiduals
The most typical example is that beneath the north- e errorvector
eastern Japan arc, where inclined P-wave low-velocity e, higher-order termof perturbation and
zones have been clearly imaged using data acquired observation errors
through microearthquake observations with dense net- II depthfrom the Earth's surface km
works. The low-velocity zones with 2-6% velocity (; travel timeresidual for ith earthquake
lows are continuously distributed from the upper crust at jth station
right under active volcanoes to a depth of 100-150 km
D.II, correction term for focal depthof ith
in the mantle wedge. their thicknesses being about earthquake
50 km. They are approximately parallel to the dip
~m column vectorconsisting of correc-
(-30°) of the underlying subducted Pacific plate and
tion terms for source andmedium
their lower edges are 30-60 km apart from the top of parameters
the subducted plate. These low-velocity zones probably
st; correction term fororigintimeof ith
reflect the pathway of magma ascent from a deeper part
earthquake
of the mantle to the Earth's surface. along which addi-
~F" correction term for seismic wave ve-
tional evidence for deep magmatic activity has been
locity at »th grid point
found.
~¢i correction term for latitude of ith
earthquake
Notation
Units ~A, correction term for longitude of ith
earthquake
C" coefficient of powerseries expressing ¢ latitude 0, degrees
configuration of ith velocity
A longitude 0, degrees
discontinuity
G matrixconsisting of partial deriva-
tives of travel time
Hi depthto ith velocity discontinuity km Introduction
Q quality factorthat is a dimensionless dimensionless
measure of anelasticity New oceanic plates are formed at mid-ocean
P-wave quality factor dimensionless ridges by the upwelling of hot mantle material,
S-wave quality factor dimensionless producing mid-ocean ridge magmatism. Conser-

Magmatic Systems Copyright © 1994 by Academic Press. Inc.

Edited hy M. P. Ryan 179 All rights of reproduction in any form reserved.
180 Akira Hasegawa and Dapeng Zhao

vation of mass suggests that an equal amount of et al., 1992). Possible evidence for deep-seated
material returns downward, back into the Earth's magmatic activity in subduction zones has been
mantle. This consumption of the oceanic plates obtained from these studies. For example, seismic
takes place mainly in subduction zones, where low-velocity zones extending from the upper crust
heavier oceanic plates subduct into the mantle be- just beneath active volcanoes to the mantle wedge
neath the lower-density continental plates. The are clearly delineated from tomographic studies.
subduction of cold, and hence relatively heavy, Taking into account the velocity-temperature re-
oceanic plates causes high seismic and volcanic lation of the mantle rocks obtained from the labo-
activity along the subduction zones. The funda- ratory measurements, this fact suggests the exis-
mental paradox of subduction zone magmatism is tence of an ascending flow of hot mantle wedge
that thus both the high heat-flow values and the material beneath active volcanoes in subduction
abundance of melt generation occur just above the zones. In the present chapter, we briefly describe
relatively cold subducting plate, which is consid- some of the results of these recent seismological
ered to be an enormous heat sink (e.g., Uyeda, studies on the deep structure of island arc mag-
1982; Davies and Stevenson, 1992). matic regions.
Seismic waves, passing through the Earth's in-
terior and finally arriving at seismic stations on
the surface, provide information on the physical Three-Dimensional Seismic Velocity Structure
properties of the materials along their ray paths
within the Earth. Many studies using seismic A number of studies of the three-dimensional seis-
waves have been conducted for estimating both mic velocity structure on both local and regional
the seismic wave velocity structure and the seis- scales have been conducted in various regions of
mic attenuation (anelasticity) structure of the the world since the pioneering works of seismic
Earth's interior. Seismic properties of rocks and tomography by Aki and Lee (1976) and Aki et al.
minerals, which compose the crust and the mantle, (1977). The most thorough studies of the three-
at high pressures and temperatures have also been dimensional seismic velocity structure of sub-
studied in the laboratory. Laboratory measure- duction zones conducted so far are those of the
ments show that seismic wave velocity in the Japanese subduction zone, partly because dense
mantle rocks decreases with increasing tempera- seismic networks have been deployed by several
ture, and in particular, it drops abruptly when the institutions in this region. On a regional scale,
degree of partial melting exceeds some threshold Hirahara (1977) and Hirahara and Mikumo (1980)
amount (e.g., Murase and Kushiro, 1979; Murase investigated the three-dimensional P-wave veloc-
and Fukuyama, 1980; Sato and Sacks, 1990). ity structure of the crust and upper mantle beneath
Seismic anelasticity in peridotite also depends the Japanese Islands by using teleseismic and lo-
on temperature, and Q values decrease rapidly cal earthquake data, and pointed out the existence
with increasing temperature even in the subsoli- of a P-wave high-velocity zone corresponding to
dus temperature range (Sato and Sacks, 1990). Of the subducting Pacific plate.
course, in some cases, decreases in seismic wave The inclined high-hvelocity zone correspond-
velocity or in the Q value are caused not by high ing to the subducting plate has also been detected
temperatures but by other factors such as chemi- by the work of Kamiya et al. (1989) and Zhou
cal composition changes. Nevertheless, estima- and Clayton (1990), who estimated the three-
tions of the detailed seismic structure are critical dimensional P-wave velocity structure to a depth
to develop an understanding of the thermal and of 1200 km beneath the Japanese Islands by using
mechanical state within the Earth, and they thus teleseismic data. On the basis of the estimated to-
yield important information on the deep structure mographic images, they discussed the prospects
of magmatic regions. for further penetration of the subducting Pacific
Recent seismological studies, and new seismic plate into the lower mantle, although the conclu-
imaging techniques such as seismic tomography, sions of the two studies opposed each other. Van
have revealed a more precise image of the seismic der Hilst et al. (1991) have imaged the P-wave
structure of the subducting plate and the mantle velocity structure of the mantle to a depth of
wedge above it (e.g., Hasegawa etal.. 1991; Zhao 1200 km beneath northwest Pacific island arcs by
 Figure 1 Horizontal section of
thefractional P-wa ve velocity per-
turbations at the 40-km-depth
level beneath northeastern Japan
(Hasegawa et al.. 1991; Zhao et
al.. 1992). Velocity perturbations
are shown as percentages by color
scale on the left. Red and blue
indicate low and high P-wave
velocities, respectively. Solid tri-
angles, crosses. and open ~q uares
are the locations of active volca-
nne s, deep low-frequency
microearthquakes. and midcrustal
S-wave reflectors. respectively.

Figure 2 Vertical cross sections of fractional P-wave velocity perturbations along the lines (a) AA', tb) BB', and (c) CC' in the
insert map (Hasegawa et al.• 1991: Zhao et al., 1992). Velocity perturbations are shown as percentages according to the same color
scale as in Fig. l , Locations of the trench axis. active volcanoes, and land area are shown atthe top by inverted triangles, red triangles,
and thick horizontal lines, respectively. The Conrad and Muho discontinuities and the top and bottom of the subducting Pacific plate
are denoted by thick lines. Open and red circles denote microearthquakes and deep low-frequency microearthquakes within a
60-km width along each line.
 8. Deep Structure of Island Arc Magmatic Regions 181
using a more realistic background earth model volcanoes in northeastern Japan. The results of
and surface-reflected seismic phase data as well as this study are described in some detail in the next
direct phase data. Their result has yielded higher- section because they provide the most physically
quality images of the subducted Pacific plate and obvious and most typical images of the crust and
has shown that the subducted plate beneath the upper mantle velocity structure beneath subduc-
northeastern Japan arc and the Izu-Bonin arc does tion zones.
not sink into the lower mantle but is deflected at the Compressional wave low-velocity zones in the
boundary between the upper and lower mantle. crust and in the mantle wedge beneath active vol-
Their images of slab deflection at the 670-km dis- canoes similar to those in the Japanese Islands
continuity are in good agreement with the result of have also been found in the Cascadia subduction
Fukao et al. (1992), who have also yielded higher- zone and in the Alaska-Aleutian subduction zone,
quality images by solving simultaneously for up- where extensive studies of three-dimensional seis-
dates of the one-dimensional background earth mic velocity structures on regional or local scales
model and for aspherical variations in P-wave have been initiated. Inversion of teleseismic P-
velocity. wave arrival time data has led to the successful
Many tomographic studies on a local scale detection of an inclined high-velocity zone that
have been carried out in several regions beneath corresponds to the subducting Juan de Fuca plate
the Japanese Islands by using arrival time data beneath Washington and Oregon (Michaelson and
from local earthquakes (e.g., Miyamachi and Mo- Weaver, 1986; Rasmussen and Humphreys, 1988;
riya, 1984; Nakanishi, 1985; Hasemi et al., 1984; Harris et al.. 1991). Harris et al. (1991) have also
Obara et al., 1986; Horie and Aki, 1982; Ishida detected a low-velocity zone in the crust and
and Hasemi, 1988; Hirahara et al., 1989). These partly In the uppermost mantle beneath the Crater
studies have also indicated the existence of the in- Lake volcano region. By inverting teleseismic
clined high-velocity zones corresponding to the P-wave travel time data, Benz et al, (1992) have
subducting Pacific and Philippine Sea plates be- imaged a steeply dipping portion of the subduct-
neath the Japanese Islands. Hirahara et al. (1989) ing Gorda plate as a +5% velocity high to a depth
investigated the P-wave velocity structure to a near 270 km and a low-velocity zone in the crust
depth of about 200 km beneath central Japan and and in the mantle wedge to a depth of about
detected low-velocity bodies in the mantle wedge 100 km beneath the active volcanic field that
beneath the active volcanoes. One of these low- includes Mt. Lassen, of northern California. In re-
velocity bodies coincides with an S-wave aniso- lated Cascadia work, Iyer et al. (1990) have im-
tropic body as estimated from shear wave splitting aged the three-dimensional distribution of low-
analyses, suggesting the presence of magma-filled velocity regions beneath the Newberry volcano in
crack alignment in the body (Ando et al., 1983). Oregon to a depth of 26 km by the tomographic
Tomographic studies by Hasemi et al. (1984) inversion of explosion-generated compressional
and by Obara et al. (1986) give a detailed three- wave travel time residuals.
dimensional P-wave velocity structure to a depth Zhao and Christensen (1992) have applied the
of about 200 km beneath northeastern Japan and tomographic approach of Zhao et al. (1992) to lo-
have revealed low-velocity zones continuously cal earthquake data in central and southern Alaska
distributed in the crust and in the mantle wedge and have estimated the three-dimensional seis-
just beneath the active volcanoes. mic velocity structure of the crust and upper man-
Zhao et al. (1992) have obtained more distinct tle to a depth of 200 km. They have clearly delin-
P-wave tomographic images of the crust and the eated the subducting Pacific plate having a 4-6%
upper mantle beneath northeastern Japan by de- P-wave velocity high and a thickness of about
veloping a new tomographic method (Figs. 1 and 50 km. Low-velocity zones are also imaged in the
2). This study has updated the works by Hasemi crust as well as the mantle wedge beneath active
et al. (1984) and Obara et al. (1986) by improving volcanoes. Although crustal low-velocity zones
the resolution. It has clearly delineated P-wave and mantle wedge low-velocity zones beneath
low-velocity zones, which are inclined to the west volcanic fields have not been clearly imaged in
and are continuously distributed from the upper- other subduction zones, future studies of seismic
most mantle to the upper crust beneath the active tomography on regional or local scales with high
182 Akira Hasegawa and Dapeng Zhao

spatial resolution wil1 open the possibility for de- polating the depths at four surrounding gnd nodal
tection, just as that described in this report. points.
Three-dimensional grids are arranged individ-
ually in every layer bounded by the discontinui-
P-Wave Velocity Structure beneath ties, and the velocity at each grid point is taken to
the Northeastern Japan Arc be an unknown parameter. A velocity V(</J, A, h)
at any point (</J, A, h) in the modeling space can
In this section, a P-wave tomographic study in the be calculated by linearly interpolating the veloci-
northeastern Japan arc (Hasegawa et al., 1991; ties V(</Ji' Ai' h k ) at eight surrounding grid points
Zhao et al., 1992) is described briefly as an ex- (</J i, Ai' hk) as
ample showing a typical image of the crust and
V(</J, A, h) =
upper mantle in subduction zones. Details of the
tomographic inversion are shown in Zhao et al.
(1992).
Most of the conventional tomographic studies
conducted on local scales have neglected the ef-
fect of the complex shape of seismic velocity dis-
continuities such as the Moho, the Conrad discon-
tinuity, and the upper boundary of the subducting
plate itself. This neglect distorts the estimated to-
mographic images, especial1y near the discontinu-
ities. Calculated ray paths of seismic waves based
on a simple one-dimensional velocity model usu-
ally adopted in the conventional tomographic stud- where h is the depth from the Earth's surface. An
ies deviate considerably from the real paths at efficient three-dimensional ray tracing algorithm
large hypocentral distances and in highly inhomo- that iteratively uses a pseudo-bending technique
geneous regions such as subduction zones. This (Um and Thurber, 1987) and Snell's law has been
deviation also seriously distorts the estimated to- developed. This algorithm calculates the ray paths
mographic images. and the travel times of seismic waves rapidly and
A new method of seismic tomography has been accurately in the general velocity structure men-
developed by Zhao et al. (1992) to solve these tioned above.
problems. This method copes with a general veloc- Starting with the initial hypocenter locations,
ity structure with complex velocity discontinuities the origin times for a set of earthquakes, and an
in the modeling space and with three-dimensional initial velocity model, the correction terms for
velocity variations in each layer bounded by the source and medium parameters are estimated it-
velocity discontinuities. Depth distributions of eratively in a way that the observed arrival time
the velocity discontinuities are expressed in two data are best explained in the least-squares sense.
ways. One is to define the velocity discontinuity The observation equation is written as
by using power series of latitude and longitude.
The depth to the ith velocity discontinuity Hi is Teal
IJ
+ (OJ)
od»}:
A'" +
'f'1
(OJ)
aA..
All.
I
expressed as If'IJ IJ

Hi(</J, A) = Cd + C2</J + C3A

.
+ C i4</J2 (1)
+ (OJ)
ah ii
s», + sr; (3)
+ C s A2 + C6</JA + ...,
aT
where </J and A are latitude and longitude, respec- + 2: -AV
n aV n
n + eij'
tively, and Ci are coefficients of power series for
the ith discontinuity. The other approach is to de- where nr is the observed arrival time for ith
fine the discontinuity by using a two-dimensional earthquake at jth station; Til is the calculated ar-
grid. Once a depth distribution of grids is given rival time based on the velocity model; </Ji' Ai, hi'
for a discontinuity, the depth to the discontinuity TOi are the latitude, longitude, focal depth, and ori-
at any location can be calculated by linearly inter- gin time of ith earthquake; A</Ji' AA i, Ah i , ATo i
 8. Deep Structure of Island Arc Magmatic Regions 183
are the correction terms for the above parameters; those of several other national universities in Ja-
(iJT/iJ<!J);j, (iJT/iJA)ij, (iJT/iJh)ij are the partial de- pan (Hokkaido, Hirosaki, Tokyo, Nagoya, Kyoto,
rivatives of travel time with respect to latitude, Kochi, and Kyushu University). The arrival time
longitude, and focal depth; Vn is the velocity at data used are 42,494 first P-wave data, 8141 first
nth grid point; Ll V n is the correction term for ve- S-wave data, and 284 P-to-S and S-to-P converted
locity at 11th grid point; (iJT/iJv n ) is the partial wave data at the velocity discontinuities from
derivative of travel time with respect to velocity 1200 shallow, intermediate-depth, and deep earth-
parameters; and eij is the higher-order term of per- quakes. The medium under study is divided into
turbation and observation errors. Travel times and four layers by the Conrad discontinuity, the Moho,
ray paths are calculated by the three-dimensional and the upper boundary of the subducted Pacific
ray tracing algorithm. Partial derivatives with re- plate. The four layers correspond to the upper
spect to hypocenters can be calculated analyti- crust, the lower crust, the mantle wedge, and the
cally. Velocity parameter derivatives are calcu- mantle below the upper plate boundary. The depth
lated by using a linear interpolation function (2) distributions of the Conrad, the Moho, and the top
expressing the velocity field. of the plate are fixed and expressed by continuous
The travel time residual tij is written as functions of spatial locations so as to coincide
with the result of previous studies (Hasegawa
(4)
et al., 1983; Matsuzawa et al., 1986, 1990; Zhao
The travel time residuals form a whole set of data et al., 1990). One, one, sixteen, and sixteen layers
with a column vector d of dimension N, where of. grid nets are arranged in the upper crust, the
N is the number of arrival time data. Correction lower crust, the mantle wedge, and the mantle be-
terms for source and medium parameters, defined low the upper plate boundary, respectively. The
by a column vector Llm, are expressed as separation between grid points is 25-33 km in
both vertical and horizontal directions.
Llm T = (Ll<!J" LlA" Sh, LlTo" ..
Two kinds of resolution tests are made to eval-
Ll<!JM, LlAM. LlhM, LlToM, (5)
uate the resolution of the tomographic images ob-
LlV" LlV2 , • • • , LlVK ) ,
tained. One is checkerboard resolution tests with
where M and K are the numbers of earthquakes various wave lengths of velocity change, where
and grid points, respectively. The data vector d the basic idea of the testing is given in Humphreys
and the unknown parameter vector Llm are re- and Clayton (1988). In the tests, positive and neg-
lated as ative velocity perturbations are assigned at regular
intervals to the three-dimensional blocks (here,
d = GLlm + e, (6) grid points) of a homogeneous velocity model.
where e is an error vector, and G is a matrix with Inversion of synthetic arrival time data calcu-
dimension N X (4M + K) whose elements con- lated from this checkerboard velocity model by
sist of the partial derivatives. There are several ap- using the inversion algorithm provides images
proaches to solving Eq. (6). Here, a conjugate gra- from which it is easy to understand where the
dient solver, the LSQR algorithm of Paige and resolution is good or poor. Figure 3 shows the re-
Saunders (1982), is used to solve the extremely sult of the checkerboard resolution test. P-wave
large and sparse system of observation equations velocity perturbations obtained by the resolution
arising from the inversion problems. The algo- test are plotted on the horizontal sections at depths
rithm has been used by several researchers (e.g., of 10, 24, 40, 65, 90, and 115 km, respectively.
Nolet, 1985; Spakman and Nolet; 1988; van der The originally assigned patterns of the checker-
Hilst et al., 1991) and is confirmed to be an effi- board are well reconstructed for the study area.
cient algorithm to solve large inversion problems. Reconstructed amplitudes of velocity anomalies
Zhao et al. (1992) have obtained P-wave to- are more than 80% of the original amplitudes for
mographic images of the crust and upper mantle most grid points.
to a depth of 200 km beneath northeastern Japan The other resolution test is to take the tomo-
by applying the method described earlier to the graphic images obtained from the actual data set
arrival time data of seismic waves observed by as the synthetic velocity model. Comparing the in-
the seismic networks of Tohoku University and verted images from the synthetic arrival time data
184 Akira Hasegawa and Dapeng Zhao

140E 142E 140E 142E 140E

0-3%
o
o
o

40N
Oo/c

•
•
•
·3%
38N

10km 24km

142E 140E 142E 140E

0-3%
o
o
o
40N
0%

•
•
•
• 3%
38N

Figure 3 Result of the checkerboard resolution test plotted on horizontal sections at the 10-,24-,40-,65-,90-, and 115-km-
depth levels (Zhao et 01., 1992). In this test alternating positive and negative velocity perturbations are assigned to the three-
dimensional grid points of a homogeneous velocity model. Inverted P-wave velocity perturbation for this synthetic checkerboard
velocity model is shown at each grid point by the scale on the right.

with the original ones, we can see how the origi- locity structure of the crust and upper mantle in
nal velocity structure is reconstructed. The result, northeastern Japan is shown in Figs, I and 2.
which is not shown here, indicates that the origi- Figure I shows a horizontal section of fractional
nal tomographic images are well reconstructed for P-wave velocity perturbations at a 40-km depth.
most parts of the study area, although the images Figures 2A-2C are three vertical cross sections of
are distorted to some degree at the edge of the area fractional P-wave velocity perturbations along the
as expected. These resolution tests indicate that a profiles of AA', BB', and CC' in the inset map,
meaningful solution is accurately obtained in the which are nearly perpendicular to the trench axis,
present tomographic study, The fractional velocity perturbation is from the
The estimated three-dimensional P-wave ve- mean value of estimated velocities at each given
 8. Deep Structure of Island Arc Magmatic Regions 185
depth level and is shown by the color scale. Red three vertical sections of velocity perturbations
and blue colors correspond to low and high ve- (Figs. 2A-2C). The vertical sections further show
locities, respectively, the perturbation scale being that the low-velocity zones in the crust and the up-
from - 6 to 6%. Microearthquakes located within permost mantle wedge beneath active volcanoes
a 60-km width along each profile are plotted as dip to the west and extend to a depth of 100-
circles in Figs. 2A-2C. Also shown by the thick 150 km. Although the work by Suyehiro and Sacks
curves in the figure are the locations of the Conrad (1983) and that by van der Hilst et al. (1991) sug-
discontinuity, the Moho, and the top of the sub- gest the possibility that the low-velocity zones
ducting plate, which are fixed in the tomographic continue to depths larger than 100-150 km, the
inversion procedure. present result shows this is not the case. The low-
An inclined high-velocity zone corresponding velocity zones are nearly parallel to the dip of the
to the subducting Pacific plate is clearly delin- subducting Pacific plate. This feature can be seen
eated in all the three vertical sections shown in in all the vertical sections along the three profiles
Fig. 2. The bottom of the high-velocity Pacific perpendicular to the trench axis. Davies (this vol-
plate can be clearly recognized, the estimated lo- ume) has computed the thermal structure and the
cation of the bottom also being drawn by a solid flow field induced in the slab and upper mantle
(and broken) curve in the figure. The thickness of wedge to 140 km depth in a general treatment of
the subducting Pacific plate can be estimated to be subduction zone magma generation.
80-90 km. A reliable estimate of the plate thick-
ness has not been made from the previous tomo-
graphic studies (e.g., Hirahara, 1977; Hirahara Three-Dimensional Seismic
and Mikumo, 1980; Kamiya et al., 1989) because Attenuation Structure
of the inherent lower spatial resolution. Umino
et al. (1990) have detected a reflected and S-to-P- The seismic attenuation (anelasticity) structure of
converted wave at the bottom of the subducting the crust and the upper mantle is estimated from
Pacific plate in seismograms of both intermediate- observed data of seismic wave attenuation, which
depth and deep earthquakes. They have estimated provides additional information on the physical
the thickness of the plate to be 80-90 km by using properties of the subduction zone magmatic sys-
arrival times of this reflected wave, which agrees tem. Highly attenuated seismic waves have been
well with the estimation from the tomographic found beneath several volcanic fields in the world,
inversion. The bottom of the subducting Pacific suggesting the presence of magmatic or hydro-
plate as imaged by the inversion (Fig. 2) has a thermal bodies beneath active volcanoes. The de-
sharp velocity contrast, which again agrees with tection and location of strongly attenuating bodies
the existence of the reflected SP wave at the bot- can be achieved by using a method similar to seis-
tom. Figure 2 also shows that the earthquakes mic velocity tomography. Attenuating bodies re-
forming the double seismic zone (Hasegawa et al., duce the amplitude or change the spectral content
1978) occur in the upper half of the subducting of seismic waves that pass through them, just
Pacific plate and that the lower half of the plate is as low-velocity bodies increase the travel time
evidently incapable of generating earthquakes. of seismic waves. Tomographic methods used in
In northeastern Japan, many active volcanoes travel time tomography can also be applied to am-
are distributed on the island arc's surface, occur- plitude data or to amplitude spectrum data to de-
ring primarily along the volcanic front that passes termine the attenuation structure of the medium.
through the middle of the land area and runs par- Highly attenuating bodies have been tomo-
allel to the trench axis. As seen in Fig. I, active graphically imaged mainly in the upper crust
volcanoes plotted at the top as red triangles are beneath active volcanoes or beneath geothermal
mostly located above the low-velocity zones in the areas in western North America (e.g., Young and
uppermost mantle wedge. Horizontal sections of Ward, 1980; Ho-Liu et al., 1988; Evans and Zucca,
velocity perturbations in the crust, although not 1988; Clawson et al., 1989; Iyer et al., 1990). For
shown here, indicate that the low-velocity zones attenuation tomography, Ho-Liu et al. (1988) used
are also distributed in the crust beneath active vol- observed S- to P-wave amplitude ratios determined
canoes. This distribution is partly seen on the from records of 16 earthquakes in the Coso-
186 Akira Hasegawa and Dapeng Zhao

Indian Wells region, southern California. The at- section of the estimated Q; 1 structure of the cen-
tenuation inversion with a small block size of 2 X tral Kanto-Tokai area. The upper boundaries of
2 X 0.2 km delineates a highly attenuating body the subducting Pacific plate and the Philippine Sea
(S-wave quality factor Qs == 30) at depths of 3- plate are indicated by large solid circles in this fig-
5 km beneath the Coso-Indian Wells region. The ure. Low-Qp values (shaded area) are clearly seen
location of this body coincides with a slow P-wave in the mantle wedge on the back-arc side of the
velocity anomaly mapped by Walck and Clayton volcanic front, which is shown by a solid triangle
(1987), suggesting the existence of a magmatic or at the top. Finally, Sato and Sacks (1990) have es-
hydrothermal body beneath this region. timated the thermal structure of the mantle wedge
Evans and Zucca (1988) determined the P-wave beneath northeastern Japan on the basis of Qs and
velocity and attenuation structure of the upper Qp values determined by Umino and Hasegawa
crust beneath Medicine Lake volcano, north- (1984). For Qs = 96 ± 36 and Qp = 210 ± 80,
ern California, by using a high-resolution active an approximate temperature of 980 ± 30° C has
source P-wave travel time and attenuation tomog- been inferred at a depth of 31-56 km.
raphy method. The P-wave attenuation structure Similar attenuation structures of subduction
was estimated by inverting observed amplitude zones have been estimated even from observed
spectral ratios between stations. A P-wave low- data of the seismic intensity, that is, the degree of
velocity low-Q body not larger than a few tens ground shaking caused by earthquakes (Hashida
of cubic kilometers in volume is tomographically and Shimazaki, 1987; Hashida, 1989; Satake and
imaged at depths 1-3 km beneath the eastern cal- . Hashida, 1989). Hashida and Shimazaki (1987)
dera, possibly corresponding to the magma cham- and Hashida (1989) applied a three-dimensional
ber feeding several of the youngest summit silicic block inversion method to observed seismic inten-
eruptions (Evans and Zucca, 1988). By using a sity data and estimated the attenuation structures
method similar to that of Evans and Zucca (1988), of the crust and the upper mantle beneath north-
Clawson et al. (1989) imaged a relatively high eastern Japan and beneath whole parts of the Japa-
P-wave attenuation body (P-wave quality factor nese Islands, respectively. Their results show the
Qp == 40) in the upper crust beneath the northeast- presence of low-Q zones in the mantle wedge (to
ern Yellowstone caldera, which also corresponds a depth of ~90 km) beneath active volcanoes and
to a slow P-wave velocity anomaly. high-Q zones corresponding to the subducting
Attenuation tomographic studies of the mantle Pacific plate. Low-Q zones in the mantle wedge
wedge also have been performed, although their beneath active volcanoes are also detected from
spatial resolution is much lower than those of the inversion of seismic intensity data to a depth of
upper crust described previously. By applying a -100 km beneath North Island, New Zealand
three-dimensional block inversion method to a set (Satake and Hashida, 1989). Figure 5 shows a
of observed S- to P-wave spectral ratios, Umino NW-SE vertical cross section of the New Zealand
and Hasegawa (1984) estimated the attenuation attenuation structure that they estimated along a
structure of the crust and upper mantle beneath profile nearly perpendicular to the trench axis. A
northeastern Japan. They obtained low-Q, values low-Q zone imaged beneath the volcanic front
in the crust beneath the active volcanoes, a low-Qs (triangle) dips to the northwest and attains a depth
value (Qs == 150) in the mantle wedge on the of about 100 km.
back-arc side of the volcanic front, an interme- Comparison of the estimated attenuation struc-
diate-Qs value (Qs == 500-800) in the mantle tures with the seismic velocity structures shows
wedge on the fore-arc side of the volcanic front, that low- (high-) Q regions generally correspond
and a high-Q, value (Qs == 1500) in the subduct- to low- (high-) velocity regions, although the im-
ing Pacific plate itself. ages obtained of low-Q zones in the mantle wedge
A similar attenuation structure in the crust and beneath the active volcanoes and the high-Q zones
in the upper mantle of the subduction zone has corresponding to the subducted slab are less clear
also been obtained beneath the Kanto-Tokai area, than those resolved by travel time tomography.
in central Japan, by Sekiguchi (1991). The method This difference is due to the lower spatial reso-
he used is similar to that of Umino and Hasegawa lution of the attenuation tomography. Consider-
(1984). Figure 4 shows an east-west vertical cross ing the velocity-temperature relation and the Q-
 8. Deep Structure of Island Arc Magmatic Regions 187

, .
0 - -' 0-
o~",,\
or
'1 _

0
I

'_1'_
.. \•...
\ '

/ I \0 <,o~
" " 0 <, , ~,'
00
13 .
B q~
'b
'........
./
I
/
,
G 8/
-------- -0.0040
-------- -0.0020
( 0";:0 0 ~ ~:
----- 0.0000
O . .' ,:. ".:,' , '." ;,'l',.. ~ ,.'/'"~.,'
~ • • •• \ i - j 0 , 0 0.0020
~2-... 0.0040
,-.. • ~/ :.....JO
, ' ..•.... ,-- J'io ,gil 0.0060
\r\\ >i ( oro oi"
\ 0.0080
·......i.. ·.·. ' '.'. './ ......i..... - ' .-..,';14,'~o,
(
.
...· \)
0
i '--

;.
' ," . • .. -<0"./0/
,
,
... / / , , - , .
'-_.-, \
"
''XI'
, " ('

\,/,. ....',

"/:
( ,
'1_)'.
(
i. /
1"
• '--~/
r', I . / /\
, , \ ' / ,/0
13 I

I
.,'---
I'
I' ...
\.
...
... '

//, ,I. ){,' ._, .

(,',' :e{_,1' . .
"" I' ,:.,/

:/ .." /1-
{

'fA / I

300 -
.~' .
.....- - - - - - - - - - - - - - - - '
137.00 141.00
Figure 4 E-W vertical cross section of the Qp I structure in the Kanto-Tokai area of central Japan. Contours indicate estimated
values of Q p, (at an interval of 0.002). Shaded areas indicate low-Qp regions (Qp I > 0.002). Large solid circles denote the tops
of the subducting Pacific and Philippine Sea plates. The triangle at the top is the location of the volcanic front. Microearthquakes
located within a 33-km width along the profile are also denoted by open circles. (After Sekiguchi, 1991.1.

temperature relation of mantle rocks, these seis- Midcrustal Magma Bodies Detected
mic observations -suggest the existence of high by Reflected Seismic Waves
temperatures and partially molten materials in the
low-velocity low-Q zones of the mantle wedge A typical example of magma bodies seismically
beneath active volcanoes. detected so far in the midcrust are those beneath

O..---.....,.....-_.--J_~~.,...,..~--=----,
-0.6 -0.1 1.9 0.6 1.0, 3.4 : 0.1 -0.5 0.6
30 "<, /" - - :. '. ,/ ~
0.0 -2.0 -0.6 2.2,- 5.3 "3.9/ -2.0 -1.1 I.S
80 <,
" "' - L- " , /
, ,
...
-0.7 ,2.5,1.9 -2.7/1.1
",'

130 -- - YH/
~~1.7
180
-~2.3 -2.6 100km
230
0.7 -3.0

o 100 500 600

Figure 5 NW-SE cross section of attenuation structure in the northeastern part of North Island, New Zealand. Numerals are
the estimated deviations from the initial values of the attenuation coefficient in units of 10- 2 S-I. and the contours are drawn at
an interval of 2 X 10 -2 S -'. Locations of the volcanic front and land area are denoted by a triangle and a thick horizontal line
at the top. (After Satake and Hashida, 1989.)
188 Akira Hasegawa and Dapeng Zhao

the Long Valley Caldera (Sanders, 1984; Hill sharp impulsive phase denoted by SxS following
et al., 1985; Dawson et al.. 1990; Iyer et al.. the direct S-wave at the two stations located just
1990). Sanders (1984) estimated the locations and above the earthquake focus. This anomalous phase
configurations of magma bodies beneath the cal- has very large amplitudes and is most clearly de-
dera from S-wave attenuation analyses. The esti- fined on the horizontal component seismograms.
mated magma bodies, i.e., anomalously high S- Phase identification by arrival time analyses
wave attenuation bodies, are located at depths of shows that this phase is a reflected S-wave (SxS
5 -15 km. The magma bodies inferred from tele- phase) from a strong velocity discontinuity exist-
seismic travel time inversions have a volume of ing in the midcrust. A ray path of this SxS phase
500-1000 krn' (Iyer etal.. 1990). Reflected waves is schematically illustrated on a vertical section by
are also found both from the top and from the bot- a bold broken line in Fig. 7. The phase identifica-
tom of the estimated magma bodies (Hill et al.. tion is confirmed by detection of a reflected and
1985; Luetgert and Mooney, 1985; Zucca et al., S-to-P-converted wave (SxP phase) from the same
1987). velocity discontinuity, a signature that can be
Unlike the mid-ocean ridge or the intraplate clearly seen on the vertical component at station
volcanoes, almost no clear evidence for the exis- GZD in Fig. 6. A ray path for the SxP phase is
tence of magma bodies in the midcrust has been also shown by a bold broken and solid line in
reported for subduction zone volcanoes from seis- Fig. 7. Arrival time analyses of the SxS phase
mic observations. An exception is the distinct observed by a dense seismic network show that
S-wave reflectors in the midcrust detected at sev- the reflector is distributed over an area of 15 X
eral locations near active volcanoes in the Japa- 10 km 2 at depths of 8-15 km and becomes shal-
nese Islands. Figure 6 is an example of three- low toward the north at an angle of about 30°, in
component short-period seismograms of a shallow the direction of the Nikko-Shirane volcano (Ma-
microearthquake that occurred beneath Nikko- tsumoto and Hasegawa, 1991).
Shirane volcano in central Japan. One can see a A distinct S-wave reflector in the midcrust,

P, SxP

UDi+~~~~*~~ iF'

~NS+~
EW-+;A~--
U::~I\!~
~NS~~
EW
_J.'fT).~_r,j\~l1~~I~U ~fVV i.,V'~'I~. my"V
I'h~~ ~H~~If~
S" o
I I I
5 sec
I

Figure 6 An example of three-component seismograms of a shallow microearthquake that occurred near Nikko-Shirane vol-
cano, Japan. These seismograms were recorded at two stations, GZD and UCH. located just above the earthquake focus. Arrival
times of direct P- and S-waves are denoted by P and S. Later arrivals. reflected waves from a midcrustal thin magma body, are
clearly seen and are indicated by SxS and SxP.
 8. Deep Structure of Island Arc Magmatic Regions 189
station tors are found in northeastern Japan, and their
locations are plotted as large squares on the hori-

* source
zontal section of P-wave velocity perturbations
(Fig. 1). As is obvious from this figure, all the
/,1 reflectors are located near active volcanoes (red
It tJ. triangles) and/or in or around the P-wave low-
velocity zones (red or yellow areas). The features
of SxS and SxP phases described previously indi-
cate that these are the reflected phases from thin
Figure 7 Schematic illustration of ray paths for a reflected magma bodies existing in the midcrust beneath
S-wave (SxS) and a reflected and S-to-P converted wave (SxP) active volcanoes, or in close proximity to the
from a thin midcrustal magma body. Solid and broken lines
P-wave low-velocity zones.
denote P- and S-waves, respectively.
Additional seismic evidence for deep-seated
magmatic activity in northeastern Japan is the
identical with that described previously was first deep low-frequency microearthquakes that occur
detected and located beneath the central Rio beneath the active volcanoes (Hasegawa et al.,
Grande Rift near Socorro, New Mexico (Sanford 1991; Hasegawa and Yamamoto, 1994). Most
et al., 1973). The large amplitude of the SxS shallow earthquakes beneath the surface in north-
phase relative to the direct S phase and the ratio eastern Japan are known to occur in the upper
of SxP to SxS amplitudes can be explained by a r'5 km of the crust, forming a brittle seismogenic
large velocity contrast across the discontinuity un- zone. Exceptionally deep microearthquakes, well
derlain by very-low-rigidity material such as a below the base of the brittle seismogenic zone,
magma body. Ake and Sanford (1988) estimated also occur at depths of 22 -40 km beneath the land
the thickness of the magma body by modeling area, although their occurrence is rather rare. All
the observed spectra of the reflected phase. The of the 151 deep events found thus far at 10 loca-
results show a thin (-70-m) layer of nonrigid, tions in northeastern Japan have extremely low
low-velocity material underlain by a second, thin dominant frequencies (1.5- 3.5 Hz) for both com-
(-60-m) layer of slightly higher-velocity mate- pressional and shear waves. Figure 8a shows an
rial, the total thickness of the low-velocity region example of three-component short-period seis-
being less than 150 m. mograms selected from the 151 events. Another
Spectral ratios of SxS to direct S phases ob- example of seismograms for an event with nearly
served in the Nikko-Shirane volcanic area have the same epicenter location but with a normal fo-
three peaks in the frequency range 3- 20 Hz, cal depth is shown in Fig. 8b. It is obvious by
which can be explained by a very thin (-I DO-m) comparing the two that the deep event has anoma-
magma body with low-rigidity material (Matsu- lously low dominant frequencies for both the P-
moto and Hasegawa, 1991). A very thin magma and the S-waves. The epicenters of all 151 deep
body model is also supported by the observations low-frequency events are plotted as crosses on the
that direct S-waves, passing vertically through the horizontal section of P-wave velocity perturba-
body and observed at stations just above it, do not tions (Fig. 1). The deep events within a 60-km
have spectra different from those at other stations. width along the profiles AN, BB', and CC' are
These observations indicate that passing vertically plotted as red circles on the vertical sections of
through the body does not attenuate S-waves sig- velocity perturbations (Figs. 2A-2C). The deep
nificantly, suggesting a very thin thickness of the events are clearly isolated from the main activity
body (Hasegawa et al., 1991). of normal focal-depth events and are located ap-
Distinct S-wave reflectors in the midcrust, simi- proximately under the active volcanoes (red tri-
lar to that described above, have been detected so angles) or around the low-velocity zones (red or
far at nine locations beneath the Japanese Islands yellow areas).
(Mizoue, 1980; Mizoue et al., 1982; Horiuchi Focal mechanisms of these low-frequency
et al., 1988; Iwase et al., 1989; Nishiwaki et al., events have not been determined since the mag-
1989; Hori and Hasegawa, 1991; Inamori et al., nitudes of these events are small. A preliminary
1992; Hasegawa et al., 1991). Five of these reflec- estimation by a moment-tensor inversion using
 a H=30km M=1.8

NS

EW

UD

p S

b H=9km M=I.8

NS

EW

UD

p
S

o 10 20 sec

Figure 8 (a) An example of three-component seismograms of a deep (focal depth H = 30 km), low-frequency microearthquake
that occurred beneath Hiuchidake volcano and (b) that of a normal focal depth (H = 9 km) microearthquake with nearly the
same epicenter location. These seismograms are recorded by short-period (I s) seismographs at two stations, MAl and YMZ,
about 45 and 75 km east of the two earthquakes, respectively. Arrival times of P- and S-waves are indicated by P and S,
respectively.
 8. Deep Structure of Island Arc Magmatic Regions 191
p- and S-waveforms prefers a non-double-couple et al., 1992), and beneath the Sakurajima volcano
mechanism to a normal double-couple mecha- in southern Kyushu (Goto et al., 1992). All these
nism, which is expected for ordinary earthquakes events are located beneath active volcanoes and
caused by fault slip (Kosuga and Hasegawa, 1992). have features similar to those detected in north-
This fact and the anomalous features of these eastern Japan. Similar deep low-frequency mi-
events described previously suggest that these croearthquakes have not been detected to date in
deep low-frequency events are generated by deep- other subduction zones of the world, possibly due
seated magmatic activity, such as the rapid move- to the lower overall resolution of seismic networks
ment of magma accompanied by a fracture of in those regions. The extensive distribution of the
crustal or mantle rocks, either in the lower crust or deep low-frequency events throughout Japan sug-
in the uppermost mantle beneath the active vol- gests that the occurrence of deep low-frequency
canoes. The distinct S-wave reflectors described events beneath active volcanoes is a phenomenon
previously are located at shallower depths in the common to all the subduction zones, and a fun-
midcrust above these low-frequency events. damental indicator of relatively deep magmatism.
Studies detecting deep low-frequency micro-
earthquakes at other locations in the Japanese Is- Deep Structure of Arc Volcanoes
lands have been reported in rapid succession.
These microearthquakes occur beneath the Toka- The cross-arc vertical cross-sectional model of the
chi volcano in Hokkaido (Suzuki et al., 1992), be- crust and upper mantle beneath northeastern Ja-
neath the Izu-Oshima volcano in central Japan pan, inferred from the recent seismic observations
(Ukawa and Ohtake, 1987), beneath the Hida described in the previous sections, is shown sche-
mountainous range in central Japan (Yamauchi matically in Fig. 9. It illustrates the generalized

seismogenic zo

upper mantle

Figure 9 Schematic illustration of across-arc vertical cross section of the crust and upper mantle beneath the northeastern
Japan arc. The subduction of the oceanic plate with a high seismic velocity (high V) and a low seismic attenuation (high Q)
generates a secondary mechanically induced convection in the overlying mantle wedge. Decompression melting within the
ascending mantle wedge flow produces low V and low Q. Magma within the ascending flow finally reaches the top of the mantle
and is then segregated from the stagnated mantle diapir, which also has low V and low Q. Magma further migrates upward into
the crust due to its positive buoyancy, again producing low V and low Q around it. The temporary storage of magma at midcrustal
levels appears as distinct S-wave reflectors. which are considered to be thin magma bodies. Repeated discharges of magma. by
its further upward migration to the Earth's surface, form the arc volcanoes.
 192 Akira Hasegawa and Dapeng Zhao

deep structure of island arc magmatic regions. of the active volcanoes are distributed on the
The subduction of the oceanic plate generates a back-arc side of the volcanic front and approxi-
secondary mechanically induced convection in the mately parallel to it. The two separate chains of
overlying mantle wedge (e.g., McKenzie, 1969; volcanoes similar to the northeastern Japan arc are
Sleep and Toksoz, 1973; Toksoz and Bird, 1977). present in most subduction zones, and systematic
We infer that the low-velocity zones that are con- differences in magma compositions between the
tinuously distributed in the mantle wedge and par- two chains have been pointed out (e.g., Sakuyama
allel to the dip of the subducting plate (clearly de- and Nesbitt, 1986; Marsh, 1979). A volcano be-
lineated in Fig. 2) are images of the ascending longing to the back-arc side volcanic chain can be
flow of hot mantle material from depth and are a seen along the profile BB' in Fig. 2; Chokai vol-
portion of the secondary subduction-induced con- cano is located near the Japan Sea coast. We can
vection. Decompression melting within the as- see from Fig. 2B that a side path of the low-
cending mantle wedge flow produces low seismic velocity zone originating in the mantle wedge ul-
velocities and high attenuations (low Q). Magma timately reaches the root region of this volcano.
within the ascending mantle flow finally reaches Many of the petrological or geochemical stud-
the top of the mantle and is then segregated from ies extensively conducted to date attribute the
the stagnated diapir. The low-velocity zones just magma generation in subduction zones to the pro-
beneath the Moho and imaged by seismic veloc- gressive dehydration of the oceanic plate during
ity tomography are the manifestation of mantle its descent (e.g., McBirney, 1969; Anderson et al..
diapirs, and their magmatic activity may gener- 1976; Kushiro, 1983; Tatsumi et al .. 1983; Ta-
ate deep low-frequency microearthquakes around tsumi, 1989), implying direct vertical ascent of
them, which are inferred to be related to the rapid magma to the volcanoes on the surface rather than
movement of magma. the inclined ascending flow. The presence of the
Magma bodies may further migrate upward two volcanic chains with systematic differences in
into the crust because of their positive buoyancy, their chemical compositions has been explained
at which time they suffer fractional crystallization by this vertical ascent of magma from the sub-
and chemical reaction with the surrounding crustal ducted plate. Strong S-wave reflections at 8 Hz
rocks. This portion of the upward migration of from the top of the subducted plate detected be-
magma in the crust again produces a seismic low- neath the southern Kanto district indicate the ex-
velocity region and the resulting high attenua- istence of a narrow zone containing many liquid
tion: effects that are clearly imaged in vertical bodies (Obara and Sato, 1988). This work offers
cross sections of the P-wave velocity perturba- seismic evidence for the presence of free water or
tions (Fig. 2) and in that of the resulting attenu- melts on the upper surface of the subducted plate.
ation structure (Fig. 5). The temporary storage of If the subduction-induced mantle wedge flow ex-
magma at midcrustal levels appears as distinct S- ists as suggested by the seismic tomographic im-
wave reflectors. This upward migration of magma ages of the mantle wedge, H 20 from the sub-
raises the temperature of crustal materials around ducted plate may not ascend vertically. Davies and
it and consequently causes the local elevation of Stevenson (1992) and Davies (this volume) pro-
the bottom of the brittle seismogenic zone. Magma pose that the combination of vertical ascent of wa-
finally reaches the Earth's surface by its further ter as a free phase and the transport of hydrous
upward migration, possibly along preexisting con- phases by the subduction-induced flow leads to
duits in the upper crust, and repeated discharges the net transport of H 20 being horizontal. across
to the surface form the arc volcanoes. the mantle wedge from the subducted plate. It is
Figure 2 shows that the low-velocity zones in quite probable that this kind of horizontal trans-
the mantle wedge penetrate into the crust and fi- port, as well as a vertical mode of H 20 migration,
nally reach the root regions of the active volca- is actually occurring in th~ source region of sub-
noes at the surface, thus providing a pathway of duction zone magmatism.
magma ascent as described previously. Beneath In any case, the seismic structure imaged by
the northeastern Japan arc, in addition to the main the tomographic studies provides an important
volcanic chain along the volcanic front itself, some constraint for understanding the processes and
 8. Deep Structure of Island Arc Magmatic Regions 193
three-dimensional structure of magma generation frequency earthquakes beneath the volcanoes in Kyushu,
in subduction zones. in "Abstracts, 1992 Japan Earth and Planetary Science
Joint Meeting," p. 243.
Harris, R, A., Iyer, H. M., and Dawson, P, B. (1991). Imaging
Acknowledgments the Juan de Fuca plate beneath southern Oregon using tele-
seismic P wave residuals, 1. Geophys. Res. 96, 19879-
19889.
We appreciate the critical reviews of M. P. Ryan, R. van der Hasegawa, A., and Yamamoto, A. (1994). Deep, low-
Hilst, and two anonymous referees who helped to improve this frequency microearthquakes in or around seismic low-
manuscript. velocity zones beneath active volcanoes in north-eastern
Japan, Tectonophysics, in press.
Hasegawa, A., Umino, N" and Takagi, A. (1978). Double-
References planed deep seismic zone and upper mantle structure in the
northeastern Japan arc, Geophys. J. R, Astron. Soc. 54,
Akc. J, P., and Sanford, A. R. (1988). New evidence for the 281-296.
existence and internal structure of a thin layer of magma Hasegawa, A., Urnino, N., Takagi, A., Suzuki, S., Motoya, Y,
at mid-crustal depths near Socorro, New Mexico, Bull. Kameya, S., Tanaka, K., and Sawada, Y (1983). Spatial
Seism. Soc. Am. 78,1335-1359. distribution of earthquakes beneath Hokkaido and north-
Aki, K., and Lee, W. H. K. (1976). Determination of three- ern Honshu, Japan, J. Seism. Soc. Japan 36, 129-150.
dimensional velocity anomalies under a seismic array us- Hasegawa, A., Zhao, D., Hori, S., Yamamoto, A., and Hori-
ing first P arrival times from local earthquakes. 1. A ho- uchi, S. (1991). Deep structure of the northeastern Japan
mogeneous initial model, J. Geophys. Res. 81, 4381- arc and its relationship to seismic and volcanic activity,
4399. Nature 352,683-689.
Aki, K., Christoffersson. A., and Husebye, E. S. (1977). De- Hasemi, A. H., Ishii, H., and Takagi, A. (1984). Fine structure
termination of the three-dimensional seismic structure of beneath the Tohoku District. northeastern Japan arc, as de-
the lithosphere, J. Geophys. Res. 82, 277-296. rived by an inversion of P-wave arrival times from local
Anderson, R. N., Uyeda, S., and Miyashiro, A. (1976). Geo- earthquakes, Tectonophysics 101, 245-265.
physical and geochemical constraints at converging plate Hashida, T. (1989). Three-dimensional seismic attenuation
boundaries. I. Dehydration in the down-going slab, Geo- structure beneath the Japanese Islands and its tectonic and
phys. J. R. Astron. Soc. 44, 333-357. thermal implications. Tectonophysics 159, 163-180.
Ando, M., Ishikawa, Y., and Yamazaki, F. (1983). Shear wave Hashida, T., and Shimazaki, K. (1987). Determination of seis-
polarization anisotropy in the upper mantle beneath Hon- mic attenuation structure and source strength by inversion
shu, Japan, J. Geophys. Res. 88, 5850-5864. of seismic intensity data: Tohoku District, northeastern Ja-
Benz, H. M., Zandt, G., and Oppenheimer, D. H. (1992). Lith- pan arc, 1. Phys. Earth 35, 67-92.
ospheric structure of northern California from teleseis- Hill, D. P., Bailey, R. A., and Ryall, A. S. (1985). Active tec-
mic images of the upper mantle, 1. Geophys. Res. 97, tonic and magmatic processes beneath Long Valley cal-
4791-4807. dera, eastern California: An overview, J. Geophys. Res. 90,
Clawson, S. R, Smith, R. B., and Benz, H. M. (1989). P wave 11111-11120.
attenuation of the Yellowstone caldera from three- Hirahara, K. (1977). A large-scale three-dimensional seismic
dimensional inversion of spectral decay using explosion structure under the Japan Islands and the Sea of Japan,
source seismic data, J. Geophys. Res. 94,7205-7222. J. Phys. Earth 25, 393-417.
Davies, 1. H. (1994). Lateral water transport across a dynamic Hirahara, K., and Mikumo, T. (1980). Three-dimensional seis-
mantle wedge: A model for subduction zone magmatism, mic structure of subducting lithospheric plates under the
in "Magmatic Systems" (M. P. Ryan, ed.), Academic Japan Islands, Phys. Earth Planet. Int. 21, 109-119.
Press, San Diego, CA. Hirahara, K., Ikami, A., Ishida, M., and Mikumo, T. (1989).
Davies, 1. H., and Stevenson, D. 1. (1992). Physical model of Three-dimensional P-wave velocity structure beneath cen-
source region of subduction zone volcanics, 1. Geophys. tral Japan: Low-velocity bodies in the wedge portion of the
Res. 97,2037-2070. upper mantle above high-velocity subducting plates, Tec-
Dawson, P. B., Evans, 1. R., and Iyer, H. M. (1990). Tele- tonophysics 163,63-73.
seismic tomography of the compressional wave velocity Ho-Liu, P., Kanamori, H., and Clayton, R W. (1988). Appli-
structure beneath the Long Valley region, California, J. cations of attenuation tomography to Imperial Valley and
Geophys. Res. 95, 11021-11050. Coso-Indian Wells region, southern California, J. Geo-
Evans.J. R., and Zucca, 1.1. (1988). Active high-resolution seis- phys. Res. 93, 10501-10520.
mic tomography of compressional velocity and attenuation Hori, S., and Hasegawa, A. (1991). Location of a mid-crustal
structure at Medicine Lake volcano, northern California magma body beneath Mt. Moriyoshi, northern Akita Pre-
Cascade Range, J. Geophys. Res. 93, 15016-15036. fecture, as estimated from reflected SxS phase, J. Seism.
Fukao, Y., Obayashi, M., Inoue, H., and Nenbai, M. (1992). Soc. Japan 44, 39-48.
Subducting slabs stagnant in the mantle transition zone, Horie, A., and Aki, K. (1982). Three-dimensional velocity
J. Geophys. Res. 97,4809-4822. structure beneath the Kanto District, Japan, J. Phys. Earth
Goto, K., Hasegawa, S., and Kanjo, K. (1992). Deep low- 30,255-281.
 194 Akira Hasegawa and Dapeng Zhao

Horiuchi, S., Hasegawa, A., Takagi, A., Ito, A, Suzuki, M., and causes of plate motions, Geophys. J. R. Astron. Soc.
and Kameyama, H, (1988). Mapping of a melting zone 18, 1-32.
near Mt. Nikko-Shirane in northern Kanto, Japan, as in- Michaelson, C. A, and Weaver, C. S. (1986). Upper mantle
ferred from SxP and SxS reflections, Tohoku Geophys. structure from teleseismic P arrivals in Washington and
J, 31, 43-55. northern Oregon, J. Geophys. Res. 91, 2077 -2098.
Humphreys, E., and Clayton, R, W (1988), Adaptation of back Miyamachi, H., and Moriya, T. (1984). Velocity structure be-
projection tomography to seismic travel time problems, neath the Hidaka mountains in Hokkaido, Japan, J. Phys.
J. Geophys. Res. 93, 1073-1085, Earth 32, 13-42.
Inamori, T, Horiuchi, S., and Hasegawa, A. (1992), Location Mizoue, M. (1980). Deep crustal discontinuity underlain by
of mid-crustal reflectors by a reflection method using af- molten material as deduced from reflection phases on
tershock waveform data in the focal area of the 1984 West- microearthquakes seismograms, Bull. Earthq. Res. lnst,
ern Nagano Prefecture earthquake, J. Phys. Earth 40, Unit'. Tokyo 55, 705-735.
379-393. Mizoue, M., Nakamura, 1., and Yokota, T (1982). Mapping of
Ishida, M" and Hasemi, A. H. (1988). Three-dimensional fine an unusual crustal discontinuity by microearthquake re-
velocity structure and hypocentral distribution of earth- flections in the earthquake swarm area near Ashio, north-
quakes beneath the Kanto - Tokai District, Japan, J. Geo- western part of Tochigi Prefecture, central Japan, Bull.
phys. Res. 93, 2076-2094. Earthq. Res. lnst. Unit'. Tokyo 57, 653-686.
Iwase, R., Urabe, S., Katsumata, K., Moriya, M., Nakamura, Murase, T, and Fukuyama, H. (1980). Shear wave velocity
1., and Mizoue, M. (1989). Mid-crustal magma body in in partially molten peridotite at high pressures, Carnegie
southwestern Fukushima Prefecture detected by reflected lnst, Washington Year Book 79,307 -310.
waves from microearthquakes, in "Programme and Ab- Murase, T, and Kushiro, 1. (1979). Compressional wave ve-
stracts ofSeism. Soc. Japan," No. I, p. 185. locity in partially molten peridotite at high pressures, Car-
Iyer, H. M., Evans, J. R., Dawson, P. B., Stauber, D. A., and negie lnst. Washington Year Book 78, 559-562.
Achauer, U. (1990). Differences in magma storage in dif- . Nakanishi, 1. (1985). Three-dimensional structure beneath the
ferent volcanic environments as revealed by seismic to- Hokkaido-Tohoku region as derived from a tomographic
mography: Silicic volcanic centers and subduction-related inversion of P-arrival times, J. Phys. Earth 33, 241-256.
volcanoes, in "Magma Transport and Storage" (M. P. Nishiwaki, M., Morita, Y., Nagare, S., Kakishita, T., Osada,
Ryan, ed.), Wiley, Chichester/Sussex, England. Y., and Nagai, N. (1989), Detection of S-wave reflector be-
Kamiya, S., Miyatake, T, and Hirahara, K. (1989). Three- neath the Matsushiro array, central Japan, in "Programme
dimensional P-wave velocity structure beneath the Japa- and Abstracts of Seism. Soc. Japan," No, I, p. 184.
nese Islands, Bull. Earthq. Res. Inst. Univ. Tokyo 64, Nolet, G, (1985). Solving or resolving inadequate and noisy
457-485. tomography system, J. Compo Phys. 61,463-482.
Kosuga, M., and Hasegawa, A. (1992). Moment tensor analy- Obara, K., Hasegawa, A, and Takagi, A. (1986). Three-
sis of low-frequency microearthquakes, "Abstracts, 1992 dimensional P and S wave velocity structure beneath the
Japan Earth and Planetary Science Joint Meeting," p. 244. northeastern Japan arc, J. Seism. Soc. Japan 39, 201-215.
Kushiro, 1. (1983). On the lateral variation in chemical com- Obara, K" and Sato, H. (1988). Existence of an S wave reflec-
position and volume of quaternary volcanic rocks across tor near the upper plane of the double seismic zone beneath
Japanese arcs, J. Volcanol. Geotherm. Res. 18,435-447. the southern Kanto district, Japan, J. Geophys. Res. 93,
Luetgert, J. H., and Mooney, W D. (1985). Crustal refraction 15037 -15045.
profile of the Long Valley caldera, California, from the Paige, C. C., and Saunders, M. A (1982). LSQR: An algo-
January 1983 Mammoth Lakes earthquake swarm, Bull. rithm for sparse linear equations and sparse least squares,
Seism. Soc. Am. 75,211-221. ACM Trans. Math. So/Mare 8, 43-71.
Marsh, B. D. (1979). Island arc development: Some observa- Rasmussen, J., and Humphreys, E. (1988). Tomographic im-
tions, experiments and speculations, J, Geol. 87,687 - 713. age of the Juan de Fuca plate beneath Washington and
Matsumoto, S., and Hasegawa, A. (1991). Characteristics of western Oregon using teleseismic P-wave travel times,
mid-crustal S-wave reflector in Nikko-Ashio region, in Geophys. Res. Lett. 15, 1417-1420.
"Programme and Abstracts of Seism. Soc. Japan," p. 204. Sakuyama, M., and Nesbitt, R. W (1986). Geochemistry of the
Matsuzawa, T, Umino, N., Hasegawa, A, and Takagi, A. quaternary volcanic rocks of the northeastern Japan arc,
(1986). Upper mantle velocity structure estimated from s. Volcanol. Geotherm. Res. 29, 413-450.
PS-converted wave beneath the northeastern Japan arc, Sanford, A. R., Alptekin, 0., and Toppozada, T R. (1973). Use
Geophys. J. R. Astron. Soc. 86,767-787. of reflection phases on microearthquake seismograms to
Matsuzawa, T., Kono, T., Hasegawa, A., and Takagi, A (1990). map an unusual discontinuity beneath the Rio Grande Rift,
Subducting plate boundary beneath the northeastern Japan Bull. Seism. Soc. Am. 63, 2021-2034.
arc estimated from SP converted waves, Tectonophysics Sanders, C. O. (1984). Location and configuration of magma
181, 123-133. bodies beneath Long Valley, California, determined from
McBirney, A R. (1969). Andesitic and rhyolitic volcanism of anomalous earthquakes signals, J. Geophys. Res. 89, 8287-
orogenic belts, in "The Earth's Crust and Upper Mantle" 8302.
(P. J. Hart, ed.), Geophys-Monogr. Ser, Vol. 13, AGlJ Satake, K., and Hashida, T. (1989). Three-dimensional attenu-
Washington, DC, pp. 501-506. ation structure beneath North Island, New Zealand, Tec-
McKenzie, D. P. (1969). Speculations on the consequences tonophysics 159, 181-194.
 8. Deep Structure of Island Arc Magmatic Regions 195
Sato. H., and Sacks, 1. S. (1990). Magma generation in the Umino, N., and Hasegawa, A. (1984). Three-dimensional Q.
upper mantle inferred from seismic measurements in pe- structure in the northeastern Japan arc, J. Seism. Soc. Ja-
ridotite at high pressure and temperature, in "Magma pan 37,217-228.
Transport and Storage" (M. P. Ryan, ed.), Wiley, Chi- Umino, N., Matsuzawa, T., and Hasegawa, A. (1990).
chester/Sussex, England. X phases after P arrivals from deep earthquakes beneath
Sekiguchi, S. (1991). Three-dimensional Q structure be- the northeastern Japan are, "Abstracts Ann. Meet. Seism.
neath the Kanto-Tokai district, Jl,pan, Tectonophysics Soc. Japan," No.2, p. 200.
195,83-104. Uyeda, S. (1982). Subduction zones: An introduction to com-
Sleep, N.. and Toksoz, M. N. (1973). Evolution of marginal parative subductology, Tectonophysics 81, 133-159.
basins, Nature 233, 548-550. Van der Hilst, R., Engdahl, R., Spakman, w., and Nolet, G.
Spakman, w., and Nolet, G. (1988). Imaging algorithms, ac- (1991). Tomographic imaging of subducted lithosphere
curacy and resolution in delay time tomography, in (N. 1. below northwest Pacific island arcs, Nature 353, 37 -43.
Vlaar, G. Nolet, M. 1. R. Wortel, and S. A. P. L. Cloetingh, Walck, M. c., and Clayton, R. W. (1987). P wave velocity
eds.), pp. 157-187, Mathematical Geophysics, Reidel. variations in the Coso region, California, derived from lo-
Suyehiro, K., and Sacks, 1. S. (1983). An anomalous low ve- cal earthquake travel times, 1. Geophys. Res. 92, 393-406.
locity region above the deep earthquakes in the Japan sub- Yamauchi, K., Ando, M., and Wada, H. (1992). Low-
duction zone, 1. Geophy. Res. 88, 10429-10438. frequency earthquakes in the northern Chubu region(2), in
Suzuki, S., Miyamura, 1., and Nishimura, Y. (1992). Low- "Abstracts, 1992 Japan Earth and Planetary Science Joint
frequency microearthquakes occurring near the Moho Meeting," p. 240.
boundary beneath Tokachidake volcano in relation to the Young, C. Y., and Ward, R. W. (1980). Three-dimensional Q
1988-89 eruption, in "Abstracts 1992 Japan Earth and model of Coso Hot Springs known geothermal resource
Planetary Science Joint Meeting," p. 242. area, J. Geophys. Res. 85, 2459-2470.
Tatsumi, Y. (1989). Migration of fluid phases and genesis of Zhao, D., and Christensen, D. (1992). Tomographic imaging
basalt magmas in subduction zones, J. Geophys. Res. 94, . of the crust and upper mantle beneath central and southern
4697-4707. Alaska, Seism. Res. Lett. 63, 64.
Tatsurni, Y., Sakuyarna, M., Fukuyama, H., and Kushiro, 1. Zhao, D., Hasegawa, A., and Horiuchi, S. (1992). Tomo-
(1983). Generation of arc basalt magmas and thermal graphic imaging of P and S wave velocity structure be-
structure of the mantle wedge in subduction zones, 1. Geo- neath northeastern Japan, J. Geophys. Res. 97, 19909-
phys. Res. 88, 5815-5825. 19928.
Toksoz, M. N., and Bird, P. (1977). Formation and evolution Zhao, D., Horiuchi, S., and Hasegawa, A. (1990). 3-D seismic
of marginal basins and continental plateaus, in "Island velocity structure of the crust and the uppermost mantle in
Arcs, Deep Sea Trenches and Back-Arc Basins" (M. Tal- the northeastern Japan are, Tectonophysics 181,135-149.
wani and W. C. Pitman, eds.), Amer. Geophys. Union. Zhou, D., and Clayton, R. W. (1990). P and S wave travel time
Ukawa, M., and Ohtake, M. (1987). A monochromatic earth- inversions for subducting slab under the island arcs of
quake suggesting deep seated magmatic activity beneath northwest Pacific, J. Geophys. Res. 95, 6829-6851.
Izu-Ooshima Volcano, Japan, J. Geophys. Res. 92, 12649- Zucca, 1. 1., Kasameyer, P. w., and Mills, 1. M., Jr. (1987).
12663. Observation of a reflection from the base of a magma
Urn, J., and Thurber, C. H. (1987). A fast algorithm for chamber in Long Valley caldera, California, Bull. Seism.
two-point seismic ray tracing, Bull. Seism. Soc. Am. 77, Soc. Am. 77, 1674-1687.
972-986.
This page intentionally left blank
Chapter 9 Lateral Water Transport across a Dynamic Mantle
Wedge: A Model for Subduction Zone Magmatism

J. Huw Davies

Overview Units

p Difference between total pressure, p, Pa

A model for subduction zone magmatism wherein the and hydrostatic pressure; i.e., P =
mantle wedge melts by hydrous fluxing from the sub- P - pgz, where z is the depth in meters
ducting slab is presented. The subducting slab induces T Temperature °C
secondary flow in the overriding mantle wedge. This
V. Mantle velocity m S-I
induced flow leads to high temperatures attained near,
V, Horizontal component of mantle m- S-l
but not at, the subducting slab. Water is released from
velocity
amphibole in the oceanic crust at a depth of around
80 km. On entering the mantle, it reacts to form am- V. Vertical component of mantle velocity m. s ' '
phibole, which is carried down by the induced flow as a Grain size m
part of the solid matrix. At a depth of around 100 km b Constant in permeability versus po- dimensionless
it breaks down and releases its water. The water ul- rosity relationship
timately rises vertically through the mantle, passing d Thickness of mechanical lithosphere m
through the amphibole-saturated mantle until it reaches
dry mantle. There it reacts to form amphibole, which is
f Porosity dimensionless
g Acceleration due to gravity m·s -,
once more carried down by the induced flow. This pro-
cess repeats, with the net effect being the lateral trans- II Thickness of amphibolitized layer m
port of water from the oceanic crust out into the mantle k Permeability m'
wedge. Finally, the water reaches the high temperatures Time
near the slab where amphibole is unstable. At this point \1' Weight percentage water content of dimensionless
the hydrous melt leads to ever greater degrees of melt- amphibolitized mantle
ing as it rises vertically. The process ultimately leads to
iJ.p Density difference between fluid and kg -rn " "
such substantial degrees of melting that some of the matrix
melts can segregate in sufficiently large cracks to em-
1) Viscosity Pa· s
place magma in the lithosphere. These processes de-
K Thermal diffusivity m2 • S-I
scribe the proposed source region for subduction zone
magmatism. Pm Density of mantle kg -rn?
This chapter discusses details, variations, and uncer- PI Density of fluid (melt or water) kg' m- 3
tainties surrounding this lateral transport mechanism.
This model can potentially explain many of the features
of subduction zone magmas, including the location of
the volcanic front and the composition of proposed pri- Introduction
mary magmas.
Subduction and subduction zone magmatism are
fundamental processes in the evolution of the
Notation Earth. They play critical roles in the present day
Units
differentiation of the Earth, where subduction
F Total mass flux of water from slab into kg . m - I • S I
zones are believed to be the major sites of genera-
wedge per meter along strike tion of continental crust. Subduction is also sig-

Magmatic Systems Copyright © 1994 by Academic Press. Inc.

Edited by M, P. Ryall 197 All rights of reproduction in any form reserved.
198 J. Huw Davies

nificant in the water and carbon cycles, and be- Thermal Model
cause it is one of the most significant geodynamic
processes, an improved understanding of its na- Brief Review of Previous Thermal Models
ture should lead to insights into the driving forces
of plate tectonics. In this chapter, we present a To place our thermal model in context we briefly
model for subduction zone magmatism that com- review previous thermal models. One of the earli-
bines slab and wedge flow kinematics, thermal est thermal models of subduction was the analytic
structure, and a magma source region consistent model of McKenzie (1969). The model addressed
with geophysical and geochemical constraints. the thermal state of the slab and its relationship to
In contrast to magmatism at mid-ocean ridges, Wadati-Benioff seismicity. It did not address the
where it is largely agreed that the responsible pro- thermal state of the mantle wedge, which in fact
cess is adiabatic upwelling, magmatism at sub- was assumed to be at a constant hot temperature.
duction zones has led to a range of proposals for Soon afterward, models tried to account for the
the source region and processes. With the advent subduction zone magmatism; initially they as-
of plate tectonics it was proposed that shear and sumed that the magmatism was the result of fric-
frictional heating led to melting of the oceanic tional/viscous heating at the slab-wedge interface
crust (Oxburgh and Turcotte, 1970; Turcotte and (Ox burgh and Turcotte, 1968, 1970; Turcotte and
Schubert, 1973). Hsui et al. (1983) suggested that Schubert, 1968, 1973). Clearly this assumption
the oceanic crust melted as a result of high tem- dominated the preceding thermal models and all
peratures arising from the hot induced mantle flow other models where it was significant (Minear and
impinging on it. Some have proposed that the Toksoz, 1970a, 1970b; Toksoz et al., 1971). The
melting is the result of adiabatic upwelling, of ei- preceding thermal models are not considered re-
ther eclogitic crust (Brophy and Marsh, 1986; Ida, alistic since frictional and viscous heating are now
1987) or mantle wedge (Plank and Langmuir, believed to be of only secondary importance. This
1988; Tatsumi et al., 1983), while many other fact is expanded upon in the section Frictional
workers have argued that the melting results from Heating at the Slab-Wedge Interface.
fluxing of the peridotitic mantle wedge by water The subducting plate induces a secondary flow
from the oceanic crust (Davies and Stevenson, in the mantle wedge, which advects heat into the
1992; Gill, 1981; Tatsumi, 1989). mantle wedge corner (Anderson et al., 1980; An-
One of the striking observations of subduction drews and Sleep, 1974; Bodri and Bodri, 1978;
zone volcanism is that the active volcanoes nearest Honda, 1985; Hsui ef al., 1983; McKenzie, 1969).
the trench seem to describe nearly straight lines Hsui et al. (1983) argued that the induced flow
parallel to the trench: the volcanic front (Marsh, could be sufficient to melt the subducting slab
1979b; Sugimura, 1960). In addition, these volca- significantly. Davies and Stevenson (1992) have
noes have been found to lie some 120 km above the shown with a similar model, but one that includes
Benioff zone of the subducting plate (Tatsumi, the slab as well as the mantle wedge, that the
1986). This constraint is quite robust and one that subducting slab is unlikely to melt extensively.
any model hoping to explain subduction zone mag- Honda (1985) modeled the Tohoku subduction
matism must be able to fulfill. zone in Japan, and he needed to propose relatively
We argue that the subducting plate induces a high shear stresses on the interface to satisfy the
flow in the mantle wedge. This flow leads to high surface heat flow. Subsequent work though has
temperatures near the subducting slab and also shown that substantial heat is generated in the con-
to a mechanism for the lateral transport of H 2 0 tinental crust in Japan, and as a result high shear
across the dynamic wedge from the subducting stresses are not required (Furukawa and Uyeda,
oceanic crust. This transport results in melting of 1989; Nagao and Uyeda, 1989). Thermal models
the hot wedge peridotite due to the lowering of the that include induced secondary mantle flow and
solidus by the influx of water. In the next section ignore other heat sources have also been used
we describe the results of thermal modeling and to investigate seismic signatures of subduction
comment on the different proposals mentioned (Creager and Jordan, 1984; Helffrich et al., 1989).
previously. The results help motivate the mecha- All the preceding models have prescribed
nism for the lateral transport of water. the subduction process kinematically. Dynamical
 9. Lateral Water Transport across a Mantle Wedge 199
modeling of subduction is only just beginning and cal lithosphere is assumed to be only 40 km thick.
it is not as robust as kinematic modeling, espe- We discuss in the section Ablation why this dif-
cially since the initiation of subduction is so poorly ference might be reasonable. No significant heat
understood. Workers have attempted to combine sources or body forces are assumed to exist. The
the kinematic description of subduction with a mantle is assumed to have a Newtonian constant
dynamic model in the mantle wedge (free convec- viscosity rheology. Given a kinematically driven
tion). Many problems with interpreting and con- flow field and the lack of body forces, temperature-
straining these models exist, as discussed in Da- dependent viscosity was found to make little dif-
vies and Stevenson (1992). First, the dynamical ference. Temperature-dependent viscosity could
effects can be very sensitive to the rheology; for have significant effects in dynamic models where
example, a large increase of viscosity at depth body forces are included.
could lead to dramatic consequences (Jurdy and The thermal results presented are for steady
Stefanick, 1983). Unfortunately, the details of the state-thus the results describe steady-state ther-
actual rheology are not well known. The second mal structures. Hence it must be remembered that
problem is more fundamental for these hybrid before the steady state is reached, temperatures in
models: the negative thermal buoyancy in the the mantle wedge will probably be higher than
mantle wedge near the slab is counted twice be- those presented here, since subduction zones at the
cause it contributes to the motion of the subduct- time of their initiation are expected to be at least
ing slab (which is prescribed kinematically), as as.hot as average mantle. Subduction zones prob-
well as the dynamic motions in the mantle wedge. ably reach thermal equilibrium relatively quickly,
Therefore we suggest that free convection involv- being largely controlled by the time it takes for
ing thermal buoyancy can be rigorously enter- heat to conduct through the mechanical litho-
tained only in dynamic models of subduction. sphere of the overriding plate. For a lithospheric
Otherwise, it is impossible to know how to modify thickness of 40 km, this fact suggests a time of
the thermal buoyancy of the mantle wedge to ac- around 5 Myr (t = d 2/1T 2K). The attainment of
count for the kinematic description of the sub- thermal equilibrium will probably take longer
ducting slab. Jurdy and Stefanick (1983) have if the ablation of the mechanical lithosphere is
demonstrated that the possible effects of free con- achieved partly by the conduction of heat. For in-
vection can be significant given low viscosities. stance, if the mechanical lithosphere were 80 km
Since the effects are currently poorly constrained, in thickness, then the equilibration time would ap-
investigators have favored the better constrained proach 20 Myr.
end-member case of forced convection. This case
assumes that sources of buoyancy in the mantle Induced Flow in the Mantle Wedge
wedge will only modulate the overall flow field
rather than dominate it (Davies and Stevenson, We first present arguments as to why the sub-
1992). Clearly this assumption needs to be checked ducting slab must induce a secondary flow in the
by a dynamic model of subduction constrained by mantle wedge. The mantle wedge must be mobile
good estimates of mantle rheology. since a stagnant mantle wedge could not provide
the source material from which to generate the
Outline of Thermal Model magma to produce some of the large volcanic
structures (e.g., the Andes) found above subduc-
In this section we present an outline of the ther- tion zones. We expect the slab to induce flow in
mal model, and in subsequent sections we discuss the mantle wedge by shear coupling across the
some of the more important assumptions in more wedge-slab interface. Even if a very weak shear
detail. Before presenting the results we also dis- zone that limited shear coupling existed between
cuss theoretical and numerical details. The funda- the subducting plate and the overriding mantle, the
mental assumptions are that the subducting plate neighboring mantle wedge would still be cooled
induces a secondary flow in the overriding mantle by the slab and its net resultant negative buoyancy
wedge and that frictional heating is not considered would drive a downflow. In addition, a stagnant
significant. In addition, the thermal lithosphere is mantle wedge would be cooled progressively by
assumed to be 100 km thick whereas the mechani- the underthrusting subducting slab and would be-
200 J. Huw Davies

o km =============

-----------------------'1
---------------------"1
-----------------. - , III
-----------------"111
~~---------.- - -, II III
///--------,,\ II/II
',,',----"\1 III/I
""--"""1/111,
", .. ·· .. "IIII/,
""'''-''IIII/'
\"----~//III/,
, ...'----r,//I/
400km

Figure I The velocity field induced by a subducting slab. Note no absolute velocity scale is provided, since the scale is
arbitrary. The longest vectors correspond to the subduction velocity of the descending slab. The descending slab and overriding
mechanical lithosphere are shaded. From Davies and Stevenson (1992).

come ever cooler. Ultimately, without the induced overriding a viscous asthenosphere. Clearly this
flow advecting in heat, the region beneath the vol- model is a major simplification but probably de-
canic front would become so cold that no mag- scribes the behavior of this system well on a
matism or heat flow anomaly could exist. Figure I timescale of 10 Myr or greater. The depth of the
illustrates an example of the flow induced in the transition is somewhat arbitrary. For the following
mantle by a subducting slab, where local thermal model we have assumed that this transition occurs
buoyancy in the mantle wedge is insignificant, and at a depth of 40 km. It is argued that a thin me-
the influence of rheology on the flow is well ap- chanical lithosphere might be expected near the
proximated by a fluid of constant viscosity. volcanic front due to the effect of ablation. The
mantle flow is focused slightly upward in this re-
Frictional Heating at the Slab-Wedge Interface gion if it is like classic corner flow (as observed
weakly in Fig. 1; Batchelor, 1967). This focusing
As a second consideration, we present arguments
will then lead to higher temperatures, which will
that support our neglect of frictional and shear
lead to less viscous rheologies, leading to a further
heating. Yuen et al. (1978) have shown that hot
focusing of the hot advective flow (Bodri and
rocks are so weak that little heat is produced;
Bodri, 1978). This ablative process will be limited
hence, frictional heating cannot play the critical
when the conduction of heat and the temperature-
role in generating subduction zone magmatism.
dependent rheology combine to ultimately pro-
In a detailed study, van den Beukel and Wortel
duce a steady-state geometry. Another contribut-
( 1987) have shown that the effect due to frictional
ing factor to the ablation is the very high stresses
heating is relatively local and can lead to tempera-
implicit in the flow through such a confined geom-
tures along the thrust interface in the subduction
etry; remember that high stresses lower the effec-
zone being some 50 to 200 0 C hotter than those
tive viscosity in materials described by power-law
derived in the model of this chapter, depending
rheologies, such as those undergoing dislocation
upon rheology. Their work also shows that ignor-
creep. Figure 2 is a schematic of the possible ge-
ing frictional heating has little or no effect on the
ometry for the mantle wedge corner, reflecting the
temperatures evaluated within the mantle wedge.
largest changes in rheology. In the thermal mod-
Peacock (1992) also argues for low shear stresses
eling of subduction zones, the mechanical bound-
at depths of 15-50 km due to the low tempera-
ary layer has been assumed to be of constant
tures recorded by blueschist-facies metamorphic
thickness, rather than thickening away from the
rocks.
wedge corner. We have used a thickness more ap-
propriate for the ablated wedge corner since we
Ablation
are more interested in modeling the temperatures
A significant aspect of the model is its rheology. in the source region of the subduction zone mag-
In this model we have assumed a rigid lithosphere mas. If the lithosphere were thickened away from
 9. Lateral Water Transport across a Mantle Wedge 201

Overriding neath or around the subducting slab, then there is

Trench
Plate " a greater impact. It is interesting to note that a
wide range of convergence velocities exist, but for
slabs that subduct deeply (>400 km), subduction
velocities seem to be largely constant (Otsuki,
1989; Otsuki et al., 1990). As seen in the follow-
ing section, relatively high subduction velocities
are required for hot mantle wedges, and hence in
Mantle Wedge this respect rollback and back-arc spreading are
critically important in raising the values of some
low convergence velocities.

Figure 2 Summary diagram of the possible geometry of Theoretical Background and Numerical
wedge corner following ablation.
Implementation

We have solved the advection-conduction heat

the trench, then one would expect higher tempera- equation assuming an incompressible material,
tures since the material being advected into the where the subduction process is enforced by means
mantle wedge would be coming from a greater of kinematic boundary conditions. The material is
depth. In addition, while a sharp wedge corner is assumed to have a Newtonian constant viscosity
common in thermal models, in reality it is likely rheology. Note the cold rigid lithosphere of the
to be smooth. A smooth corner in the model pre- overriding plate is enforced by internal boundary
sented would lead to slightly lower temperatures conditions where the appropriate nodes are held
at the slab-wedge interface at shallow depths, fixed. Similarly, the mechanical lithosphere of the
since the induced flow would not be able to im- downgoing plate is modeled kinematically by pre-
pinge as strongly on the shallow slab. scribing the subduction velocity to appropriate
nodes. The velocity boundary conditions are illus-
Trench Migration trated in Fig. 3. The following equations were
solved.
There is much evidence that subducting slabs gen-
The equation for the conservation of mass (con-
erally do not subduct in a direction parallel with
tinuityequation),
their dip, but rather move at an angle between
their angle of dip and the downward vertical (Gar- V· V = 0, (I)
funkel et al., 1986; Hamilton,1988). In all ther-
where V is the velocity of the mantle, and V is the
mal models to date, the subducting slabs have been
gradient operator.
modeled as "sliding down slots," i.e., moving
The equation for the conservation of momen-
parallel to their dip. The difference can be largely
tum (Stokes equation),
accounted for by considering the subduction ve-
locity (the velocity at which the subducting plate VP = V . (7]VV), (2)
enters the asthenosphere) rather than the conver-
where P is the difference between the total pres-
gence velocity (the velocity at which the subduct-
sure and the hydrostatic pressure, and 7] is the vis-
ing plate approaches the rigid part of the overrid-
cosity of the mantle.
ing plate-the rigid part is usually beyond the arc
The equation for the conservation of energy
and any back-arc basin). Clearly how the mantle
(heat equation),
flow accommodates rollback (trench migration) is
potentially quite significant, but if, during thermal aT/at + V . VT,= KVq, (3)
assimilation at depth, the subducting slab is incor-
where T is temperature, t is time, and K is the ther-
porated into the mantle on the overriding plate
mal diffusivity.
side, then the thermal models are probably only
weakly affected. On the other hand, if the mantle We have used the Boussinesq approximation (Trit-
accommodates the process by flowing from be- ton, 1977), which implies we ignore all variations
 202 J. Huw Davies
Zero Velocity (Pinned) Fixed Constant Velocity
I I Iii I

Ul
Ul
e
en
iii
...oE
z if
e
CII II
N

Zero Normal
Stress
I "
Zero Normal
Stress

Figure 3 The velocity and stress boundary conditions for the momentum equation, for the 60 0 dipping slab; similar boundary
conditions applied for the 300 dipping slab model. From Davies and Stevenson (1992).

in density other than as a body force in the Stokes respond to nondimensionalized quantities. Drop-
equation; this assumption leads to the continuity ping the primes we get the nondimensionalized
equation reducing to a requirement for incom- equations
pressibility. Since we solve for steady-state tem-
'VP (5)
perature structures, the heat equation becomes
and
(4)
The equations were nondimensionalized as fol-
v . 'VT = 'V 2 T. (6)
lows: x' = xld, z' = zld, t' = «tld", P = TJKP' /d 2 , The thermal boundary conditions are illustrated
where d is a characteristic length scale, e.g., the in Fig. 4, together with the finite-element grid
depth of the model (400 km). The primes cor- used to solve the preceding equations. A compari-

T=O

Linear
gradient II11111I11111111 '~iliI5Iiil l '~Error function
gradient

Zero Heat Flux

Figure 4 The temperature and heat flow boundary conditions, and the finite-element grid used to solve both the energy and
momentum equations. This grid is for the 60 0 dipping slab; the grid for the 300 dipping slab model had even higher resolution.
From Davies and Stevenson (1992).
 9. Lateral Water Transport across a Mantle Wedge 203
son of the thermal boundary equations in Fig. 4 Results of Thermal Modeling
and the velocity boundary conditions in Fig. 3
shows the difference in the assumed thicknesses In Fig. 5 we illustrate the results of the thermal
of the thermal and mechanical lithospheres al- modeling (Davies and Stevenson, 1992). What is
ready discussed. The finite-element grid allows us demonstrated most clearly is how the subduction
to reduce the influence of the boundary conditions zone is appreciably cooler than the surrounding
selected by keeping the boundaries far from the mantle. This difference is expected since subduc-
region of interest. It also allows good resolution in tion zones involve the advection of oceanic crust
the region of interest without incurring unneces- and mantle that has cooled near the Earth's surface
sary additional computational expense. and is brought back into the mantle. The obvious
The momentum equation was solved using question and a seeming paradox is, Why should
a penalty method to enforce incompressibility there be melting at all in subduction zones? We
(Hughes et al., 1979a), whereas the energy equa- now try to describe the physical processes that
tion is solved implicitly (Hughes et al., 1979b) us- control some of the details of the results.
ing a streamline upwind Petrov-Galerkin method In thermal models where local heat sources, as
(Brooks, 1981). An explicit version of the code well as advection of heat by migration of aqueous
has been adapted for vectorizing machines and is fluids or melts, have been ignored, the only pro-
described in King et al. (1990). cesses left are advection by flow of solid matrix
The equations solved are nondimensional, but and conduction. Frictional heating was argued
to present specific results we assumed the follow- previously to be of only limited importance and
ing constants. The thermal diffusivity K was as- the migration of fluids (Peacock, 1987b) has been
sumed to be 10 -6 m? . s - I (Fujisawa et al., 1968) shown to be unimportant for the large-scale ther-
and T = 1 (the temperature at the base of the ther- mal structure of subduction zones. Advection is
mal lithosphere ) was assumed to be 1325°C (Me- the transport of heat by the movement of material;
Kenzie and Bickle, 1988). The actual value of the and in the subduction zone case, the advection of
matrix viscosity was not required since it does not the ocean floor into the mantle is of the greatest
appear in the nondimensionalized equations, other importance. Since heat conduction in rocks is very
than as a scaling for the pressure, which is never inefficient, relatively cool temperatures may ex-
interpreted. This simplification is a consequence tend deep into the upper mantle (McKenzie, 1969).
of the fact that the velocity field is kinematically Since a secondary flow is induced in the mantle
prescribed. wedge, there is also significant advection of heat

Figure 5 Thermal field of a subduction zone with induced flow of the mantle wedge. The dashed lines outline the mechanical
lithospheres. The coolest and hottest temperatures have been shaded for emphasis, with the contours 200° C apart. Note how
cool temperatures extend down into the mantle. The contour between the two shaded regions is the 600°C isotherm. Note how
the cold region of the slab grows inward away from the interface, and the hot region also moves slowly away from the interface.
These regions are termed "boundary layers" in the text. Adapted from Davies and Stevenson (1992).
 204 J. Huw Davies

toward the wedge corner. The induced flow though If the subducting plate is very young, then it is
must ultimately turn and flow down-dip alongside possible for the oceanic crust to be heated by its
the slab, and the heat flow across the slab-wedge underlying hot mantle in such a way that the cold
boundary is by conduction. Hence, due to the in- tongue disappears by the time the oceanic crust
efficiency of conduction, the slab-wedge bound- reaches a depth of 100 km. In this case, clearly the
ary remains relatively cool, while at the same time mantle wedge is not cooled as dramatically, and
the mantle wedge can remain hot, even though it both the slab and the wedge are much hotter. Since
is close to the subducting slab. From a close in- local heat sources such as frictional heating and
spection of Fig. 5 one can observe two thermal dehydration playa role in deciding at what age
"boundary layers" developing very slowly from a this effect is significant, a firm answer cannot be
point just below where the slab enters the wedge; given from this model, but we are confident that
one grows out into the wedge, and the other grows the thermal structure in the mantle wedge is prac-
into the slab, as we go deeper. They can be thought tically independent of the age of the subducting
of as thermal boundary layers since the flow of plate for ages greater than 30 Myr, and probably
heat perpendicular to them is nearly totally by much younger.
conduction, with little or no contribution by ad-
vection. The presence of two boundary layers re- Dependence on the Dip of the
duces the cooling effect of the slab on the wedge, Subducting Plate
and the heating effect of the wedge on the slab; The dip of subduction was found to playa neg-
in this way very steep thermal gradients can be . ligible role, provided it was not very shallow. The
maintained. dip only begins to affect the thermal structure in
the mantle wedge when it is so shallow that the
Dependence on the Slab Velocity thermal boundary layer under the overriding plate
Since advection is all important, the velocity and the thermal boundary layer growing out from
at which slab material is subducted clearly plays the subducting plate into the mantle wedge begin
a dominant role in controlling the thermal struc- to interact and squeeze out the hot tongue from the
ture through the thicknesses of the thermal bound- corner of the mantle wedge.
ary layers. The effect of subduction velocities
on the thermal structure is illustrated in Fig. 6. Implications of the Results
Slower velocities lead to less advection of heat,
and hence, conduction is relatively more signifi- From the results of the thermal modeling, one can
cant, leading to weaker thermal gradients. Hence, see that the induced secondary flow does not lead
the subducting slab is hotter and the mantle wedge to high temperatures at the surface of the sub-
is cooler, for slower subduction velocities. Otsu- ducting oceanic crust as proposed by Hsui et al.
ki's (1989) observation, noted previously, that sub- (1983). This result occurs because the velocity of
duction velocities are practically constant and high the induced flow normal to the oceanic crust must
at 7 em . yr ~ I even though convergence velocities be zero, and it approaches zero over an appre-
vary widely is quite significant in this respect. ciable distance (~1 0 km). Therefore, the subduct-
ing oceanic crust is not likely to be the major
Dependence on the Age ofSubducting Plate source region of subduction zone magmas.
Irrespective of its age, the surface of the oce-
anic crust is always near 273. K just before sub- Adiabatic Upwelling
duction. Therefore, material this cool is always
. advected into the mantle wedge. Hence, one al- Adiabatic upwelling could play a role in sub-
ways expects a relatively cool tongue to penetrate duction zone magmatism. Rather than address
somewhat into the mantle. Clearly, the age of the individually the specific, models that have been
subducting plate plays little role in the tempera- proposed, we shall concentrate on presenting ar-
tures in the overriding mantle wedge since it is guments that show the general difficulty encoun-
largely a balance between the advection of heat by tered by all such processes. Although extension
induced secondary wedge flow and conduction combined with decompression melting in ascend-
into the cold oceanic crust of the downgoing slab. ing mantle flow is the cause of melting at mid-
 9. Lateral Water Transport across a Mantle Wedge 205

Plate Velocity = 1.8 cm . y(1

400 km I.-L --L-L..L.."'--'~.L__L____./ Plate Velocity = 4.5 cm • yr· 1

Plate Velocity =9.0 em- y(1

Figure 6 Effect of subduction velocities on thermal structures. (a) 1.8 cm· yr~l. (b) 4.5 cm· yr " ', (c) 9.0 cm· yr I. The
0
mechanicallithospheres are shaded, and the thermal contours are spaced 200 0 C apart. The heavy contour corresponds to 800 C.
Note the higher subduction velocity leads to the hotter mantle wedge. From Davies and Stevenson (1992).

ocean ridges, in back-arc basins, and in the split- Melting can potentially produce major sources of
ting of some volcanic fronts, it generally is not the buoyancy, the buoyant residue and melt. Clearly
dominant mode of behavior in subduction zones. though, adiabatic upwelling driven by melting
Adiabatic upwelling is also the probable explana- cannot initiate melting. Other potential sources of
tion for melting at "hotspots," where hotter than buoyancy are the presence of water and hydrated
average mantle upwells in plumes due to its own mantle; their impact though is much reduced by
intrinsic thermal positive buoyancy. Subduction their presence in cooler material, which is both
zones are cooler than surrounding mantle and more rigid and intrinsically negatively buoyant.
hence exhibit only negative thermal buoyancy. Subduction zones not only lack the enabling
206 J. Huw Davies

factors of adiabatic upwelling of other tectonic en- is of debatable but probably minor importance,
vironments but also possess characteristics that whereas melting with induced flow in the mantle
intrinsically make generating a substantial upflow wedge is always significant and probably all im-
difficult. This trend is fundamentally because sub- portant. If substantial matrix flow exists parallel
duction zones are regions of large-scale down flow to the arc (along-strike), then the preceding state-
of mantle material. The slab couples to the over- ment of an expected closed cell is not rigorous.
riding mantle wedge leading to an induced down- Such a complicated matrix flow pattern should
flow in the mantle wedge near the slab. This result lead to nearly regularly spaced alternate regions
will occur to some degree even if they are rheo- of active (upflowing) and inactive (downflowing)
logically decoupled by a weak zone because, as arc. Little evidence for such a complicated model
mentioned earlier, the dramatic cooling effects exists.
would extend across the decoupling zone and give
the surrounding mantle an intrinsic negative buoy- Phase Equilibria for Hydrous Peridotite
ancy. Davies and Stevenson (1992) have shown
that it would require large amounts of positive In Fig. 7 we present a simplified phase diagram
buoyancy, a low-viscosity mantle wedge, and a for both the peridotite and the basalt systems,
decoupling of the mantle wedge to allow for sub- based on the work of Green and Wyllie (Green,
stantial upwelling. The geometrical constraints 1973; Wyllie, 1979). The diagram presents the
also make the development of an effective upflow solidus for anhydrous systems, i.e., no water pres-
difficult since a cold rigid subducting slab exists ent: the solidus for hydrated systems with no free
at a depth of around 120 km below the volcanic fluid, i.e., amphiboles present; and the solidus for
front, and at around 30-50 km a rigid mechani- water-saturated systems, i.e., both amphiboles and
cal lithosphere tends to inhibit further upwelling free aqueous phase present. The most striking fea-
closer to the surface.
The upflow that complements the downflow in
the mantle wedge for corner flow is much more 100
diffuse and more nearly horizontal than vertical
(Fig. 1), so that it is not adiabatic. As seen from
the results of the thermal modeling, this upflow
leads to only a modest rising of isotherms in a re- ~
stricted region (Figs. 5 and 6). This limited effect 50 E.
Ql
o
is important for the model proposed, but is not
sufficient to lead to melting directly by adiabatic
upwelling.
Given that magma and a residue of melting o
600 800 1000 1200 1400
are ultimately produced, positively buoyant body Temperature ('C)
forces and hence some net upwelling will clearly
Figure 7 Phase diagram of the peridotite and basalt systems.
exist. We argued previously that its influence will
The second letter identifies the bulk composition: P, peridotite;
probably be only a modulation. Even if the local B, basalt. The first letter describes the amount of water present.
buoyancy forces were to become dynamically sig- D represents dry conditions, i.e.. anhydrous. W represents wet
nificant, its influence would wane since the result- conditions; i.e. water-saturated. free aqueous fluid is present in
ing upflow in the mantle wedge would form a addition to amphiboles. Am represents conditions where there
is no free water, but the system is hydrated with the water held
small enclosed convection cell that would become
in amphiboles. A possible path of a water molecule through
progressively less fertile and colder, until finally the phase diagram is illustrated by the thin dashed line; it pro-
no melt could be formed (Davies and Stevenson, gresses from low to high temperatures, where it is ultimately
1992). Hence, even in the scenario of buoyancy- carried away in melts. The up-pressure parts of the path are
driven upwelling, adiabatic upwelling could not where the water is carried in the solid matrix fixed in hydrated
minerals (probably the amphibole pargasite), while the down-
continue once initiated; rather it would be only the
pressure parts of the path are where water moves as a free
second and dependent one-half of a cycle, initi- phase, either by fracture or by porous flow, either as an aque-
ated by melting in a regime of induced flow in ous fluid or as a water-rich melt. Based on the experimental
the mantle wedge. Therefore, adiabatic upwelling work of Green (1973) and Wyllie (1979).
 9. Lateral Water Transport across a Mantle Wedge 207
ture of this diagram is the substantial reduction in points out that diffusion alone is too slow to allow
the solidus due to the presence of water. It has sufficient migration of water from the slab into the
been argued for some time that it is the presence mantle.
of water that is responsible for the generation
of melt in subduction zones. What has generally
been glossed over is the means by which the water Basic Mechanism of Lateral Water Transport
reaches that part of the mantle wedge that is suf-
ficiently hot to produce melt. This omission has From the results of the thermal modeling (pre-
been used to cast doubt on the whole process of sented in more detail in Figs. 8a and 8b), it is clear
melting by fluxing; e.g., Marsh (I979a) rightly that the subducting oceanic crust is relatively cool

-----,r-
-----------AP-----~

---------------,,
1200 C

------------- -"
----------- -
\
\

-----------"
....... \

--------- ....... ,,\

14 ak m L- --'-----'--'--'----'---'--'-.L.L.L.L..l~

1------- 800 0 c--=-~~-='---'=--="'-='-===...-:

--:WY"':

Figure 8 Computed thermal and flow structures (a) for a 60° dipping slab and (b) for a 30° dipping slab. The heavy contour
represents 800 0 C. An interpretation of the phase boundaries is provided and labeled as in Figure 7, but with AP representing
AmP above the wet solidus, where we get dehydration melting. The lines with squares are the breakdown limits of amphibole in
peridotite (AmP) and in basalt (AmB), leading to subsequent dehydration. The lines with circles are the breakdown limits of
amphibole leading to dehydration melting. The line with diamonds represents water-saturated melting (wet peridotite, WP). The
isotherms are separated by 200 0 C. The directed arrows show the induced flow in the mantle wedge as well as the outboard upper
mantle. Note that the lateral transport mechanism is effective out to the thermal stability maximum of amphibole; i.e., the
secondary induced flow still has a horizontal component away from the wedge corner, at the base of the amphibole ledge. The
largest vectors correspond to the subduction velocity 7.2 ern- yr '. From Davies and Stevenson (1992).
208 J. Huw Davies

and the mantle wedge is also cooler than the sur-

rounding mantle. Hence no obvious reason exists Mantle Wedge

why there should be magmatism. We have further

argued earlier that although adiabatic decompres-
sion might contribute, it cannot be a factor until Melting
there is significant melting. Hence, the only well-
Amphibole
accepted mechanism left for initiating melting is Dehydration
the input of water into the mantle wedge. Tatsumi
(1986) was one of the first to address the problem
of how water reaches the hot part of the mantle.
He points out that amphiboles dehydrate at shal-
lower depths in the basaltic bulk composition than
/ Transport of amphibole-fixed water
in a peridotitic bulk composition. He proposes
that the amphiboles in the oceanic crust dehydrate
Movement of free water / water saturated
by a depth of 80 km (see curve AmB in Fig. 7) melt through amphibole saturated mantle
and that the water migrates into the mantle wedge
where it reacts with the anhydrous peridotite to Figure 9 Synopsis of the lateral water transport mechanism.
We schematically follow one water molecule. To emphasize
form amphibole again. He then proposes that this
the two main components of the mechanism, we have assumed
hydrated mantle is carried down by the induced that the water can always migrate rapidly vertically. From Da-
flow until it reaches a depth of around 100 km, at vies and Stevenson (1992).
which point the water is released and generates
melting (see curve AmP in Fig. 7). From the re-
sults of the thermal models in Fig. 8 we see that corner. Once more, it reaches a pressure at which
the mantle next to the subducting plate is not hot the water is released, and the cycle is repeated. As
enough to melt. Tatsumi (1989) envisages that the seen in Fig. 9, a schematic of the fundamental pro-
water migrates vertically, inducing ever increasing cess, the net effect is the lateral transport of wa-
melting as it encounters higher temperatures. He ter across from the slab and into the hot mantle
then requires that diapirs be spawned off the top, wedge. Watson et al. (1990) have shown that wa-
rise, and be heated to yet higher temperatures ter establishes an interconnected network with oli-
(Tatsumi et al., 1983). Ida (1987) points out that vine under higher pressures and temperatures (e.g.,
the diapirs cannot be small enough to be effec- = 11000 C at I OPa, and probably around 10000 C
tively heated, yet still be large enough to over- at 3 OPa), allowing the vertical transport of fluids
come the induced flow. We extend Tatsumi's ideas by intergranular porous flow. They note that sur-
for the migration of water, which avoids the re- face energy anisotropies seem to give partial con-
quirements for diapirism. nectivity for water at temperatures as low as 9000 C
As shown in Fig. 8, a hot wedge of mantle ex- (at 1 OPa). Connectivity at such low temperatures
ists not far from the oceanic crust. Clearly, if water is also suggested in Watson (1991). At some
were transported into this region of the mantle point, the water is not released as a free phase
wedge, then it would be possible to generate large from the amphiboles on their high-pressure break-
amounts of melting even given the downward in- down, but is dissolved in the melts formed by their
duced flow. As Tatsumi argues, the water is re- breakdown in dehydration melting (Beard and
leased from the amphiboles at a depth of about Lofgren, 1990). These melts similarly rise through
100 km in the mantle wedge (as seen in Fig. 7, and the amphibole-saturated mantle and react above
interpreted in Fig. 8). Provided the water moves with dry mantle to yield amphibole. Fujii et al.
vertically more quickly than it is carried back (1986) have shown that hydrous basaltic melts
down by the induced secondary flow, it will rise readily lead to permeable interconnections along
through the amphibole-saturated mantle until it grain boundaries in peridotite, so even if crack
reaches higher dry mantle. Here it reacts with the propagation is not dominant, the melts in this
hot country rocks to form amphibole. It then again region are still expected to rise rapidly by po-
becomes part of the induced flow and is carried rous media flow. At some distance into the wedge
back downward and away from the mantle wedge though, amphibole is no longer stable. and in such
 9. Lateral Water Transport across a Mantle Wedge 209
a case one expects appreciable melting. If suffi- low the mechanical lithosphere and freeze, react-
cient melting does occur, then some melt can seg- ing with its surrounding mantle; then it would be
regate from the source region and be emplaced in entrained back into the wedge corner where it
the lithosphere, leading ultimately to volcanism. would have to be recycled quickly. Morris et al.
Clearly, important requirements of the water (1990) have argued from the variation of IOBe/Be
transport mechanism are (i) that the water or versus B/Be that the mantle wedge is uncontami-
water-rich melt can at some point migrate rapidly nated with boron; i.e., its residence time in the
in a near-vertical direction and (ii) that the in- subarc mantle is comparable to, or less than, the
duced flow has a significant component of ve- lOBe decay time, -5 Myr. Since boron would
locity away from the mantle wedge corner, at least probably be found in any melts that could not es-
out as far as the point at which amphibole is no cape the mantle wedge, this constraint suggests
longer stable. that either all melts do leave the mantle wedge or
Hence, this mechanism suggests that the loca- they are quickly recycled. Melts that do not reach
tion of the source region is well out in the interior the crust but are emplaced in the mantle part of
of the mantle wedge, initiating at the depth corre- the mechanical lithosphere also satisfy this boron
sponding to the maximum thermal stability of am- constraint.
phibole, and rising more or less vertically with
probably the bulk of the melting occurring at the
shallowest regions of the source region. It is un- Discussion of the Lateral Transport
clear where the top of the source region will be. Mechanism
Davies and Stevenson (1992) assumed that the top
of the source region corresponds to the maximum In this section we expand on our mechanism,
temperatures encountered by the rising melts. considering the process in more detail and then
One question that must be addressed is, Why clouding it somewhat by considering potential
do the very hydrous melts (which are part of the complications.
lateral transport mechanism) rehydrate the mantle
when they reach dry mantle, but the melts leaving Transition from a Silica-Rich Fluid
the source region do not all stop and rehydrate the to a Hydrous Melt
shallow mantle? A very important aspect of this
apparent selectivity must be that for melts to reach As mentioned previously, at some point the mo-
the near surface in subduction zones, the magmas bile phase changes from a silica-rich hydrous
must segregate into large cracks. Only then can phase near the subducting plate to a hydrous sili-
they propagate sufficiently rapidly so that they do cate melt near the source region. At pressures of
not cool, freeze, and react with their mantle wall 3 GPa, they are two separate phases and cannot be
rock to form hydrous minerals. On the other hand, changed continuously from one to the other (Eg-
for the very hydrous melts not to segregate away gler, 1987). The melt can hold up to 25 wt% water
from the mantle wedge, they must move suffi- (Green, 1973) and the water can hold more than
ciently slowly in small cracks or more slowly still 30 wt% silica (Walther and Helgeson, 1977).
in small veins or pores by porous flow. In this case Due to the potential two-phase nature of the
chemical equilibrium is apparently sufficiently fluid in this transition region, permeability could
maintained so that the hydrous melts react with drop and lead to a concomitant increase in the
their wall rocks to form hydrous minerals (as- amount of fluid present. The water-rich phase is
suming that they are not already saturated), while expected to have a very low viscosity, and the hy-
migrating only a short distance. The potentially drous melt phase to have a higher viscosity. Three
different styles of behavior must reflect the appre- factors compete to control its viscosity: most im-
ciably lower porosity away from the source re- portantly the high water content that will lower
gion, which may not allow the development of the viscosity and the high silica content that will
major dikes and cracks. increase the viscosity, and less importantly the
Clearly it is possible that even some of the melt relatively low temperatures that will lead to higher
formed in the source region is not emplaced in the viscosities. Clearly the lower permeability and the
mechanical lithosphere. Rather it might stop be- possibility of higher viscosity could lead to de-
210 J. Huw Davies

creased velocities and hence a buildup in fluid- that the bulk of the water released at 100 km depth
filled porosity at the transition. This buildup could on breakdown of pargasite is not incorporated in
lead to a substantial body force, but its impact hydrous minerals that are stable to much greater
may well be limited by the fact that permeability depths and temperatures, then it is not critical if
is such a strong function of porosity (proportional the water does not interconnect immediately fol-
to the second or third power; Cheadle, 1989). lowing dehydration.
Even though we do not know the detailed behav- The presence of CO 2 reduces the possibility of
ior of the system during this potentially funda- interconnection (Watson et aI., 1990). The amount
mental transition, it is plausible that the perme- of CO 2 relative to H 20 in fluids released from the
ability is not greatly reduced due to the ability of slab is probably small (Giggenbach, 1992). This
heat conduction at grain scales to produce only a assumption is also suggested by the fact that cal-
single phase in any portion of the porous network. cite is stable under the conditions present shallow
In this case, the two phases will not interfere and in the subduction zone (Huang et al., 1980).
hence the permeability might not be reduced dra- Unfortunately no textural experiments have
matically. In addition, the high water contents of been undertaken between water and garnet, am-
the melts probably lead to relatively low viscosi- phibole, or pyroxenes, so we do not know the
ties, such that the increase in porosity as we cross depth at which an interconnected network will be
the transition will be so small that the buildup of formed in basalt/eclogite. Water has been specu-
porosity and positive melt-enhanced buoyancy is lated to have a smaller dihedral angle with garnet
insufficient to control the overall dynamics of the' than with olivine (Davies and Stevenson, 1992).
mantle wedge. This speculation is based solely on the observa-
tion that garnet tends to float in pools of melt in
Transport of Water-Dominated Fluid amphibolite melting experiments, which suggests
strongly that at least basalt readily interconnects
Watson et al. (1990) have shown that water exhib- with garnet. It must be noted that the lateral trans-
its lower dihedral angles with olivine (i.e., is more port mechanism does not require that all the water
likely to set up an interconnected fluid network actually enter the wedge at a depth of 80 km or
along crystal edges) under higher pressures and shallower. We speculate as follows on potential
temperatures. The water probably does not form mechanisms and pathways for water mobility in
an interconnected network upon initial release the mantle wedge.
into the mantle wedge due to relatively low tem- If olivine is a good analog for eclogite miner-
peratures. It does at some point however form an als, then the cold temperatures of the slab suggest
interconnected network since the water is continu- that there is no interconnection at textural equilib-
ally taken to conditions of higher pressure and rium. The cold temperatures could prevent the at-
higher temperature. The texture at equilibrium tainment of textural equilibrium, but it must be
depends upon the nature of both the matrix and remembered that the slab is undergoing phase
the fluid, through the surface interactions. Since changes, and releasing water by dehydration; both
the dissolved contents of the hydrous phase can processes lead to more rapid kinetics.
change the nature of these surface interactions If experiments with eclogite minerals show no
they could be very significant. Fluids in subduc- interconnection, even at higher pressures, then we
tion zones have been interpreted to have appre- need to consider other processes that are currently
ciable amounts of dissolved solutes (Philippot poorly understood. The existence of magmatism
and Selverstone, 1991), which would probably in- itself is the strongest argument for the mobility
crease the probability of interconnection (Watson of water from the cold slab and across the cold
et al., 1990). Experimentally Watson etal. (1990) mantle wedge because it is easier to accept the hy-
have found only a small effect for halides with oli- pothesis of water mobility under these conditions
vine as opposed to the large effect of halides with than to discard the hypothesis of hydrous fluxing
quartz. Dissolved silica might be more effective of the mantle wedge as the cause of subduction
because it would make the fluid more melt-like zone magmatism. Clearly, this aspect of the mi-
(melts do interconnect with olivine), and the level gration of water from the slab to the wedge is the
of dissolved silica could be very high. Provided most poorly understood and is the weakest link to
 9. Lateral Water Transport across a Mantle Wedge 211
the hydrous fluxing hypothesis. For those who do temperature, they suffer less from blockage by the
not see hydrous fluxing as a strong hypothesis and precipitation of solutes. Watson et al. (1990) state
believe the mobility of water under cold condi- the difficulties inherent in the consideration of
tions to be impossible, we present some further volatiles transported by cracks. Experiments by
evidence for the mobility of water. Brenan and Watson (1988) suggest that crack
Large serpentine diapirs have been observed in propagation is a plausible mechanism for fluid mi-
the Marianas fore-arc (Fryer, 1992), veining in the gration at high P and T. at least at the grain scale.
Catalina schist fore-arc melange (Bebout, 1991; Clearly of significance is the associated defor-
Bebout and Barton, 1989), and pervasive hydra- mation, which could lead to aligned textures and
tion and metasomatism in the ultramafic hanging also to the syntectonic migration of fluid (Philip-
wall of the Trinity thrust (Peacock, 1987a), as well pot and Selverstone, 1991). For example, the con-
as appreciable flow deep in accretionary prisms tinuous increase of water pore fluid pressures may
(Vrolijk et al., 1988); all point to the mobility of ultimately produce hydrofracturing. It is possible
water under these conditions. Mysen et al. (1978) then that the volume increase produced by the de-
have experimentally measured an aqueous infil- hydration of some minerals could lead to higher
tration rate in peridotite of 23 m . yr -I, at 2 GPa pressures if the fluid cannot readily escape. This
0
and 850 C, although the identity of the active mi- situation is unlikely for amphibole dehydration
cromechanisms and their relevance for the mantle in its high-pressure breakdown region since that
remain uncertain. involves a volume decrease. The increase in tem-
An additional process is probably some form perature experienced by the fluids as they are
of microcracking, but first we mention factors dragged down in the wedge flow field and es-
that might make the development of a porous net- pecially as they propagate upward could lead
work more likely than suggested solely by dihe- to increases of pressure, which could hydraulic-
dral angle data. Interconnection could result from ally open fractures. Are intermediate-depth earth-
aligned, needle-shaped amphiboles with aniso- quakes related to the presence of overpressured
tropic surface energies, such that there is intercon- fluids (Blanpied et al., 1992)? If so, it may be
nection in a special direction (Wolf and Wyllie, possible that some fluids are differentially com-
1990). Alternatively, interconnection could be the pressed, and perhaps because of the temporary
result of heterogeneous and localized hydration presence of a seal, for example, cannot readily mi-
along fractures at mid-ocean ridges, leading natu- grate, leading to a pore pressure increase, and
rally to interconnection at dehydration, assuming facilitating subsequent faulting. The possibilities
that the heterogeneous hydration was not homog- are numerous. Pressurization could reach a level
enized during the subsequent metamorphism as where one would get hydrofracturing in several
the result of subduction. One interesting possibil- plausible scenarios. The direction of least com-
ity in the wedge is that the water sets up an inter- pressive stress (Davies and Stevenson, 1992) is
connected network by forming pathways around such that a hydro fracture would be expected to
amphibole grains that form during mantle hydra- be directed outward and upward into the mantle
tion. This process requires sufficient volume frac- wedge.
tions of amphibole grains so that the resulting net- From the phase diagram in Fig. 7, we see that
work could cross the percolation threshold. If so, the solidus of water-saturated basalt is between
then the water could be generating its own per- about 650 and 750 0 C. Even though the tempera-
meability by amphibolitizing the mantle wedge. tures at the boundary between the subducting oce-
Microcracking must be seriously considered anic crust and the mantle wedge in the models
since water at these temperatures and pressures is presented are generally lower than this, it must
very reactive and hence could initiate subcritical be remembered that frictional heating is ignored,
0
cracking by means of stress corrosion (Anderson which will increase temperatures by up to 50 C
and Grew, 1977; Atkinson, 1984). Ultimately the (Peacock, 1990) or possibly by as much as 200 0 C,
microfractures coalesce to form larger cracks that depending upon rheological conditions (van den
could propagate efficiently and temporarily leave Beukel and Wortel, 1987). In addition, tempera-
a porous network in their wake. Since the cracks tures are higher for younger subducting crust and
are always propagating into conditions of higher for younger subduction zones, which have not yet
 212 J. Huw Davies

established a steady-state thermal structure. The Estimate of the Water Flux That Enters
slab-wedge interface gets hotter as we go deeper. the Mantle Wedge
Therefore, there are certain situations in which the
water leaves the slab as a hydrous melt and not as The water starts its journey from the ocean to the
a water-rich fluid. Since most silicate melts exam- mantle wedge during hydrothermal activity at the
ined experimentally to date form interconnected mid-ocean ridge. Alt et al. (1986) have logically
networks with most silicate minerals, it is prob- argued that the hydrothermal alteration of oceanic
able that this melt would also have no problem mi- crust is strongly controlled by permeability. Fol-
grating from the subducting crust (Cooper and lowing this logic, the greatest alteration should be
Kohlstedt, 1982; Waff and Bulau, 1979, 1982). expected in the extruded basalts, decreasing as
Given that most processes of migration are we descend through the sheeted dikes, and be-
probably favored by higher temperatures and ing dramatically lower in the gabbros. Peacock
pressures, two other possible geometries for the (1990) assesses the degree of alteration and sug-
lateral water transport mechanism are illustrated gests 2 wt% water in the upper 2.5 km of basalt
in Fig. 10. and I wt% water in the lower 5 km of gabbro.

Hydrated mantle Mantle with water that can

~ migrate vertically

El Mantle with free but

stationary water
Hydrated mantle, with mobile

m Source region
~ water or a mobile hydrous melt

Figure 10 Two possible paths for the lateral transport of water out into the mantle wedge. (a) Water leaves the slab easily, but
cannot migrate vertically in the mantle wedge until it is dragged to higher temperatures and pressures. (b) Water cannot leave the
slab until it is dragged to higher temperatures and pressures, but can migrate easily upon entering the mantle wedge.
 9. Lateral Water Transport across a Mantle Wedge 213
Adjusting the figures to allow for the rapid drop in the source of some of this fluid may be the thrust
permeability with depth, we suggest that the up- itself (Vrolijk et al., 1988). Further evidence for
per 1.2 km might be widely altered, containing the mobility of water in subduction zones and ac-
1.5 wt% water, as observed in hole 504B (Becker cretionary prisms is provided by abundant evi-
et al.. 1989), and the rest of the sheeted dikes and dence of pore fluid expulsion (Kastner et al..
gabbros might be negligibly altered, only suffer- 1991) and the alteration of the Catalina schist (Be-
ing alteration locally. Other deep sea drilling holes bout and Barton, 1989), which was in the over-
quoted by Peacock (332B and 418A; 1990) aver- hanging wall of a subduction zone.
aged I wt% water in their drilled sections «600 m Clearly, as the oceanic crust proceeds to higher
of pillow basalt); the only substantial drilling of pressures and temperatures, so the metamor-
gabbro (a fracture zone where more hydration phic facies change from blueschist to greenschist
might be expected because of greater access) in through amphibolite to eclogite. Blueschists have
hole 735B gave 0.8 wt% water, averaged over been estimated to hold 3 wt% water, so it is pos-
500 m. This estimate implies substantially less sible that some of the water released in the sedi-
water than Peacock's estimate, which was already ments is incorporated into the blueschists below,
less than many previous estimates (Ito et al., since the initial degree of hydration is probably
1983, 2 wt% throughout the oceanic crust; Fyfe, less than 3 wt%. Generally the water content of
1978, and Anderson et al.. 1976, who estimate each succeeding higher pressure facies is lower.
3 wt% through the whole thickness of the oceanic We suggest that all the water released in the tran-
crust). sition from the blueschist facies to greenschist fa-
Gregory and Taylor (1981) found in the Oman cies, and a proportion of water released during the
ophiolite evidence from oxygen isotopes for ex- greenschist to amphibolite facies transition, es-
tensive water-rock interaction to a depth greater capes back up the thrust or into the fore-arc since
than 5 km. It is possible that the Oman ophiolite it occurs at a shallower level than the base of the
corresponded to a back-arc basin rather than an thrust (around 60 km; Tichelaar and Ruff, 1989).
oceanic ridge, and clearly that it was obducted The amount of water released into the mantle
rather than subducted implies that it is not typi- wedge is therefore probably much lower than the
cal. Since ophiolites are affected by later regional estimate of water fixed in hydrous minerals at the
metamorphism, doubts exist about whether they ridge. Ultimately most of the water is held in am-
are truly representative of alteration at mid-ocean phiboles, probably pargasite.
ridges.
As the oceanic crust subducts beneath the over- Importance of Hydrous Minerals Other
riding plate in the trench, most of the sediments Than Amphiboles
are added to the foot of the accretionary prism
or underplated beneath. The free pore water and Clearly it is possible that a hydrous mineral more
weakly bound water are released and driven to the stable than amphibole could take some water to
surface by compaction processes. Hence, most of an appreciable depth and that some of this water
the water in the sediments and the oceanic crust might reappear in back-arc basin volcanism. Can-
does not reach the thrust; in fact, it might be the didates include the nominally anhydrous min-
presence of large amounts of water that prevents a erals (e.g., garnet, clinopyroxene, olivine, ortho-
brittle thrust extending all the way to the surface. pyroxene; Bell and Rossman, 1992), as well as
We assume that all water released at a shal- K-amphibole and phlogopite (Niida and Green,
lower level than the bottom of the interplate thrust 1990; Sudo and Tatsumi, 1990). They all could
returns to the near surface up the thrust or enters take the water deep into the mantle and balance
the fore-arc (Fryer, 1992; Peacock, 1987a). Then the output from the mantle at mid-ocean ridges
we need only consider the water that is released and at hotspots. The amounts of phlogopite, K-
below this depth. The question of whether flow is amphibole, and K-richterite all depend on the
allowed along a fault is difficult to answer; but the amount of potassium available. Since the amount
evidence for substantial flow along the decolle- is likely to be very low in a depleted mantle,
ment beneath accretionary prisms suggests that phlogopite and the other K-hydrous minerals are
214 J. Huw Davies

unlikely to be sufficiently common to clear the the cornposition of the wedge but also on the
subduction zone wedge of all its water. They amount of water, the composition of its carrier
might be responsible for the water in back-arc ba- phase, and the temperature and pressure (Tatsumi,
sin basalts though (Stolper and Newman, 1992). 1989; Thompson, 1992). From experiments with
Other candidates proposed to remove water peridotite, pargasite was found to be the primary
from the subduction zone should be discussed. hydrous mineral formed (Green, 1973). Green
The dense hydrous magnesian silicates, or "al- suggests that one could get up to 30 wt% pargas-
phabet minerals," can hold large amounts of water ite, i.e., a fixed water content in the wedge (11') of
and are stable to high pressures. It seems probable 0.4 wt% H 2 0 . By equating the flux from the oce-
though that they are not common in hydrated arc anic crust into the mantle wedge (F) with the flux
peridotites (Wunder and Shreyer, 1992). Serpen- carried horizontally by the induced flow of the
tine is stable to high pressures, but not to high amphibole-saturated mantle wedge, we can evalu-
temperatures, so it could be very important locally ate the thickness of the amphibolitized layer (for
in cold regions of the mantle wedge nearest the further details see Appendix B, Davies and Ste-
slab due to its high water content, but is unlikely venson, 1992); i.e.,
to be important in removing water from the sub-
(7)
duction zone region. Talc is another important hy-
drous mineral, with relatively high water content; where F is the flux of H 2 0 , V, the horizontal com-
it is also stable to high pressures (beyond 5 GPa), ponent of mantle velocity, w the weight percent-
but its thermal stability is limited to less than age of H 2 0 in the amphibolitized mantle, p.; the
0
900 C. Chlorite stability is temperature depen- mantle density, and h the thickness of the amphi-
dent, and at 5000 C chlorite is stable to 5 GPa, but bolitized layer. We have ignored the water in the
at 8000 C it is stable to only 3 GPa (Goto and Ta- free fluid porosity, which is also carried horizon-
tsumi, 1990). Like serpentine, it is improbable tally and is shown as follows to be less than w.
that talc or chlorite will carry much water out of Assuming V, = 2 cm· yr- I (6 X 1O-lOm· S-I),
the subduction zone but again they could be im- 11' = 0.4%, and a flux (F) of water entering the
portant in the cooler regions of the mantle wedge mantle wedge, which is the result of the release of
and could carry some water beyond the source re- 1 wt% water from a I-km thickness of oceanic
gion of the volcanic front. For further discussion crust subducting at j crn-yr v'{Z X 10 9m ' s-I),
of the distribution of water in the upper mantle, then the thickness (h) is 8.75 km. Since the hori-
see Thompson (1992) and Tatsumi (1989). zontal component of the induced mantle wedge
velocity decreases with distance from the slab, the
How Much Amphibole Is Formed? thickness of amphibolitized mantle increases with
distance from the slab. This trend might be coun-
If the mantle is dry next to the oceanic crust, then tered by the fact that the mantle nearest the oce-
the water reacts to form amphibole (probably par- anic crust possibly cannot hold as much water in
gasite). The mantle entrained nearest the slab is hydrous minerals since it might be more depleted
very shallow, possibly directly beneath the crust. due to its shallower origin (i.e., w < 0.4 wt%). If
The probable depth from which this mantle is the flux instead had been 2 wt% water released
ablated is a strong function of the composition from 7 km of oceanic crust then the thickness of
and temperature of the mantle, and also the pre- the amphibolitized mantle at a point at which its
vious tectonic history in terms of stress release, horizontal velocity is 2 em . yr - 1 would be around
work hardening, etc. Further from the subducting 120 km. Since the mantle wedge is not this thick,
oceanic crust, the entrained mantle comes from the mechanism is not able to transport all the wa-
greater depths and is probably more fertile. Hence, ter laterally, and the excess water returns to the
it is possible that little amphibole can be formed near surface. As stated previously, we prefer lower
in the harzburgite close to the slab, but it increases fluid fluxes entering the mantle wedge than pre-
volumetrically as we progress away from the slab viously assumed, so we expect the thickness of
into more lherzolitic compositions. Clearly, what amphibolitized mantle to be closer to 10 km than
hydrous phases are formed depends not only on 100 km.
 9. Lateral Water Transport across a Mantle Wedge 215
How Much Free Water Is There? a liquid phase. This porosity is sufficiently high to
have a limited influence on the phase diagram.
The amount of free phase can be calculated by as- Amphibole in this case is not stable to as high
suming that the system is steady state and by temperatures as those for the case with no excess
equating the vertical fluxes (Appendix B, Davies water, which is the assumption for the amphibole
and Stevenson, 1992). The vertically downward stability curve in the phase diagram of Fig. 7. The
flux, carried by the hydrous minerals in the in- peridotite here would have a water content of
duced wedge flow, must be balanced by the verti- 0.6 wt% (0.4 wt% (11') + 0.2 wt% (20% of f));
cal flux in the upward direction resulting from the Green (\ 973, see his Fig. 3) suggests that amphi-
flow of the free positively buoyant phase. If the bole should still be stable until around 1100° C.
ascent and vertical transport are the result of hy- Hence, we expect the source region to be initi-
drofracture, then they are very rapid and very little ated at slightly lower temperatures (compared to
free phase is required. To evaluate an upper bound around 1150-1175° C), closer to the slab, which
we assume that the upward transport is a result of also implies that the peak temperature of the
the slower porous flow. The relevant equation is source region is proportionally lower. Many un-
certainties evident in these calculations, including
the permeability versus porosity relationship, the
where p.; is the mantle density, V, is the vertical viscosity of the fluid, and the grain size, would
component of mantle velocity, w is the weight affect estimates of the porosity of the free water-
percentage of H 2 0 in the amphibolitized mantle, carrying phase in the amphibole-saturated re-
f is the fluid-filled porosity, p, is the density of the gions. Provided the porosity does not become so
fluid, a is the grain size, !::"P is the density contrast high as to change the phase diagram, then the ba-
between the fluid and the matrix, g is the gravita- sic mechanism will still be effective. The flux of
tional acceleration, TJ is the viscosity of the fluid, water into the wedge and the water content of
and b is the permeability-porosity constant. Here amphibolitized mantle are also uncertain and af-
we have assumed that the permeability (k) is pro- fect the height (thickness) of the region that is
portional to the square of both the grain size (a) amphibolitized.
and porosity (f), i.e., k = a 2f 2/b (as predicted in
detailed calculations by Cheadle (1989) for low Chemistry of Amphiboles
porosity, with b = 3 X 103 ) . If the water has a
viscosity (TJ) of 10- 4 Pa . s then we find that the We have continuously assumed that pargasite is
porosity (f) is 6 X 10 -5. We have assumed a ver- the dominant amphibole and hydrous mineral at
tical velocity (VJ of 3 cm . yr : ', a grain size (a) these depths. It is critically important (as men-
of I mm, and a density difference (!::"p) between tioned previously) that there is no other hydrous
the fluid and the matrix of 2.0 X I0 3 kg . m 3. mineral that is sufficiently stable and so common
Clearly such a low porosity has no effect on the that it can take all the water released by pargasite.
phase diagram; i.e., it is unclear whether even the The amphibole family is a very broad family with
experimental charges of nominally anhydrous a range of cation sites, enabling it to accommodate
runs could demonstrate that they had less than a wide range of ions. This flexibility leads to the
0.0 I wt% water present, due to surface-adsorbed possibility that slight variations exist in the chemi-
water. Applying the same calculation to the region cal composition of the amphiboles as we go to
where wet basalt is the free phase carrying water higher temperatures and pressures so that their
vertically upward, we find a value for the porosity stability is increased somewhat. Can this increase
(f) of 10 -2. Here we have assumed a viscosity of help explain the occasional observation that the
10 Pa . s, a grain size of 3 mm, a density difference potassium contents of lavas tend to increase as we
of 0.5 k . g . m -3, a vertical velocity downward progress away from the trench (Dickinson and
for the induced flow of 3 cm . yr r and a solu-
'. Hatherton, 1967)? The assumption would be that
bility of 20 wt% for water in the melt. A larger more K is present at greater depths since amphi-
grain size might be expected due to Ostwald rip- boles richer in potassium may be more stable than
ening at the higher temperatures in the presence of Na-rich amphiboles.
 216 J. Huw Davies

Model Predictions are poorly understood. It is unclear what differ-

ences free convection in a dynamic thermal model
Presence and Location of the Volcanic Front would make relative to the present forced convec-
From thermal modeling we have seen that suffi- tion in a kinematic model of subduction; it clearly
ciently high temperatures can be achieved quite would lead to a modulation and possible tempo-
close to the slab at a depth of 60-100 km. The ral variability (Davies and Stevenson, 1992). The
1200° C isotherm can get to within 20 km of the thermal effects of melting were not included; they
oceanic crust. The lateral H 2 0 transport mecha- should not have a large effect on the total ther-
nism would then suggest a source region located mal structure, but could lead to significant local
vertically above the subducting slab at a depth of changes in and around the source region. For ex-
around 120-140 km. If the melts were to rise ample, they might lead to differences in the source
vertically then the volcanic front would be 130- width.
ISO km above the Benioff zone, which is some- Our lack of understanding with regard to the
what deeper than estimates. Davies and Steven- depth of the asthenosphere-lithosphere boundary
son (1992)-assuming that the least compressive (thickness of the mechanical boundary layer) af-
stress controls the direction of crack propagation fects the accuracy of the shallow thermal struc-
in the asthenosphere-show that the stress regime ture. This uncertainty probably has a larger effect
induced by a corner flow focuses the cracks to- on our lack of understanding of the location of the
ward the mantle wedge corner, leading to esti- top of the source region than does our uncertainty
mates for the height of the volcanic front above . with regard to magma segregation. The top of the
the Benioff zone that agree well with observa- source region is at least as shallow as the maxi-
tions. Another possibility is that higher tempera- mum temperature achieved vertically and no shal-
tures are achieved closer to the slab. Such tem- lower than the solidus. Since these levels are not
peratures could occur if the mantle surrounding very far apart « 10 km), the uncertainty in the lo-
subduction zones is hotter than the mantle sur- cation of the top of the source region resulting
rounding mid-ocean ridges, or because the flow is from uncertainty in the control of magma segre-
more focused than that produced by the uniform gation is small.
viscosity assumed in this model, due to variable There is some uncertainty about whether the
rheology, or comes from greater depth due to local stress field controls the direction of crack
thickening of the mechanical lithosphere of the propagation. The difference between crack propa-
overriding plate away from the mantle wedge gation controlled by the local stress field and ver-
corner. The thermal stability of pargasite in the tically propagating cracks limits the possibilities.
mantle is probably slightly lower than assumed in The resulting uncertainty is a small but systematic
the phase diagram of Fig. 7 due to the limited difference in the prediction of the location of the
presence of free water. as discussed in the pre- volcanic front. As mentioned earlier, the vertically
ceding section. Hence, the presence and location propagating cracks lead to volcanic fronts about
of the volcanic front can be rationalized by this 20 km too high above the Benioff zone for this
model for the lateral transport of water to the model.
source region. The high heat flow at the volcanic The uncertainty in the phase diagrams is diffi-
front is partly attributed to heat flow resulting cult to judge, given the large number of variables
from advection by the magma, but such advection and the differences between some studies and simi-
can have no more than a local effect. We ascribe larities between others (Basaltic Volcanism Study
this anomaly in general to the thinness of the me- Project, 1981; Olafsson and Eggler, 1983; Wal-
chanical lithosphere in this region due to ablation lace and Green, 1988). The maximum dehydra-
leading to a much higher conductive geothermal tion melting temperature of amphibole (pargasite)
gradient. from 2-3 GPa, which exerts strong control of the
location of the source region, seems relatively ro-
Uncertainties in Geometric Predictions bust, e.g., 1050°C(Wyllie, 1979), 1150°C(Green,
1973), 1080° C (Wallace and Green, 1988), and
The uncertainties in our model are very difficult to 1070° C (Olafsson and Eggler, 1983).
assess because many of the component processes Many of the uncertainties inherent in the lateral
 9. Lateral Water Transport across a Mantle Wedge 217
· water transport mechanism have been discussed nism that is the focus of this chapter. They consid-
previously. The uncertainty in the velocity of ver- ered only the part of the transport involving a fluid
tical propagation of the fluid depends first on phase, assumed chemical equilibrium, and used
whether it is by porous flow or crack propagation. the partition coefficients of Brenan and Watson
If it is by porous flow, what is the permeability, (1991). The study ignored the difference between
and what is the viscosity of the fluid? These ques- the spinel and garnet facies of the peridotite. The
tions are very difficult to answer, but clearly the study demonstrated that expected trace element
slower the velocity the higher the fluid content. inputs from the subducting slab followed by per-
Provided the water content does not reach levels colation through the mantle wedge could produce
such that the phase diagrams considered are in- an arc trace element signature, provided the parti-
applicable, then the remaining predictions, e.g., tion coefficients were two to three orders of mag-
source width, source region position, are not nitude less than those currently suggested by ex-
greatly affected. In conclusion, many of the in- perimental data, or that the water fluxes were
dividual components of this model are very un- around three orders of magnitude higher than cur-
certain; even so, the major processes are surpris- rently estimated. Other possibilities include an
ingly robust. The melting of a water fluxed mantle imperfect chemical equilibrium since the water
wedge is the result of a two-component lateral possibly migrates by high-velocity fracturing pro-
transport mechanism for the water; it combines cesses. In addition, it should be noted that the ac-
the near-vertical transport of a free phase with tual partition coefficients are very sensitive to the
transport in hydrated minerals carried down by exact nature of the fluid and solid matrix.
the induced secondary flow.
Other Predictions
Major Element Composition of Primary Magmas
The hydrous fluxing model of Davies and Bickle
Davies and Bickle (1991) have undertaken a
(1991) based on this lateral transport mechanism
simple and crude model for melting by fluxing,
suggests reasonable segregation temperatures and
where it was assumed that the source region was
reasonable water contents. Their predictions of
in hydrous equilibrium; i.e., the water content of
arc growth compared to previous estimates (Rey-
the melt is the water content of the whole local
mer and Schubert, 1984) suggest low fluid fluxes
system. With that model they generated predic-
from the oceanic crust or that not all the melt
tions of primary magmas that fall in the middle of
reaches the arc crust. The average degree of melt-
current estimates. They found that the volume of
ing predicted is 2 to 8%, which again may not re-
magma was sensitive to the water flux and to the
flect the melts observed at the surface since prob-
thickness of the melting column (i.e., the thickness
ably not all the primary magmas reach the arc
of mechanical lithosphere). Stolper and Newman
crust. The assumption of the input of water from
(1992) have measured the water contents of back-
the slab allows one to explain the enhanced Sr iso-
arc basin magmas and found a strong correlation
topes (from the Sr input into the oceanic crust by
between water content and composition. They
hydrothermal circulation) and the lOBe and B sig-
have extended the study to include volcanic front
nature (Morris et al.. 1990). It also explains the
magmas in the Marianas and argue that the mag-
more hydrous and explosive nature of subduction
mas are the result of hydrous fluxing, following
zone volcanics.
percolation through the mantle wedge. Fluids flux-
This process is expected to produce a locally
ing the back-arc basins have traveled a longer path
heterogeneous seismic velocity and attenuation
such that they are in equilibrium with the mantle,
structure in the source region and the strongly hy-
while fluids fluxing the mantle to produce the vol-
drated regions. The excellent tomographic seis-
canic front magmas have not equilibrated com-
mic studies of Hasegawa et al. (1991) and Hase-
pletely with the mantle.
gawa and Zhao (1994) seem to image the hot
mantle tongue predicted by the thermal models.
Slab Trace Element Signature
There is also a weak suggestion of a feature at
Hawkesworth et al. (1993) have assumed a perco- around 100 km depth near the slab surface, which
lation model based on the lateral transport mecha- one might speculate could be related to the release
218 J. Huw Davies

of water. We probably need to be conservative in fluxed by a fluid component carrying water in-
interpreting all tomographic images of deep struc- ward from the subducting oceanic crust. The only
tures, especially due to the difficulty of correcting exceptions are possibly the subduction of very
for shallow poorly resolved heterogeneous struc- young, hot oceanic crust or where subduction has
tures. In addition, interpretation of seismic signa- just initiated.
tures is nonunique and difficult, but the preceding The subducting slab induces a corner flow in
study seems largely to support the mechanism the mantle wedge that advects heat into the wedge
presented. Limited spatial resolution (>25 km) corner, leading to relatively high temperatures
though makes it unlikely that the individual com- close to the subducting slab. Most water released
ponents (source region, hydrated layer, region with at shallower levels than the bottom of the inter-
water-rich low-porosity melts) of the mechanism plate thrust is assumed to migrate back up the
can be imaged clearly. The seismic attenuation thrust or into the fore-arc, which is part of the
study of Umino and Hasegawa (1984) also sug- mechanical lithosphere. The remaining water re-
gests a similar thermal structure, with Sato (1992) leased, largely from amphiboles, is argued to en-
and Sato and Sacks (1990) interpreting the inter- ter the mantle wedge where it reacts to form
mediate attenuation in the fore-arc as evidence of additional amphiboles. The induced secondary
hydration. flow carries these amphiboles down through their
breakdown pressure (depth) range, where the
Discussion of Estimated Magma Segregation
water is subsequently released. It rises vertically
Temperatures
through the amphibole-saturated mantle until it
Not all observations can be explained easily by reaches dry mantle, with which it again reacts to
this model. One observation concerns the pressure form additional amphibole. The cycle thus repeats
and temperature conditions under which estimated with the net integrated result being the lateral
primary basalts are in equilibrium with olivine, transport of water away from the oceanic crust and
orthopyroxene, and clinopyroxene (Nye and Reid, inward to the hot regions of the mantle wedge.
19,86; Tatsumi et al., 1983). Tatsumi et al. (1983) The process is assumed to continue even when it
suggest that the magmas segregate at a tempera- is a near water-saturated melt that is rising verti-
ture of 1320° C and at pressures ranging from II cally since at low water contents, amphiboles are
to 23 kbar. It is argued that these temperatures are stable to some extent above the water-saturated
so high that they are due to a diapir, and due to the solidus.
absorption of the latent heat of melting in the dia- In this chapter, we hope we have enforced the
pir, it must have passed through a region with tem- robustness of the general mechanism while dis-
peratures higher than 1400°C. A temperature of cussing the uncertainties in critically assessing it.
1400° C is higher than those envisaged in this The major question is, How does water leave the
model. Interpretation of the results of these equili- slab and move across the cold region of the mantle
bration experiments in terms of conditions in the wedge? There is little question in my mind that
mantle wedge assumes that the primary magmas such movement occurs, and a selection of possible
on segregation were in equilibrium with the three processes have been identified, with microcrack-
residual minerals. The primary magma is pro- ing a probable candidate; however, as long as this
duced over a depth interval and not at one point uncertainty remains, models of hydrous fluxing of
(McKenzie and Bickle, 1988) and probably does mantle wedge will be incomplete. Such models
not continue to reequilibrate as it migrates up can explain the presence and location of a vol-
through the melting column (Spiegelman and Ken- canic front and also give a reasonable estimate for
yon, 1992). Hence, the assumptions implicit in the the composition of the primary magma.
interpretation of Tatsumi et al. (1983) are ques-
tionable and the constraints provided by the re-
sults are equally questionable. Acknowledgments

I acknowledge discussions with Prof. Dave Stevenson and

Conclusions Dr. Mike Bickle, who have contributed to aspects of this work.
The reviews of two anonymous referees helped improve this
The source areas of subduction zone magmatism chapter. The NSF (US) and NERC (UK) supported much of
are regions within the mantle wedge that are the work presented here.
 9, Lateral Water Transport across a Mantle Wedge 219
References Brophy, J. G., and Marsh, B. D. (1986). On the origin of high-
alumina arc basalts and the mechanics of melt extraction,
Alt, 1. C, Honnorez, 1.. Laverne, c., and Emmermann, R. 1. Petrol. 27, 763-789.
(1986). Hydrothermal alteration of Ikm section through Cheadle, M. 1. (1989). "Properties of Texturally Equili-
the upper oceanic crust, Deep Sea Drilling Project Hole brated Two-Phase Aggregates," Ph.D. Thesis, Univ. of
504B: Mineralogy, chemistry and evolution of seawater- Cambridge.
basalt interactions, 1. Geophys. Res. 91, 10309-10336. Cooper, R. E, and Kohlstedt, D. L. (1982). Interfacial energies
Anderson, O. L., and Grew, P. C. (1977). Stress corrosion in the olivine-basalt system, in "High Pressure Research
theory of crack propagation with applications to geophys- in Geophysics," (S. Akimoto and M. N. Manghnani eds.),
ics, Rev. Geophys. Space Phys. 15,77 -104. Adv. Earth Planet. Sci. 12, pp. 217-228, Center for Aca-
Anderson, R. N., Delong, S. E., and Schwarz, W. M. (1980). demic Publications, Tokyo.
Dehydration, asthenospheric convection' and seismicity in Creager, K. c., and Jordan, T. H. (1984). Slab penetration into
subduction zones, J. Geol. 88, 445-451. the lower mantle, 1. Geophys. Res. 89, 3031-3049.
Anderson, R. N., Uyeda, S., and Miyashiro, A. (1976). Geo- Davies, 1. H., and Bickle, M. 1. (1991). A physical model for
physical and geochemical constraints at converging plate the volume and composition of melt produced by hydrous
boundaries. Part I. Dehydration in the downgoing slab, fluxing above subduction zones, Philos. Trans. R. Soc.
Geophys. J. R. Astron. Soc., 44, 333-357. London A 335, 355-364.
Andrews, D. 1., and Sleep, N. H. (1974). Numerical modelling Davies, 1. H., and Stevenson, D. 1. (1992). Physical model of
of tectonic flow behind island arcs, Geophys. 1. R. Astron. source region of subduction zone volcanics, J. Geophys.
Soc. 38, 237-251. Res. 97, 2037-2070.
Atkinson, B. K. (1984). Subcritical crack growth in geological Dickinson, W. R., and Hatherton, T. (1967). Andesitic vol-
materials, J. Geophys. Res. 89,4077-4114. canism and seismicity around the Pacific, Science 157,
Basaltic Volcanism Study Project (1981). "Basaltic Volcanism 801-803.
Study Project, Basaltic Volcanism on the Terrestrial Plan- Eggler, D. H. (1987). Solubility of major and trace elements
ets," Pergamon, New York. in mantle metasomatic fluids: Experimental constraints
Batchelor, G. K. (1967). "An Introduction to Fluid Dynam- in "Mantle Metasomatism" (M. A. Menzies and C. J.
ics." Cambridge Univ. Press, Cambridge. Hawkesworth, eds.), pp. 21-41, Academic Press, San
Beard, 1. S., and Lofgren, G. E. (1990). Dehydration melting Diego.
and water-saturated melting basaltic and andesitic green- Fryer, P. (1992). Mud volcanoes of the Marianas, Sci. Am.
stones and amphibolites, 1. Petrol. 32, 365-401. Feb., 26-32.
Bebout, G. E. (1991). Field-based evidence for devolatiliza- Fujii, N., Osamura, K., and Takahashi, E. (1986). Effect of
tion in subduction zones: Implications for arc magmatism, water saturation on the distribution of partial melt in the
Science 251, 413-416. olivine-pyroxene-plagioclase system, J. Geophys. Res.
Bebout, G. E., and Barton, M. D. (1989). Fluid flow and meta- 91,9253-9260.
somatism in a subduction zone hydrothermal system: Cat- Fujisawa, H., Fujii, N., Mizutami, H., Kanamori, H., and Aki-
alina schist terrane, California, Geology 17, 976-980. moto, S. (1968). Thermal diffusivity ofMg,Si0 4 , Fe,Si0 4 ,
Becker, K., et al. (1989). Drilling deep into young oceanic and NaCI at high pressure and temperature, J. Geophys.
crust at hole 504B, Costa Rica Rift, Rev. Geophys. 27, Res. 73, 4727-4733.
79-102. Furukawa, Y., and Uyeda, S. (1989). Thermal state under the
Bell, D. R., and Rossman, G. R. (1992). Water in Earth's Tohoku Arc with consideration of crustal heat generation,
mantle: The role of nominally anhydrous minerals, Sci- Tectonophysics 164, 175-187.
ence 255, 1391-1397. Fyfe. W. S., Price, N. 1., and Thompson, A. B. (1978). "Fluids
Blanpied, M. L., Lockner, D. A., and Byerlee, J. D. (1992). An in the Earth's Crust." Elsevier, Amsterdam.
earthquake mechanism based on rapid sealing of faults, Garfunkel, Z., Anderson. C. A., and Schubert, G. (1986).
Nature 358, 574-576. Mantle circulation and the lateral migration of subducted
Bloxham, 1., and Jackson, A. (1991). Fluid flow near the sur- slabs, J. Geophys. Res. 91,7205-7223.
face of Earth's outer core, Rev. Geophys. 29, 97-120. Giggenbach, W. E (1992). Isotopic shifts in waters from
Bodri, L., and Bodri, B. (1978). Numerical investigation of tee- geothermal and volcanic systems along convergent plate
tonic flow in island-arc areas, Tectonophysics 50, 163-175, boundaries and their origin, Earth Planet. Sci. Lett. 113,
Brenan,1. M., and Watson, E. B. (1988). Fluids in the litho- 495-510.
sphere, 2. Experimental constraints on CO, transport in Gill, 1. (1981). "Orogenic Andesites and Plate Tectonics,"
dunite and quartzite at elevated poT conditions with impli- Springer-Verlag, New York.
cations for mantle and crustal decarbonation processes, Goto, A., and Tatsumi, Y. (1990). Stability of chlorite in the
Earth Planet. Sci. Lett. 91,141-158. upper mantle, Am. Mineral. 75, 105-108.
Brenan, 1. M., and Watson, E. B. (1991). Partitioning of trace Green, D. H. (1973). Contrasted melting relations in a pyrolite
elements between olivine and aqueous fluids at high PoT upper mantle under mid-oceani~' ridge, stable crust and is-
conditions: Implications for the effect of fluid composition land arc environments, Tectonophysics 17, 285-297.
on trace element transport. Earth Planet. Sci. Lett. 107, Gregory, R. T., and Taylor, H. P. (1981). An oxygen isotopic
672-688. profile in a section of Cretaceous oceanic crust, Samail
Brooks, A. (1981). "A Petrov-Galerkin Finite-Element For- ophiolite, Oman: Evidence for 1'0 buffering of the oceans
mulation for Convection Dominated Flows," Ph.D. The- by deep (>5km) seawater-hydrothermal circulation at
sis, California Institute of Technology. mid-ocean ridges, J. Geophys. Res. 86, 2737 -2756.
220 J. Huw Davies
Hamilton. W. B. (1988). Plate tectonics and island arcs. Geol. McKenzie, D. P., and Bickle, M. 1. (1988). The volume and
Soc. Am. Bull. 100, 1503-1527. composition of melt generated by extension of the litho-
Hasegawa, A.. and Zhao, D. (1994). Deep structure of island sphere, J. Petrol. 29, 625-680.
arc magmatic regions as inferred from seismic observa- Minear,1. w., and Toksoz, M. N. (l970a). Thermal regime of
tions in "Magmatic Systems" (M. P. Ryan, ed.), Academic a downgoing slab, Tectonophysics 10,367 -390.
Press, San Diego. Minear.J, w., and Toksoz, M. N. (l970b). Thermal regime of
Hasegawa, A., Zhao, D., Hori, S., Yamamoto, A.. and Hori- a downgoing slab and new global tectonics, J. Geophys.
uchi, S. (1991). Deep structure of the northeastern Japan Res. 75,1397-1419.
arc and its relationship to seismic and volcanic activity. Morris, 1. D., Leeman, W. P., and Tera, F. (1990). The sub-
Nature 352, 683-689. ducted component in island arc lavas: Constraints from Be
Hawkesworth, C. 1., Gallagher, K, Hergt, 1. M., and Mc- isotopes and B-Be systematics, Nature 344,31-36.
Dermott, F. (1993). Trace element fractionation processes Mysen, B. 0., Kushiro, I., and Fujii, T. (1978). Preliminary
in the generation of island arc basalts, Phil. Trans. R. Soc. experimental data bearing on the mobility of H,O in crys-
London A., 342, 179-191. talline upper mantle, Carnegie lnst. Washington Year
Helffrich, G. R., Stein, S., and Wood, B. 1. (1989). Subduction Book, 793-797.
zone thermal structure and mineralogy and their relation- Nagao, T., and Uyeda, S. (1989). Heat flow measurements in
ship to seismic wave reflections and conversions at the the northern part of Honshu. Northeast Japan, using shal-
slab/mantle interface, J. Geophys. Res. 94, 753- 763. low holes, Tectonophysics 164, 301-314.
Honda, S. (1985). Thermal structure beneath Tohoku, North- Niida, K, and Green, D. H. (1990). Pargasitic dehydration soli-
east Japan-A case study for understanding the detailed dus of peridotites hydrated in subduction wedge mantle,
thermal structure of the subduction zone, Tectonophysics EOS Trans. Am. Geophys. Union 71, 949.
112,69-102. Nye, C. 1., and Reid, M. R. (1986). Geochemistry of primary
Hsui, A. T.. Marsh, B. D., and Toksoz, M. N. (1983). On the and least fractionated lavas from Okmok volcano, central
melting of the subducted oceanic crust: Effects of subduc- Aleutians: Implications for arc magmagenesis, J. Geo-
tion induced mantle flow, Tectonophysics 99, 207 - 220. phys.Res.91,10271-10287.
Huang, w.i., Wyllie, P. 1., and Nehru, C. E. (1980). Subsoli- Olafsson, M., and Eggler, D. H. (1983). Phase relations of am-
dus and liquidus phase relationships in the system CaO- phibole, amphibole-carbonate, and phlogopite-carbonate
SiO,-CO, to 30kbar with geological applications, Am. peridotite: Petrologic constraints on the asthenosphere,
Mineral. 65, 285-301. Earth Planet. Sci. Lett. 64, 305-315.
Hughes, T. 1. R., Liu, W. K., and Brook, A. (1979a). Finite Otsuki, K (1989). Empirical relationships among the conver-
element analysis of incompressible viscous flows by the gence rate of plates, rollback rate of trench axis and island-
penalty function formulation, J. Comput. Phys. 30, 1-60. arc tectonics: "laws of convergence rate of plates," Tec-
Hughes, T. J. R., Pister, K S., and Taylor, R. L. (I 979b). tonophysics 159, 73-94.
Implicit-explicit finite elements in nonlinear transient Otsuki, K, Heki, K, and Yamazaki. T. (1990). New data
analysis, Compo Methods Appl. Mech. Eng. 17, 159-182. which support the "laws of convergence rates of plates"
Ida, Y. (1987). Structure of the mantle wedge and volcanic proposed by Otsuki, Tectonophysics 172, 365-368.
activities in the island arcs, in "High-Pressure Research Oxburgh, E. T.. and Turcotte, D. L. (1968). Problems of high
in Mineral Physics" (M. H. Manghnani and Y. Syono, heat flow and volcanism associated with zones of descend-
eds.), pp. 473-480, Terra Scientific Publishing Co., To- ing mantle convective flow, Nature 216,1041-1043.
kyo; American Geophysical Union, Washington, DC. Oxburgh, E. T., and Turcotte, D. L. (1970). Thermal structure
Ito, E. I., Harris, D. M., and Anderson, Jr, A. T. (1983). Alter- of island arcs, Geol. Soc. Am. Bull. 81, 1665-1688.
ation of oceanic crust and geologic cycling of chlorine and Peacock, S. M. (l987a). Serpentinization and infiltration meta-
water, Geochim. Cosmochim. Acta 47, 1613-1624. somatism of the Trinity peridotite, Klamath province,
Jurdy, D. M., and Stefanick, M. (1983). Flow models for back- northern California: Implications for subduction zones,
arc spreading, Tectonophysics 99,191-206. Contrib. Mineral. Petrol. 95,55-70.
Kastner, M., Elderfield, H., and Martin, 1. B. (1991). Fluids in Peacock, S. M. (I 987b). Thermal effects of metamorphic
convergent margins, What do we know about their com- fluids in subduction zones, Geology 15, 1057-I 060.
position, origin, role in diagenesis and importance for oce- Peacock, S. M. (1990). Fluid processes in subduction zones.
anic chemical fluxes? Philos. Trans. R. Soc. London A 335, Science 248, 329-337.
243-259. Peacock, S. M. (1992). Blueschist-facies metamorphism, shear
King, S. D., Raefsky, A., and Hager, B. H. (1990). ConMan: heating and P-T-t paths in subduction shear zones, J.
Vectorizing a finite element code for incompressible two- Geophys. Res. 97, 17693-17707.
dimensional convection in the Earth's mantle, Phys. Earth Philippot, P., and Selverstone, 1. (1991). Trace-element-rich
Planet. Int. 59,195-207. brines in eclogitic veins: Implications for fluid composi-
Marsh, B. D. (I 979a). Island arc development: Some ob- tion and transport during subduction, Contrib. Mineral.
servations, experiments and speculations, J. Geol. 87, Petrol. 106, 417 -430.
687-713. Plank, T., and Langmuir, C. H. (1988). An evaluation of the
Marsh, B. D. (l979b). Island-arc volcanism, Am. Sci. 67, global variations in the major element chemistry of arc ba-
161-172. salts, Earth Planet. Sci. Lett. 90, 349-370.
McKenzie, D. P. (1969). Speculations on the consequences Reymer, A.. and Schubert, G. (1984). Phanerozoic addition
and causes of plate motions, Geophys. J. Roy. Astron. Soc. rates to the continental crust and crustal growth, Tectonics
18,1-32. 3,63-77.
 9. Lateral Water Transport across a Mantle Wedge 221
Sato, H. (1992). Thermal structure of the mantle wedge be- structure in the northeastern Japan arc, .I. Seismol. Soc. .Ia-
neath northeastern Japan: Magmatism in an island arc pan (Zisin) 37, 217-228.
from combined data of seismic anelasticity and velocity Van den Beukel, J., and Wortel, R. (1987). Temperatures and
and heat flow,.I. volcanol. Geotherm. Res. 51, 237-252. shear stresses in the upper part of a subduction zone, Geo-
Sato, H., and Sacks, I. S. (1990). Magma generation in the phys. Res. Lett. 14, 1057-1060.
upper mantle inferred from seismic measurements in pe- Vrolijk, P., Myers, G., and Moore, 1. C. (1988). Warm fluid
ridotite at high pressure and temperature, in "Magma migration along tectonic melanges in the kodiak accretion-
Transport and Storage" (M. P. Ryan. ed.), pp. 277-292, ary complex. Alaska. .J. Geophys. Res. 93, 10313-10324.
Wiley, Chichester/Sussex, England. Waff, H. S., and Bulau, J. R. (1979). Equilibrium fluid distri-
Spiegelman, M., and Kenyon, P. (1992). The requirements for bution in an ultramafic partial melt under hydrostatic stress
chemical disequilibrium during magma migration. Earth conditions,.I. Geophys. Res. 84, 6109-6114.
Planet. Sci. Lett. 109,611-620. Waff, H. S., and Bulau, 1. R. (1982). Experimental determina-
Stolper, E.. and Newman, S. (1992). The role of water in the tion of near-equilibrium textures in partially molten sili-
petrogenesis of Mariana trough magmas, submitted for cates at high pressures, ill "High Pressure Research in
publication. Geophysics" (S. Akimoto and M. H. Manghnani, eds.),
Sudo, A., and Tatsumi, Y. (1990). Phlogopite and K- Adv. Earth Planet. Sci., Vol. 12, pp. 229-236, Center for
amphibole in the upper mantle: Implication for magma Academic Publications, Tokyo.
genesis in subduction zones, Geophys. Res. Lett. 17, Wallace, M. E., and Green, D. H. (1988). An experimental de-
29-32. termination of primary carbonatite composition. Nature
Sugimura, A. (1960). Zonal arrangement of some geophysical 335,343-345.
and petrological features in Japan and its environs, .J. Fac. Walther. 1. V" and Helgeson, H. C. (1977). Calculations of the
Sci. Univ, Tokyo Sec. 212(2), 133-153. thermodynamic properties of aqueous silica and the solu-
Tatsumi, Y. (1986). Formation of the volcanic front in subduc- bility of quartz and its polymorphs at high pressures and
tion zones. Geophys. Res. Lett. 13,717-720. temperatures, Am . .I. Sci. 277, 1315-135 I.
Tatsumi, Y. (1989). Migration of fluid phases and genesis of Watson, E. B. (l99\). Diffusion in fluid-bearing and slightly-
basalt magmas in subduction zones, .J. Geophys. Res. 94, melted rocks: Experimental and numerical approaches il-
4697-4707. lustrated by iron transport in dunite, Contrib. Mineral. Pet-
Tatsumi, Y., Sakuyama. M.. Fukuyama, H.. and Kushiro, I. 1'01.107,417-434.
(1983). Generation of arc basalt magmas and thermal Watson, E. B., Brenan, 1. M., and Baker, D. R. (1990). Distri-
structure of the mantle wedge in subduction zones, .I. Geo- bution of fluids in the continental mantle, in "Continental
phys. Res. 88, 5815-5825. Mantle" (M. A. Menzies. ed.), pp. II i -125. Clarendon
Thompson, A. B. (1992). Water in the Earth's upper mantle, Press, Oxford.
Nature 358, 295-302. Wolf. M. B.. and Wyllie. P. 1. (1991). Dehydration-melting of
Tichelaar, B. w., and Ruff, L. (1989). Variability in depth of solid amphibolite at 10 Kbar: Textural development, liquid
seismic coupling along the Chilean subduction zone, EOS interconnectivity and applications to the segregation of
Trans. Am. Geophys. Union 70, 398. magmas, Mineral. Petrol. 44, 151-179.
Toksoz, N., Minear, 1. w.,and Julian, B. R. (1971). Tempera- Wunder, B., and Shreyer, W. (1992). Metastability of the lOA
ture field and geophysical effects of a down-going slab, .I. phase in the system MgO-Si0 2-H,O (MSH): What about
Geophys. Res. 76, 1113-1138. hydrous MSH phases in subduction zones? .J. Petrol. 33,
Tritton, D. 1. (1977). "Physical Fluid Dynamics," Van Nos- 877-889.
trand Reinhold, New York. Wyllie, P. 1. (1979). Magmas and volatile components, Am.
Turcotte, D., and Schubert, G. (1968). A fluid theory for the Mineral. 64, 469-500.
deep structure of dip-slip fault zones, Phys. Earth Planet. Yuen, D. A., Fleitout, L., Schubert, G., and Froidevaux, C.
Jill. 1,381-386. (197l\). Shear deformation zones along major transform
Turcotte, D.. and Schubert, G. (1973). Frictional heating of the faults and subducting slabs, Geophys. .I. R. Astron. Soc.
descending lithosphere, .I. Geophvs. Res. 78, 5876-5886. 54,93-119.
Urnino, N., and Hasegawa, A. (1984). Three-dimensional Q.
This page intentionally left blank
 Chapter 10 Buoyancy-Driven Fracture and Magma Transport
through the Lithosphere: Models and Experiments

Moritz Heimpel and Peter Olson

Overview Notation
Units
Magma generated in partially molten regions in the up-
E Young's modulus Pa.MPa
per mantle may ascend to crustal magma chambers and
K, mode I (purely tensile) stress Pa. m'?
ultimately to the Earth's surface via fracture propaga-
intensity
tion alone or by fracture-assisted diapirs. We present the
results of experiments on buoyancy-driven fluid trans- Kif mode I critical stress intensity Pa·m '12
port by fracture propagation, using various fluids in- (fracture toughness)
jected into gelatin. The buoyancy and volume of fluids L lesser fracture toughness of two Pa·m '12
have been varied over three orders of magnitude, and crack tips
several concentrations of gelatin, which correspond to K+ greater fracture toughness of two Pa m-?v

a range of shear velocities and fracture toughnesses, crack tips

are used. The resulting crack propagation velocities P internal overpressure Pa.MPa
. vary over four orders of magnitude. Two regimes of overpressure near crack tip Pa, MPa
Po
propagation are identified: (a) slow propagation char-
Q, source flux m~. S-l
acterized by cracks with a subcritical stress intensity;
UE elastic wave velocity m· S·-I
and (h) fast propagation characterized by a supercriti-
cal stress intensity and a fracture profile that narrows UF fracture resistance velocity m· s -I
smoothly to a thin conduit at the tail. Crack propagation Up Poiseuille velocity m· s
velocities are found to depend on the fluid buoyancy, crack or dike half-thickness
a m
the yield strength and fracture toughness of the solid
al crack or dike half-thickness at m
medium, and the size of the fluid-filled fracture. Be-
s = I'
cause none of the existing models of fracture propaga-
tion predict the observed velocities, we derive a new G mu maximum crack or dike half- m
thickness
model, in which fracture at the tip and closure at the tail
are coupled by elastic wave propagation down the crack a, crack or dike conduit half- m
surface. The experimental data bound fracture veloci- thickness
ties derived from this model and the predicted velocities h crack half-length m
of buoyancy-driven magma fracture in the lithosphere c fracture resistance velocity dimensionless
are consistent with velocities inferred for magmatic coefficient
systems. In addition, crack-tip fracture and the conse- g gravity m. S-2

quent production of seismic radiation' from propagating

g' buoyancy m's-
dikes are consistent with our model. Also in this chap-
ter, the role of the fracture toughness on magma fracture I' crack-tip radius m
is briefly reviewed. Fracture toughness estimates ob- I' crack or dike head radius m
tained from the dimensions of igneous dikes are shown 2h crack or dike height m
to be two to three orders of magnitude higher than labo- s, .r, y, z Cartesian coordinates m
ratory fracture toughness estimates, indicating that the
lag time for crack closure
field-derived and laboratory-derived estimates are
probably appropriate for magma fracture under differ- u average fluid velocity m- S-I

ing conditions. sr, hydrostatic pressure drop Pa, MPa

Magmatic Systems Copyright © 1994 byAcademic Press. Inc.

Edited by M. P. Ryan 223 Allrights of reproduction in anyform reserved.
224 Moritz Heimpel and Peter Olson

Units the lithosphere, the issue of what controls the shape

and the velocity of buoyant propagating fractures
viscous pressure drop Pa, MPa
remains unresolved. To address this issue, we
crack geometric parameter dimensionless
have performed experiments designed to investi-
YJ fluid viscosity Pa· s
gate the mechanics of buoyancy-driven fluid frac-
J1. shear modulus Pa, MPa
ture and test existing models of this phenomenon.
IJ Poisson's ratio dimensionless While some elements of each of these existing
P solid density kg m
v 1
models are identified in the experiments, none of
PI fluid density kg' m :" the models explains the propagation velocities we
extensional stress at crack tip Pa, MPa observe.
CT, extensional yield strength of solid Pa, MPa
Magma Fracture Models

Buoyancy-driven magma fracture has been inves-

tigated from several points of view. First. there is
Introduction the "quasi-static fracture" model, which utilizes
the two-dimensional shape of a static volume of
Fracture propagation is a mechanism by which fluid embedded in an elastic solid (Weertman,
positively buoyant magmas, generated in mantle 1971; Pollard and Muller, 1976). According to
upwellings, are transported through the litho- .this model, when a crack is sufficiently large and
sphere. Magma injection (and dike formation) is the fluid sufficiently buoyant, the stress intensity
of primary importance in the building of intra- at the leading edge exceeds the fracture toughness
plate oceanic islands, in continental rifting, and in of the country rock, and the fracture propagates
the generation of the oceanic crust at mid-ocean upward. In order to maintain the shape predicted
ridges. While mechanisms of magma transport by elasticity theory, the fracture must close at the
such as diapirs of partial melt or magma (White- tail. This static model does not predict propaga-
head and Luther, 1975; Marsh, 1982) and solitary tion velocities.
wave propagation within partially molten regions A second model is based on the stress corrosion
(McKenzie, 1984; Scott and Stevenson, 1984; Ol- theory of materials science, as applied to magma
son and Christensen, 1986) are likely to be impor- fracture (Anderson and Grew. 1977). In this model,
tant in the viscous asthenosphere, a mechanism cracking is controlled by environmental factors
involving fracture propagation seems necessary to such as temperature, chemical reaction kinetics,
transport magma through the elastic lithosphere. and the fluid viscosity near the crack tip, and the
Most dikes that intersect the surface of the stress intensity is assumed to be less than the frac-
Earth have their intermediate origin in crustal ture toughness. The resulting crack propagation
magma chambers a few kilometers in depth. How- velocities are typically less than 0.1 mm . s -I. The
ever, the generation of melt takes place in the up- importance of this process in magma fracture is
per mantle below the base of the lithosphere at probably limited to special circumstances, such as
depths over the range of approximately 100 km in dike initiation and the slow cracking of high frac-
intraplate provinces to approximately 30 km at ture toughness barriers that separate melt volumes
mid-ocean ridges. The low yield strength of par- at depth.
tially molten rock suggests that magma fractures Investigators have also considered the role of
can readily initiate in partially molten regions fluid (magma) viscosity in limiting the propaga-
(Fowler, 1990; Sleep, 1988). Crosscutting dikes in tion velocity (Spence et al., 1987; Lister, 1990;
the mantle peridotites of ophiolites, which are in- Turcotte, 1990; Lister and Kerr, 1991). In this ap-
ferred to have formed while the peridotites were proach, fracture mechanics determines the crack-
partially molten, provide field evidence of magma tip shape, but the effect of the crack-tip fracture
fracture nucleation in partial melt regions (Nico- resistance on the crack propagation velocity is as-
las, 1986, 1990). sumed to be negligible. Models based on this con-
Although fracture propagation is the generally cept predict a magma fracture with a slightly bulg-
accepted mechanism for magma transport through ing head that tapers to a narrow conduit. With a
 10. Buoyancy-Driven Fracture and Magma Transport 225
constant source flux the propagation velocity in Table 1
these models is equivalent to the Poiseuille ve- Experimental Gelatin Properties
locity of the conduit fluid.
The flow of magma directly into dikes in par- Concen- Shear Shear Yield Fracture
tration modulus velocity strength toughness
tial melt regions is a complex process involving (rn- s -I)
(%) (Pa) (Pa) (Pa'm/2 )
concurrent porous flow, matrix compaction, and
the fracture of the rock matrix so that flux into the 1.4 190 0.44 980 15
dike is limited by rock permeability, rock vis- 1.6 276 0.53 1300 19
cosity, and the melt viscosity (Sleep, 1988). In 1.8 355 0.60 1650 23
such a process, two source conditions are plau- 4.0 2150 1.47 8lO0 114
sible idealizations: (a) the constant volume source
condition, where the magma source is depleted
rapidly compared to the time required for magma 1966). Table I shows the concentrations and cor-
transport through the newly formed magma frac- responding material properties of the gelatin
ture, and (b) the constant flux source condition. mixtures used. High-clarity, 250-bloom, pigskin-
appropriate if flow into the magma fracture pro- derived gelatin in granular form was supplied by
ceeds long after fracture nucleation. In this chap- Kind and Knox Company. Table 2 shows the
ter we discuss the shape and velocity of buoyant properties of the fluids injected into the solidified
fractures with constant fluid volume that close at gelatin mixtures.
the tail and of fractures with constant fluid flux in . The yield strengths of the gelatin gels were ob-
which the leading pulse of fluid tapers to a conduit tained directly by measuring the force required
of uniform thickness. pull an object of known surface area out of the
gelatin samples. Shear modulus data were obtained
by comparing the theoretical and experimental
Experiments on Buoyancy-Driven Fracture shapes of fluid-filled cracks, and gelatin fracture
toughness values were obtained from plots of
Among the previous experimental investigations stress intensity versus crack velocity.
of buoyancy-driven fracture propagation (Fiske The apparatus shown in Fig. I consists of a
and Jackson, 1972; Maaloe, 1987; Takada, 1990), transparent right circular cylinder, filled with gela-
only Takada presented propagation velocity data. tin, inside a transparent fluid-filled jacket. The
Our experimental results extend the original ex- jacket controls the gelatin temperature and also
perimental results of Takada. We also propose a corrects for optical distortion. The cylinder can be
new model for buoyancy-driven crack propaga- rotated to observe the fractures in any orientation.
tion, in which the velocity depends on the fracture For each experiment the gelatin was allowed to set
resistance of the solid. From dimensional analysis for 24 h at 8° C. Known volumes of several types
based upon the experimental results we derive a of positively buoyant fluid were injected into the
fracture resistance velocity U F and show that it gelatin at the bottom of the cylinder. The injected
collapses the velocity data for our experiments.
Also, by applying U F to rock properties, we show
that the fracture resistance is likely to be impor- Table 2
tant for buoyancy-driven magma fracture propa- Experimental Fluid Properties
gation in the lithosphere.
Buoyancy Shear viscosity
Working fluid (m -s 2) (Pa's)
Experimental Method and Results
Air +9.8 10- 5
Gelatin is a clear, brittle, viscoelastic solid with Hexane -+:3.4 10- 4
a low rigidity and a Poisson's ratio of nearly 0.5. 1 Cst Si oil + 1.8 10- 3
It has been useful as a modeling material in Mineral oil 0.1
+ 1.5
geotechnical engineering applications because its
Corn syrup solution -4.1 5.0
low rigidity allows gravity to be significant in
Mercury (Hg) -126.0 10-'
laboratory-scale models (Richards and Mark,
226 Moritz Heimpel and Peter Olson

10- 1

"7 10-2
en
1.4,
1:I air
g o 1.6, air
+ 1.6, hexane
::> 10-3 x 1.6, min. oil
c
·u • 1.8, air
0 " 4.0, air
~ 10-4
e 1.6, Hg

10-1 100 101

Volume (ml)
Figure 2 Summary plot of crack propagation velocity versus
volume of injected fluid for various gelatin concentrations and
fluids.
1'--20 cm-..\

Figure 1 Schematic sketch of the experimental apparatus.

. The raw data from the constant volume experi-
Cylinder of height 50 em and diameter 20 cm contains gelatin ments consist of the observed crack propagation
and may be rotated within the cooling tank. Constant volumes velocities versus injected fluid volume. These data
of positively buoyant fluids are injected at the bottom. Con- are shown in Fig. 2.
stant flux of negatively buoyant fluid is injected at the top of
In those experiments using low-viscosity flu-
the cylinder. Cross-sectional (left) and in-plane (right) shad-
ows of a typical constant volume crack are shown projected
ids, we find that the crack propagation velocity
onto the tank walls by parallel light. is evidently limited by the fracture resistance of
the solid. Accordingly, an appropriate measure of
the loading parameter for crack propagation is the
fluid temperature matched the gelatin tempera- purely tensile, or mode l , stress intensity factor
ture. In some experiments, negatively buoyant K,. A circular or "penny-shaped" crack of ra-
fluid was injected at constant volume and at con- dius I' is a good approximation for the shape of the
stant flux at the cylinder top. constant volume fractures near the crack tip. The
The positively buoyant fluids ascended (nega- crack tail tends to extend so that the crack height
tively buoyant fluids descended) through the solid 2h from tail to tip is somewhat greater than 21' (see
gelatin by fracture. Most of the fractures propa- Figs. 7a and 8). Hence the appropriate formula-
gated with smooth fracture surfaces. Many frac- tion for K 1 is
tures ascended with step-like motion of the tail,
which resulted in roughly evenly spaced notches
on an otherwise smooth fracture surface. The min-
eral oil-filled fractures tended to display rough where Po = pg'2h is the overpressure near the
fracture surfaces, possibly indicating chemical re- leading edge of the crack, 2h is the crack height,
activity between the mineral. oil and the gelatin ris the crack head radius, g' = g(p - p,)/pisthe
gel. Fluid viscosities were sufficiently low so that fluid buoyancy, g is the gravitational acceleration,
the volume of fluid lost to the trailing conduit p is the solid density, and p, is the fluid density.
from the propagating crack through the tail was Figure 3 shows the observed crack propagation
small compared to the injected volume. We ob- velocity U versus K, forthe constant volume ex-
tained images of the ascending cracks by project- periments. We identify two regimes of propaga-
ing parallel light through the tank onto a gridded tion. In regime I, U is small «0.7 em . s -1) and
screen and photographing the shadowgraph im- the crack velocity is highly sensitive to the load-
age. Crack dimensions and velocities were mea- ing (U ClC Ks, where 5 < 11 < 8). This type of
sured from the resulting photographic images. dependence of crack propagation velocity on the
 10. Buoyancy-Driven Fracture and Magma Transport 227
10-1 the three regions of subcritical crack growth de-
c 1.4, air
0 1.6, air
+ 1.6, hexane
~ scribed by Atkinson and Meredith (1987; see
Fig. 3). In our experiments, however, the transi-
x 1.6. min. oil
10-2 tion from 5 < n < 8 at low propagation velocity
•..
1.8, air
4.0, air (regime 1) to n = 2 at high propagation velocity
0 1.6, Hg
"7
C/O
(regime 2) is not followed by a third regime analo-
10-3 gous to region 3 for subcritica1crack growth where
s
~
n increases again (Aktinson and Meredith, 1987).
In fact Fig. 3 shows that regime 2 in our experi-
10-4 ments extends to propagation velocities approach-
ing the shear velocity of the medium (U/U s =
0.2). We therefore arrive at the hypothesis that
the transition from regime 1 to regime 2 marks the
1 2 onset of dynamic fracture propagation (see the
10 10
next two sections for dimensional and physical
K, (Pa. m li 2)
justification of this assertion).
Figure 3 Plot of the stress intensity factor K, versus the ex- The critical stress intensity factor K,c associ-
perimentally observed crack propagation velocity U for the ated with the transition from subcritical to dy-
various gelatin concentrations and fluids used in the constant namic propagation can be obtained graphically
volume experiments. Two regimes of crack propagation are
from the U versus K I data. Determination of the
distinguishable. In regime I, cracks propagate slowly and
U C( K;', where 5 < n < 8. In regime 2, fast propagation oc- K Ic values of the gelatin preparations used in our
curs and U C( Ki, experiments is more clearly obtained from Fig. 4,
which shows the observed crack propagation ver-
sus K I on linear axes. We define KIcfor our gelatin
loading is analogous to that exhibited by other gel experiments as the value K, (Eq. (1)) where
brittle solids such as ceramics and rocks during the transition from slow propagation in regime I
subcritical crack growth (Evans and Langdon, to fast propagation in regime 2 is observed. We
1976; Atkinson and Meredith, 1987). In regime 2, have graphically determined K Ic from Fig. 4 at the
U is large (>0.7 em . S·I) and U ex K? This re- intersection of the best-fit line of velocity vs stress
gime of crack propagation resembles region 2 of intensity factor with the horizontal axis.

10 20 30 10 20 30 40 50
0.025 D 1.4%, air
D 0 1.6%, air 0.10
+ 1.6%, hexane
0.020 K Ie =15Pa.m'l2 o 1.6%, Hg 0.08
= 19 Pa. m l12

I
0.Q15 KIe 0.06
0.010 0.04
"7
en 0.005 0.02

E- 0.000 0.00
0 0.03
TJ 0.Q15
.£
~ 0.02
0.010

0.01
0.005

0.00
0.000
20 30 40 50 40 80 120 160 200
Stress intensity (Pa . m 112)
Figure 4 Crack propagation velocity versus stress intensity factor for four gelatin concentrations. For each concentration, the
critical stress intensity K" is determined by extrapolating the curve to zero crack velocity.
228 Moritz Heimpel and Peter Olson

Dimensional Analysis of Results

A major goal of our experiments is to characterize

the functional dependence of the buoyancy-driven
crack propagation velocity on the physical parame-
ters involved. In the previous section, we argue
the crack propagation velocity V F is a function of
¥ D 1.4, air
the applied loading K" which depends on the fluid I o 1.6, air
I
buoyancy. V F may also depend on the viscosity 71 I + 1.6, hexane
in the crack cavity, the rigidity (characterized by I x 1.6, min. oil
/ • 1.8, air
the shear velocity V E), the local stress a y required .J. A 4.0, air
to fracture the solid medium near the crack tip, , / o 1.6, Hg
and length 2h and thickness 2a of the propagating
fracture. Hence,
10-3 10-2 10-1
V = F(K" 71, V E , up 2h, 2a). (2)
Calculated U (m-s')
F
Dimensional analysis then yields the following set
of four dimensionless groups, Figure 5 Experimental crack propagation velocity versus
calculated fracture resistance velocity given by Eq. (8) for
(TIl> II 2 , II), II 4 ) c = I. Most of the data collapse onto a line of varying slope,
(3) 'indicating that U o: U~. The data for constant volumes of min-
= (!!...- :!. ~
V E ' h' «, V2h'
71
V
E)
u y2h '
erai oil plot away from the line. The mineral oil fractures
tended to have anomalously rough fracture surfaces. pos-
sibly indicating chemical reaction between the oil and the
such that gelatin gel.
(4)
Letting Figure 5 plots V F given by Eq. (8) versus the ob-
served crack propagation velocity V and shows
II 3
III ex:--
m (5) that the most of the data collapse onto a line of
II 4 varying slope indicating that V = f( V F)'
and combining Eqs. (I), (3), and (5) yield the The velocity proportionality given by Eq. (8)
Poiseuille velocity as illustrated in Fig. 5 relates the crack propaga-
tion velocity to crack parameters for both the slow
pg'a? and fast regimes of propagation in our experi-
Vpex:--. (6)
71 ments. However, Eq. (8) does not include the criti-
This is the appropriate crack propagation velocity cal stress intensity factor K lc associated with tran-
when the resistance to crack propagation is con- sition between these regimes. Since we interpret
trolled by the fluid viscosity. In our experiments, K lc for our experiments as the fracture toughness
however, we find that Eq. (6) does not fit the data. of gelatin, the fracture resistance velocity V F for
If the crack propagation velocity is limited by the crack propagation in regime 2 should include KIc
fracture resistance then a velocity appropriate to so that V F approaches zero as K, approaches K lc
our experiments should be a function of the crack from above. To accomplish this, we introduce a
loading and the properties of the solid. Figure 3 factor into Eq. (8) and obtain a corrected expres-
implies that in the fast propagation regime the sion for the fracture resistance velocity V F in the
crack propagation velocity V ex K¥. Hence we fast propagation regime of our experiments,
choose
VF = cVEKr! - I,
- - ( 1 -Kr) (9)
III ex: m, (7) 2h u~ " Kt;
which yields a fracture resistance velocity where c is a dimensionless constant of proportion-
ality and K, is given by Eq. (I).
V ex: V
F
---.!SL ex: V (pg':h)2
E
(8) Figure 6 shows a comparison of all of the ex-
E u~2h uy perimental propagation velocity data and the frac-
 10. Buoyancy-Driven Fracture and Magma Transport 229
It is usually assumed in dynamic fracture me-
c 1.4%. air
o 1.6%. air chanics that the crack propagation velocity very
+ 1.6%. hexane rapidl y approaches the elastic wave velocity of the
)( 1.6%. min. oil
• 1.8%. air solid once K, exceeds K Ic • During dynamic propa-
1& 4.0%, air
gation, the dynamic stress intensity factor Kid is a
o 1.6%. Hg 00
1ib function of the loading and the crack extension
1lP
c .. velocity Kid = f(? r, dr/dt) (Freund, 1990). Mo-
c 0
"'l rita (1993) has pointed out that for the dynamic,
C
• viscoelastic fracture problem the stress intensity
"'0 A factor approximated by an elastostatic fracture
x
• equation (termed a "pseudo-stress intensity fac-
tor") varies by several-fold exceeding the fracture
toughness depending upon the loading rate.
10-3 10-2 10-1 It is important to note that the crack propaga-
Calculated U (rn-s') tion velocity in the formulation of the dynamic
F
stress intensity factor is generally equal to the rate
Figure 6 Experimental crack propagation velocity versus of crack growth dr/dt. This means that the loading
the calculated fracture resistance velocity given by Eq. (9) for geometry associated with KId varies as the crack
c = I. Only fractures with supercritical stress intensity are
propagates in the case where the load is applied
shown since subcritical fractures are undefined in Eq. (9).
externaIly to the solid medium (e.g., in a double
torsion experiment tractions are applied to the
ture resistance velocity given by Eq. (9). The crack faces at one end of the specimen). In con-
value of c is shown to be of order one, indicating trast, for constant volume buoyancy-driven crack
good agreement between the observed crack prop- propagation the load on the crack tip is due to
agation velocity and UF' Fractures with stress in- the body force pg'V, which moves along with the
tensity factor K , < K lc , which we interpret as propagating crack so that drldt = 0 even when
propagating in the subcritical regime, are not de- the propagation velocity U "* O. Therefore, for
scribed by Eq. (9) and hence do not appear on any volume V of fluid in the crack, we expect (and
the plot. we observe experimentaIly) that the loading con-
figuration reaches a time-averaged steady state
A New Model for the Crack Propagation Velocity in which the crack-tip propagation velocity is
matched by the crack-tail closure velocity.
In the previous sections we obtained a crack ve- For the case where the fluid volume in the
locity that fits the experimental data using dimen- buoyancy-driven fracture is sufficient for K, to ex-
sional analysis. We can now ask, What physical ceed Kin unstable dynamic propagation of the
process results in this crack velocity? The empiri- crack tip results, but is constrained by the condi-
cally derived crack velocity U F is a function of the tion that the loading configuration must maintain
loading term K I , crack dimensions rand h, and a nearly steady state.
the material properties of the solid U E, O"y, and This process is best illustrated by considering
K Ic • Perhaps surprisingly, Eq. (9) implies that the an incremental crack propagation model. Given a
crack loading given by K, routinely exceeds the crack initiaIly at rest in a brittle solid that is then
critical stress intensity K Ic required for dynamic fiIled with fluid so that K, slightly exceeds K Ico
crack propagation. Furthermore U F implies K ta > how will propagation proceed? Since K , > K Ic at
K Ic for cracks propagating at velocities that can the crack tip, unstable dynamic propagation (ac-
be several orders magnitude less than the elastic companied by elastic wave radiation) should oc-
wave velocity of the solid medium. cur so that the crack tip propagates upward at a
Here we present an argument that the loading velocity U - U E, where U E is the elastic wave
configuration for buoyancy-driven fracture leads velocity of the solid. However, the entire length of
to a dynamical constraint on the crack propaga- the crack cannot instantaneously respond to mo-
tion velocity. The resulting formula for the crack tion of the crack tip. As a result, the stress inten-
propagation velocity U F is then identical to Eq. (9). sity K, at the crack tip drops below K Ic after some
230 Moritz Heimpel and Peter Olson

distance d of crack-tip propagation, until the load- cracks (e.g., magma fractures) buoyancy-driven
ing configuration readjusts. Since the crack is ini- fractures may propagate at velocities V F < < V E
tially at rest, crack-tail closure can proceed only even when the loading gives a stress intensity
after information from the crack-tip propagation K1>K Ic •
event, in the form of elastic waves, arrives at the
crack tail. This requirement is equivalent to the Shape of Fractures with Constant Fluid Volume
constraint that the crack maintain a time-averaged
constant shape and that the crack tail close upward Our experiments indicate that the shape of the
at a rate equivalent to the crack-tip propagation propagating constant fluid volume fractures ap-
velocity. Hence the loading configuration for con- proximates the fluid-filled crack shape predicted
stant fluid volume buoyancy-driven cracks leads by elastostatic theory (see Fig. 8). This result veri-
to the requirement that propagation events at the fies the assertion that the fluid pressure gradient is
crack tip and closure at the tail have an associated nearly hydrostatic and is consistent with the find-
lag time t = 2h/V E , which is the time required for ings of Takada (1990). In terms of the fracture
information to travel from the crack tip to the propagation model just presented, it implies that
crack tail. (for V F « VE ) the crack adjusts dynamically to
For this incremental model the average veloc- maintain an equilibrium shape that approximates
ity of crack propagation is then the elastostatic shape.
Figure 7a shows the geometry considered in the
d dVE
VF - - - --. (10)
t 211
Here we made the approximation that the lag a
time t» diVE, where dltl; is the crack-tip dy-
namic propagation time for one cycle of incre-
mental crack propagation. The incremental crack-
tip propagation length d is the distance ahead of
s
i
Zr
the crack tip where the stress exceeds the local ef-
fective yield strength of the solid. This is similar
to the fracture mechanics definition of the pro-
cess zone radius (e.g., Pollard, 1987) or the Baren-
1
blatt cohesive zone size (Barenblatt, 1962; Rubin,
x
1993), d ex KU(T~. However, a definition for the
b -----------t
process zone size relevant to this incremental
propagation model must include the result that dy- r
namic propagation occurs only when K, > Ki;
Hence the modified definition for the process zone
+ J
size becomes

where the local effective yield stress of the solid

(T y is the sum of the cohesive strength of the solid
Za
and the regional stress acting on the crack-tip pro- .-00
cess zone (Rubin, 1993). Combining Eqs. (10)
and (11) we recover our empirically derived frac-
ture resistance velocity

V F = cV-
E
K?I ( 1 - K?) I
zo: -2 • (12) Figure 7 Shape of propagating buoyancy-driven fluid-filled
fractures. Idealized in-plane crack contour used (left) in the
211 K Ic
calculation of the cross-sectional shape (right) resulting from
Equation (12) implies that for materials with very (a) a buoyant fluid with negligible viscosity (Eq, (14)) and
low shear velocity (e.g., gelatin) or for very long (b) a buoyant fluid with significant viscosity (Eqs. (l8)-(20c)).
 10. Buoyancy-Driven Fracture and Magma Transport 231
following analysis. According to Sneddon (1946), ture in y and z is always positive and goes to zero
the thickness 2a of a static pressurized circular only at y = 0, z = - r.
crack with radius r in an elastic medium is Figure 8 compares the experimental shape of
propagating fluid-filled fractures and the theoreti-
cal shape given by Eq. (14). The fact that the ob-
(13)
served cross-sectional shape closely matches the
shape for a static crack shows that dynamic vis-
cous forces do not significantly alter the hydro-
where P is the fluid overpressure, v is Poisson's
static pressure in the fracture. This is another in-
ratio, j1, is the shear modulus of the solid, r is
dication that fluid viscosity plays a negligible role
the crack radius, and y, z are Cartesian coordi-
in controlling the overall propagation velocity of
nates. We compare the cross-sectional experimen- these fractures.
tal shape of the propagating cracks to the theoreti-
cal shape of an internally pressurized fracture
with a linear pressure gradient. Since the fracture Shape of Fractures with Constant Fluid Flux
closes at the tail as it propagates upward, the con-
dition of zero stress intensity at the tail (da/dz = Constant flux experiments were performed using
Oat z = r) is thus appropriate. This requirement negatively buoyant corn syrup, Cadmium chloride
results in the overpressure being zero at the tail. and water solution (see Table 2 and Fig. 9). The
Hence the thickness along y = 0 of a static, cir- experiments show that, after broadening rapidly
cular crack filled with buoyant fluid is near the source, a crack propagating away from a
constant source of buoyant fluid consists of a nar-
row conduit that feeds a crack head of slowly
changing size, shape, and velocity. The condition
of constant crack head size requires that the crack
head velocity is matched by the average fluid ve-
where g' = g(p - PI)/P is the fluid buoyancy, g locity in the conduit. In this section we add the
is the gravitational acceleration, P is the solid den- effects of the fluid viscosity to the pressure gradi-
sity, and PI is the fluid density. The maximum ent in the tail of the propagating fracture. Fig-
width occurs at z = r/2 and is given by ure 7b shows the (idealized) geometry considered.
The analysis is similar to that given by Lister
(1991) for a density stratified lithosphere. The
(15)
crack conduit is of uniform breadth (2a x ) and is
smoothly attached to the circular leading edge of
Except for a factor of 2/7T, the result given by the crack. It is convenient to place the origin of
Eqs. (14) and (15) is identical to the shape for coordinates (z = 0) at the tip of the crack. We as-
a two-dimensional crack in the presence of a lin- sume that the horizontal pressure gradient is neg-
ear pressure gradient and with zero stress inten- ligible so that horizontal cross sections of the frac-
sity at the tail (Weertman, 1971; Pollard and Mul- ture are ellipses. The gradient in overpressure for
ler, 1976). the laminar flow of buoyant fluid in a planar crack
It is interesting to note that contrary to the two- arises from the sum of the buoyant force of the
dimensional case, the fluid overpressure near the fluid and the viscous stresses transmitted into the
tail of a static circular crack is never negative. fluid from the sides of the crack:
This is a consequence of the finite breadth of the
circular crack. The curvature of the crack face 3T/Qs
dP = -pg'
- +- - (16
a)
indicates the relative pressure. A convex out- ds 2a
3r'

ward shape (positive curvature) indicates a posi-

tive fluid overpressure and a concave (negative where
curvature) crack face indicates a fluid pressure Qs = U7Tar (16b)
less than the ambient pressure in the solid me-
dium. Even though the curvature in z given by and U is the average fluid velocity and T/ is the
Eq. (14) is negative near the tail, that total curva- dynamic shear viscosity of the fluid in the crack.
232 Moritz Heimpel and Peter Olson

Figure 8 Two comparisons of the experimental versus theoretical elastostatic crack shape. showing the in-plane shadow pho-
tograph in the first column. the cross-sectional photograph in the second column. and a plot of the theoretical shape given by
Eq. (14) in the third column. The upper row shows 0.8 ml air in 1.4% gelatin. The lower row shows 10 ml air in 4.0% gelatin.

Since horizontal pressure gradients are assumed near the crack head (where the crack contour is
to be negligible, the elastic normal stress acting on circular) to e= I for the section of the crack far
the fluid is downstream of the crack tip (Rubin and Pollard,
1987). Since we are interested in the details of
P(s)
e jW(S)
(17)
the shape near the head of the crack we will take
(1 - v)r(s)' o = 7T/2. The average vertical fluid velocity. is
assumed to be constant so that the pressure gradi-
e
where is a dimensionless parameter that ranges ent in the crack head is approximately hydrostatic
e
in value from = 7T/2 for the section of the crack when the maximum thickness of the crack head

Figure 9 Comparison of the experimental versus theoretical crack shape with constant source flux. showing the in-plane
shadow photograph in the first column. the cross-sectional photograph in the second column. and a plot of the theoretical shape
given by Eqs. (I8)-(20c) in the third column. all from the same experiment using a corn syrup. cadmium chloride. and water
solution of specific gravity IAI in 1.80/c concentration gelatin.
 234 Moritz Heimpel and Peter Olson

2a max is a few times greater than the crack conduit Application to Magma Transport
thickness 2a oo • Reasoning thus, we approximate
the thickness 2a(s) in the crack head by Eq. (14), The relevance of fluid-filled fracture models to
rewritten in terms of s, magma fracture depends strongly on both the
fracture properties of the country rock and the
2a(s) == ;pg'C : v) buoyancy and viscosity of the ascending magma.
(18) Buoyancy-driven magma fracture is important at
(2r - s)(2rs - S2)1/2
depths in the lithosphere ranging from the par-
for 0< s < rand aoo/a m • x < 1/3. For a crack head tially molten source region where dike nucleation
of slowly changing volume, the breadth 2r of the most likely occurs to the level of neutral buoyancy
crack is nearly constant for s > r. Differentiat- (LNB) where magma accumulates to form crustal
ing Eq. (17) with respect to s for constant rand magma chambers (Ryan, 1987). We are therefore
combining the result with Eqs. (16) we obtain for particularly interested in the country rock and the
s> r, magma properties at depths ranging from approxi-
mately 100 to 2 km, and thus in confining pres-
da = 2r(~)(317Qs _ pg')' (19) sures from 3500 to 60 MPa.
ds 7T JL 2a 3r
Note that da/ds = a when 317QJ2a3r = pg', That Fracture Properties of Country Rock
is, for constant flux conditions the width of the
conduit is constant as s tends to 00 where the fluid The theoretical yield strength of a solid with no
buoyancy is balanced by the viscous stress on the flaws is based on the bond strength between atoms
crack walls. Equation (19) is easily integrated to and is approximately E/IO, where E is the Young's
yield an analytical expression relating s and a. modulus (Lawn and Wilshaw, 1975). This gives
The constant of integration is evaluated by impos- an upper limit of (J"y = 5000 MPa for the exten-
ing the condition of continuous width of the crack sional yield strength in the lithosphere at sub-
at s = r. The result for s > r is crustal depths.
The experimental value of (J"y under high con-
fining pressures is difficult to test directly and es-
timates are thus generally obtained from triaxial
tests. A yield criterion that relates the shear failure
of the triaxial test to the extensional yield strength
is then used (Paterson, 1978). In the following
where sections we use values of (J"y = 100-1000 MPa as
representative of igneous rocks under confining
317Q ,)1/3 J37T17 U (20b) pressures of 60 to 1000 MPa. For active magmatic
( 2pg'r 2pg' systems, where magma rises frequently (with re-
spect to the magma-channel solidification time)
and
along preexisting fracture planes, the extensional

al = :;2(1-JL--
- v) pg'r? (20e)
strength will be negligible and will probably not
exceed 10 MPa (Ryan, 1988).
The estimation of the fracture toughness of
is the half-width of the crack at s = r determined rocks appropriate to magma fracture in the litho-
from Eq. (18). sphere has been a subject of debate in the rock
Figure 9 shows an example of the constant mechanics and magma transport literature. The
source flux experiments at three different times theoretical value of the fracture toughness is ob-
in the development of the buoyancy-driven fluid- tained by considering a crack in an elastic solid as
filled fracture and compares the experimental a pressurized slit of half-length b. The mode I
cross-sectional shape of the propagating fluid- stress intensity factor is given by
filled fractures near the crack head and the
theoretical shape given by Eqs. (18)-(20c). (21)
 10. Buoyancy-Driven Fracture and Magma Transport 235
where P is the uniform remote tensional load or region or the "process zone" ahead of the crack tip
the uniform internal overpressure on the crack. between laboratory cracks and dikes is consistent
Lawn and Wilshaw (1975) give an estimate of the with the difference in scale between the microcrack
stress concentration factor at the tips of an ellipti- zone and the jointing zone, respectively. Compari-
cal "crack" (a« b) of thickness a and crack tip son of laboratory fracture toughness test results
radius of curvature r: and field studies of dikes indiates that the magni-
tude of microcracking and jointing is dependent on
(22) the size of the crack. High-temperature (diffusion-
driven or dislocation-driven) crack-tip plasticity
where O"b is the extensional stress at the crack is likely to be another important scale-dependent
tip. The fracture toughness K rc is defined as the mechanism in mantle magma fracture.
critical value of the stress intensity required for As a result of one or both of the above pro-
crack propagation. Letting O"b = O"y = EIIO when cesses, the fracture toughness of country rock ob-
K 1 = Krcand combining Eqs. (21) and (22) result tained from the dimensions of dikes is generally
in the theoretical estimate of the fracture tough- two or three orders of magnitude greater than those
ness for an ideal solid with an atomically sharp obtained from laboratory experiments. This sug-
crack: gests that, in addition to being a material property,
the fracture toughness is a parameter that depends
(23) upon the length scale of the crack and the crack-
tip process zone size. Thus, fracture toughness val-
For r = 10- 9 m and E = 1011 Pa, the resulting ues obtained from laboratory tests on centimeter-
estimate for the ideal (mode I) fracture toughness scale samples are probably not appropriate as a
is K rc = 105 Pa ml/2. propagation criterion for kilometer-scale magma
As defined above, the fracture toughness is a fractures.
material property. However, microcrack forma- Because of the great lateral extent of dikes
tion in the region of extensional stress ahead of (10 km is typical) few can be traced along their
the crack tip may complicate the crack-tip en- entire length and it is not surprising that little data
ergy balances. Off-axis microcracking increases exist on the dimensions of dikes in the field.
the fracture energy and hence the fracture tough- Table 3 is a compilation of data collected by vari-
ness and has been shown to accompany the propa- ous authors of the dimensions of dikes. The value
gation of laboratory scale cracks (Kobayashi and of the apparent fracture toughness is estimated as
Fourney, 1978). In a field study of mafic dikes, De-
f-L a
laney et al. (1986) described the similarity be- Ki; = (1 _ v) yr' (24)
tween microcracking in laboratory experiments
and dike-parallel joints in the field. They showed where f-LI(1 - v) is the rock rigidity, a is the av-
that the difference in magnitude of the high stress erage dike half-thickness, and r is the dike half-

Table 3
Country Rock Fracture Toughness Estimates from Dike Dimensions

fL/(I- v) 2a 2r K"
Host rock type Magma type (GPa) (m) (m) (MPavrn) Reference

Granite Rhyolite 22 7 11,000 1,000 Reches and Fink (1988)

Igneous Basalt/andesite 4 25 10,000 700 Macdonald et al. (1988)
Shale Minette I 2 346 -100 Delaney and Pollard (1981)
Sedimentary Lamprophere -2 12 11,550 1,300 Pollard and Muller (1976)
Basalt Tholeiite 10 12 4,000 1,300 Gudmundsson (1983)
Pyroclastic flow Basalt/andesite 5 I 480 200 Fedatov et al. (1978)
236 Moritz Heimpel and Peter Olson

breadth. The form of the dike or fissure is as- the crack propagation velocity. However, even
sumed to be elliptical. The resulting laboratory when the fluid viscosity is sufficient to allow a
and field estimates of K Ic lie in the range I < fracture conduit to form, UF < Up for a fracture
K Ic < 1000 MPa . m 1/2. We consider 100 < «; < with head width significantly greater than conduit
1000 MPa . m 1/2 an appropriate estimate for magma width. In this case the fracture resistance must still
fracture in the lithosphere. control the crack propagation velocity. Hence the
results of our experiments apply to the elastically
Melt Properties deforming head of a magma-filled fracture that is
propagating in the lithosphere when UF < Up in
The driving force for magma fracture is controlled the dike head even though Up defines the fluid ve-
by the gradient in overpressure of the magma. locity in the magma fracture conduit. Since UF is
Flow is driven by the density difference between proportional to the process zone size and inversely
the country rock and magma in the presence of proportional to the elastic wave travel time down
gravity (buoyancy forces) and is resisted by frac- the length of the crack, a process zone size of I m
ture near the crack tip and by viscous shear trans- with a magma-filled fracture head length of sev-
mitted into the flowing magma from the crack eral kilometers implies a propagation velocity of
faces. about l m- s - I. This velocity is of the same order
The density of basaltic magma (Kilauea 1921 of magnitude as that inferred from geophysical
olivine tholeiite) at confining pressures corre- observations of magma fracture (Spera, 1980; Aki
sponding to depths as great as 50 km in the litho- et al.. 1977). It is also similar to estimates of the
sphere has been studied by Fujii and Kushiro Poiseuille velocity for the same fracture size (Lis-
(1977). They found an approximately linear in- ter and Kerr, 1991). Figure 10 compares the frac-
crease in the melt density as a function of pressure ture resistance velocity and the Poiseuille veloc-
from 2600 kg . m -3 at atmospheric pressure to ity for a range of rock fracture toughnesses and
2900 kg . m -3 at 1.5 GPa. The density of litho- magma viscosities. For high-viscosity magma or
spheric rocks from the Moho to 100 km depth var-
ies from 2900 to 3300 kg . m -3. Hence a typical
value for the magma buoyancy g' = g(p - PI)/P Dike head width (m)

is l m s ",
v
10.2 10-1 10°
Magma viscosities for a variety of melts have
been compiled by Ryan and Blevins (1987). The 101
viscosity is sensitive to the absolute temperature. "7
~

Since the basalt liquidus increases with pressure 5 10°

from about I230°C at the surface to about 1500°c ~

at 100 km depth, the melt viscosity ranges from

'g 10-1
'ii
>
about 100 Pa . s at the surface to I Pa . s at the c
0
base of the lithosphere, respectively, assuming '::1
os 10-2
OIl
os
that the magma temperature is near the liquidus c.. Fracture
temperature during ascent. £ 10-3 resistance
velocity

10-4 ill IV V
Implications for Magma Ascent Velocities
4
Our experiments indicate a new regime for propa- 103 10
gating fluid-filled cracks, characterized by a frac- Dike head length (m)
ture resistance velocity UF. Does U F apply to Figure 10 Comparison of the fracture resistance velocity UF
magma fracture propagation velocities in the litho- to the Poiseuille velocity Up applied to magma fracture in
sphere as well? To answer this question we must the lithosphere. Numerals I through V refer to the following:
first assess the relevance of our experiments to (I and II) Up with viscosities 100 and I Pa . s, respectively.
(III-V) UF with fracture toughnesses 10', 10 8 • and 109 Pa .
magma fracture propagation in the lithosphere. In
m ' 12, respectively. In all cases /L!(l - v) = 5 X 10'0 Pa, u y =
our constant flux experiments the low fluid vis- 10" Pa, g' = 1.0 m· s -', and U. = 4.0 krn- s - I (the value for
cosities ensured that the fracture resistance of the U. is chosen as a typical value of the shear velocity in the
solid, rather than the fluid viscosity, determined elastic lithosphere),
 10. Buoyancy-Driven Fracture and Magma Transport 237
for weak, near-surface crustal rocks, U F can be shape slowly, the average fluid velocity in the
greater than the Poiseuille velocity in the head of conduit approximates the fracture propagation ve-
the forming dike. Hence, very near the surface, locity. The average fluid velocity is related to the
the magma viscosity probably limits the crack source flux and the conduit dimensions by
propagation velocity. On the other hand, for low-
Qs
viscosity basaltic magmas and for elastic and frac- U = --. (25)
ture properties representative of the lithosphere, tta.r
the Poiseuille velocity Up is typically greater than Dividing Eq. (20b) by Eq. (15) yields an expres-
U F' This suggests that the fracture resistance ve- sion for aJa.;
locity U F limits the magma fracture velocity at
depth in the lithosphere.
How is the fracture resistance model of
buoyancy-driven magma fracture propagation dis-
An approximate expression for the newly formed
tinguishable from models that assume that the
dike breadth 2,. is obtained by combining Eqs. (12)
fracture resistance is negligible once the fracture
and (25) and rearranging,
toughness has been exceeded? Models in which R
only the Poiseuille velocity limits the propaga- 2,. == [008( Q;(T~ )3/
tion speed imply that all of the gravitational po- . 7JU~(pg')5
tential energy lost during magma ascent goes into 1/3
1iKrc
+ _---'0.:...-
frictional (viscous) heating of the magma. The • (27)
]
16(pg'F
energetics of these models do not allow for seis-
mic wave radiation accompanying crack propa-
Figures 11 and 12 are plots of Eqs. (27) and (26),
gation, except by postulating the existence of a
respectively. Both a Ja ; and 21' are plotted versus
vapor phase present in the dike tip (Anderson,
the source flux Qs for three viscosities represen-
1978; Sammis and Julian, 1987). In the fracture
tative of basaltic magma and for two values of the
resistance model for magma fracture propaga-
fracture toughness. Figure 12 shows that for a
tion presented here, the shape of the dike head and
the propagation velocity are coupled by seismic
waves. Hence seismicity is an integral element of ::.1000
our model.
7000
The average fluid velocity for a fracture with a
constant shape and constant source flux is uniform 4000
over the entire fracture length and equal to the
2000
propagation velocity. In the magma fracture con-
duit, the fluid buoyancy is balanced by the viscous 1000
stresses (see Eqs. (19)-(20c). However, the vis-
cous stress contribution to the total pressure gra- ...-. 11000
dient !::.P v is strongly dependent on the local crack E
......., 7000
width (!::.P v :x 1Ia 3) whereas the hydrostatic com- ..c
ponent of the pressure gradient !::.P h is constant
.:0
C':l
4000
Q)
(!::.P h = pg' :x lla x \ see Eq. (16a)). This means ....
..0 2000
Q)
that where the magma fracture head thickness 2a .;.:
is a few times greater than the conduit thickness CS 1000
2a x , the pressure gradient in the crack head is
nearly hydrostatic (a.]« = 0.3 gives !::.Pj!::.P h = 0 1 2 3 4
10 10 10 10 10
0.03). We can use this to check the applicability
Source flux (m3 . S·I)
of our model to magma fracture propagation (and
dike formation) in the lithosphere by determining Figure 11 The magma fracture head radius is plotted as a
function the source flux Q, for I < Q, < 10" m ' . s -I, for 1/ =
the conditions under which the ratio of the conduit
1,10, and 100 Pa· s, and for K, = 107 and 109 Pa- m '/2 (see
thickness to the maximum dike head thickness Eq, (27)). All other material properties are constants chosen to
axla m is less than 0.3. When the head of an be representative of the lithosphere; f./.!O - II) = 5 X 10 '0 Pa,
ascending buoyant fracture changes volume and CTy = 109 Pa,p = 3300 kg' m " ', g' = Lrn- S-2.
238 Moritz Heimpel and Peter Olson

magma density is known as the level of neutral

buoyancy (LNB; Ryan, 1987). As a parcel of
magma approaches the LNB from below, the driv-
ing force for vertical propagation becomes de-
pleted and further fracture propagation proceeds
in blade-like magma fractures that spread laterally
away from the magma source and may result in vol-
canic rift zones (Rubin and Pollard, 1987; Ryan,
1987; Lister and Kerr, 1991).
When two or more magma fractures ascend in
1O-1~ close proximity to each other the stress fields in-
duced by the individual parcels will mechanically
perturb each other depending on their crack-tip

10
-2 ~
KIc = 109Pa. m l12
offsets, overlaps, and mutual distances (Ryan,
1990; Takada, 1994). Acceleration associated
with crack-tip splitting and coalescence is also a
0 1 2 3 4
10 10 10 10 10 likely source of seismic radiation (Sammis and
Source flux (rrr' . S·I) Julian, 1987).
Figure 12 The ratio of the dike conduit thickness to the Heat exchange between flowing magma and the
maximum dike head thickness ax/am is plotted as a function . wallrock is another important process in magma
the source flux Q, for I < Q, < 10 4 m? . s I, for 1J = I, 10, transport. Whether magma tends to solidify in a
and 100 Pa . s, and for K, = 10' and ]0' Pa . min (see magma conduit or whether the wallrock melts
Eq, (26)). All other material properties are constants chosen to
depends on the relative contributions of con-
be representative of the lithosphere; J-L/(\ - II) = 5 X 10 10Pa,
CTy = 10' Pa.p = 3300 kg . m : ', R' = 1m· So,. ductive heat transfer by the wallrock and advec-
tive heat transfer by the flowing magma, respec-
tively (Bruce and Huppert, 1990; Delaney and
fracture toughness value of K lc = 103 MPa . m 1/2 Pollard 1982).
(representative of estimates from the dimensions
of dikes in the field), a.Ja.; is always less than 0.3.
This implies that the fracture resistance velocity Summary
U F will generally control the buoyancy-driven
magma fracture propagation speed if the fracture Crack propagation (in our experiments) is found
toughness of the lithosphere is about 103 MPa . to be controlled by the fracture resistance of the
m \/2. If the fracture toughness is much lower than solid. We define the fracture toughness of the
103 MPa . m 1/2 then Fig. 12 shows that the Poi- gelatin gels to be the critical stress intensity asso-
seuille velocity is probably appropriate for low- ciated with the transition between two observed
source-flux and high-viscosity magma and the crack propagation regimes; in regime I slow crack
fracture resistance velocity is appropriate for rela- propagation is characterized by a strong velocity
tively high-source-flux and low-viscosity magma. dependence on KI ; in regime 2 fast propagation is
characterized by velocity U ox K], Dimensional
Additional Processes analysis based on the experimental results yields
an expression (Eq. (9)) for the propagation ve-
In this chapter we have presented experimental locity, in regime 2, that collapses the data (Fig. 6).
work on isolated, vertically propagating fluid- To provide physical justification of the di-
filled fractures and applied the results to magma mensional analysis we introduce a model for
fracture in the lithosphere. There are several im- buoyancy-driven fracture in which the loading
portant processes associated with magma trans- configuration provides a dynamical constraint on
port in the lithosphere that were not addressed by the crack propagation velocity. The idea behind
these experiments. the model is that the crack tip of a buoyancy-
In most igneous environments the near-surface driven fracture with stress intensity exceeding the
crustal density falls below the magma density. The fracture toughness cannot undergo unlimited un-
horizon at which the crustal density equals the stable (dynamical) propagation since the rest of
 10. Buoyancy-Driven Fracture and Magma Transport 239
the crack takes a finite time to respond to changes Anderson, O. L. (1978). The role of magma vapors in volcanic
tremors and rapid eruptions, Bull. Volcanol. 41, 341-353.
in the crack geometry.
Anderson, O. L., and Grew, P. C. (1977). Stress corrosion
The observed shape of the propagating con- theory of crack propagation with applications to geophys-
stant volume fractures with low-viscosity fluids ics, Rev. Geophys. 15,77-103.
closely approximates the theoretical elastostatic Atkinson, B. K., and Meredith, P. G. (1987). The theory of
shape of an internally pressurized crack. This subcritical crack growth with applications to minerals and
rocks, in "Fracture Mechanics of Rock" (B. K. Atkinson,
agreement indicates (a) the pressure gradient
ed.), pp. 111-166, Academic Press. London, England.
in the experimental cracks is nearly hydrostatic Barenblatt, G. I. (1962). The mathematical theory of equilib-
and propagation is independent of the fluid vis- rium cracks in brittle fracture. Adv. Appl. Mech. 7, 55-
cosity, and (b) propagation in regime 2 is "quasi- 129.
dynamic" in the sense that the crack shape adjusts Bruce. P. M., and Huppert. H. E. (1990). Solidification and
melting along dykes by the laminar flow of basaltic magma,
dynamically to the elastostatic shape in order to
in "Magma Transport and Storage" (M. P. Ryan, ed.),
maintain crack-tip loading. pp. 87 -101, Wiley, Chichester/Sussex, England.
The observed in-plane shape of the propagat- Delaney, P. T., and Pollard, D. D. (1981). Deformation of host
ing fractures resulting from the experiments using rocks and flow of magma during growth of minette dikes
a constant flux of viscous fluid is used in con- and breccia-bearing intrusions near Ship Rock, New Mex-
ico, "U.S. Geol. Surv. Prof. Paper," 1202.
junction with elasticity theory to derive a semi-
Delaney, P. T., and Pollard, D. D. (1982). Solidification of ba-
empirical formula for the cross-sectional shape. saltic magma during flow in a dike, Am. J. Sci. 282,
The cross-sectional shape consists of a crack head 856-865.
that tapers at depth to a thin conduit. Delaney, P. T., Pollard, D. D., Ziony, 1. I., and McKee, E. H.
Application of U F to magma fracture implies (1986). Field relations between dikes and joints: Emplace-
ment processes and paleostress analysis, J. Geophys. Res.
propagation velocities of order 1 m . s - I in the
91, 4920-4938.
lithosphere. Scaling of the relative contribution of Evans, A. G., and Langdon, T. G. (1976). "Structural Ceram-
the buoyancy and viscous stress to the pressure ics," "Progress in Materials Science," Vol. 21, pp. 174-
gradient in a fluid-filled fracture shows that when 441, Pergamon Press, Oxford, England.
the fracture head is a few times thicker than the Fedotov, S. A., Enman, V. B., Maguskin, M. A., Levin, V. Y.,
Zharinov, N. A., and Enman, S. V. (1978). Deformations
fracture conduit, the pressure gradient in fracture
of the Earth's surface in the vicinity of the New Tolbachik
head is nearly hydrostatic. This indicates that even Volcanoes (1975-1976), in "The Great Tolbachik Fissure
when the Poiseuille velocity controls flow in the Eruption, Geological and Geophysical Data 1975-1976"
conduit, the fracture resistance velocity U F can (S. A. Fedotov and Yeo K. Markhinin, eds.), pp. 267-282,
still control the crack propagation velocity. Ap- Cambridge Univ. Press, Cambridge.
Fiske, R. S., and Jackson, E. D. (1972). Orientation and
plying this scaling to magma fracture in the litho-
growth of Hawaiian volcanic rifts: The effect of regional
sphere we show that U F is likely to control the structure and gravitational stresses, Proc. R. Soc. London
propagation of low-viscosity magma fractures at Ser. A 329, 299-326.
depth. Fowler, A. C. (1990). A compaction model for melt trans-
port in the Earth's asthenosphere. Part II. Applications,
in "Magma Transport and Storage" (M. P. Ryan, ed.),
pp. 16-32, Wiley, Chichester/Sussex, England.
Acknowledgments Freund, L. B. (1990). "Dynamic Fracture Mechanics," Cam-
bridge Univ. Press, Cambridge.
Fujii, Too and Kushiro. (1977). Density, viscosity, and com-
This research has been supported by NSF Grant EAR8916152.
pressibility of basaltic liquid at high pressures, Carnegie
M. Heimpel thanks Paul T. Delaney for a thorough review
Inst. Washington Yearbook, 419-424.
and Dr. Michael Ryan for all his advice and support. Thanks
Gudmundsson, A. (1983). Form and dimensions of dykes in
also to Allan Rubin for some lively discussions on fracture
eastern Iceland, Tectonophysics 95, 295-307.
mechanics.
Irwin, G. R. (1958). Fracture, in "Handbuch der Physik"
(S. Flugge, ed.), Vol. 6, pp. 551-590, Springer-Verlag,
New York.
Kobayashi, T., and Fourney, W. L. (1918). Experimental char-
References acterization of the development of the microcrack process
zone at a crack tip in rock under load, "Proceedings, 19th
Aki, K., Fehler, M., and Das, S. (1977). Source mechanisms Natl. Symp. Rock Mech.," pp. 243-246.
of volcanic tremor: Fluid-driven crack models and their Lawn, B. R., and Wilshaw, T. R. (1975). "Fracture of Brittle
application to the 1963 Kilauea eruption, 1. Volcanol. Solids," Cambridge Univ, Press, Cambridge.
Geotherm. Res. 2, 259-287. Lister, 1. R. (1990). Buoyancy-driven fluid fracture: The ef-
240 Moritz Heimpel and Peter Olson

fects of material toughness and of low-viscosity precur- pressure: Implications for dike propagation, J. Geophys.
sors, J. Fluid Mech. 210,263-280. Res .. 98, 15,919-15,935.
Lister, J. R. (1991). Steady solutions for feeder dykes in a den- Ryan, M. P. (1987). Neutral buoyancy and the mechanical evo-
sity stratified lithosphere, Earth Planet. Sci. Lett. 107, lution of magmatic systems, in "Magmatic Processes:
233-242. Physicochemical Principles," (B. O. Mysen, ed.), Spec.
Lister, 1. R, and Kerr. R. C. (1991). Fluid-mechanical models Pub!. I, pp. 259-287, The Geochemical Society, Univer-
of crack propagation and their application to magma trans- sity Park, PA.
port in dykes, J. Geophys. Res. 96, 10049-10077. Ryan, M. P. (1988). The mechanics and three-dimensional in-
Macdonald, R., Wilson, L., Thorpe, R. S., and Martin, A. ternal structure of active magmatic systems: Kilauea Vol-
(1988). Emplacement of the Cleveland Dyke: Evidence cano, Hawaii, J. Geophys. Res. 93,4213-4248.
from geochemistry, mineralogy. and physical modelling, Ryan, M. P. (1990). The physical nature of the Icelandic
J. Petrol. 29, 559-583. magma transport system, in "Magma Transport and
Marsh, B. D. (1982). On the mechanics of igneous diapirism, Storage" (M. P. Ryan, ed.), Wiley, Chichester/Sussex,
stoping, and zone melting, Am. J. Sci. 282, 808-885. England.
Maaloe, S. (1987). The generation and shape of feeder dykes Ryan, M. P., and Blevins, J. Y. K. (1987). The viscosity of
from mantle sources, Contrib. Mineral. Petrol. 96,47 -55. synthetic and natural silicate melts and glasses at high tem-
McKenzie, D. P. (1984). The generation and compaction of peratures and I bar (10' Pascals) pressure and at higher
partially molten rock, J. Petrol. 25, 713-765. pressures, "U.S. Geo!. Surv. Bul!.," 1764,
Morita, N. (1993). Personal communication. Sammis, C. G., and Julian, B. R. (1987). Fracture instabilities
Nicolas, A. (1986). A melt extraction model based on struc- accompanying dike intrusion, J. Geophys. Res. 92, 2597-
tural studies in mantle peridotites, J. Petrol. 27, 999- 2605.
1022. Scott, D. R., and Stevenson, D. 1. (1984). Magma solitons,
Nicolas, A. (1990). Melt extraction from mantle peridotites: Geophys. Res. Lett. 11, 1161-1164.
Hydrofracturing and porous flow, with consequences for Sleep, N. H. (1988). Tapping of melt by veins and dikes,
oceanic ridge activity, in "Magma Transport and Storage" J. Geophvs. Res. 93, 10255-10272.
(M. P. Ryan, ed.), pp. 159-173, Wiley, Chichester/Sussex, Sneddon, 1. N. (1946). The distribution of stress in the neigh-
England. bourhood of a crack in an elastic solid, Proc. R. Soc. Lon-
Olson, P., and Christensen, U. (1986). Solitary wave propaga- don 187, 229-260.
tion in a fluid conduit within a viscous matrix, J. Geophys. Spence, D. A., Sharp, P. W, and Turcotte, D. L. (1987).
Res. 91, 6367-6374. Buoyancy-driven crack propagation: A mechanism for
Paterson, M. S. (1978). "Experimental Rock Deformation- magma migration, J. Fluid Mech. 174; 135-153.
The Brittle Field." Minerals and Rocks, Vo!. 13, Springer- Spera, F. (1980). Aspects of magma transport, in "Physics of
Verlag, New York. Magmatic Processes" (R. B. Hargraves, ed.). pp. 265-
Pollard, D. D. (1987). Elementary fracture mechanics applied 323, Princeton Univ. Press, Princeton, N1.
to the structural interpretation of dykes. in "Mafic Dyke Takada, A. (1994). Accumulation of magma in space and time
Swarms" (H. C. Halls and W H. Fahrig, eds.), Geo!. As- by crack interaction, in "Magmatic Systems" (M. P. Ryan,
soc. Canada Spec. Paper 34, pp. 5-24. ed.), Academic Press, San Diego.
Pollard, D. D., and Muller, O. H. (1976). The effect of gradi- Takada, A. (1990). Experimental study on propagation of
ents in regional stress and magma pressure on the form of liquid-filled crack in gelatin: Shape and velocity in hydro-
sheet intrusions in cross section, J. Geophys. Res. 81, static stress condition, J. Geophys, Res. 95, 8471-8481.
975-984. Turcotte, D. L. (1990). On the role of laminar and turbulent
Reches, Z., and Fink, 1. (1988). The mechanism of intrusion of flow in buoyancy driven magma fractures, in "Magma
the Inyo Dike, Long Valley caldera, California, 1. Geo- Transport and Storage" (M. P. Ryan, ed.), pp. 103-111.
phys. Res. 93, 4321-4334. Wiley, Chichester/Sussex, England.
Richards, R., and Mark, R. (1966). Gelatin models for photo- Weertman, J. (1971). Theory of water-filled crevasses in gla-
elastic analysis of gravity structures, Exp, Mech. 6, 30-38. ciers applied to vertical magma transport beneath oceanic
Rubin, A. M., and Pollard, D. D. (1987). Origins of blade-like ridges, J. Geophys. Res. 76,1171-1183.
dikes in volcanic rift zones, "U.S. Geo!. Surv. Prof. Pa-. Whitehead, 1. A., and Luther, D. S. (1975). Dynamics of labo-
per," 1350, pp. 1449-1470. ratory diapir and plume models, 1. Geophys. Res. 80,
Rubin, A. M. (1993). Tensile fracture ofrock at high confining 705-717.
Chapter 11 Accumulation of Magma in Space
and Time by Crack Interaction

Akira Takada

Overview times, one after another. The volume of a magma-filled

crack after the nth coalescence amounts to (the initial
Magma accumulates in space and time during its as- volume of a stationary crack) X 2". If a magma volume
cent. Crack interactions play a significant role in the is accumulated of the order of 10" m 3 (0.1 km 3 ) , 23-10
process of magma accumulation. Gelatin experiments coalescences are needed.
on crack interactions with a range of initial geometries When the differential stress in the horizontal plane
are summarized. For example, two vertical, collinear, is small at a large supply rate of magma, cracks may
buoyancy-driven cracks; two vertical, buoyancy-driven coalesce with relative ease. Magma concentrates in
cracks whose planes are oriented through a variety of space in its ascent process, so that a magma path may
angles to each other; and two vertical, parallel, offset, be formed. Such a case plays a role in the formation of
buoyancy-driven cracks can coalesce. Buoyancy-driven a polygene tic volcano. When the differential stress in
cracks coalesce with relative ease, compared with non- the horizontal plane is large, cracks coalesce with diffi-
buoyant internally pressurized cracks. The stress fields culty. Such a case plays a role in the formation of a
near two collinear, pressurized cracks under hydrostatic monogenetic volcano. On the other hand, at a small
conditions are evaluated, using the stress intensity fac- supply rate of magma, the possibility of crack interac-
tors at a crack tip. The stress fields near two parallel. tions in space and time decreases. A monogenetic vol-
offset, pressurized cracks under various regional stress cano is formed whether the differential stress is large
conditions are evaluated. using the expansions of Wes- or small.
tergaard stress functions. On the basis of the results of
gelatin experiments and stress analyses, a qualitative Notation
diagram of crack height vs horizontal distance between Units
two cracks, indicating the conditions for coalescence of
F1(w) complete elliptic integral of the
vertical, buoyancy-driven cracks, is proposed. When
first kind
the height of the upper crack is smaller than that of the
F,(w) complete elliptic integral of the
lower main crack, the main crack can catch up with the
second kind
upper crack, and there is a minimum coalescence height
of the upper crack at a certain horizontal separation dis- 1m imaginary part
tance. When the height of the lower crack is larger than K, fracture toughness Pa m-"
v

that of the upper main crack, the lower crack can catch K1 stress intensity factor (mode I) at Pa·m l12
up with the main crack, and there is a maximum co- a crack tip
alescence height of the lower crack. When the horizon- M volume of the stationary crack m'
tal separation distance is reduced, the range of crack H(h) relation between crack width and m
heights for coalescence increases. An increase of the height
applied differential stress on remote boundaries pre- Pa
P(x) excess pressure in a crack
vents buoyancy-driven cracks from interacting.
P/x) excess pressure in crack j Pa
The efficiency of magma-filled crack coalescence in
(j = 1,2)
the Earth's mantle and in the crust is discussed. Magma-
filled cracks can coalesce and increase in volume during r.: maximum excess pressure Pa

their ascent as long as a range of crack sizes and ascent Po excess liquid pressure in a crack Pa
velocities exists. One simple model for crack coales- po} excess liquid pressure in crack j Pa
cence is that cracks of nearly the same size coalesce n (j = 1,2)

Magmatic Systems Copyright © 1994 by AcademicPress. Inc.

Edited by M. P. Ryall 241 All rightsof reproduction in any form reserved.
 242 Akira Takada

Units Units

Re real part rotation angle radians

v ascending velocity m's-' w number dimensionless
2,(z) complex function
2,,(z) complex function
a half-height of a crack m
Introduction
0, half-height of crack j (j = I, 2) m
half length of the smaller crack m
Beneath a polygenetic volcano, magma concen-
d horizontal distance between two m trates in space and through time so that an approxi-
cracks
mately steady-state magma path may be formed.
e distance between two cracks nor- dimensionless
To supply discrete volcanic centers, magma must
malized by 0,
accumulate in space when it is otherwise scattered
g acceleration of gravity m . S-2
in the partially melted mantle (Fig. 1). On the
h crack height m
other hand, a volcano erupts intermittently. Thus,
crack height m magma must accumulate in time when it is other-
n half-length of the larger crack dimensionless wise believed to be produced at a constant rate in
normalized by 0,
the partially melted mantle. The role of magma
no number dimensionless accumulation is therefore vitally important for
distance from crack j (j = 1,2) m magma ascent and eruption.
s; /Icomponent of the regional Pa, MPa There is a well-understood currently active
stress (u = .r, y, x)') magmatic system in Hawaii where it has been
/I component of the stress from Pa, MPa demonstrated that magma ascends great distances
crack j Iu = x,y,xy;j = 1,2) by a magma fracture mechanism (e.g., Shaw,
time 1980; Ryan et al., 1981, 1987a, 1988). Seismic
crack width m data with well-located hypocenters, for example,
x Cartesian coordinate m define a primary conduit region for Kilauea that is
X Oj x coordinate of the center of the m most consistent with a side-by-side ascent mode
crackj(j= 1,2) of buoyant fractures (Ryan, 1988). This chap-
)' Cartesian coordinate m ter deals with magma transport through a crack.
Yo, y coordinate of the center of the m Magma ascent by a buoyancy-driven crack in the
crack j (j = J. 2)
complex number
density difference Accumulation
in space
Cartesian coordinate m
TJ viscosity Pa· s

J-L shear modulus Pa, MPa, GPa

/J Poisson's ratio dimensionless
1T Pi = 3.1415 dimensionless
p density kg' m-'
(T m ;\, maximum principal stress Pa,MPa
(T m " . minimum principal stress Pa, MPa 100%
(T,,(X.Y) u component of stress around a Pa, MPa
crack (u = .r, y, xy)
Figure 1 Magma accumulation in space and time. The left
(r",(x, y) /Icomponent of stress originated Pa. MPa
by crackj (/I = x, y, xy;j = 1,2) diagram shows magma accumulated from the partial melting
rock. The right diagram shows a large magma body produced
O""i2(x, y) u component of stress originated Pa, MPa from the total melting rock. I and x represent time and space,
by crack I and crack 2 (u = x, y, respectively. The boundary conditions are composed of the
xy) magma production rate as input, the density of the host rocks,
0", principal stress (j = I, 2, 3) Pa, MPa the stress condition, and the geotherm.
 11. Accumulation of Magma in Space and Time 243
Earth's lithosphere has been studied by Weertman addition, the efficiency of crack coalescence in
(1971a, 1971b), Spence and Turcotte (1990), Ta- the mantle and in the crust, and the implications
kada (1989), Lister and Kerr (1991), and others. of crack interactions on volcano evolution, are
Takada (1989) proposed a magma transport model discussed.
by considering an extensive crack propagation
system. Unlike soliton wave transport (e.g., Scott
Crack Interactions
and Stevenson, 1984; Scott et al., 1986; White-
head, 1986; Whitehead and Helfrich, 1990), crack
Approach to Crack Interaction
interactions were found to play a significant role
in the process of magma accumulation in space It is difficult to calculate the detailed patterns of
and through time. three-dimensional crack interactions. The effect of
Fundamental theories of crack interaction have positive fluid buoyancy on crack interactions, for
been studied with reference to the mechanics of example, may be inferred from analog experiments
magma transport and the earthquake process. on the interactions of liquid-filled cracks in gela-
Two-dimensional open crack interactions under tin. The two-dimensional interaction of buoyancy-
a tensile stress field have been developed by us- driven cracks is evaluated using stress analyses.
ing dislocation theory and fracture mechanics There are two approaches to evaluate the two-
(e.g., Yokobori et al., 1965, 1971; Yokobori dimensional crack interaction: one is an approach
and Ichikawa, 1967). Analog experiments of utilizing stress intensity factors at a crack tip and
two-dimensional crack interaction problems have the other is an approach that examines stress tra-
been carried out by using a photoelastic material jectories near cracks. The first is a fracture me-
(Lange, 1968) and a soda-lime glass (Swain and chanics approach and the latter employs classical
Hagen, 1978). Delaney and Pollard (1981), Pol- continuum mechanics.
lard et al. (1982), and Pollard and Aydin (1984) The former approach is applied to the interac-
applied the two-dimensional interaction of paral- tion of collinear cracks. Whether or not a crack
lel, offset, internally pressurized cracks to dike in- will propagate (the crack propagation criteria) de-
trusions and spreading mid-ocean ridge segments. pends on the crack extension force; the direction
A pressurized crack is one with no buoyancy of in which a crack will propagate depends on the
the fluid in the crack relative to the surrounding local stress field at the crack tip as measured by
medium; a buoyancy-driven crack is one with the ratio of stress intensity factors, for example,
buoyancy of the fluid relative to the surrounding mode II/mode I (Lawn and Wilshaw, 1975; Pol-
medium. Stoyanov and Dabovski (1986) studied lard, 1987). Delaney and Pollard (1981), Pollard
the development of echelon cracks under oblique et al. (1982), and Pollard and Aydin (1984) stud-
extension. Olson and Pollard (1989) estimated the ied the interaction of internally pressurized cracks
crack paths by the interactions of the two pressur- using linear fracture mechanics. Erdogan and Sih
ized cracks under a remote differential stress. Ryan (1963) evaluated the stress fields represented by
(1990) evaluated the displacement fields near three stress intensity factors at a crack tip and estimated
parallel, offset, pressurized cracks to study the in- the direction of crack extension under plane load-
ternal mechanics of the Icelandic magma transport ing and traverse shear.
system. Takada (in press) performed gelatin ex- The crack extension paths near two parallel
periments on the three-dimensional crack interac- offset cracks under various stress boundary con-
tions between buoyancy-driven cracks and applied ditions are inferred from the latter approach. It
the results to the development of the subvolcanic is difficult to estimate the interaction between
structure including a dike and sheet complex. In buoyancy-driven cracks. However, the stress tra-
contrast to previous studies, the crack interactions jectories approach has an advantage in knowing
were considered from the viewpoint of magma ac- the approximate crack extension pathways. The
cumulation in space and through time. maximum principal stress (0' ~~x) trajectories con-
This chapter summarizes gelatin experiments trol the approximate direction of crack exten-
on fluid-filled crack interactions and then evalu- sion. The gradient of the minimum principal
ates the mechanics of crack interactions between stress (0' min) along a 0' max stress trajectory governs
buoyancy-driven cracks, using stress analyses. In whether the crack will extend or not.
244 Akira Takada

There are two types of crack interactions: the trolling the fracture dynamics than that of buoy-
interaction of liquid-filled cracks and the interac- ancy, once a magma-filled crack begins to extend.
tion between a liquid-filled crack and a solid-filled However, when stress fields are discussed, the
crack (Takada, in press). The latter type of inter- effects of the fracture resistance at a crack tip is
action will not induce two cracks to coalesce. This very small, in comparison with that of the velocity
chapter deals with the interaction of liquid-filled of crack propagation. Thus, gelatin experiments
cracks. are valid for constructing an approximate view of
Takada (1990) reported the existence of both a the complex pattern of three-dimensional crack
growing crack and an isolated crack in an elastic interaction.
medi um. Lister and Kerr (1991) concl uded that a Several experiments on crack growth under
growing crack driven by positive buoyancy with a nonhydrostatic conditions, using gelatin or soft
constant flow rate has a bulbous nose at the crack organic polymers, have been performed: the inter-
tip and a growing feeder tail connected with the action between parallel offset cracks in a photo-
source "feeder dike." However, the source must elastic material (Lange, 1968), crack growth in
have sufficient magma volume to maintain a rela- gelatin under the gravitational stress field (Fiske
tively constant flow rate. Prior to crack ascent, and Jackson, 1972), crack growth in acethylcel-
magma accumulation has already occurred. It is lulose jelly (Dabovski and Stojanov, 1981), and
thus important to study the interactions of isolated the development of echelon cracks (Stoyanov and
cracks in the process of transporting the magma Dabovski, 1986). On the other hand, a few, quan-
accumulation from the partially melted mantle. .titative gelatin or agar experiments on the propa-
This chapter deals with the spatial accumulation gation of a buoyancy-driven crack under hydro-
of otherwise isolated cracks and adds some con- static conditions were carried out: on crack shape
siderations toward understanding the accumula- (Maalee, 1989), on crack shape and crack propa-
tion of buoyancy-driven, growing cracks with a gation (Takada, 1990), and on the lateral growth
source operating at a constant flow rate. of a crack at the neutral buoyancy level (Lister
A quasi-static isolated crack is approximately and Kerr, 1991). Takada (in press) studied the
equivalent to one end member of cracks such that interaction of two buoyancy-driven, liquid-filled
K1 = K; at the upper dike tip and K ~ 0 at the cracks.
lower tip. For example, the additional length scale The methods of experiments and physical prop-
of a quasi-static, isolated crack in gelatin is a few erties of gelatin were reported in detail in Takada
to several centimeters (Takada, 1990). As crack (1990; in press). E-290 gelatin powder prepared
coalescences repeat, an isolated crack increases by Miyagi Chemical Industrial Company Ltd.,
in volume and in length so that it departs from Sendai, Japan, was used. A 1.25% gelatin solution
the quasi-static case. It is difficult to evaluate the was solidified in a rectangular container 3 em in
stress field near a long crack that is not quasi- length, 30 cm in width, and 50 em in height. The
static. In the stress analyses and in the discussions density of gelatin is 1007 ± I kg . m -\ the
of this chapter, it is assumed that isolated cracks Young's modulus is about 0.6-1.3 X 103 Pa, and
with various lengths are quasi-static. the Poisson's ratio is nearly 0.5. The stress state in
gelatin is nearly hydrostatic before the injection of
Summary of Gelatin Experiments liquids. As a working fluid, I cSt-silicon oil (den-
sity 822 kg' m 3, viscosity 0.8 X 10 -3 Pa . s) was
Analog experiments employing gelatin suggest injected into the tank gelatin from the bottom and
that buoyancy-driven cracks are relatively easy the wall of the container by a syringe or a micro-
to coalesce with each other; that is, buoyancy pump, to generate two vertical cracks (Fig. 2).
is effective for crack coalescence (Takada, in Both the inclination and the trend of a crack in
press). In gelatin experiments, one can observe gelatin can be controlled by the obliquely cut tip
the three-dimensional interactions of buoyancy- of a needle. If the lower 'crack is larger than the
driven. liquid-filled cracks. The fracture resis- upper crack, the lower crack ascends at a faster
tance of a crack tip is relatively large in gelatin speed to catch up with and coalesce with the upper
(Takada, 1990). Lister and Kerr (1991) pointed crack that is ascending at a lower velocity. The as-
out that the fracture resistance effect by a magma- cent velocity is proportional to (the crack height)"
filled crack in rocks is far less important in con- in gelatin (Takada, 1990). The results of the gela-
 II. Accumulation of Magma in Space and Time 245
GELATIN EXPERIMENTAL METHODS loads do not appear to coalesce without a prior in-
crease in the liquid volume within the cracks. Two
Crack interactions
vertical, parallel, offset, buoyancy-driven cracks
gelatin can coalesce if the horizontal distance between two
cracks is not large: for example, about a separation
E distance of one-half the vertical height of the upper
~t o
o
It)
crack. On the other hand, two externally loaded
cracks have never been observed to coalesce.
syringe
In Fig. 3B, the lower crack was initiated verti-
cally, and the upper crack was inclined and dip-
ping to the right. After injection, the lower crack
propagated upward with positive buoyancy and
was unperturbed by the stress field from the the
upper buoyancy-driven crack (Fig. 3B, I). Under
SUMMARY OF THE INTERACTION the influence of the stress field of the upper crack,
OF BUOYANCY- DRIVEN CRACKS
the lower crack began to interact with the lower
tail of the upper crack, turned to the left, and fi-
nally coalesced (Fig. 3B, 2-4).
In Fig. 3C, liquid was supplied from the bottom
of. the gelatin block. Cracks of the same size as-
cended one after the other at a fairly regular inter-
val through the preexisting fracture. Once a sub-
stantially larger crack was formed, the two cracks
Collinear Normal, twist coalesced. The lower larger crack could thus catch
up with the upper smaller crack and coalesce.

Interaction of Two Collinear Cracks

Delaney and Pollard (1981) evaluated stress inten-

Vertical, horizontal Parallel offset sity factors at the tips of two collinear, internally
pressurized cracks of the same size. They reported
Figure 2 Gelatin tank experimental methods for the study of that collinear cracks are attracted to each other.
crack interactions (upper), and a summary of the types of in- Yokobori et al. (1965) studied the stress intensity
teractions of buoyancy-driven cracks (lower).
factors at the tips of collinear open cracks of dif-
ferent sizes under a uniaxial stress field. To study
tin experiments are summarized in Fig. 2. This two collinear pressurized cracks under hydrostatic
chapter adds two new cases shown in Fig. 3 to the conditions, this study expands the stress intensity
results of Takada (in press), The additional length factors (mode I) of Yokobori et al. (1965):
scales of cracks in Figs. 2 and 3 are in the range
of the quasi-static case. 1T(2n + e)
K, Po
Two vertical, collinear, buoyancy-driven cracks 4ea s
can coalesce as illustrated in Fig. 3A. Two vertical
cracks whose planes are oriented through a vari-
ety of angles to each other can also coalesce. A (2a, - (2a, + ea s ) (1 a)
lower vertical buoyancy-driven crack can coalesce
with an upper horizontal crack, inducing the lower
crack to close as its fluid is discharged into the .(1 _F,(~) F 2 (w)
)
upper crack and inducing the upper crack to start
to grow horizontally; however, some liquid is re-
tained in the lower crack. On the other hand, two 4n
w= (Ib)
cracks subjected to a set of external boundary (2n + e)(e + as)
Figure 3 Photographs showing gelatin experiments on the interactions of collinear buoyancy-driven cracks (A), the interaction of two buoyancy-driven cracks that approach in an oblique
position (B), the coalescence involving fluid exchanges and changes in the crack size (C), respectively. The larger crack at the lower position catches up with the smaller one at the upper
position, and they coalesce (A, I ~ 4; B, I ~ 4). As the lower crack approaches the upper crack, it begins to tum toward the upper crack. The lower crack develops a C-shaped, curved
structure in the horizontal section at its tip, which appears as differential thickness in the photograph (B, 3). The lower small cracks, marked with open triangles, are ascending rapidly one
after the other through the preexisting fracture JUSI below the upper slowly ascending large crack (C, I), and begin 10 coalesce by fluid exchanges and changes in crack size (C, 2-4).
248 Akira Takada

K1/Po evaluated the crack extension paths by the inter-

crack 1.2 1.6 2 actions of two parallel offset pressurized cracks
under a remote differential stress, using the bound-
ary element method. For a small remote differen-
2 tial stress, the closely spaced, two parallel offset
cracks exhibit an asymptotic approach to each
4 other. A large remote differential stress produces
a nearly planar path. Takada (in press) evaluated
6 the stress fields near an internally pressurized
crack under various regional stress fields, and the
8 stress fields near a buoyancy-driven crack under
hydrostatic conditions.
crack 10 This chapter evaluates the stress fields near
e
two internally pressurized cracks under various
Figure 4 The stress intensity factor K, (mode 1) at the tip of regional stress conditions to determine the ap-
the smaller crack due to the interaction of two collinear pres- proximate patterns of the stress trajectories under
sured cracks (after Yokobori et at.. 1965). Po is the excess
various remote stress conditions and to discuss the
pressure. The distance between two cracks, and the half-length
of the larger crack. are normalized by the half-length of the interactions of buoyancy-driven cracks. Wester-
smaller crack. a" respectively. gaard stress functions (Westergaard, 1939) were
adopted to expand the two-dimensional stress
analysis. Complex stress functions, Z,(:) and
F,(w), F 2(w) are complete elliptic integrals of the Zll(Z), represent the stress field near a crack with
first and second kind, respectively. Po is the excess the symmetrical excess pressure distribution P(x)
liquid pressure against the crack wall, as> e, and n along the long axis (Tada et al.. 1973),
are the half-height of the smaller crack, the dis-
tance between two cracks normalized by as> and
z = z + yi (2)

the half-height of the larger crack normalized by

as> respectively. Figure 4 shows that the smaller 1Ty z2 - a?
crack is attracted by the larger crack and that two
cracks with shorter distances are attracted more
.f·a
a P(~)Ya2
Z -
r-;:-----,""

~
~2 d~
(3)

strongly. The attraction force is the stress gradient

induced by the larger crack. The gelatin experi- Zll(Z) = 0 (4)
ments suggest that collinear, buoyancy-driven,
liquid-filled cracks are easy to coalesce compared
P(x) = Po (internally pressurized crack) (5)
with two remotely loaded cracks (Figs. 3A, P(x) = Po + t:.pg(x + a)
and 3C). (tluid-buoyancy-driven crack), (6)
where z is a complex number, a is the half-height
Interaction of Two Parallel Offset Cracks
of the crack, ~ is a variable in the range - a to a.
Pollard (1973) estimated the stress fields around and x and yare the coordinates on the long axis
a two-dimensional, internally pressurized crack and the short axis of a crack. The x component of
under hydrostatic conditions. Roberts (1970) and stress, /TAx, y), the y component, /Ty(x, y), and the
Dabovski (1979) studied the stress fields around shear (xy) component, /Txy(x, y), are
an elliptical magma body under regional stress /TAx. y) Re[Z,(z)] - y Im[dZ,(z)/dz] (7)
conditions. Delaney and Pollard (1981), Pollard
et al. (1982), and Ryan (1990) calculated the dis- /Ty(x. y) Re[Z((z)] + y Im[dZ,(z)/dz] (8)
placement fields near two offset cracks. Pollard
/Tx/ .r, y) - y Re[dZ,(z)/dz], (9)
and Aydin (1984) estimated the stress field near
two internally pressurized cracks under exten- where Re[Z(z)], Im[Z(z)] are the real and imagi-
sional stress conditions. Olson and Pollard (1989) nary parts of the complex function Z(z), respec-
 11. Accumulation of Magma in Space and Time 249
Sx 2 \ \ I II 3
A \I
,
I II
C

,
\ I I I I
I I /I
I I I 1/
II 1//
2

0
........
\ \ \
, ...........
\
I I 11\
I I I 1\
I I , I I

-2
/I
I I ," I
I I

Omin Omax trajectories Omin

Sy , 3
B 2 " I
II I I I ' 0
,

,
I /I
"
I I I I
I II I /

...I" -'//
/"'-'",
111-" ....... " " " ,
2 1111 \\\,',,"
I , I I I \ I \ \ \ \ \ \
I I \ \ \ " 1 1\ \ , \ \
0 \ \ \ \ \ I I I I \ \ \ I I
,~~ ................... \ \ \ \ \ \ \ I I \ I I I
................ v v v v v v v v v III/
,...................
, \ II I
,,,
I I \ \
I I I I
II I
II I I I
-2 /I I I I
3
omex trajectories

Figure 6 Contour map of the minimum principal stress

Figure 5 The two-dimensional stress field near two Iiquid- and the maximum principal stress (IT m,,) trajectories
(IT mio)
filled cracks subject to the regional stress field. P, (x), P,(x), near a remotely loaded internally pressurized crack (A), those
S,. and s; are the excess pressures of crack 1 and crack 2. and near a buoyancy-driven crack (B), the contour map of ITmin (C).
the regional stresses, respectively. and the IT max trajectories near two internally pressurized cracks
(D). The excess internal pressure of an equally pressurized
crack, and the maximum excess pressure of a buoyancy-driven
crack, are assumed to be 1. The compressional stress is posi-
tively. (T,j(X - Xo j, )' - )'OJ), (Tyj(X - X Oj, )' -
tive. The stress in the stippled area represents extensional
YOj), (T,y)(x - X Oj, )' - are the stresses near
)'OJ) components. The compressive component fields are unstippled.
crack j centered at the coordinate (XOj, YOj) (j = The contour interval of the stress is 0.2. A short bar represents
I, 2) (Fig. 5). The stresses near two offset cracks the calculated IT m" trajectory at each mesh point.
(crack I and crack 2) under the remote regional
stress fields, (T,IZ, (TyIZ, and (T,yIZ, are, based on the
method of superposition, crack interactions of two buoyancy-driven cracks.
The compressional stress is positive. The princi-
(TdZ = (T,I(X - XOI> )' - )'01) (10) pal stress axes are given, rotating the x-Y coordi-
+ (T,z(x - X 02 , )' - )'OZ) + s, nate through the angle ¢J:
O"Y12 (Tyl(x - XOI. Y - )'01) ¢J = 0.5 Arctan[2(T,yl2/((T,lz - (TyIZ)]' (13)
(II)
+ (TyzCX - .\"02' )' - )'OZ) + S"
¢J depends on the liquid pressure and the
O"xy12 a; 1(x - X 0 1 , )' - YOI) differential regional stress defined as "the maxi-
(12)
+ (Tn-z(x - X02, )' - )'OZ) + Sxy, mum principal stress-the minimum principal
stress." The pattern of stress trajectories depends
where s .. , s., s ,yare x,)'. and x)' components of the only on the differential regional stress, if the dif-
regional stress, respectively. ferential stress is normalized by the constant ex-
Solving Eqs. (2)-( 12), the (Tmax trajectories and cess liquid pressure (Takada, in press).
the values of the (Tmin near an internally pressur- Parallel, offset, internally pressurized cracks
ized crack, near a buoyancy-driven crack, and may interact under hydrostatic conditions. The
near two offset, internally pressurized cracks are detailed paths of two parallel offset pressurized
obtained as shown in Figs. 6 and 7. The results are cracks under an isotropic remote stress were simu-
approximate, but they are useful for knowing the lated by Olson and Pollard (1989). Such paths are
250 Akira Takada

( 0, 8 (0.25, " 1)
" 3 r---+---'---i
-----_
-
" ....
... -"'''''''\\
2 ""I''''' """'''''''
/111 v vv vv v v vv 2 ~~~; ~~~~~~~~~
I 1 I I \ I I I \ \ \ \ \ \ J /I 1 I I " \ \ \ \ \
I 1 \ \ \ II II \ \ \ \ \ I 1\ \ \ I I I I ' \ \ \ I

\\\\ \\\I' \
\ \ '" I I I I \ \\ II
\ \. \. \. \. \ \ \ \ I
III
I I I
, \ 1\ I t i l l \ \ I I I
1\ \ \ \ \ 111
'I.' \\\ \\ \\
II J I
111/
... .......... v v vv vv vv
--,,,...
"""
-, "" '- ",,"'<I'I'
............... ,
-------- ...-_--
\\
"
...
2 3 2 3

c 0.5, 1, 1) o ( 1, " 1)

3 "/I
I I I I
11"\\\11
I \ \ \. \ \ \ \ I
III/ \ \ \\\ \

2 :~~~
I f II II I \\
~~~l~ ~
,
2
[ /I \ , I I I I I
I 1\ , 1 I I I I I
1\ \ \ \ I I I I
1\ \ \ \ II \ ,
1\ \ \ \ \ \ \ I
, \ \ \ \ \ \ \ I
1 \ \ \ \ \ \ \ I
I \ \ \. \. \. \. \. I I f /

2 3 2 3
Figure 7 The (T max trajectories near two internally pressurized cracks. Numerals in parentheses are the differential stress nor-
malized by liquid pressure, the crack height contrast, and the distance between two cracks normalized by the half-height of the
left crack. (A) Hydrostatic conditions; (B), (C), and (D) differential stresses of 0.25,0.5, and I. respectively. The pattern is the
same as in Figure 6.

observed at various scales in the field (Fig. 13 in host rocks and magma is small, the effect cannot
Pollard et al.. 1982; Fig. 8). Figures 8A and 8B be found. For small cracks such as the veins in
may be equivalent to the cross sections of parallel, Fig. 8B, this effect also cannot be detected.
offset, internally pressurized cracks. For example, The (T max trajectories near two cracks curve
the offset dikes in Fig. 8A outcrop on a river bed with a sigmoidal pattern as shown in Fig. 6D. The
as the horizontal cross section, so that the effect contour map of the (T min value near two internally
of buoyancy on crack pattern cannot be found. pressurized cracks in Fig. 6C indicates that the
Moreover, if the density difference between the contour interval on a (T max trajectory decreases to-
ward the cracks, that is, the stress gradient on the
(T max trajectory increases. The stress near the crack

A B Quartz- becomes more compressional toward the crack.

Basalt dike feldspar Thus, it is difficult for a crack tip to extend from
vein
the right crack to the left crack in this illustration.
This tendency is supported by the crack pattern in
Fig. 8. However, the contour interval on a (T max tra-
jectory near the tail of a buoyancy-driven crack
(Fig. 6B) is larger than that near an internally pres-
surized crack (Fig. 6A). The compressional stress
near the tail of a buoyancy-driven crack does not
20 em 20 em increase. If it is assumed that another buoyancy-
driven crack propagates from the right side toward
the crack tail of Fig. 6B, the two cracks can
Figure 8 Plan views of interacting dike tips. Two basalt
coalesce. The buoyancy' of the buoyancy-driven
dikes in the welded tuff of the Shitara igneous complex, in the
Shitara district, central Japan (A), and two quartz-feldspar crack prevails over the stress gradient, in that (T min
veins in the granitic rocks of the Cottonwood stock, Utah (after becomes more compressional toward the crack
Pollard et al.. 1982) (B). tail of Fig. 6B. If we exchange internally pres-
 11. Accumulation of Magma in Space and Time 251
surized cracks for buoyancy-driven cracks in Conditions for Crack Coalescence
Figs. 6C and 6D, it is possible that the buoyancy-
driven crack on the right extends approximately The gelatin experiments combined with the stress
on the [T max trajectory to coalesce with the crack analyses indicate that the conditions for offset
on the left. This supports the results of gelatin crack coalescence depend on the crack heights,
experiments on the interactions of two parallel, the separation distance between the two cracks,
offset, buoyancy-driven cracks. Parallel, offset, and the differential stress. The maximum excess
buoyancy-driven cracks are relatively easy to co- pressure caused by buoyancy in a vertical crack,
alesce under hydrostatic conditions. P rnax e is proportional to the crack height, h, if
As the differential stress increases from 0 to I, Po = 0 in Eq. (6):
the contour patterns of the [T max trajectories be- P m ax = !:ipgh. (16)
come parallel (Figs. 7A and 7B). This suggests
that it is more difficult for cracks to coalesce at Thus, a parameter, the crack height, includes the
larger differential stresses. The detailed paths of excess pressure. The condition shown in Fig. 9 is
two parallel offset pressurized cracks under sev- inferred from Figs. 2-4, 6, and 7.
eral remote stresses were reported by Olson and The ascent velocity of an isolated crack, V, in
Pollard (1989). This tendency is valid for the the case of laminar flow, is presented in several
case near two buoyancy-driven cracks. The criti- previous works (e.g., Weertman, 1971b; Takeuchi
cal value of the differential stress for coalescence et al., 1972; Takada, 1990),
is around 0.5 (Fig. 7C). 2
V = !:ipgw (17)
On the other hand, the possibility of coales- 37] ,
cence decreases with the increasing size contrast
of two cracks. The far-field stress perturbations
from a crack are in an inverse relationship to (the
crack separation distance normalized by the crack
size)? (Pollard and Segall, 1987). The far-field
stress from crackj, Suj, is o d
o
(14) ...
Q)
Q.
Q.
where P Oj , aj' and Ij are the maximum excess :::J
pressure of crack i, the height of crack j, and the
distance from crack j, respectively (j = 1,2). It is
assumed that the maximum excess pressure in a
crack is in proportion to the crack height. For a2 =
(I/l1 o)al> P 02 = (I/l1 o)Po" the stress from crack 2
at the far stress field is given by Eq. (14):

Su2 cc P02(a2/r2)2 (I/110)p0\(a,/l1 or 2)2

+-'
= (1/no) 3Pol(a l/r2)2. (15) ..c
0)
'Q5
As the size of crack 2 becomes smaller than ..c
~.
crack 1, that is, 110 becomes larger, the stress de-
rived from crack 2 decreases abruptly (Eq. (15)).
~
o
Moreover, Eq. (15) indicates that the interaction
between two cracks decreases as the distance be-
tween the two cracks increases, as intuitively ex- Figure 9 Crack height vs the horizontal separation distance
between two crack diagrams showing conditions promoting
pected. If the horizontal separation distance be-
crack coalescence. The height of the main crack is /0' In the
tween two cracks is 11 0 times as large as the case stippled regions of the diagrams, two cracks may coalesce.
in Fig. 7A, the stress from crack 2 is (1/110) 2 times The upper region «/0) and the lower region (>/0) represent
as large as that case. coalescence with the upper and lower cracks. respectively.
252 Akira Takada

where YJ and ware the viscosity of injected liquid ily on the conditions of the coalescence from the
and the crack width, respectively. The crack width lower growing crack into the upper isolated crack.
is given by the function H(h). Substituting into
Eq. (17), the ascent velocity is
Applications to Magma Accumulation
V ex H(h)2. (18)
in Space and Time
For example, the ascent velocity of an isolated
crack in gelatin was obtained by Takada (1990): Magma Accumulation
(19) If a magma reservoir is not initiated in the partially
It is assumed that the ascent velocity becomes melted mantle, magma may accumulate through
larger with increases in the height of an isolated magma-filled cracks. It is possible that a magma-
crack. Thus, when the height of the upper crack is filled crack is generated in the partially melted
smaller than that of the lower main crack, the main mantle or around this region (e.g., Nicolas, 1986,
crack can catch up with the upper crack. When the 1990; Sleep, 1988). According to Lister and Kerr
height of the lower crack is larger than that of (1991), the maximum thickness and the maximum
the main crack, the lower crack can catch up with height of an isolated stationary crack are
the main crack. When the horizontal separation !5.0...)2/3
distance is smaller, the range of crack heights for ( Spg (20)
coalescence becomes wider, which is inferred from
Figs. 3 and 6, and Eqs. (14) and (15). The concave w
max
_ (A K~ )1/3 (21)
curve of Fig. 9 represents the minimum coales- upgm
cence height of the upper crack at a certain hori-
zontal separation distance, d, when interactions
m = JL/(I - v),
occur between the main crack and the upper crack. where Kc ' JL, and u are the fracture toughness, the
The convex curve represents the maximum co- shear modulus, and the Poisson's ratio, respec-
alescence height when interactions occur between tively. The volume of the crack (assuming equal
the main crack and the lower crack. In the stippled height and lengths) is given by Eqs. (20)-(21):
regions between two curves, two cracks can co-
M - (hmax)2wmax. (22)
alesce. 10 is the height of the main crack. The up-
per «1 0 ) and lower regions (>10 ) represent the For K c = 4 MPa . m l /2 (e.g., Atkinson, 1984),
coalescence with the upper and lower cracks, re- D.p = 400 kg' m :", JL = 20 GPa, and v = 0.3,
spectively. In the case of collinear cracks, the h max = 100 m, W max = 1.4 mm, and M = 14 m '
main crack can coalesce with the upper crack with are obtained. The apparent resistance tends to in-
a height in the range 0-1 0 , or with the lower crack crease due to the cooling of magma, or due to
with a height greater than 10 (Figs. 3 and 4). If the plastic deformation at a crack tip. The estimated
differential stress increases, the region where the fracture toughness at very low confining pressures
two cracks can coalesce is reduced (Fig. 7). from the shape of a dike in the field (e.g., Delaney
The elastic stress perturbations from a buoy- and Pollard, 1981) may be useful. For example,
ancy-driven, growing crack of Lister and Kerr h max = 8 X 10 2 m, w max = 0.1 m, and M =
(1991) originate mainly from a bulbous nose at 8 X 10 4 m ' are obtained for K, = 100 MPa· m 1/2•
the crack tip and a feeder tail. The height of the If a magma volume of 108 m ' (0.1 km 3) is ac-
growing crack from the source may be equivalent cumulated, 7 X 10 6 isolated stationary cracks for
to the height of an isolated crack. The crack height K; = 4 MPa· m 1/2 or 1.2 X 103 stationary isolated
from the source should be used for evaluating the cracks for K c = 100 MPa . m 1/2 must coalesce. One
crack interactions. The feeder tail with constant simple model for crack coalescence is adopted here
width generates a local stress perturbation, in con- as an example. Cracks of nearly the same size co-
trast with the closed tail of an isolated crack alesce n times one after another (Fig. 10). Two
(Fig. 6). It prevents coalescence from the lower cracks of a volume M coalesce with each other so
isolated crack to the upper growing crack. The that the volume of a new crack becomes M X 2
condition for crack coalescence focuses primar- (first coalescence). Two cracks of a volume M X 2
 11. Accumulation of Magma in Space and Time 253
Small differential stress K, = 100 MPa . m 1/2 are needed (Eqs. (22) and
(23». Moreover, fluctuations of the crack size can
produce significant magma accumulations during
their ascent (Fig. 3C). It is more efficient that
cracks with various orientations (Fig. 2) or those
with a network of cracks coalesce.
Magma-filled cracks can coalesce and increase
in volume during their ascent (Takada, 1989) as
long as a range of crack sizes and ascent velocities
exists. Both the size and ascent speed are con-
trolled by the physical properties of the crust, such
as the density difference between the magma and
Input the host rocks, and the stress conditions. A larger
A crack at a lower velocity is formed as the den-
sity difference decreases. As basalt magma ap-
Large differential stress
proaches the depth where the density difference is

:.',' ,:t :,'::,'

.
. .:
. .
: :. ::.
. .. .
zero (the level of neutral buoyancy, e.g., Ryan,
1987a, 1987b; Rubin and Pollard, 1987; Takada,
1989; Lister and Kerr, 1991, and Ryan, 1994 (this

.f t':, t:, t
volume», the ascent velocity approaches zero .
The stress gradient with depth is proportional
to the apparent density. The stress gradient adjusts
, t t: , t:·, the crack propagation driven by buoyancy (Ta-
kada, 1989), especially the level of neutral buoy-
ancy (Lister and Kerr, 1991). A constant supply of
magma-filled cracks promotes the coalescence of
magma-filled cracks, that is, the magma accumu-
Input lation in time at this level.
B
Lister and Kerr (1991) proposed that, once a
Figure 10 Schematic diagram showing magma accumula-
magma-filled crack starts to rise, it never stops un-
tion in space. The upper diagram (A) represents the case of a
small differential stress; the lower diagram (B) represents the til it reaches the neutral buoyancy level. It is an
case of a large differential stress. In (A and B) the production ideal liquid-filled crack. The apparent resistance is
rate of magma is large enough for magma-filled cracks to in- generated due to the solidification of magma at a
teract. Magma-filled cracks are generated in the input block. crack tip or due to the ductile deformation at a
In the case of basalt magma, the input block means the as-
crack tip. This effect promotes the efficiency of
thenosphere. At a small magma production rate, the possibility
of crack interactions in space and time decreases. the magma accumulation during magma ascent in
the crust. It is possible that magma-filled cracks
coalesce to increase in volume during magma as-
coalesce so that the volume of a new crack be- cent (Takada, 1989).
comes M X 2 2 (second coalescence). The volume
of a magma-filled crack after the /lth coalescence Variation of a Volcano
amounts to
The arrangement of the principal stress axes and
M X 2". (23)
the value of the regional stress govern the magma
The stationary crack increases in volume so that accumulation in space. If (T1 is in the horizon-
the crack becomes unstable to propagate. Crack tal plane, a vertical crack isinitiated (Anderson,
coalescence occurs during crack propagation un- 1951). In this case, there are two types in the ar-
der the conditions envisioned in Fig. 9. If a magma rangement of the principal stress axes: One is that
volume is accumulated of the order of 10 8 m' (TI is in the vertical plane; the other is that (Tl is

(0.1 km '), 23 coalescences (2 21 times) for K; = in the horizontal plane (Fig. 11). The former in-
4 MPa . m 1/2 or 10 coalescences (2 10 times) for cludes the case in which a normal fault occurs; the
254 Akira Takada

(a) (b) (c)

0'1> 0'2> 0'3 0'1>0'2;:;0'3 0'1;:; 0'2;:;0'3
0'2
0'1
0'3~ ~ ~
/
O'~
ct0'3 0'1
0'3[[[]]J] [I]]]]] ~
I
0'1
10'2 , 0'3$////// $////U ~
I
/ /0'
_I- 3 0'2

~
7 I
/ 0'3.
~ ~
Figure II Schematic three-dimensional stress trajectories under various arrangements of the magnitudes of the principal stress
axes. The upper row shows IT, is in the vertical plane and IT, and IT., lie in the horizontal plane. The lower row shows IT, is in the
vertical plane and IT, and IT, are in the horizontal plane. (a-c) IT, - 'IT, is large (large differential stress). IT, - IT, is small, and
IT, - (T, is small (small differential stress), respectively.

latter includes the case in which a strike-slip fault vantage in the formation of a polygenetic volcano
occurs, Nakamura (1986) proposed that a mono- with parallel fissure eruptions (Fig. ,lOB).
genetic volcano occurs in the region with crustal Column (c) in Fig. II illustrates the case in
extension. The former is equivalent to the condi- which (J', - (J', is small. The differential stresses in
tion proposed by Nakamura (1986), that is, (J',lies the horizontal plane and in the vertical plane are
in the horizontal plane. On the other hand, Fedo- smallerthan (J', - (J',. The radial (J', trajectories are
tov (1981) proposed that the activity of monoge- formed in both planes. The coalescence of magma-
netic volcanoes such as cinder cones occurs at filled cracks is accelerated in the crust. This con-
a small magma supply rate. First, the case of a dition advances the formation of a polygenetic
magma supply rate sufficient for cracks to interact volcano with radial fissure eruptions (Fig. 10).
is discussed. Two parallel cracks are difficult to coalesce un-
Column (a) in Fig. II illustrates the case der regional stress with the normalized differential
(J', - (J'" that is, the differential stress is large and stress more than around 0.5. If the production or
both (J', - (J' 2 and (J' 2 - (J', are also large. Accord- supply rate of magma-filled cracks increases, sev-
ing to crack interaction theory, parallel stress tra- eral parallel cracks will interact with one another.
jectories develop in the horizontal plane and in the Such multi crack interactions tend to cancel the ef-
vertical plane. This condition prevents cracks from fects on the regional extensional stress, that is, the
coalescing. This condition also promotes the for- differential stress. The cracks can coalesce with
mation of a monogenetic volcano.(Fig. 10). one another. The increase in magma production or
Column (b) in Fig. II illustrates the case in the supply rate of magma-filled crack promotes
which (J'I - (J' 3 is large and (J' 2 - (J' 3 is small. the concentration of magma in space, that is, the
When (J', is in the vertical plane, vertical stress tra- formation of a magma path (Takada, in press).
jectories parallel to (J', develop. This condition This condition tends to promote the formation of
lends itself to the formation of a monogenetic vol- a polygenetic volcano. If the crustal extension is
cano (Fig. lOA). When (J', is in the horizontal large enough at a large magma production rate,
plane, the radial stress trajectories develop in the such as an oceanic ridge, a parallel dike swarm
vertical plane, and the parallel (J', trajectories de- develops and may grow into a magma reservoir in
velop in the horizontal plane. This case has an ad- the lower crust (Takada, 1989). On the other hand,
 II. Accumulation of Magma in Space and Time 255
for the formation of a monogenetic volcano with The interactions of isolated cracks may apply
primary magma, special conditions such as the in- to magma accumulation in the Earth's mantle and
crease in buoyancy by volatile components, or the in the crust. The generation of apparent fracture
imposition of a stress gradient such that the tensile toughness, crack coalescences with various ar-
stress increases upward, may be needed (Takada, rangements, and fluctuations on crack size causes
1989). At a small production or supply rate of an increase in the efficiency of the magma accu-
magma, the possibility of crack interactions in mulation process.
space and time decreases. A monogenetic volcano Magma accumulation due to the three-dimen-
will be formed whether the differential stress is sional crack interactions on various arrangements
large or small. The theory of crack interactions of the principal stress axes are considered, if the
can explain the case proposed by Fedotov (1981). production rate of magma is large enough for
The basic crack interaction concept has been magma-filled cracks to interact. When the differ-
discussed. However, these are ideal cases. A ential stress in the vertical plane is small, cracks
magma-filled crack with a small volume cannot tend to coalesce relatively easily with one another.
reach the surface because of heat losses and so- Magma concentrates by crack interactions in space
lidification. For a monogenetic volcano, magma- in its ascent process so that a magma path may be
filled cracks must coalesce with one another to in- formed. This case has an advantage in the forma-
crease in volume. In general, the differential stress tion of a polygenetic volcano. When the differen-
in the mantle and the lower crust is not large com- tial stress in the vertical plane is large, cracks are
pared with that of the upper crust. For a monoge- difficult to coalesce. This case advances the for-
netic volcano, magma-filled cracks may coalesce mation of a monogenetic volcano. On the other
in the mantle and the lower crust. hand. at a small production rate of magma-filled
cracks, a monogenetic volcano will be formed
whether the differential stress is large or small.
Conclusions

Crack interactions playa significant role in the pro- Acknowledgments

cess of magma accumulation in space and time.
Gelatin experiments on crack interactions with The author is grateful to H. Koide of the Geological Survey of
various geometric initial conditions suggest that Japan for valuable comments and critical readings of this
positive fluid buoyancy is effective for crack co- manuscript. The author has benefited from discussions with
K. Kurita and A. Shinya of Tsukuba University. The author
alescence. First, the interaction of collinear, re-
appreciates critical reviews by L. Wilson of University of Lan-
motely loaded cracks was evaluated by the stress caster. A. M. Rubin of Princeton University. and M. P. Ryan of
intensity factors at a crack tip. Second, the stress the U.S. Geological Survey.
fields near two parallel, offset, remotely loaded
cracks subjected to various regional stress fields
were calculated, using the the expansion of Wes- References
tergaard stress functions. The coalescence effects
of the height contrasts of cracks and of that on the Anderson. E. M. (1951). "The Dynamics of Faulting and Dyke
separation distance between two cracks are also Formation." 2nd ed .. Oliver and Boyd. London. England.
considered. Whether two cracks can coalesce or Atkinson. B. K. (1984). Subcritical crack growth in geologic
not depends on the height contrast of two neigh- materials, J. Geophys. Res. 89,4077-4114.
Dabovski, C. (1979). Stress and faulting around sheet-like
boring cracks, the horizontal separation distance
magmatic chambers. Geotect. Tectonophys. Geodyn. 8-9,
between two cracks. and the differential stress 17-38 (in Russian with English abstract).
field. The increase in the differential stress level Dabovski, C., and Stojanov, S. (1981). Fracturing around mag-
prevents cracks from interacting. A qualitative matic chambers as observed in laboratory models, Geo-
diagram of the crack height vs the horizontal dis- teet. Tectonophys. Geodyn. 13,3-18 (in Russian with En-
glish abstract).
tance between two cracks, indicating the condi-
Delaney, P. T.. and Pollard. D. D. (1981). Deformation of host
tions of crack coalescence. was proposed on the rocks and flow of magma during growth of minette dikes
basis of the results of gelatin experiments and and breccia-bearing intrusion near Ship Rock. New Mex-
stress analyses. ico, "U.S. Geol. Surv. Prof. Paper," 1202.
256 Akira Takada

Erdogan, E, and Sih, G. C. (1963). On the crack extension in T. L. Wright, and P. Stauffer, eds.), U.S. Geol. Surv. Prof.
plates under plane loading and transverse shear, 1. Basic Paper, 1350. pp. 1395-1447.
Eng. Trans. ASME Ser. D 35, 519-527. Ryan, M. P. (l987b). Neutral buoyancy and the mechanical
Fedotov, S. A. (1981). Magma rate in feeding conduits of dif- evolution of magmatic systems. in "Magmatic Process:
ferent volcanic centers. J. Volcanol. Geotherm. Res. 9, Physicochemical Principles" (B. O. Mysen, ed.), Geo-
379-394. chemical Soc. Spec. Publ. No. I, pp. 259-287.
Fiske, R., and Jackson, E. D. (1972). Orientation and growth Ryan, M. P. (1988). Mechanics and three-dimensional internal
of Hawaiian volcanic rifts: The effect of regional structure structure of active magmatic systems: Kilauea volcano.
and gravitational stresses, Proc. R. Soc. London. Ser. A Hawaii, J. Geophys. Res. 93,4213-4238.
329,299-326. Ryan. M. P. (1990). The physical nature of the Icelandic
Lange, E E (1968). Interaction between overlapping parallel magma transport system. ill "Magma Transport and Stor-
cracks: A photoelastic study, Int. J. Fracture Mech. 4, age" (M. P. Ryan, ed.), pp. 176-224. Wiley, Chichester/
287-294. Sussex, England.
Lawn, B. R., and Wilshaw, T. R. (1975). "Fracture of Brittle Ryan, M. P. (1994). Neutral buoyancy controlled magma
Solids," Cambridge Univ. Press, New York. transport and storage: A summary of basic relationships,
Lister, 1. R., and Kerr, R. C. (1991). Fluid-mechanical model in "Magmatic Systems" (M. P. Ryan. ed.), Academic
of crack propagation and their application to magma trans- Press, San Diego.
port in dykes, J. Geophys. Res. 96, 10049-10077. Ryan. M. P., Koyanagi, R. Y.. and Fiske. R. S. (1981) Model-
Maaloe, S. (1989). The generation and shape of feeder dykes ing the three-dimensional structure of macroscopic magma
from mantle sources, Contrib. Mineral. Petrol. 96, 47 -55. transport systems: Application to Kilauea volcano, Ha-
Nakamura, N. (1986). Volcanology and the plate tectonics. waii.1. Geophys. Res. 86, 7111-7129.
Bull. Volcanol. Soc. Japan Special number "What is a vol- Scott. D. R., and Stevenson. D. 1. (1984). Magma solitons,
cano," S I-S 16 (Japanese with English abstract). Geophys. Res. Lett. 11, 1161-1164.
Nicolas, A. (1986). A melt extraction model based on struc- Scott, D. R., Stevenson, D, J., and Whitehead. 1. A.. Jr. (1986).
tural studies in mantle peridotites, J. Petrol. 23, 568-582. Observation of solitary waves in a viscously deformable
Nicolas, A. (1990). Melt extraction from mantle peridotites: pipe, Nature 319, 759-761.
hydro fracturing and porous flow, with consequences for Shaw, II. R. (1980). The fracture mechanism of magma trans-
oceanic ridge activity, ill "Magma Transport and Storage" port from the mantle to the surface, in "Physics of Mag-
(M. P. Ryan, cd.), pp. 159-173. Wiley, Chichester/Sussex, matic Processes" (R. B. Hargraves. ed.), pp. 201-264.
England. Princeton Univ. Press. Princeton, N1.
Olson, 1., and Pollard, D. D. (1989). Inferring paleostress Sleep. N. H. (1988). Tapping of melt by veins and dikes,
from natural fracture pattern: A new method, Geology 17, J. Geophys. Res. 93, 10255-10272.
345-348. Spence, D. A.. and Turcotte. D. L. (1990). Buoyancy-driven
Pollard, D. D. (1973). Derivation and evaluation of a mechani- magma fracture: A mechanism for ascent through the
cal model for sheet intrusion, Tectonophysics 19, 233-269. lithosphere and the emplacement of diamonds, 1. Geo-
Pollard, D. D. (1987). Elementary fracture mechanics applied phvs. Res. 95,5133-5139.
to the structural interpretation of dykes, in "Mafic Dyke Stoyanov, S., and Dabovski, C. (1986). Morphology of frac-
Swarm" (II. C. Hall and W. E Fahrig, eds.), Geol, Assn. turing in zones of oblique extension: Experimental results
Canada Spec. Paper, 34. pp. 5-24. and geological implications, Geotect, Tectonophys. Geo-
Pollard, D. D., and Aydin, A. (1984). Propagation and linkage dyn. 19,3-22 (in Russian with English abstract).
of oceanic ridge segments, J. Geophys. Res. 89, 10017- Swain, M. V., and Hagan, 1. T. (1978). Some observations of
10028. overlapping interacting cracks, Eng. Fracture Mech. 10,
Pollard, D. D., and Segall, E (1987). The theoretical displace- 299-304.
ments and stresses near fractures in rock: With applica- Tada, H., Paris, P. C., and Irwin. G. R. (1973). "The Stress
tions to faults, joints, veins. dikes. and solution surfaces, Analysis of Cracks Handbook." Del Research Corp.. Hell-
in (B. K. Atkinson, ed.), Fracture Mechanics of Rock, ertown, PA.
pp. 277-349. Academic Press, London. Takada, A. (1989). Magma transport and reservoir formation
Pollard, D. D., Segall, E, and Delaney. E T. (1982). Formation by a system of propagating cracks, Bull. Volcanol. 52,
and interpretation of dilatant echelon cracks, Geol. Soc. 118-126.
Am. Bull. 93,1291-1303. Takada, A. (1990). Experimental study on propagation system
Roberts, J. L. (1970). The intrusion of magma into brittle of liquid-filled crack in gelatin: Shape and velocity in
rocks. in "Mechanism of Igneous Intrusion" (G. Newall. hydrostatic stress condition, J. Geophys. Res. 95, 8471-
and N. Rast, eds.), Geol. J. Spec. Issue, 2, 287-338. 8481.
Rubin, A. M., and Pollard, D. D. (1987). Origins of blade-like Takada, A. (1994). Development of a subvolcanic structure by
dikes in volcanic rift zones, in (R. W. Decker. T. L. Wright. the interaction of liquid-filled cracks, J. Volcanol. Geo-
and P. Stauffer, eds.), "Volcanism in Hawaii," U.S. Geol, thermo Res. (in press). .
Surv. Prof. Paper, 1350. pp. 1449-1470. Takeuchi, H., Fujii, N., and Kikuchi, M. (1972). How magma
Ryan, M. P. (l987a). The elasticity and contractancy of Ha- goes up? J. Seism. Soc. Japan 25, 266-268 (in Japanese).
waiian olivine tholeiite, and its role in the stability and Weertman,1. (l97Ia). Theory of water-filled crevasses in gla-
structural evolution of sub-caldera magma reservoirs and ciers applied to vertical magma transport beneath oceanic
rift systems, in "Volcanism in Hawaii" (R. W. Decker, ridge, J Geophys Res. 76, 1171-1183.
 II. Accumulation of Magma in Space and Time 257
Weertrnan, J. (1971b). Velocity at which liquid-filled cracks aileI elastic cracks and parallel slip bands respectively
move in the Earth's crust or in glaciers, J. Geophys. Res. based on the concept of continuous distribution of dislo-
76, 8544-8553. cations. II. Rep. Res.Inst. Strength Fracture Materials To-
Westergaard, H.· M. (1939). Bearing pressures and cracks. hoku Univ, 3, 15-37.
Trans. ASME J. Appl. Meeh. 6,49-53. Yokobori, T., Ohashi. M., and Ichikawa. M. (1965). The inter-
Whitehead. J. A. (1986). Buoyancy-driven instabilities of low- action of two collinear asymmetrical elastic cracks, Rep.
viscosity zone as models of magma-rich zone, J. Geophys. Res. Inst. Strength Fracture Materials Tohoku Unlv. 1,
Res. 91, 9303-9314. 33-39.
Whitehead, J. A.. and Helfrich. K. R. (1990). Magma waves Yokobori, T.. Uozurni, M.. and Ichikawa. M. (1971). Interac-
and diapiric dynamics, in "Magma Transport and Stor- tion between non-coplanar parallel staggered elastic cracks,
age" (M. P. Ryan. ed.), pp. 53-76, Wiley. Chichester/Sus- Rep. Res. Inst, Strength Fracture Materials Tohoku Univ.
sex. England. 7,25-47.
Yokobori, T.• and Ichikawa, M. (\967). The interaction of par-
This page intentionally left blank
Chapter 12 Generalized Upper Mantle Thermal Structure of the
Western United States and Its Relationship to Seismic
Attenuation, Heat Flow, Partial Melt, and Magma
Ascent and Emplacement
Hiroki Sato and Michael P. Ryan

Overview sistent with the temperatures estimated from the ve-

locity data. The thermal structures thus inferred from
The crust and upper mantle of the western United States seismic velocities are generally consistent with those
are characterized by low seismic velocities and high from seismic anelasticity studies. However, fine veloc-
seismic attenuation values, as well as high heat flow and ity structures may determine small-scale low velocity
high electrical conductivity values, Generalized tem- anomalies (and therefore small-scale, high-temperature
perature distributions inferred for the region provide anomalies) that are not always resolved from the more
partial constraints for understanding these anomalous generalized seismic anelasticity analysis. Such regional
and apparently fundamental features. Laboratory seis- anomalies may correspond to discrete magma genera-
mic measurements of rocks from the upper mantle at tion sites and to local concentrations of upper mantle
high pressures and high temperatures are compared magma, ascending diapirs, and local areas of astheno-
with seismic observations in an effort to estimate the spheric upwelling. Peaks in partial melt content occur
overall thermal structure of this portion of the upper at =165 km (=3%) and at =145 km (=10%) beneath
mantle. We report these estimated temperature distri- the Intermountain (1M) and Western Margin (WM) re-
butions beneath the western United States and further gions, respectively. Elastic dislocation treatments of
estimate the laterally averaged degree of melting as a veins and deep dikes yield magma-filled fracture heights
function of depth. Beneath the eastern Rockies, the re- that range from 500 m to II km, suggesting that the
sulting upper mantle temperature is below the dry soli- magma ascent pathway is highly disconnected. The as-
dus, and no melting is generally expected. From the cent of positively buoyant melt batches across the Mo-
Rocky Mountains to the far west, anomalously low seis- horovicic discontinuity is treated as a composite mate-
mic velocities and low Q-in regions such as the Rio rial interface characterized by paired Young's modulus
Grande rift, the Basin and Range province and the Cas- ratios (E2/E,). The modulus defect induced by melt in
cade volcanoes-yield temperatures higher than the low aspect ratios in the semi-consolidated gabbro com-
dry solidus, and partial melting (to =10 vol.%) is in- plexes near the Moho produces pronounced crack-tip
ferred. The relatively high gradient temperature-depth enlargements for the magma-charged fractures that pen-
profiles of the western United States are higher than etrate this interface. These bulbous enlargements may
normal mantle geotherms and are comparable to the thus combine with the interface rheological contrasts to
oceanic and hot continental geotherms. We summarize locally inhibit further crack advance, modify the trans-
available heat flow data in the western United States port pathway, and serve as nuclei for deep sills.
and calculate conductive temperatures within the litho-
sphere. The temperatures inferred from surface heat Notation
flow values are consistent with those from laboratory Units
seismic data for a dry peridotite. I
B Medium I crack flank height m
Seismic velocity structures have permitted estima-
tions of anomalously high temperatures in magmatic E Isotropic Young's modulus MPa,OPa
regions. Beneath the Rio Grande rift, the Basin and E* Effective aggregate Young's modu- MPa,OPa
Range province, and the Salton Trough, velocity pertur- lus of microporous solid .
bations below - 6% indicate the presence of partial £, Young's modulus surrounding the MPa,OPa
melt and temperatures above the dry solidus. Relatively crack flanks (medium I)
high heat flow regions coincide with these low velocity E2 Young's modulus surrounding the MPa,OPa
regions, and have yielded conductive temperatures con- crack-tip (medium 2)

Magmatic Systems Copyright © 1994 by Academic Press, Inc.

Edited by M. P. Ryan 259 All rights of reproduction in any form reserved.
 260 Hiroki Sato and Michael P. Ryan

Units Units

E,IE , Bimodulus contrast dimensionless Dihedral angle degrees

AG Difference in the ill situ stress gradi- MPa· km " ' Grain boundary energy between ma- Nm I
ent and the magma pressure gradient trix grains
g Gravitational acceleration m· s Grain boundary energy between ma- Nm I

h Height above subjacent magma m trix and melt

along a completely interconnected
fluid pathway
h* Fracture serniheight m
H, Buoyancy-driven crack critical m
height Introduction
K* Effective aggregate bulk modulus of MPa, GPa
microporous solid Many geophysical surveys have indicated anoma-
P Uniform fracture magma pressure MPa lous features in the crust and upper mantle be-
Q Seismic quality factor dimensionless neath the western United States: low seismic ve-
locities (e.g., Holbrook, 1990; Hearn et al., 1991),
Qp Seismic quality factor of compres- dimensionless
sional waves high seismic wave attenuation (e.g., Patton and
Taylor, 1984; AI-Khatib and Mitchell, 1991), rela-
Qprn Seismic quality factor of compres- dimensionless
sional waves at solidus temperature tively high electrical conductivities (e.g., Wanna-
Q" Seismic quality factor of shear dimensionless
maker et al., 1989; Klein, 1991), high heat flow
waves values (e.g., Sass et al., 1981; Blackwell et al.,
T Temperature °c, K 1990), and a relatively thin crustal and litho-
spheric thickness (e.g. Soller et al., 1982; Bechtel
Tm Solidus temperature of dry -c, K
peridotite et al., 1990). Seismic studies have further indi-
cated the deep presence of regional magmatism as
TITm Homologous temperature dimensionless
inferred from observed low velocities, high wave
T; Potential temperature "C
attenuation, and the occurrence of wave reflec-
Ii Volume of melt per unit crack height m'lm
tions from magma reservoir margins (e.g., Rine-
Ii, Seismic velocity of the km . s hart and Sanford, 1981; Kissling, 1988; Iyer et al.,
asthenosphere
1990; Holbrook, 1990). From a geodynamic and
\', Seismic velocity of the lithosphere km ' s - I regional tectonic perspective, in the northern por-
\ om Seismic velocity at solidus km s ° I
tion of the western United States, the Juan de Fuca
temperature
and the Gorda plates are being subducted beneath
\ 0p Elastic compressional wave velocity km . s I
the Cascade volcanic system, whereas in the south-
\' Elastic shear wave velocity km· s I western portion, the subduction of the oceanic
W Magma fracture width m crust from the East Pacific Rise tectonically loads
Z Vertical Cartesian coordinate m the San Andreas fault. In addition, there is a re-
a = Bih" dimensionless gion of continental extension and shallow igneous
activity: the Basin and Range province. These fun-
f3 = \/0 - 71l"!0 - aI)
damental structural relationships are well-known
7} = Zih" dimensionless
characteristics of the region and provide geologic
J-L '" Effective aggregate shear modulus MPa, GPa contrasts with other cooler and more stable areas
of a microporous solid
of North America (e.g., Robertson, 1972; Der
J-L Elastic shear modulus MPa, GPa
et al., 1982; Gough, 1984; Nathenson and Guf-
IJ Poisson's ratio dimensionless fanti, 1988).
¢ Porosity dimensionless Temperature distributions determined in the
p; In situ country rock density /; . cm ' upper mantle beneath the western United States
p.; Melt density g . crn :" may provide important constraints for increasing
AIT H Difference in the horizontal cornpo- MPa
our understanding of the anomalous features out-
nents of boundary loads along the lined previously. Peridotites are considered the
crack semiheight dominant lithology of the upper mantle. The labo-
 12. Generalized Upper Mantle Thermal Structure 261
ratory seismic (acoustic) data for a dry peridotite temperatures at depth (e.g., Murase and Fuku-
at high pressures and high temperatures have been yama, 1980; Kampfmann and Berckhemer, 1985;
compared with in situ seismic observations, and Sato et al., 1989b). Detailed descriptions of the
the generalized temperatures of the upper mantle experimental techniques and laboratory measure-
have thus been estimated. In previous papers (Sato ments have been reported by Murase and Kushiro
and Sacks, 1989, 1990; Sato et al., 1989a), we (1979) and Sato et al. (I 989b). Here we briefly
have investigated the thermal structure of, for ex- describe the laboratory results.
ample, the oceanic upper mantle from seismic Seismic velocities in a dry peridotite (a spinel
anelasticity and velocity data. Temperatures so lherzolite) have been determined by Murase and
derived were consistent with the temperatures in- Kushiro (1979) and Murase and Fukuyama (1980)
ferred from surface heat flow data; therefore, labo- as a function of temperature (1000-1300°C) at
ratory seismic results have been demonstrated to confining pressures to 1 GPa (10 kbar). Since the
be applicable to the Earth. temperature range of their measurements extended
In this chapter we estimate the generalized above the solidus of dry peridotite, the effect of
temperature profile and the degree of partial melt partial melting on seismic velocities was exam-
in the upper mantle beneath the western United ined. The partial melt fraction of the sample was
States from published seismic velocity and an- also determined over the same pressure and tem-
elasticity data. These inferred temperatures are perature ranges by Murase and Fukuyama (1980),
then compared with conductive geotherms de- using the f3-track method described by Mysen
rived from heat flow. Our procedure thus yields andKushiro (1977) and thin-section-based petro-
temperature profiles from the surface to -300 km graphic observations.
depth beneath the western United States. In addi- The experimental results have shown that the
tion, there are many small-scale anomalies (low seismic velocities in partially molten peridotite
seismic velocity, low Q, and high heat flow) in re- decrease sharply with increasing temperature; i.e.,
gions such as the Cascade volcanoes (e.g., Hearn the velocity at hypersolidus temperatures may be
et al., 1991; Harris et al., 1991), the Rio Grande used to constrain reliable mantle temperatures.
rift (e.g., Carpenter and Sanford, 1985; Halder- Importantly, the pressure dependence of both the
man and Davis, 1991), the Long Valley caldera velocity and the partial melt fraction is e...sentially
(e.g., Ryall and Ryall, 1981; Dawson et al., 1990), accounted for by that of the solidus temperature
the Coso hot springs area (Reasenberg et al., of peridotite, i.e., the homologous temperature
1980; Young and Ward, 1980), and the Salton dependence (Sato et al., 1989b). The homolo-
Trough (Humphreys and Clayton, 1990). Abnor- gous temperature is a dimensionless number that
mal thermal structures are to be expected be- is formed from the ratio TiTm , where Tm is the
neath these regions. We have also estimated tem- solidus temperature at a given composition and
perature distributions in these anomalous regions. pressure. It expresses the dimensionless thermal
The temperatures so derived are then discussed relationship of a rock to its melting point. The
with previous tectonic and magmatic models de- laboratory velocities, V, normalized by the ve-
rived from seismic, petrologic, and geothermal locity, Vme at the solidus temperature, Tme for
studies. both compressional and shear waves determined
at 0.5 and 1 GPa plot as a single trend as a func-
tion of the homologous temperature T/Tm (T is
Laboratory Seismic Data absolute temperature in kelvins), as shown in
for Mantle Peridotite Fig. 1A. The melt fractions determined at 0.5 and
1 GPa also depend on a single homologous tem-
Seismic Velocity of Mantle Peridotite perature (Fig. 1B). These observations, therefore,
allow one to extrapolate the experimental results
In the laboratory, the seismic properties of a dry to higher pressures (greater depths), by simply
mantle peridotite have been determined at high knowing the solidus as a function of pressure. The
pressures and temperatures in order to provide a solidus of dry peridotite has been determined up
comparative basis with seismic observations from to 14 GPa (140 kbar and =400 km depth) by Ta-
regional surveys and to enable the estimation of kahashi (1986). The solidi, T"" used here were
 262 Hiroki Sato and Michael P. Ryan

A I I

1.1 I- -
° 0 ODe! •
1.0
0
• ~
-.0.
0.9 - •0· -
•
E
~
0.8 f-- • -

1.0 GPa P-wave •

•°
0
0.7 - 0.5 GPa • -

• 1.0 GPa S-wave °•

0.6 - 0 0.5 GPa
• -

- •
0.5 l- I DRY PERIDOTITE
I -

I I
0.8 0.9 1.0 1.1
TITm

B
25
PRESSURE

1.0 GPa
°
'0::R.
20
• 0.5 GPa

"0
Z
Z 15
0
~
o
<X:
a:
u,
f- lO
...J
W
~

OL---'-_---L-_-'--------"_-'-_---L-----"_-'-_---L-_=_ _-'-_-'--------"_-'-_...L.....J
0.8 0.9 1.0 1.1
T/T m

Figure I Diagrams for determining temperature and degree of partial melt from seismic velocity data. (A) Normalized velocity
as a function of the homologous temperature (T!T~,) in dry peridotite; solid (open) squares and open (solid) circles at 0.5 and
1.0 GPa. respectively, for compressional (shear) waves (original velocity data from Murase and Kuxhiro (1979) and Murase and
Fukuyama (1980)). Note that T and T., are given in kelvins in this figure. Within the experimental errors in the velocity and the
solidus determinations, both compressional and shear velocities have a single trend in this diagram. (B) Fraction of partial melt
as a function of the homologous temperature (TIT.,). The degree of partial melt determined at 0.5 (squares) and 1.0 GPa (circles)
again depends on the single homologous temperature (after Sato et 11/., 1985).
 12. Generalized Upper Mantle Thermal Structure 263
1180 and 1220°C (1450 and 1490 K) at 0.5 and Table 1
I GPa, respectively. Example Thermal Structure of a Low Velocity Zone"
The normalized velocity in Fig. IA has been
compared with the seismic velocity structure, i.e., Fraction of
Age Depth Temperature partial melt
the ratio of asthenospheric velocity, V., to the lith-
(Ma) (km) (OC) (vol%)
ospheric velocity, VL • Fig. IA shows that the tem-
perature dependence of velocity becomes smaller 0-5 30 1260
at lower temperatures and that the velocity ratio 45 1310 7"
approaches 1.05. The velocity increase at even 60 1370
lower temperatures seems to be fairly small. We 70 1410
may therefore define lithospheric velocity, VL , by 85 1470 5"
(Sato et al., 1989a)
100 1530
VLIVm = 1.05. (I)
"Constructed for the Iceland Plateau. after Sato et al. (l989a)
"Because of the limitedresolution in seismic surface wave studies,
This equation is equivalent to
the melt fraction is determined only for the average depth of the low
velocity zone.
(2)
We calculate Va/Vmfrom Eq. (2) and obtain an es- perature at this depth is estimated to be 1580 K
timate of TITm by using Fig. IA. Tm as a function (J 310°C). We also estimate temperatures at vari-
of depth is obtained from the solidus determined ous depths using the known TITm (= 1.04) and the
by Takahashi (1986) (cf. Fig. 8A). Therefore the Tm from Takahashi (1986). The thermal structure
temperature T is calculated as a function of depth. thus obtained is summarized in Table I. The ac-
For T> Tm , we employ Fig. IB to estimate the tual temperature and melt fraction could both be
volume fraction of partial melt. higher at 0 Ma than at 5 Ma in the asthenosphere.
Figure IA indicates that the velocity drops Because of the limited resolution in seismic sur-
by about 6% at subsolidus temperatures (0.85 < face wave studies, the temperature and partial melt
TITm < 1). This implies that some low velocities fraction in Table I show bulk average values. For
in the asthenosphere could be explained by sub- asthenosphere older than 5 Ma, the relative veloc-
solidus temperature increases without invoking ity drop in the low velocity zone is less than 6%
partial melting (Sato et al., 1989a). The more and no melting is expected in the region. These
gradual drop of the velocity at T < Tm (as con- results indicate that extensive partial melting ex-
trasted with T > Tm) does not allow one to esti- ists in the asthenosphere younger than 5 Ma, but
mate accurate temperatures within the subsolidus that melting is not necessarily required in astheno-
region. Only in the region where the melt fraction sphere older than 5 Ma under the slowly spreading
is not less than 2 vol%, may we constrain a ther- (=20 mm . yr - I) Iceland Plateau. More detailed
mal structure from Fig. I. descriptions of the temperature estimates from
We illustrate with some examples of thermal seismic velocities have been reported by Sato
structure determined from laboratory and seismic et al. (1989a).
velocities (Fig. I). For the Iceland Plateau, Evans
and Sacks (1979) determined the seismic velocity Seismic Anelasticity of Mantle Peridotite
structure (VL = 4.5 km . s I) from the inversion
of both Rayleigh and Love wave data. In the re- The thermal structure of the upper mantle has also
gion of 0-5 Ma asthenosphere, an anomalously been estimated from seismic anelasticity (Q 1)
low velocity (V. = 3.9 km . s I) is found at a data of mantle peridotite by Sato and Sacks
depth of 25-65 km. Using Eq. (2), VaIVL = 0.87 (1989). We briefly describe the laboratory results
(=3.9 km . s - \14.5 km . s I) and gives V.IVm = and the method of constructing a temperature-
0.91, which then yields TITm = 1.04 (Fig. IA) depth profile.
and the melt fraction 7 vol% (Fig. IB). An aver- Laboratory measurements show, importantly,
age depth for this low velocity zone is 45 km that the seismic attenuation of compressional
where Tm is 1520 K from high pressure experi- waves (Q p - I) again depends on the ratio of rock
ments by Takahashi (1986). Therefore the tem- temperature to the solidus temperature, that is, at-
264 Hiroki Sato and Michael P. Ryan

I I I T -r-r-r I I I I ,
'PRESSURE'l
-- -
10 f-
• 0.73 GPa -

••
· O.48GPJ
l·~OGPa

E
0..
a
•• •
Ii.
a
•••
••
• ••

I DRY PERIDOTIT~
.- #
•••

I t t I , t t t t ~

0.85 0.90 0.95 1.00 1.05 1.10

T/Tm

Figure 2 Normalized QI' as a function of the homologous temperature (TT.,) in dry peridotite. T and T'; are given in kelvins
in this figure. The data taken at three different pressures have a single trend within the experimental errors in the Q" and the
solidus determinations. This diagram is used to estimate the temperature from seismic anelasticity data. Data from Sato ct al.
(I989b).

tenuation shows a homologous temperature depen- sufficiently high temperatures (0.86 ~ Tl'I'; ~
dence (Fig. 2). We may thus extrapolate the labo- 1.05)-are used to estimate the thermal structure
ratory results to higher pressures by simply using of the asthenosphere. In the lithosphere where
the known solidus as a function of pressure. The temperatures are low (TlTm < 0.7-0.8), the
parameter Q pm (the Q p value at the solidus tem- physical mechanism of attenuation may differ
perature) is expressed as (Sato and Sacks. 1989) from that at higher temperatures. We therefore de-
termine the temperature from seismic Q at asthen-
(3)
ospheric depths. Observations of heat flow have
where P is pressure and Q <p and P<p are 3.5 and yielded conductive geotherms in the lithosphere
0.073 GPa, respectively. From the seismic Q struc- (e.g., Chapman and Pollack, 1(77). We discuss
ture, therefore, both Qp and Qpm are known. The later the generalized lithospheric geotherm from
ratio Qp/Qpm then yields an estimate of TITm from heat flow data.
Fig. 2. The value T; is obtained from melting ex- The comparison of the laboratory Qp results
periments by Takahashi (986) (cf. Fig. 8A), and (Fig. 2) with the in situ seismic Q structure has
thus the temperature T may be determined. If the provided estimates of mantle temperatures (Sato
quality factor of shear waves, Q" is determined and Sacks, 1989). In the asthenosphere beneath the
from seismic studies, we employ the assumption Iceland Plateau, for example, Chan et al. (1989)
Qp/Q, = 2.25 (assuming that the fundamental loss have determined Q, = 7 at 45 km depth in the
of seismic energy occurs not in compression but 0-5 Ma region and Q, = 50 at 65 km depth in the
in shear; e.g., Anderson et al., 1965) to make 5-10 Ma region. Employing Qp/Q, = 2.25 and
comparisons with the laboratory Qp. If the tem- then comparing with Fig. 2, we obtain a tempera-
perature so derived is higher than Tm e we may es- ture of 1290°C and 3 vol% melt in the 0- 5 Ma
timate the melt fraction by employing Fig. lB. asthenosphere .(Sato and Sacks, 1989). In 5-
The laboratory results (Fig. 2)-determined at 10 Ma asthenosphere, however, the temperature is
 12. Generalized Upper Mantle Thermal Structure 265
below the dry solidus (I 140°C), and no melting neath the western United States, detailed Q struc-
is expected. It has been shown (Sato and Sacks, tures have been reported (e.g., Patton and Taylor,
1989) that the temperature and the fraction of melt 1984; AI-Khatib and Mitchell, 1991; Halderman
determined from seismic anelasticity studies are and Davis, 1991). Patton and Taylor (1984), for
consistent with those from seismic velocity and example, showed that both the Rayleigh-wave and
heat flow data. More detailed descriptions of the the short-period Love-wave attenuation data are
temperature estimates from seismic Q have been consistent with a frequency-independent Q model,
reported by Sato and Sacks (1989). with low Q, (= 100) in the lower crust and with
the lowest values (Q, = 30) in the upper mantle
at a depth of 60 km beneath the Basin and Range
Generalized Thermal Structure province. In this region, Q is generally low and
beneath the Western United States from extensive partial melting in the upper mantle has
Seismic Anelasticity Data been indicated (e.g., Patton and Taylor, 1984;
Lay and Wallace, 1988). AI-Khatib and Mitchell
Upper Mantle Anelasticity Structure (1991) have also obtained low Q values in the
upper mantle starting at about 50- 70 km depth
We first describe generalized upper mantle temper- beneath the western United States (Fig. 3). The
ature profiles inferred from seismic Q data. Be- studies by AI-Khatib and Mitchell (1991) have de-

50 !tfjt===_

100

~
~
I 150
I-
a.
w
0 ER

200 -

250

300 L J....JL--f:====~l-. LL ---L -----.)

50 17 13 10

Figure 3 The Q structure for the Eastern Rockies (ER), the Intermountain (IM) region, and the Western Margin (WM) region.
Horizontal bars denote one standard deviation of the Q value (from AI-Khatib and Mitchell, 1991).
266 Hiraki Sato and Michael P. Ryan

termined the lateral variations of the upper mantle cade centers of Three Sisters and the Newberry
anelasticity structure in some detail from recorded volcano and then through the High Lava Plains of
surface waves. They divided the western United south central Oregon and north of the Railroad
States into three regions: the Eastern Rockies Point and Fish Creek volcanic centers of Nevada.
(ER), the Intermountain (1M) region, and the Transect LON-DUG begins just north of MI. SI.
Western Margin (WM) region. In each region, Helens and then passes beneath the Cascade vol-
they have determined the Q s structure to 300 km canic center of Mt. Adams and on to the western
depth. Their results are consistent with previous Snake River Plain. Transect LON -ALQ effectively
studies and show a good overall correlation with resamples the same volume as LON-DUG, but or-
the broad-scale tectonic and magmatic activity in thogonally cuts the Jemez volcanic zone as well.
the western United States. The highest Q, values The Jemez volcanic zone is defined by the NE-
are obtained in the Rocky Mountain region where SW trending San Carlos - Springerville - Que-
tectonic activity has died out, whereas the lowest mado - Zuni Bandera - MI. Taylor - Jemez -
upper mantle Qs values are obtained beneath the Brazos - Ocate and Raton volcanic centers of
western margin of the United States where the oc- Arizona, New Mexico, and southern Colorado
currence of tectonic activity is more recent. We (Figs. 4-6; Table 2). Transect TUC-LUB cuts
here employ the Q structure determined by AI- the Rio Grande rift at a high angle near the Jor-
Khatib and Mitchell (199 I) to estimate tempera-
ture-depth profiles.
Table 2
Centers of Quaternary Volcanism" in Relation to Their State,
Their Location in the Index Map of Figure 5, and the
Volcanic Selling
Principal Seismic Transects for Anclasticity Determinations
Seismic transect CO-BKS of AI-Khatib and
Arizona Nevada
Mitchell (1991) passes through southwestern Or-
GC Grand Canyon CD Carson Desert
egon and northern California, just west of the Cas- P Pinnacate FC Fish Creek
cades and then on beneath the Geysers-Clear S Springerville LC Lunar Crater
Lake volcanic center. Transect CO-GSC passes SA Sentinel-Arlington RP Railroad Point
through that portion of the Cascades that in- SB San Bernardino TM Timber Mountain
SC San Carlos Mexico
cludes Crater Lake and the Medicine Lake vol- ~ew
SF San Francisco A Albuquerque
cano, then on southward beneath the Long Valley,
California BR Brazos
Big Pine, Coso, and Lava Mountains volcanic AP Amboy-Pisgah CV Cerro Verde
regions of California. Both transects enclose a BP Big Pine CZ Carrizozo
slender wedge-shaped region that includes the CA Cima J Jemez Mountains
southern Cascades centers of Mt. Shasta and Mt. CH Cinder Hill JM Jornado del Muerto
CL Clear Lake MT Mount Taylor
Lassen. The relationship of these transects to the
CM Cargo Muchacho Mtns. o Ocate
distribution of upper Cenozoic volcanic centers is CO Coso PO Portrillo
given in Fig. 4, and their relationship to Quater- E\1 Eagle Mountains Q Quemado
nary and active volcanic centers is given in Fig. 5 G Gaffs R Raton
and in Table 2. The physiographic nature of the K Kearn Z Zuni-Bandera
L\1 Lava Mountains Ltah
crust above the transects is illustrated in Fig. 6.
LV Long Valley BD Black Rock Desert
Collectively these traverses include portions of SP Sonora Pass KA Kanab
the Cascades and Pacific Coast Ranges, the Sierra SS Salton Sea KO Kolob
Nevada, and a westernmost sliver of the Basin and T Truckee-Donner Pass SG SI. George
Range province. They arc here referred to as the U Ubehebe
Wyoming
Western Margin (WM) region. Colorado LH Lcucite Hills
The 1M region seismic transects of AI-Khatib AS Aspen YP Yellowstone Plateau
E Eagle
and Mitchell (1991) have great cumulative length
Idaho
and diagonally cut across the Cascades and then
B Blackfoot
cut the grain of the Basin and Range province.
Seismic transect CO-DUG passes beneath the Cas- "After Luedke and Smith (1991).
 12. Generalized Lpper Mantle Thermal Structure 267

\
\

~.
~
I. .A
'§
,
I

I . ... "~,'

~r_ .

·: ~ ~ ' ~ . I , ~ · J ~.~
~ -
~

~
~~.
¥
'.

\d . -----
~~

.
AlQ
.

JeT

Figure 4 Seismic ray path distributions used for anelasticity evaluations in relationship to the distribution of upper Cenozoic
volcanic rocks erupted over the interval 0-5 Ma. Endpoints on the seismic transects are: EDM, Edmonton, Alberta, Canada;
JCT, Johnson City, Texas; LCB, Lubbock, Texas; TCC, Tuscan, Arizona; ALQ, Albuquerque. New Mexico; LON. Longview.
Washington; DUG, Dugdale, Utah; CO, Corvallis, Oregon; BKS, Berkeley, California; GSC, Goldstone, California. Seismic
station locations from Al-Khatib and Mitchell (1991). Volcanic eruptive distribution from Luedke and Smith (1991) as modified
from Smith and Luedke (1984).

nado del Muerto center north of Truth or Conse- cisco volcanic field of central. Arizona (Figs. 4
quences, New Mexico (Fig. 5). Finally, along and 5), Figure 7 illustrates the major physio-
transect DUG-TUC, the Black Rock Desert, Ko- graphic provinces of the western United States in
lob, and Kanab volcanic centers of southwestern relation to the digital shaded relief topography of
Utah are traversed, in addition to the San Fran- Thelin and Pike (1991).
 268 Hiroki Sato and Michael P. Ryan

EDM

---~---

e--,
--iJ"-
120°
r-

- r
- -.--------l
105°

I
\
~~ ~
~
';::::>
",,0

'LON
,
\
I
- -- (
'. I
45°..- •
I 1

r:
---, \

JeT

Figure 5 Distribution of volcanic centers active over the interval 0-1.6 Ma (solid pattern) in relationship to (a) the distribution
of volcanic activity over the period 0-5 Ma (dashed lines); and (h) the seismic transects used to infer the generalized anelasticity
structure of the deep crust and upper asthenosphere. Quaternary volcanic centers are listed hy state in Tahle 2 and are after
Luedke and Smith (1991). The seismic transects are based on the study of Al-Khatib and Mitchell (1991). Endpoint seismic
station codes are provided in Fig. 4.

Upper Mantle Temperature ture by Al-Khatib and Mitchell (1991) is at the

and Partial Melt Fraction depth 50- 70 km. Therefore, we determine tem-
peratures beneath that depth: Temperature uncer-
We compare the Q, structure of Fig. 3 with the tainties are also shown in Fig. SA, allowing for a
laboratory data of Fig. 2 by using the Q p/Q, ratio ± l51Jc error in the experimental measurements. If
(=2.25) described earlier, and obtain the general- one considers errors in the seismic Q structure, the
ized geotherms as shown in Fig. SA. The litho- temperature uncertainty would be larger. How-
sphere-asthenosphere boundary from the Q struc- ever, some of these errors are removed in the cali-
 12. Generalized Upper Mantle Thermal Structure 269

Figure 6 Digital shaded relief topography of the western United States in relation to the principal seismic transects for the
anelasticity study of the upper asthenosphere. Seismic station identities are given in the legend to Fig. 4. Relief topography from
Thelin and Pike (1991). '

bration process by using temperatures from other assessment of the uncertainty in the seismic Q is
data sources such as heat flow. This is discussed beyond the scope of this study. Errors in Tm may
later (see also Sato and Sacks. 1989). A complete also cause errors in estimated temperature. The
270 Hiroki Sato and Michael P. Ryan

Figure 7 Digital shaded relief topography of the western United States in relation to physiographic regions discussed in the
text. Region codings are: BR, Basin and Range; CA, Cascade Mountains; CO, Colorado Plateau: CP, Columbia Plateau; GP,
Great Plains; MR, Middle Rocky Mountains; NR, Northern Rocky Mountains; PE, Pacific Border; SN, Sierra Nevada: SR,
Southern Rocky Mountains; WB. Wyoming Basin. Physiographic provinces from Fenneman (1928). Relief topography from
Thelin and Pike (1991).
 12. Generalized Upper Mantle Thermal Structure 271
TEMPERATURE (OC)
600 1000 1400
I

I
l
I
I
100~ 100 r --- -----
I

l
E WM
~
J: J
~
Cl
I
I
I
I

~
200 200
(Kmll
V

Melt Fraction, vol. %

10
\

t
II
I
\
\
1
i
I
\"
\ \
\
, \\ 1
\ \

1\ J
L-'--__---'-__--'-__~ _'___~ ____'_ ____'I
A
Figure 8 Thermal structure of the upper mantle beneath the western Lnited States from the seismic anelasticity structure.
(A) Temperature profile, (solid lines) for the Eastern Rockies (ER), the Intermountain (1M). and the Western Margin (WM)
regions. Horizontal bars denote uncertainties in temperature from ~ 15'/, errors in the laboratory Q data. Dashed and dotted
lines arc dry peridotite solidus from Takahashi (1980) and mantle adiabat (0.4°(' . km - "), respectively. Although the tempera-
ture, in the ER region are slightly less than the dry solidus. those in the 1M and W\1 regions cross the solidus and partial melt is
expected. (8) Partial melt fraction as a function of depth. A relatively large amount of melt (to 10 vol'I,) is expected in the
WM region, where the temperature i, above the solidus at ·120 200 km depth.

dry peridotite solidi determined by Takahashi and separates the convective adiahatic interior of the
Kushiro ( 1(83). Scarfe and Takahashi (1986), and upper mantle from the conductive lid above. The
Takahashi (1986), however, arc found to be con- mechanical boundary layer of the upper litho-
sistent (within ±20°C or so) despite some com- sphere contains the crust-mantle interface that
positional variations between their studies. may experience brittle elastic fracture and creep
An extrapolation of the conductive geotherrns rupture during the dike and sill emplacement epi-
upward in temperature until they intersect the sodes that underplate the crust.
mantle adiahat provides an estimator of litho- The Q structure by Al-Khatib and Mitchell
spheric thickness (c.g., Mckenzie and Bickle, (1991 ) is noted to be an average over a wide area
1988). Such estimators applied to the WM and of the western United States (Figs. 3 and 7). Fig-
1M geotherms of Fig. 8 yield lithospheric thick- ure 7 illustrates the major physiographic provin-
nesses of = 140 km and = 150 km, respectively, ces of the western United States in relation to the
These transitions, from lithosphere to astheno- digital shaded relict' topography of Thelin and
sphere, fall within the thermal boundary layer that Pike (1991). The 1M region, for example, includes
272 Hiroki Sato and Michael P. Ryan

the Basin and Range province and the Colorado tures, the melt fraction increases and the seismic
and Columbia Plateaus, where the local geol- velocity then drops rapidly with increasing tem-
ogy and the geophysical characteristics differ. The perature (cf, Fig. I). This may also cause a rapid de-
Q structure in the ER region is averaged along a crease of Q. We may not, therefore, be able to ex-
seismic ray path as long as ~3000 km. Therefore, trapolate our Q data to temperatures above 1.05Tm •
the temperatures in this study represent only broad A rapid Q decrease (that is expected at TITm >
and generalized averages. 1.05) gives lower temperature estimates (as the
Beneath about 120 km depth, the temperatures adiabat in the 1M and WM regions; Fig. 8A) than
derived are well above the dry peridotite solidus the temperatures determined from the extrapola-
in the 1M and WM regions, and therefore partial tion of experimental data. The scatter of Q values
melting is inferred. By using Fig. IB, the degree in the 1M and WM regions becomes larger be-
of this melting for the dry upper mantle is deter- neath about 150 km depth (Fig. 3). Fairly small
mined and is shown in Fig. 8B. A relatively large error bars in the ER region should give much
amount of melt (up to =10 vol%) is obtained be- more reliable temperature estimates than those in
neath the WM region. The partial melting is con- the 1M and WM regions. Here we simply employ
sistent with high heat flow values, high electrical the mantle adiabat at the greater depths derived
conductivities and the surface manifestation of from the temperature profile of the ER region
magmatic activity in the 1M and WM regions as (Fig.8A).
discussed later. On the other hand, beneath the ER
region, temperatures are slightly lower than the Comparison with Previous
solidus and no melting is generally required. Temperature Estimates
The lateral variations of Q yield lateral tem-
perature changes, indicating horizontal gradients Figure 9 compares the geotherms in this study
in T and an increasing upper mantle temperature with those from other geophysical or geological
and melt fraction from the east to west between data sources. Ito and Sato (1992) have given a
the Rocky Mountains and the Pacific coast. At model mantle geotherm that is determined by
65- and 160-km depths, there are about 190 and combining heat flow data, seismic anelasticity data
l30°C differences, respectively, between the ER and detailed phase diagrams of mantle minerals.
and 1M regions (Fig. 8A), which correspond to a Sclater et al. (1980) have used estimates of the
lateral temperature gradient of about 0.3°C . km -I. mean heat flux and the radioactive heat production
At 125 km depth, the temperature difference is in the crust to compute the temperature as a func-
about 90°C between the ER and 1M regions and tion of depth beneath the continents and oceans.
between the 1M and WM regions. These lateral A thermal history of the Rio Grande rift has
temperature variations are comparable to the adia- been numerically calculated by assuming a con-
batic gradient. The vertical temperature variation ductive cooling of the lithosphere after the cessa-
at shallower (50-80 km) depths is, however, tion of magmatic activity and rift extension at
much larger (10-15°C· km I), and is compara- about 25 Ma (Morgan et al., 1986). Phase equilib-
ble to the gradient of the conductive geotherm de- ria and compositions of mineral assemblages have
rived from surface heat flow, as described later (cf. been used to estimate equilibration temperatures
Fig. 11). and pressures in the upper mantle (e.g., Mercier
At greater depths, the geothermal gradient ap- and Carter, 1975; Finnerty and Boyd, 1984). In
proaches the mantle adiabat (~O.4°C . km 1, addition, Herrin (1972) and Griggs (1972) have
Fig. 8A). This adiabat coincides with the tempera- used constraints from surface heat flux, seismic
ture profile determined from the Q structure of the data, melting curves and thermal conduction mod-
ER region. However, some of the low Q values at els to estimate mantle geotherms. These previous
depths in the 1M and WM regions yield higher temperature estimates for the oceans and the hot
temperatures (i.e., TITm > 1.05) than the adia- continent are comparable to the high tempera-
bat, by simply extrapolating the laboratory data to tures found beneath the western United States as
higher temperatures. The laboratory Q measure- shown in Fig. 9. The geotherms inferred in this
ments, however, have been carried out at tempera- study (as well as others compared in Fig. 9) sug-
tures to T= 1.05Tm (Fig. 2). At higher tempera- gest higher potential temperatures, Tp , than have
 12. Generalized Upper Mantle Thermal Structure 273
TEMPERATURE (Oe)
1200 1600
I I

100 . . . <,.:::.:.:::.:.:.,.
E
~
. . :>.:::>\ . .
::I: 1&5..... ., ..
I-
0..
W
0

200
M::>\\0.. \ 0••
ER

\.' :.

300
H\ .... G

Figure 9 Geotherms of the upper mantle. Solid line; are temperatures in this study for the ER. 1.'\1, and WM regions. Dashed
lines are a model mantle geotherm (I&S) by Ito and Sato (1992), a Rio Grande rift geotherm (YOby Morgan et al. (1986), a hot
continental geotherm (S) by Sclater et al. (1980), a high temperature oceanic geotherm (M&C) by Mercier and Carter (1975),
an oceanic geotherm (G&R) by Green and Ringwood (1907), a Basin and Range geotherm (H) by Herrin (1972). and a theoreti-
cal steady-state geotherrn (G) by Griggs (1972). The temperatures beneath the western United States in this study are comparable
with oceanic or hot continental geotherms,

been suggested by McKenzie (1984) and by Me- estimate the average heat flow values in the ER,
Kenzie and Bickle (1988). At the higher tem- 1M, and WM regions. The presence of Cascade
perature end, for example, McKenzie (1984) sug- magmatism as well as many geothermal areas
gests convective diapirs with core temperatures have produced a high heat flux in the region. The
of =1550°C rising through ambient surrounding heat flow map of the United States of Sass et al.
mantle with a temperature of = 1350°C and pro- (1981) has shown values above 100 mW . m 2 in
ducing melt at P == 3.5 GPa, or about IIO-km the Cascade Range, the Rio Grande rift, and the
depth. The physical context considered is sub- northern part of the Basin and Range province.
oceanic mantle melting, a dry peridotite solidus The high heat flow in the Rio Grande rift averages
(Takahashi and Kushiro, 1983), and the very low ~95 mW . m 2 south of Socorro and is 77 mW .
mantle retention of the melts produced, i.e., vol- m 2 to the north of Socorro (Reiter et al.. 1986).
umes of basalt observed at the surface are used as Blackwell et al. (1982, 1990) have measured the
a constraint on the melting process. heat flow of the Cascade Range, and show a major
change from 40 mW . m ~2 in the Western Cas-
Surface Heat Flow in the Western United States cades to 100 mW . m -2in the High Cascades. The
heat flow averages 88 mW . m 2 in the High
We here briefly describe the heat flow measure- Cascades and 100 mW . m -2 for the Western
ments reported in the western United States and Cascade-to-High Cascade anomaly. Mean heat
274 Hiroki Sato and Michael P. Ryan

flow values in the Basin and Range and in the Si- (Blackwell et al., 1991) that about 75% of the area
erra Nevada provinces are 92 and 39 mW . m 2, of the Cordillera is characterized by heat flow val-
respectively (Sass and Lachenbruch, 1978). Ex- ues in the range 70 to 90 mW· m -2.
tension may have been an important factor in pro- The regional heat flow of the western United
ducing or maintaining the high heat flow in the States in relation to the digital shaded relieftopog-
Basin and Range province (Morgan and Gosnold, raphy is shown in Fig. 10. The ER region consists
1989). On the other hand, a deep heat sink caused mainly of Mesozoic orogenic belts, where an av-
by downward convection of heat associated with erage heat flow is 65 mW . m -2. The heat flow
subduction has been indicated for the relatively is high in the Basin and Range province (86-
low heat flow in the Sierra Nevada (Roy et al., 92 mW . m -2: Sass and Lachenbruch, 1978;
1972). In the California Coast Ranges, the sub- Blackwell et al., 1978) and in the Rio Grande rift
ducting slab has been cut off east of the San An- zone (95 mW· m- 2: Reiter et al.. 1986). Low
dreas fault, allowing the growth of a "slabless heat flows occur, however, in the Columbia Pla-
window" (e.g., Dickinson and Snyder, 1979; La- teau (62 mW . m -2: Blackwell et al., 1990) and
chenbruch and Sass, 1980). The relatively high in the Colorado Plateau (~60 mW . m 2: Gos-
heat flow (-80 mW . m -2) in this region is there- nold, 1990; and -68 mW . m -2: Morgan and
fore thought to be related to a "hole" in the litho- Gosnold, 1989). An average heat flux could be
sphere beneath which hot asthenosphere rises, thus about 80 mW . m -2 for the seismic ray path
resulting in an increase in surface heat flow by distributions in the Basin and Range, the Colum-
a factor of 2 within 4 Myr of the passage of bia Plateau, and the Colorado Plateau of the 1M
the Mendocino triple junction (Lachenbruch and region, where many ray paths pass through the
Sass, 1980; Zandt and Furlong, 1982). Although Basin and Range province (Figs. 6 and 7). The
heat flow is generally low (49 mW . m -2) in the WM region includes the Pacific coast, the Cas-
Precambrian platforms, the Mesozoic orogenic cades, the Sierra Nevada, and the western part of
belts including the Pacific coast and the Rocky the Basin and Range province. Although heat flow
Mountains show a moderate value of 65 mW . values are only about 40-54 mW· m -2in the Si-
m - 2 (Sclater and Francheteau, 1970). A mean erra Nevada (e.g., Sass and Lachenbruch, 1978;
heat flow of 76 mW . m - 2has been reported in the Sass et al., 1981; Morgan and Gosnold, 1989) and
western United States by Sass et al. (1971) and by in the Coast Range west of the Cascades (Black-
Nathenson and Guffanti (1988). A comprehensive well et al., 1982, 1990), other areas produce high
summary of heat flow data of the United States by heat flows. Examples of such high thermal fluxes
Morgan and Gosnold (1989) has reported 68 mW . are an average 75 mW . m - 2 in the southern
m 2 in the Colorado Plateau, 86 mW . m - 2in the Washington Cascades, and range from -80 to
Basin and Range province, 101 mW . m 2in the > 100 mW· m -2 in the High Cascades (Blackwell
Cascade Range, 54 mW . m -2 in the Sierra Ne- et al., 1982, 1990). Heat flows are generally higher
vada, and 141 mW . m -2 in the Salton Trough. than 60 mW . m -2 along the Pacific coast of Cali-
The preceding heat flow values have also been fornia (Sass et al., 1971, 1981). An average heat
described clearly in a heat flow map of the west- flow could be thus about 75 mW . m -2 for the
ern United States constructed by Blackwell et al. seismic ray path distributions in the WM region.
(1991). Their map shows values below 60 mW .
m -2 in the Sierra Nevada, the northern Pacific Comparison with Heat Flow
coast, the Great Plains, the Wyoming Basin, and Temperature-Depth Profiles
parts of the Columbia and Colorado Plateaus. A
relatively high heat flow region (80-100 mW . We next estimate temperatures from the surface
m 2) covers the Basin and Range province, the Rio heat flow and compare them with the geotherm of
Grande rift, and the southern Rockies. Heat flow this study. Even though there are some uncertain-
values above 100 mW . m -2 occur in the High ties in the seismic anelasticity structure, experi-
Cascades, the Snake River Plain, the Yellowstone mental data, or the chemical composition of the
area, the Battle Mountain high, the Salton Trough, upper mantle, geotherms may be well constrained
and the southern Rockies. It has been pointed out by such a comparison (cf. Sato and Sacks, 1989).
 12. Generalized Upper Mantle Thermal Structure 275

Figure 10 Regional heat flow in relation to the digital shaded relief topography of the western United States. Heat flow contours
are from Morgan and Gosnold (1989). Relief topography from Thelin and Pike (1991).

Conductive geotherms for 65,75, and 80 mW· k I throughout the crust and a temperature-
m -2 are shown in Fig. 11, following the calcu- dependent conductivity above 500°C based on ex-
lations by Chapman and Pollack (1977). They perimental data by Schatz and Simmons (1972).
have used a thermal conductivity of 2.5 W . mI. Their model comprises an upper crustal region en-
276 Hiraki Sato and Michael P. Ryan

TEMPERATURE (OC)
400 800 1200 1600
I I I i

WM

200

I -----_ from heat flO~

300

Figure 11 Comparison with geotherms derived from heat flow. The temperature profiles in this study for the ER, 1M, and WM
regions (solid lines) are consistent with the conductive geotherms of65, 80, and 75 m'W> m - '(dashed lines). respectively. Since
laboratory anelasticity measurements were carried out for a dry peridotite. and a dry solidus was employed to estimate the
temperatures, this consistency suggests a generally dry upper mantle in the western United States (within the resolution of the
seismic anelasticity structures, i.e., excluding regions of local enrichment of volatiles). The temperature distributions indicate a
gradual change from a conductive to an adiabatic regime in the upper mantle. A dotted line (M) is a Rio Grande rift geotherm
by Morgan et al. (1986). A hatched area (L&S) is a Basin and Range geotherm by Lachenbruch and Sass (1977). (See also the
curves accounting for crustal extension labelled C and D for "characteristic Basin and Range" in Fig. 9- I8 of Lachenbruch and
Sass, 1978).

riched in radioactive sources, a granulite-facies the geotherms of this study. The temperature pro-
lower crust (heat production A of 0.25 jJ,W1m :1), file of the ER region also coincides with the
and a depleted ultrabasic zone (A = 0.01 jJ,W 1m 3 ) geotherm for the mean heat flow of 66 mW . m 2
overlying pyrolite mantle (A = 0.084 jJ,W 1m'). in the Great Plains calculated by Morgan and Gos-
Because the laboratory Q data measured at high nold (1989). A similar geotherm is suggested for
homologous temperatures (Fig. 2) are used to es- the heat flow of 68 mW . m -2 in the Colorado Pla-
timate the temperature of the asthenosphere, we teau region. Since experimental anelasticity mea-
compare it with the conductive geotherm at the surements were carried out for a dry peridotite,
top of the asthenosphere. The anelasticity struc- and a dry solidus was employed to estimate the
ture by Al-Khatib and Mitchell (1991) has shown temperatures, this consistency suggests that the
an increase of Q 1 with depth in the lithosphere upper mantle beneath the. western United States
and a decrease of Q - 1 at the lithosphere-astheno- may be generally dry (within the resolution of the
sphere boundary at 50-70 km depth. We there- seismic anelasticity structures). However, because
fore make temperature comparisons within this of the limited resolution of seismic anelasticity
depth range. As shown in Fig. II, the tempera- studies, volatiles localized in regions less than
tures inferred from heat flow are consistent with several tens of kilometers may not be resolved. In-
 12. Generalized Upper Mantle Thermal Structure 277
deed, a local enrichment of volatiles may certainly the limited resolution of the seismic tomographic
occur (e.g., beneath regional geothermal areas method, melting localized in less than a few tens
such as the Long Valley caldera, the Coso hot of kilometers may not be resolved from the ve-
springs, and the Cascade volcanoes). locity structures of Hearn et al. (199 I) and Harris
Large vertical temperature gradients at shallow et al. (199 I). We here report, therefore, bulk and
depths are also consistent with the steep tempera- averaged values of the partial melt fraction and
ture increases with depth as inferred by conduc- the temperature. Local enrichments of melt, how-
tion heat transfer (Fig. 11). The temperature gra- ever, are certainly expected to occur. A fairly high
dient becomes comparable to the mantle adiabat heat flow region (> 100 mW . m -2) is in fact a
beneath ~ 150-220 km depth. Thus the geotherm relatively localized occurrence (e.g., in the Cas-
derived from seismic Q data indicates a generally cade volcanoes, the Basin and Range province,
conductive regime of heat transfer at the top of the and the Rio Grande rift; Blackwell et al., 1982,
asthenosphere and a gradual change from a con- 1990; Lachenbruch et al., 1985).
ductive to an adiabatic regime with depth. AI-Khatib and Mitchell (1991) have determined
the shear velocity structure of the ER, 1M, and
WM regions in the western United States, as well
Generalized Thermal Structure as the anelasticity structure described earlier (d.
beneath the Western United States Fig. 3). In the ER region, the velocity (Va) de-
from Seismic Velocity Data creases from 4.6 km . s - I at 140 km depth to the
minimum 4.3 km- S' 1 at 240 km depth. Using VL
General Temperature Distributions = 4.55 km . s 1 in Eq. (2), Va = 4.3 km . s - 1 gives
near-solidus temperatures, and a state of widely
We now compare the laboratory velocity data distributed partial melt may not exist in the re-
(Fig. 1) with the seismic velocity structure, and gion. In the 1M and WM regions, however, the
estimate temperatures. Overall, the P n velocity low velocity zones show Va = 4.2 to 4.0 km . s 1,
structure of the western United States has shown and I to 6 vol% melt is expected (TlTm = 1.02-
a velocity variation from 7.6 to 8. I km . s I in the 1.04). The low velocity zone is located in the 90-
uppermost mantle (Hearn et al., 1991). Low P; to 150-km-depth interval in the WM region, and
velocities beneath the Yellowstone region and the the temperature is estimated to be about l580°C at
Basin and Range province have been interpreted 120 km depth. The melt fractions and tempera-
as resulting from anomalously hot upper mantle, tures determined here from the velocity structure
but those beneath the Sierra Nevada have been as- by Al-Khatib and Mitchell (1991) are comparable
cribed to a deep crustal root. A normal continental with those determined from the anelasticity struc-
lithosphere has an average Pn velocity of 8. I km . ture (Fig. 8).
s I (e.g., Mooney and Braile, 1989). We therefore
employ VI = 8.1 km . s - I in Eq. (2). The lowest Regional Temperature Distributions
P n velocity of 7.6 km . s - 1 (nearly a 6% velocity
drop) yields Va/Vm = 0.99 from Eq. (2). This in- A detailed model of the Basin and Range velocity
dicates that only a minor amount of melt may exist structure by seismic array studies shows a small-
in the region, and thus, the temperature is ex- scale low velocity zone (less than a 5-km-thick Vp
pected to be slightly above the solidus (about = 7.5 km . s - I zone) in the uppermost mantle be-
1250°C at 40 km depth). In the Juan de Fuca plate neath northwestern Nevada (Holbrook, 1990). Us-
subduction zone, Harris et al. (1991) have deter- ing VL = 8.1 km· S-I and Eq. (2), this low ve-
mined a dipping high velocity anomaly that cor- locity suggests = I vol% melt and a temperature
responds to the subducting plate, and low velocity of 1260°C. A fairly low P-wave velocity (7.4 km .
anomalies to 6% in the mantle wedge beneath the s - I) has also been determined beneath the Salton
Cascade volcanoes. The 6% velocity drop again Trough from a relatively high ;esolution velocity
suggests the presence of minor amounts of melt. structure (a block size of 15 km width and 30 km
The amount of melt, however, could be larger in depth) by using the southern California seismic
the cores of these low velocity regions, where the array (Humphreys and Clayton, 1990). We infer
velocity may be less than 7.6 km s I. Because of up to 3-4 vol% melt in the core of this low ve-
278 Hiroki Sato and Michael P. Ryan

locity region (less than 50 km width), and the cor- or geothermal root regions as resolved by fine-
responding temperatures are up to 1300, 1410, scale velocity inversions may have higher core
and 1520°C at 45-, 75-, and 105-km depths, re- temperatures than the bulk averages resolved by
spectively. A similar amount of partial melt has anelasticity studies. The equilibration tempera-
been estimated by Humphreys and Hager (1990), tures of mantle xenoliths as calculated from ele-
although they have used the numerical calcula- ment partitioning geothermometers have also
tions of Mavko (1980) to provide this estimate. shown relatively high temperatures (e.g., 860-
In the asthenosphere beneath the Coast Ranges 980°C at 30-40 km depth and about I 100°C at
of northern California, shear velocity studies by -60 km depth beneath southern British Colum-
Levander and Kovach (1990) have determined bia; Brearley et al., 1984; Canil et al., 1987; com-
VL = 4.55 km· s land Va = 4.1 km· S-I. This pare with Fig. 8A.).
yields 3 vol% melt, and T = 1280 and 1360°C
at 40- and 60-km depths, respectively. Lateral A Remark on Melt Retention and the Effective
P-wave velocity variations determined in detail in Melt Content of a Source Region
northern California from 9383 travel time residu-
als by Benz et al. (1992) showed ( - 6%) low ve- The suggestion that melt fractions may approach
locity regions, indicative of magma bodies within = 10% beneath the WM region and may be near
the crust beneath the active volcanoes. They also =3% below the 1M region deserves comment in
imaged the steep 70° east-dipping Gorda plate to light of the tendency for low viscosity basaltic
270 km depth as having a + 5% high velocity melts to wet grain-edge intersections and to thus
anomaly, and the shallow asthenosphere at 30- escape the deforming matrix. Using the liquid-
100 km depth beneath the northern Coast Ranges phase sintering results of Beere (1975) and the
as having an average of - 4% low velocity. In two-phase flow modeling approach of Drew (e.g.,
the core of these low velocity regions (less than 1983) and Didwania and Homsy (e.g., 1981),
-40 km wide), one may expect melting (up to McKenzie (1984, 1985) has treated the matrix
4 vol%) and temperatures to 1320°C (TlTm = compaction-melt expulsion problem with appli-
1.03) at a depth of 50 km. A few percent partial cation to decompression melting. From the per-
melt and temperatures higher than the mantle soli- spective of matrix melt retention, a critical rela-
dus are also estimated from the 8% velocity drop tionship is
and the low Q (Q p = 57) in the asthenosphere
(4)
beneath the Rio Grande rift (Halderman and Da-
vis, 1991). From Eq. (2), our estimates for the 8% where (j is the dihedral angle defined by the melt-
drop (Va/VL = 0.92) are Tl'Ii; = 1.02 and 2 vol% crystal interfaces, ')'" is the interfacial free energy
melting. The asthenospheric Qp of 57 yields com- of the crystal-crystal contacts and ')',1' is the inter-
parable values as TIT", = 1.01 and 0.5 vol% melt- facial free energy of the melt-crystal contacts.
ing from Fig. 2. Beneath the east African rift zone, Importantly, the value of (j controls the geometry
by comparison, Halderman and Davis (1991) de- and stability of the pores and if (j < 60, the pore
termined a 12% velocity drop, yielding T/Tm = space at all grain edge intersections remains in-
1.04 and about 5 vol% melt. The corresponding terconnected for all values of the porosity <p. If,
temperatures are 1500 and 1530°C at 100 km however, due to mineralogical and pore fluid het-
depth beneath the Rio Grande and east African rift erogeneities, (j > 60°, then there is a threshold <p
zones, respectively. In every geothermal area, one required for the establishment of interconnected
expects partial melting whatever the details of its melt tubules in three dimensions. Results of ex-
three-dimensional distribution. periments on texture-equilibrated olivine-rich ba-
The temperatures derived here from small-scale saltic partial melts show typically 20° :5 (J :5 47°
low velocity anomalies at relatively shallow depths (Waff and Bulau, 1979, 1982, but Cheadle, 1989,
« 100 km) are generally higher than the geo- presents a more complete compilation) however
therms from seismic Q as shown in Fig. 8. This is dry melts in an OPX-rich environment may lead
consistent with the idea that small-scale magma to values of (j in excess of 60°.
generation sites, magma reservoirs, small diapirs, Fujii, Osamura, and Takahashi (1986) have
 12. Generalized Upper Mantle Thermal Structure 279
carried out experiments on sintered orthopyrox- averaged result of hoth grain-scale and vein +
ene + basalt mixtures in efforts to determine the dike-scale melt distributions.
dihedral angle e, and make inferences about the
relative interfacial free energies and the extent of Comparison with Heat Flow-Derived
interconnected permeable pathways at the grain Temperatures
scale. In melts of low water content, the high dihe-
dral angles between dry melts and orthopyroxene Temperatures determined from seismic velocity
(median e = 70°) suggest low degrees of grain data are again generally consistent with the tem-
edge wetting. This implies isolated melt pockets peratures derived from surface heat flow studies.
and a local disruption of the melt network, thus The low velocity regions in fact coincide spatially
inhibiting permeability. For water-saturated melts, with the high heat flow regions. The low velocity
however, the pyroxene-melt-pyroxene e is =52°, zone beneath the Salton Trough, for example, is
suggesting the restoration of complete fluid con- located beneath the Imperial Valley (Humphreys
nectivity (see also von Bargen and Waff, 1988). and Clayton, 1990), where the heat flow is above
Collectively, the dry melt results suggest that 100 mW· m zand averages about 140 mW· m 2
pyroxene-rich microdomains may contain melt (Lachenbruch et al., 1985). Such a high heat
fractions that are notably in excess of those ex- flow yields the notably high temperature of about
pected from applications of the McKenzie (1984, 1300°C at 45 km depth (Chapman and Pollack,
1985) treatment. 1977), which is consistent with our estimate. The
The magma generation regions suggested in northwestern Nevada region studies by Holbrook
Fig. 8 are expected to contain swarms of melt- (1990) and the Rio Grande rift area studied by
filled veins as well as deep dikes. These are the Halderman and Davis (1991) are also charac-
structures that play an important role in transport- terized by high regional heat flow levels above
ing magma out of the generation volume and up- 105 mW· m -Z(e.g., Sass et al., 1981; Blackwell
ward toward the Moho. In this sense they form a et al., 1991), which again give conductive tem-
bridgework or set of connecting links between the peratures (Chapman and Pollack, 1977) compara-
grain-scale melt flow regimes in the heart of the ble to this study. These results indicate that the
generation volume and the igneous structures that seismic velocity determined in the laboratory for
underplate the continental keel. They are not in- a dry peridotite is appropriate to the upper mantle
corporated in the McKenzie (1984, 1985) treat- lithology beneath.
ment but have been discussed by Sleep (1988) and If, however, local enrichments of water in vol-
have a fundamental role to play in upper mantle umes that are less than a few tens of kilometers
magmatism. As briefly sketched below, the sta- occur in these regions (not resolved from seismic
bility criteria for melt-pressurized cracks include velocity studies), the inferred temperature should
considerations of the elastic moduli of the host be lower than those determined from the dry soli-
rock and the in situ density contrasts between the dus, since water depresses the solidus of mantle
melt and matrix. Stability is also a sensitive func- rocks. In the case of water-saturated (or water-
tion of melt volume and the correlative crack undersaturated) upper mantle, the derived tem-
widths. For example. basaltic melt-filled cracks peratures would then be reduced by an amount up
embedded in peridotite that form veins less than to (or less than) about 250°C (Sato, 1992).
roughly 20 mm in width tend to be gravitation-
ally stable when their heights are less than about
480 m, and they will thus tend to remain in the On Discontinuous Magma Ascent and the
vicinity of the source region. If they do remain at Mechanics of Basaltic Underplating
depth, they will therefore contribute to the region-
ally averaged low Q values inferred seismically. Establishing the linkage between the generalized
These deep veins will thus add to the Q reductions depth intervals for partial melt in the WM and
induced by trapped grain-scale melt. It is then 1M regions (as illustrated in Fig. 8) and the basal-
expected that low Q values associated with re- tic and then the andesitic, dacitic, and rhyolitic
gions ofmelting upper mantle may he the volume- volcanic centers of Figs. 4 and 5 is fundamental
280 Hiroki Sato and Michael P. Ryan

for understanding the magmatism of the western recognizing that it is the finite strength of fluid-
United States. In turn, this requires constraining weakened mantle rocks at high temperatures that
the likely connections between the ascent of basic restricts the fully fluid-connected magma column
magmas and the mechanics of basaltic underplat- segments to finite heights.
ing, since basic melts pond at or near the base Employing the mixed boundary value solutions
of the continental crust prior to migration into of Muskhelishvilli (1953) and Mikhlin (1957),
the mid-crust or eventual eruption (e.g., Rivalenti Weertman (1971 a, 1971b) has used elastic dislo-
et al., 1981; Shervais, 1979; Hamilton, 1989), cation theory to determine the outward displace-
and playa substantial role in providing the heat ments of the fracture's walls as functions of the
sources for subsequent silicic magmatism. Ber- internal liquid pressure and the relevant material
gantz and Dawes (this volume) discuss aspects of properties. These studies have produced relation-
the fluid dynamics and the relative roles of convec- ships that relate the heights of individual crack
tion and conduction in lower crustal melt genera- segments to the rock strength-as reflected in the
tion events induced through basaltic underplating. shear modulus. The critical crack height, H c, may
This section reviews aspects of the discontinuous be defined as the maximum vertical extent of a
magma ascent pathways and the elastic fracture buoyancy-driven fluid-filled crack that is embed-
morphology as buoyancy-driven fractures rise ded in an elastic solid. In the Weertman (1971 a,
from the upper mantle and penetrate the crust- 1971b) relations,
mantle interface, nucleating melt-ponding epi-
H, = 2 [2 VI1- /1T (I - v)
sodes at the base of the crust. We leave for later, (5)
the important problems of andesitic, dacitic, and . g(Pi' - Pm)] 1/"
rhyolitic melt migration in higher portions of the where 11- is the shear modulus, l' is the Poisson's
continental crust-but remark that looking at as- ratio, V is the volume of magma per unit crack
pects of the mechanics of magma emplacement height, g is the gravitational acceleration constant,
across the Mohorovicic Discontinuity is a logical and Pi' and Pm are the in situ and magma densities,
first step in developing a physical process frame- respectively. As is evident in (5), low shear mod-
work for the generation of higher level magmas. uli and sizable in situ density contrasts promote
The process of upward melt migration begins the development of modest crack heights, as do
within the depth intervals indicated in Fig. 8B. modest or narrow dike and vein widths. When II
For the WM region this interval i'i centered at 2:: H c- the crack tip becomes unstable, and frac-
= 145-km depth, whereas for the 1M region it is ture at the leading (upper) edge induces upward
centered at = 165-km depth. The generalized melt melt flow into the newly created fracture open-
fractions indicated for the WM and 1M regions are ing. Finite rock strength and conservation of mass
up to about 10% and =3%, respectively, as indi- within the fluid then promote the withdrawal of
cated in Fig. 8B. Considerable local variations are magma from the crack tail, and the lower fracture
expected, as discussed below, both in terms of surfaces move toward each other and close. This
melt fractions and depth intervals. The mechanics process promotes an increment of magma ascent.
of vein and dike nucleation within the region of Vein and dike widths relevant to upper mantle
partial melting is discussed by Sleep (1988). melt migration are only roughly constrained at
present, but scale with the overall fracture height
Inherently Discontinuous and tend to be maximized as H ~ H,. Vein
Nature of the Ascent Path widths observed in the clinopyroxenite and basal-
tic glass infillings of fractures in the xenoliths of
One of the most fundamental attributes of the the Cima volcanic field, California (Fig. 5), are
magma ascent pathway is its disconnected nature. variable (Wilshire, 1990), but about 20 mm may
For basaltic and ultrabasic melts there is a second be taken as their approximate upper limit. Toward
fundamental attribute: it is a fluid-filled fracture. the other end of the width range are feldspathic
This section briefly reviews the basis for these as- and gabbroic veins and dikes observed in upper
sertions and outlines the process of magma migra- mantle peridotites. Typically these are in the 10-
tion from the regions of partial melt shown in to 20-cm range, but widths may also range up to
Fig. 8. We anticipate the results of this section by = 100 cm (e.g., Nicolas, 1986, 1990). Within a re-
 12. Generalized Upper Mantle Thermal Structure 281
constructed subcontinental setting, observations fraction, changes in the melt + crystal microtex-
of feldspathic veins and gashes in the spinellher- ture in partially consolidated portions of the gab-
zolite massif of Liguria are = 1- =3 em in width, bro complex (=the residual melt aspect ratios),
for example, while the Lanzo massif of north- and gradients in temperature. The transition zone
western Italy has clustered feldspathic veins of itself will be defined by the length and steepness
comparable width (Nicolas, 1986). The critical of these gradients, and by the dominant source of
fracture height for average vein widths of 20 mm the modulus defect.
is He = 480 m [as estimated by Weertman's Fountain (1976) and Fountain and Christensen
(1971a, 1971b) dislocation theory], whereas the (1989) review the compressional wave velocity
upper end of the dike widths (=95 em) produces structure of the upper mantle-lower crust transi-
a critical fracture height of He = 10.6 km. Hence, tion of the Ivrea-Verbano zone of northern Italy
about 11 km appears to be a rough upper limit to and beneath the western United States. In addi-
the (completely fluid-connected) vertical extent of tion, Furlong and Fountain (1986) have computed,
fracture-bound upward-migrating basaltic magma for example, the Vp (Z) for olivine gabbro and
beneath the western United States. Aspects of the quartz tholeiite compositions through the gabbro-
positively buoyant fluid ascent process have been garnet eclogite transition region. For all rock types
outlined by Secor and Pollard (1975) and by Pol- considered, an appreciable (=7.4 km- s -1:5 Vp:5
lard (1976) with application to geothermal fluids. =8.1 km . s 1) range in velocity occurs. Recol-
Lister (1990) and Lister and Kerr (1991) have lecting that the compressional wave velocity of
used similarity solutions to discuss the propaga- basaltic melt is 2.6 km . s 1, and the Voigt-Reuss-
tion of positively buoyant and neutrally buoyant Hill (VRH)-averaged Vp of olivine (Fo lOo ) is
fractures, and Takada (1990; and this volume) and 8.59 km . s -1 (Graham and Barsh, 1969), while
Heimpel and Olson (this volume) have used gela- clinopyroxene is 7.22 km . s -1 (Aleksandrov
tin modeling to study the mechanics of ascending et al., 1964) and Plagioclase (An S6 ) is 6.70 km .
fluid-filled cracks with application to the process s -1 (Aleksandrov and Ryzhova, 1962; Ryzhova,
of magma ascent. Nakashima (1993) provides a 1964), a consideration of weighted-average mix-
recent application of Weertman's (1971a, 1971b) ing model approaches to constructing polycrystal-
approach to the fracture-assisted migration of met- line-polyphase bulk estimates of Vp for the par-
amorphic fluids. tially molten margins of deep gabbroic intrusions
Considering the region of primary underplat- will not produce unique inversions of melt con-
ing to be a laterally extensive layered gabbro tent. In addition, the aggregate bulk (K*) and
complex that surmounts the peridotites of the up- shear (f.L*) moduli (and thus the bulk Vp and Vs)
permost mantle and is incrementally replenished are known to be sensitively dependent on melt as-
by the rising buoyancy-driven fractures discussed pect ratios (Walsh, 1969; Ryan, 1980). Despite
previously, permits a consideration of this region these general levels of uncertainty, dramatic re-
as a bimodulus elasticity transition zone. In this ductions in the aggregate Young's Modulus, E*,
section we ask if the concept is plausible. In the as a function of the melt percent and melt aspect
following sections we ask how it works mechani- ratio have been shown for melt-weakened basalt
cally, and if the information gained can lead to (Ryan, 1980), and broadly comparable weakening
greater insights into how regions of underplating is expected for semi-consolidated gabbro near the
can nucleate, and why rising fractures may have Mohorovicic discontinuity. Therefore both com-
difficulty penetrating the Moho. These sections position and melt fraction conspire with melt ge-
make use of a review and reapplication of the ometry to make the transition from the peridotites
work of Morita et al. (1988). Note in particular of the upper mantle to the gabbroic cumulates of
that the buoyancy contrasts imparted by vertical the lower crust a composite material interface.
gradients in in situ density are a fundamental part
'of the Moho penetration story, but are reserved for Penetration of the Mantle-Crust Transition Zone
a separate discussion and not treated here. by a Buoyancy-Driven Crack
Contributors to contrasting aggregate elastic
moduli on either side of the transition zone in- A rising buoyancy-driven crack does work on
clude changes in mineralogy, gradients in melt its environment in an effort to reduce the gravi-
282 Hiroki Sato and Michael P. Ryan

1.33
l-
I
C!) 2
LU
I
C/)
C/)
LU
-I
1
Z
o 1/2
C/)
Z 0.66 1/3
LU
~
o
MEDIUM 1
0.33

DIMENSIONLESS MAGMA
FRACTURE WIDTH
Figure 12 Dimensionless magma fracture width vs dimensionless fracture height (8/11*) for a buoyancy-driven crack pene-
trating a bimodulus interface. The interface is defined by the ratios of the Young's moduli (E ,fE I) above and below the transition
region (modified after Morita et al .. 1988. and Ryan 1993, published hy the American Geophysical Union).

tational potential energy induced through the In (6), E is the (isotropic) Young's modulus, ,)"G
magma/country rock density contrasts, This work is the difference between the magma fracture pres-
occurs through crack opening displacements, and sure gradient and the in situ stress gradient, h* is
in geologic sections. it correlates with the result- the fracture semiheight. 11 = Z/ h* is the dimen-
ing dike widths. For rising fractures that cross the sionless fracture semiheight, and \' is the Poisson's
mantle-crust transition zone, the bielastic inter- ratio. In Fig. 12, the isotropic case is shaded, and
face of Fig. 12 will be' approximately horizontal the nonisotropic cases are given by the nonshaded
and normal to the crack centerline. The outward profiles. These profiles are determined by the ra-
displacements along the fracture length determine tios of the transition zone elastic anisotropy (e.g.,
the cross-sectional crack profile, and thus the over- £2/£ I = 3). The end members of this grouping
all geometry. For given levels of magma pressure, are profiles -k and 3. Profile i- is produced by
the magnitudes of the displacements are a sensi- the crack-tip widening induced by softer material
tive function of local aggregate elastic moduli- above the transition. Profile 3, however, shows
for fixed remote boundary conditions. For the case relative crack-tip constriction, induced by the now
of elastic isotropy (£2/£ J = I) the fracture width higher modulus rock above. In Figs. 12, 13, and
is given by 14, the expressions for the isotropic cases are
given by expressions (6), (7), and (8), respec-
W = 4 (I - \'2) h*
tively, whereas the nonisotropic curves have been
E (6) computed by a finite element approach (Morita
. (/)"Gh*) '!l VI - 11 2 • et al., 1988).
2
 1.33 MEDIUM 2

:2:
L.U
(I)
(I)
(I) 3
L.U
....J
Z
o 2
(I) 0.66
Z
L.U
:2:
o
1/2
0.33

1/3

DIMENSIONLESS MAGMA
FRACTURE WIDTH
Figure 13 Dimensionless magma fracture width vs dimensionless semiheight (BIh*) for a constant fluid pressure-driven crack
expansion across a bimodulus interface. Curves describe the cross-sectional profile over the range of Young's modulus ratios: l
:s; (E'/£I) :s; 3 (modified after Morita et al., 1988).

1.33 MEDIUM 2
l-
I
(!J
L.U
I ,
:2:
L.U 1.00
(I)
(I)
(I)
L.U
....J
Z
0
(I)
0.66
z
L.U
:2:
0 MEDIUM 1
0.33 1/2

1/3

DIMENSIONLESS MAGMA
FRACTURE WIDTH
Figure 14 Dimensionless magma fracture width vs dimensionless semiheight (Blh*) for a constant fluid pressure crack with
differences in boundary loads in media I and 2 (modified after Morita et al.. 1988).
284 Hiroki Sato and Michael P. Ryan

Penetration of the Mantle-Crust Transition Zone Rubin (1993) computes dike aspect ratios for fluid-
by a Constant Fluid Pressure Crack pressured cracks embedded in isotropic Maxwell
viscoelastic media.
Magma pressure is a function of depth, h, beneath
the Earth's surface, P = pgh, and increases with
depth h, in surroundings of higher density. Over Penetration of a Material Transition Zone by
restricted vertical intervals, however, the magma a Fluid-Pressurized Crack with Differential
pressure within a fracture may be idealized as ap- Boundary Loading
proximately constant, and the parametric response
As magma-driven fractures rise through the upper-
of the crack wall displacements (the crack profile)
most mantle and penetrate the base of the crust,
can be viewed as a function of the extent of elastic
they may enter environments where substantially
anisotropy. Thus, if one focuses on cracks that
different gradients in the horizontal component of
penetrate and straddle the Moho, but have a yet
confining pressure, LlITH' exist. In many ways this
rather restricted vertical extent, the ratios of the
is expected, since the transition separates an envi-
elastic moduli above and below the crust-mantle
ronment dominated by the density of harzburgites
interface can be examined for their contributions
and less depleted peridotites (densities: p=3.2-
to crack-tip swelling.
3.3 g . em -3) from layered gabbros (densities:
The crack tip itself is embedded in rock with
p=2.8-2.9 g . em -1) and their associated cumu-
an aggregate Young's modulus E 2, whereas the
lates. If we idealize this transition as a step change
crack flanks are contained in material of modulus
(a 'reduction) in lateral confining pressure that is
E I and have a semiheight B. The fracture width
applied normal to the crack walls as one moves
for the special case of elastic isotropy (E 21E 1 = 1)
vertically upward across the transition, the rela-
is given by
tive roles of elastic anisotropy can be explored in
4 (1 - v 2 ) h*P terms of influences on the crack-tip profiles. For
W = VI - YJ2, (7) the special case of elastic isotropy, the fracture
E
width is given by
where the approximately uniform magma pres-
sure is denoted by P, and the other symbols have 4 (1 - IJ2) h*LlIT H
W = [VI - YJ2
been given previously. In Fig. 13, the isotropic E
profile is shaded, whereas the anisotropic profiles - 3:. VI - YJ 2 sin - I Q' .
are not. For a given set of far-field boundary con- 1T (8)
ditions, fluid pressure and elastic moduli have
competing effects, as expected. Higher elastic +~(Logl~1
1T 13+1
moduli favor crack profile constriction, whereas
higher fluid pressures produce crack enlargements. _ !l Log 1 13 - YJIQ' I)],
Q' 13 + YJIQ'
For profile 1, magma rising from the higher modu-
lus upper mantle has penetrated the lower modu- where Q' = Blh* is the ratio of medium 1 (crack-
lus fluid-weakened rocks above, and significant flank) height to the total fracture semiheight, and
crack-tip widening has occurred. The widening 13 = V(1 - YJ)21 (1 - Q'2).Expressedinother
at the crack-tip region, if substantial, may pro- terms, LlIT H is the difference in the horizontal
mote the further enlargement of the fracture in the components of boundary loads in the crack-tip
higher modulus rocks below, but this will occur and crack-flank regions. In Fig. 14, the isotropic
only when there is complete mechanical bonding case has been shaded. For all profiles, the sudden
across the interface, that is, no lateral slippage. In reduction to lower values of IT H in the crack-tip
profile 3, the extreme crack-tip position is rela- region promotes some level of swelling just be-
tively pinched, whereas the crack flanks below fail hind the crack terminus. When these lower lev-
to achieve the dilation of profile 1 due to the as- els of crack-tip confining stress are coupled with
sumed material bonding across the transition zone. appreciably lower crack-tip elastic moduli (as il-
Note that the requirement of complete material lustrated by profile 1), the crack-tip region expe-
bonding across the interface prevents discontinui- riences considerable swelling throughout "me-
ties in lateral displacement (horizontal slippage). dium 2." However, as also illustrated by profile ],
 12. Generalized Upper Mantle Thermal Structure 285
the crack flanks beneath (in medium 1) experi- Range province suggests Moho temperatures close
ence significant reductions in width: a result of to or exceeding the lower crustal solidus (Morgan
the higher modulus ultramafic lithologies beneath. and Gosnold, 1989). High heat flow (> 100 mW .
The three crack-tip deformation modes illus- m 2) in the Battle Mountain high in northern Ne-
trated above suggest that the penetration of the vada and extending into the Snake River Plain of
semi-consolidated margins of a gabbroic body Idaho implies hypersolidus temperatures in the
may promote significant crack-tip widening. This, lower crust (e.g., Lachenbruch and Sass, 1978).
in tum, suggests potential sets of (i) loading con- Mantle-derived magmas appear to be the source
ditions; Uf) boundary conditions, and Uff) mate- of the high heat flow in the active arc volcanism
rial property mixtures that may ultimately stimu- of the Cascade Range (e.g., Morgan and Gosnold,
late still nucleation. It has been long appreciated 1989). From gravity gradient data and the width
that the in situ density reductions of the upper of the heat flow transition zone in the Cascades,
mantle-lower crust transitions promote the stag- Blackwell et al. (1982) have postulated the pres-
nation of rising magmas by reducing the rock-melt ence of a hot, low density region of about 60 km
density contrast. The crack-tip swelling evident in in width in the crust below a depth of 7 to 10 km,
this discussion suggests an additional mechanism suggesting, in tum, the potential for a zone of tem-
for modulating ascent. porary residence for magmas derived from the
subduction zone. This inference has been chal-
lenged by Blakely (1994) who shows that the
Closing Discussion source depth of the western Cascades gravity gra-
dient can not be deeper than 2.5 km. Magmatic
Geophysical and petrological data have suggested additions to the crust due to extension are sug-
the existence of partial melting beneath the west- gested for the source of the high heat flow values
ern United States. Shankland and Waff (1977), in the Salton Trough (Lachenbruch et al., 1985;
for example, have estimated 5-8 vol% melt for Morgan and Gosnold, 1989). High heat flow val-
the relatively high electrical conductivity of 0.2- ues in young calderas such as Yellowstone, Long
0.4 S . m - 1 beneath the western United States, us- Valley, and the Valles caldera have been thought
ing numerical conductivity calculations and the to be the consequences of subjacent mantle mag-
effective medium theory for a model rock consist- matism. Upward convective heat transfer into the
ing of a basaltic partial melt within a dominantly crust by mantle-derived magmas has commonly
olivine matrix. Their model melt fractions are been indicated from high heat flow geotherms in
comparable to our present results. The high elec- the western United States (e.g., Blackwell, 1978;
trical conductivities in the crust beneath the Basin Morgan and Gosnold, 1989).
and Range province (Klein, 1991) and the West- A high seismic attenuation (Q p = 40) in the
ern and High Cascades (1iracek et al., 1989; Stan- upper crust beneath northeastern Yellowstone in-
ley et al., 1990) have also suggested the presence dicates unusual hydrothermal activity, a shallow
of melts and/or hydrous solutions. From the ob- magmatic presence, and/or a relatively porous
servations of topography, regional seismology, steam-saturated body (Clawson et al., 1989). The
heat flow, electromagnetic structure, petrochem- presence of a magma chamber has been indi-
istry, and the stress orientations within the west- cated in the crust beneath Long Valley caldera,
ern United States, Gough (1984) concluded that California, from, among other evidence, the lack
there is widespread partial melting in the upper of S-wave transmission by shallow earthquakes
mantle (at a depth of roughly 150 km) with mantle (Ryall and Ryall, 1981) and from the relatively
upflow inferred beneath. Wilshire (1990) has re- low P-wave velocities (Dawson et al., 1990). Be-
ported extensive in situ partial melting in xenol- neath the Valles caldera, New Mexico, a region of
iths brought to the surface of the Basin and Range magma storage is indicated from relatively low
province. Partial melting has been commonly ob- P-wave velocities over the depth range 6 to 14 km
served along grain boundaries in mantle xenoliths (Roberts et al., 1991). Detailed P-wave velocity
from southern British Columbia (e.g., Brearley studies by Iyer et al. (1990) have shown a magma
et al., 1984; Canil et al., 1987). chamber (7% below normal velocities) of 10 km
Relatively high heat flow in the Basin and width extending from 7 to 28 km depth beneath
286 Hiraki Sata and Michael P. Ryan

the western half of the Long Valley caldera, and a terface. Significant crack-tip swelling for dikes
small magma chamber of a few cubic kilometers that penetrate this interface suggests that dike-tip
in the upper crust beneath the Newberry volcano, inflation processes may act as nuclei for sill initia-
Oregon. The lower crust of the southwestern Ba- tion in regions of basaltic underplating.
sin and Range province has low velocities, despite
the observation that lower crustal xenoliths have
mafic compositions and high densities (Wilshire, Acknowledgments
1990). This low velocity again indicates the pres-
ence of partial melt, high temperatures, and (fluid- The authors thank C. Herzig. I.S. Sacks, T. Watanabe, E. Ito,
filled) fracture systems. Low Q values indicating and E. Nakamura for valuable discussions on the subject of
magmatic activity in the crust have also been re- this chapter. The chapter benefited greatly from careful and
thoughtful reviews by David D. Blackwell (Southern Meth-
ported beneath the Rio Grande rift (Carpenter and
odist University), Arthur H. Lachenbruch, and John H. Sass
Sanford, 1985) and the Coso hot springs region (U.S. Geological Survey). This work was partially supported
(Young and Ward, 1980). by Grant 05231221 from the Ministry of Education, Science
Beneath the Cascade volcanoes, a high con- and Culture, Japan, and by the C.S. Geological Survey's Geo-
ductivity layer on the top of the descending Juan thermal Research Program. Robert Luedke helped with origi-
nal photobases for select figures. Mary Woodruff prepared the
de Fuca plate has been observed and lends itself
typescript.
to the inference of water-charged sediments that
are partially subducted with the oceanic crust (e.g.,
Wannamaker et al., 1989; Young and Kitchen,
References
1989; Kurtz et al.. 1990). Hydrous emanations
from the subducted sediments have been found to
Aleksandrov, K. S., and Ryzhova, T. V. (1962). Elastic prop-
be consistent with a vertical conductive zone from erties of rock-forming minerals, 3, Feldspars, Bull. Acad.
the slab to the mantle wedge at ~ 15-40 km depth Sci" USSR Geophys. Ser., 2, 1129.
beneath the Willamette Basin (Young and Kitchen, Aleksandrov, K. S., Ryzhova, T. V., and Belikov, B. P. (1964).
1989), and by a subhorizontal conductor from the The elastic properties of pyroxenes. SOl'.Phys-Crystallog.•
8, (5), 589-591.
Willamette Basin to the Cascade Range at about
Al-Khatib, H. H.. and Mitchell. B. 1. (1991). Upper mantle
30 km depth (Wannamaker et al., 1989). anelasticity and tectonic evolution of the western United
Materials that ascend adiabatically in the States from surface wave attenuation, J. Geophvs. Res. 96,
Earth's mantle have a potential temperature T; at 18129-18146.
the Earth's surface of about 1300°C for the normal Anderson. D. L., Menahem, A. B.• and Archambeau, C. B.
(1965). Attenuation of seismic energy in the upper mantle,
mantle (e.g., McKenzie and Bickle, 1988). T; of
J. Geophys. Res. 70, 1441-1448.
an upper mantle diapir is 200-300°C higher than Bechtel. T. D.• Forsyth, D. W, Sharpton. V. L., and Grieve,
this normal value (Courtney and White, 1986; R. A. F. (1990). Variations in effective elastic thickness of
Wyllie, 1988). This is compatible with our esti- the North American lithosphere, Nature 343, 636-638.
mate from Fig. 8A, which is T; = 1650°C, and is Beere, W (1975). A unifying theory of the stability of pene-
trating liquid phases and sintering pores. Acta Metal! .. 23,
1590°C if one uses an adiabatic gradient of 0.6°C .
131-138.
km ~ 1. Hot mantle upflow may be a possible cause Benz, H. M" Zandt, G., and Oppenheimer, D. H. (1992). Lith-
of the high temperatures and partial melt be- ospheric structure of northern California from teleseismic
neath the WM and 1M regions. The melt fractions images of the upper mantle, J. Geophys. Res. 97, 4791-
to ~ 10 vol% are also compatible with those of 4807.
Bergantz, G. W, and Dawes. R. (1994). Aspects of magma
an ascending diapir experiencing decompression
generation and ascent in continental lithosphere, in "Mag-
melting. The basaltic products of the supersolids matic Systems" (M. P. Ryan, ed.), Academic Press, San
intervals resolved in this study migrate toward the Diego.
Moho as positively buoyant magma-pressurized Blackwell, D. D. (1978). Heat flow and energy loss in the
fractures. In geologic sections, they form veins western United States, in "CeJ.1ozoic Tectonics and Re-
gional Geophysics of the Western Cordillera" (R. B. Smith
and deep dikes. Ascending magma fractures pene-
and G. P. Eaton, eds.), Memoir 152, pp. 175-208, Geol.
trate the Moho and enter the basal sections of Soc. Am., Boulder, Co.
semi-consolidated gabbroic intrusives that may be Blackwell, D. D., Steele, 1. L.. and Carter. L. S. (1991). Heat-
mechanically treated as a composite material in- flow patterns of the North American continent: A discus-
 12, Generalized Upper Mantle Thermal Structure 287
sion of the geothermal map of North America, in "Neotec- Evans, 1. R., and Sacks, I. S. (1979). Deep structure of the
tonics of North America" (D. B. Slemmons, E. R. Engdahl, Iceland Plateau, J. Geophys. Res. 84, 6859-6866.
M. D. Zoback, and D. D. Blackwell, eds.), Decade Map Fenneman, N.M. (1928). "Physical divisions of the United
Vol. I, pp. 423-436, Geol. Soc. Am., Boulder, Co. States," national atlas map, sheet 59, scale I: 17,000,000,
Blackwell, D. D., Hull, D. A., Bowen, R G., and Steele, 1. L. C.S. Geol. Surv., Washington, DC.
(1978). "Heat flow in Oregon," Oreg. Dep. Geol. Miner. Finnerty, A. A., and Boyd, F. R (1984). Evaluaton of thermo-
Ind. Spec. Paper 4. barometers for garnet peridotites, Geochim. Cosmochim.
Blackwell, D. D" Steele, 1. L., Kelly, S" and Korosec, M. A. Acta 48, 15-27.
(1990). Heat flow in the State of Washington and thermal Fountain, D. M. (1976). The Ivrea-Verbano and Strona-
conditions in the Cascade Range, 1. Geophys. Res. 95, Ceneri zones, Northern Italy: A cross-section of the conti-
19495-19516. nental crust-New evidence from seismic velocities of
Blackwell, D. D., Bowen, R. G., Hull, D. A., Riccio, 1., and rock samples, Tectonophysics, 33,145-165.
Steel, 1. L. (1982). Heat flow, arc volcanism, and subduc- Fountain, D. M., and Christensen, N. L (1989). Composition of
tion in northern Oregon, J. Geophys. Res. 87, 8735-8754. the continental crust and upper mantle: A review, in "Geo-
Blakely, R. 1. (1994) Extent of partial melting beneath the Cas- physical Framework of the Continental United States"
cade Range, Oregon: Constraints from gravity anomalies (L. C. Pakiser and W D. Mooney, eds.), Memoir 172, Geo-
and ideal-body theory. 1. Geophys. Res. 99, 2757 -2773. logical Society of America, Boulder, Co.
Brearley, M., Scarfe, C. M., and Fujii, T. (1984). The petrology Fujii, N., Osamura, K" and Takahashi, E. (1986). Effect of wa-
of ultramafic xenoliths from Summit Lake, near Prince ter saturation on the distribution of partial melt in the
George, British Columbia, Contrib. Mineral. Petrol. 88, olivine-pyroxene-plagioclase system. 1. Geophys. Res.,
53-63. 91,9253-9259.
Canil, D" Brearley, M., and Scarfe, C. M. (1987). Petrology Furlong, K. P., and Fountain, D.M. (1986). Continental crustal
of ultramafic xenoliths from Rayfield River, south-central underplating: Thermal considerations and seismic-petro-
British Columbia, Can. 1. Earth Sci. 24, 1679-1687. logic consequences, J. Geophys. Res. 91, 8285-8294.
Carpenter, P. 1., and Sanford, A. R. (1985). Apparent Q for Gosnold, W D. (1990). Heat flow in the Great Plains of the
upper crustal rocks of the central Rio Grande rift, J. Geo- United States, J. Geophys. Res. 95, 353-374.
phys. Res. 90, 8661-8674. Gough, D. I. (1984). Mantle upflow under North America and
Chan, W W, Sacks, I. S., and Morrow. R J. (1989). Anelastic- plate dynamics, Nature 311, 428-433.
ity of the Iceland Plateau from surface wave analysis, Graham, E. K., and Barsch, G. (1969). Elastic constants of
1. Geophys. Res. 94,5675-5688. single-crystal forsterite as a function of temperature and
Chapman, D. S., and Pollack, H. N. (1977). Regional geo- pressure,1. Geophys. Res. 74, 5949-5960.
therms and lithospheric thickness, Geology 5, 265-268. Green, D. H., and Ringwood, A. E. (1967). The stability fields
Cheadle, M. J. (1989), Properties of texturally equilibrated of aluminous pyroxene peridotite and garnet peridotite and
two-phase aggrevates. Ph.D. dissertation, University of their relevance in upper mantle structure, Earth Planet.
Cambridge. Sci. Lett. 3,151-160.
Clawson, S. Roo Smith, R. B., and Benz, H. M. (1989). P wave Griggs, D. T. (1972). The sinking lithosphere and the focal
attenuation of the Yellowstone caldera from three- mechanism of deep earthquakes, in "The Nature of the
dimensional inversion of spectral decay using explosion Solid Earth" (E. C. Robertson, ed.), pp. 361-384, Mc-
source seismic data, 1. Geophys. Res. 94,7205-7222. Graw-Hill, New York.
Courtney, R. C, and White, R. S. (1986). Anomalous heat flow Halderman, T. P., and Davis, P. M. (1991). Qp beneath the Rio
and geoid across the Cape Verde Rise: Evidence for dy- Grande and east African rift zones, J. Geophys. Res. 96,
namic support from a thermal plume in the mantle, Geo- 101l3-1O128.
phys. JR. Astron. Soc. 87, 815-867. Hamilton, W B. (1989). Crustal geologic processes of the
Dawson, P. B., Evans, 1. R., and Iyer, H. M. (1990). Teleseis- United States, in "Geophysical framework of the Conti-
mic tomography of the compressional wave velocity struc- nental United States" (L. C. Pakiser and W D. Mooney,
ture beneath the Long Valley region, California, 1. Geo- eds.), Memoir 172, Geological Society of America, Boul-
phys. Res. 95, 11021-11050. der, Co.
Der. Z. A., McElfresh. T. W, and O'Donnell, A. (1982). An Harris, R. A., Iyer, H. M., and Dawson, P. B. (1991). Imaging
investigation of the regional variations and frequency de- the Juan de Fuca plate beneath southern Oregon using tele-
pendence of anelastic attenuation in the mantle under the seismic P wave residuals, J. Geophys. Res. 96, 19879-
United States in the 0.5-4 Hz band, Geophvs.JR. Astron. 19889.
Soc. 69,67-99. Hearn, T., Beghoul, N" and Barazangi, M. (1991). Tomogra-
Dickinson, W R, and Snyder, W S. (1979). Geometry of sub- phy of the western United States from regional arrival
ducted slabs related to San Andreas transform, J. Geology times, J. Geophys. Res. 96, 16369-16381.
87,609-627. Heimpel, M., and Olson, P. (1994). Buoyancy-driven fracture
Didwania, A. K., and Homsy, G. M. (1981). How regimes and and magma transport through the lithosphere: Models and
flow transitions in liquid fluidized beds. Int. J. Multiphase experiments, in "Magmatic Systems" (M. P. Ryan, ed.),
Flow 7, 563-580. Academic Press, San Diego.
Drew, D. A. (1983). Mathematical Modeling of two-phase Herrin, E. (1972). A comparative study of upper mantle mod-
flow. Annu. Rev. Fluid Mech .. 15,261-291. els: Canadian shield and Basin and Range provinces, in
288 Hiroki Sato and Michael P. Ryan

"The Nature of the Solid Earth" (E. C. Robertson, ed.), Lay, T.. and Wallace. T. C. (1988!. Multiple ScS attenuation
pp. 216-231, McGraw-Hill, New York. and travel times beneath western North America. Bull.
Holbrook, W. S. (1990). The crustal structure of the north- Seismol. Soc. Am. 78, 2041-2061.
western Basin and Range province, Nevada, from wide- Levander. A. R., and Kovach, R. L. (1990). Shear velocity
angle seismic data, J. Geophys. Res. 95,21843-21869. structure of the northern California lithosphere, 1. Geo-
Hughes, S. S. (1990). Mafic magmatism and associated tecton- phys. Res. 95,19773-19784.
ism of the central High Cascade Range, Oregon, J. Geo- Lister. 1. R. (1990). Buoyancy-driven fluid fracture: Similarity
phys. Res. 95,19623-19638. solutions for the horizontal and vertical propagation of
Humphreys, E. D., and Clayton, R. W. (1990). Tomographic fluid-filled cracks,.I. Fluid Mech. 217. 213-239.
image of the southern California mantle, J. Geophys. Res. Lister. J. R., and Kerr, R. C. (1991). Fluid-mechanical models
95,19725-19746. of crack propagation and their application to magma trans-
Humphreys, E. D., and Hager, B. H. (1990). A kinematic model port in dikes. J. Geophys. Res. 96.10049-10077.
for the late Cenozoic development of southern California Luedke, R. G., and Smith, R. L. (1984). Map showing distri-
crust and upper mantle, J. Geophys. Res. 95, 19747- bution, composition, and age of late Cenozoic volcanic
19762. centers in the Western conterminous United States, U.S.
Ito, E.. and Sato, H. (1992). Effect of phase transformations on Geological Survey, MI Series, MAP 1-1523.
the dynamics of the descending slab, in "High-Pressure Luedke. R. G., and Smith. R. L. (1991). Quaternary volcanism
Research: Application to Earth and Planetary Sciences" in the western conterminous Lnited States in the Geology
(Y. Syono, and M. H. Manghnani, eds.), pp. 257-262. of North America, vo!. K-2, "Quaternary Nonglacial Ge-
Terra Sci. Pub!. Co., Tokyo. ology: Conterminous I;.S." (R. B. Morrison. ed.), pp. 75-
Iyer, H. M., Evans, 1. R., Dawson, P. B., Stauber, D. A., and 92, The Geological Society of America.
Achauer, C. (1990). Differences in magma storage in differ- Mauko, G. M. (1980). Velocity and attenuation in partially
ent volcanic environments as revealed by seismic tomog- molten rocks, J. Geophys. Res. 85, 5173-5189.
raphy: silicic volcanic centers and subduction-related vol- McKenzie, D. P. (1984). The generation and compaction of
canoes, in "Magma Transport and Storage" (M. P. Ryan. partially molten rock, J. Petrol. 25, 713-765.
ed.), pp. 293-316, Wiley, Chichester/Sussex, England. McKenzie, D. P. (1985). The extraction of magma from the
Jiracek, G. R., Curtis, 1. H., Ramirez, J.. Martinez. \1 .. and crust and mantle, Earth and Planet. Sci. Lett. 74, 81-91.
Romo,1. (1989). Two-dimensional magnetotelluric inver- McKenzie, D., and Bickle, M. 1. (1988). The volume and com-
sion of the EMSLAB Lincoln line, 1. Geophys. Res. 94, position of melt generated by extension of the lithosphere,
14145-14151. .I. Petrol. 29, 625-679.
Kampfmann, w.. and Berckhemer, H. (1985). High tempera- Mercier, 1.-C., and Carter, N. L. (]975). Pyroxene geotherms,
ture experiments on the elastic and anelastic behaviour of J. Geophys. Res. 80, 3349-3362.
magmatic rocks, Phys. Earth Planet. Inter. 40, 223-247. Mikhlin, S. G. (1957). "Integral equations" (A. H. Armstrong,
Kissling, E. (1988). Geotomography with local earthquake transl.), Pergamon Press. New York.
data. Re\'. Geophys. 26, 659-698. Mooney. W. D., and Braile, L. W. (1989). The seismic struc-
Klein, D. P. (1991). Crustal resistivity structure from magne- ture of the continental crust and upper mantle of North
totelluric soundings in the Colorado Plateau-Basin and America. in "The Geology of North America-An Over-
Range provinces, central and western Arizona, 1. Geo- view" (A. W. Bally and A. R. Palmer, eds.), pp. 39-52.
phys. Res. 96,12313-12331. Geol, Soc. Am.. Boulder. Co.
Kurtz; R. D.. DeLaurier, 1. M., and Gupta. J. C. (1990). The Morgan, P., and Gosnold, W. D. (1989). Heat flow and thermal
electrical conductivity distribution beneath Vancouver Is- regimes in the continental United States, in "Geophysical
land: A region of active plate subduction, J. Geophys. Res. Framework of the Continental United States" (L. C. Pak-
95, 10929-10946. iser and W. D. Mooney, eds.), pp. 493-522, Memoir 172,
Lachenbruch. A. H.. and Sass, 1. H. (1977). Heat flow in Geo!. Soc. Am., Boulder, Co.
the United States and the thermal regime of the crust, in Morgan, P.. Seager. W. R., and Golombek, M. P. (1986). Ce-
"The Earth's Crust" (1. G. Heacock, ed.), pp. 626-675, nozoic thermal, mechanical and tectonic evolution of the
Geophys. Monogr. 20, Am. Geophys. Union. Washing- Rio Grande rift, J. Geophys. Res. 91. 6263-6276.
ton. DC. Morita, N.. Whitfill, D. L. and Wahl, H. A. (1988). Stress-
Lachenbruch, A. H.. and Sass, 1. H. (1978). Models of an ex- intensity factor and fracture cross-sectional shape predic-
tending lithosphere and heat flow in the Basin and Range tions from a three-dimensional model for hydraulically in-
province, in "Cenozoic Tectonics and Regional Geophys- duced fractures, J. Petrol. Technol. Oct., 1329-1342.
ics of the Western Cordillera "(R. B. Smith, and G. P. Ea- Murase, T., and Fukuyama, H. (1980). Shear wave velocity in
ton, eds.). pp. 209-250, Memoir 152, Geo!. Soc. Am., partially molten peridotite at high pressures, Carnegie
Boulder, Co. Inst. Wash. Yrhk. 79, 307-310.
Lachenbruch, A. H., and Sass, 1. H. (1980). Heat flow and en- Murase, T.. and Kushiro, r. (1979): Compressional wave ve-
ergetics of the San Andreas fault zone, J. Geophys. Res. locity in partially molten peridotite at high pressures, Car-
85,6185-6222. negie Inst. Wash. Yrbk. 78, 559-562.
Lachenbruch, A. H., Sass, 1. H., and Galanis, S. P., Jr. (1985). Muskhelishvilli. N.r. (1953). "Singular Integral Equations"
Heat flow in southernmost California and the origin of the (1. R. M. Radok, trans!.), Noordhoff, Groningen, The
Salton Trough, 1. Geophys. Res. 90, 6709-6736. Netherlands.
 12. Generalized Upper Mantle Thermal Structure 289
Mysen, B. 0., and Kushiro, I. (1977). Compositional varia- Ryzhova, TV, (1964). The elastic properties of plagioclase,
tions of coexisting phases with degree of melting of perid- Izv, Geophys. Srr. 7, 1049-1051.
otite in the upper mantle, Am. Mineral. 62, 843-865. Sass, 1. H.. and Lachenbruch, A. H. (1978). Thermal regime
Nakashima, Y. (1993). Buoyancy-driven propagation of an of the Australian continental crust, in "The Earth, Its Ori-
isolated fluid-filled crack in rock: Implication for fluid gin, Structure and Evolution" (M. W. McElhinny, ed.),
transport in metamorphism, Contrib. Mineral. Petrol. 114, pp. 301-351, Academic Press, New York.
289-295. Sass, 1. H., Lachenbruch, A. H., Munroe, R. J.. Greene, G. w.,
Nathenson, M.. and Guffanti, M. (1988). Geothermal gradi- and Moses, T. H., Jr. (1971). Heat flow in the western
ents in the conterminous United States, J. Geophvs. Res. Lnited States. J. Geophys. Res. 76, 6376-6413.
93, 6437 -6450. Sass, 1. H., Blackwell, D. D., Chapman. D. S.. Costain, 1. K.,
Nicolas, A. (1986). A melt extraction model based on structural Decker, E. R.. Lawver, L. A., and Swanberg, C. A. (1981).
studies in mantle peridotites, J. Petrol. 27, 999-1022. Heat flow from the crust of the United States, in "Physi-
Nicolas, A. (1990). Melt extraction from mantle peridotites: cal Properties of Rocks and Minerals" (Y. S. Touloukian,
Hydrofracturing and porous flow with consequences for W. R. Judd, and R. E Roy, eds.), pp. 503-548, McGraw-
oceanic ridge activity, in "Magma Transport and Storage" Hill, New York.
(M. P. Ryan, ed.), pp. 159-173, Wiley, Chichester/Sussex, Sato, H. (1992). Thermal structure of the mantle wedge be-
England. neath northeastern Japan: Magmatism in an island arc
Patton, H. 1.. and Taylor, S. R. (1984). Q structure of the Basin from the combined data of seismic anelasticity and ve-
and Range from surface waves, J. Geophys. Res. 89, locity and heat flow, J. Volcanol. Geotherm. Res. 51,
6929-6940. 237-252.
Pollack, H. N., and Chapman. D. S. (1977). On the regional Sato, H., and Sacks, I. S. (1989). Anelasticity and thermal
variation of heat flow, geotherms, and lithospheric thick- structure of the oceanic upper mantle: Temperature cali-
ness, Tectonophysics 38,279-296. bration with heat flow data, J. Geophys. Res. 94, 5705-
Pollard, D. D. (1976). On the form and stability of open hy- 5715.
draulic fractures in the Earth's crust, Geophys. Res. Lett. 3, Sato, H., and Sacks, I. S. (1990). Magma generation in the
513-516. upper mantle inferred from seismic measurements in pe-
Reasenberg, P., Ellsworth, w., and Walter, A. (1980). Teleseis- ridotite at high pressure and temperature, in "Magma
mic evidence for a low-velocity body under the Coso geo- Transport and Storage" (M. P. Ryan, ed.), p. 277-292,
thermal area, 1. Geophys. Res. 85, 2471-2483. Wiley. Chichester/Sussex, England.
Reiter, M.. Eggleston, R. E., Broadwell, B. R., and Minier, J. Sato, H., Sacks, I. S., and Murase, T. (I 989a). The use of labo-
(1986). Estimates of terrestrial heat flow from deep petro- ratory velocity data for estimating temperature and partial
leum tests along the Rio Grande rift in central and southern melt fraction in the low velocity zone: Comparison with
New Mexico, 1. Geophys. Res. 91, 6225-6245. heat flow and electrical conductivity studies, J. Geophys.
Rinehart, E. 1., and Sanford, A. R. (1981). Upper crustal struc- Res. 94, 5689-5704.
ture of the Rio Grande rift near Socorro, New Mexico, Sato, H., Sacks, I. S., Murase, T., and Scarfe, C. M. (1988).
from inversion of microearthquake S-wave reflections, Thermal structure of the low velocity zone derived from
Bull. Scismol. Soc. Am. 71, 437 -450. laboratory and seismic investigations. Geophys. Res. Lett.
Rivalenti, G.. Garuti, G., Rossi, A., Siena, E and Sinigoi, S. 15,1227-1230.
(1981). Existence of different peridotite types and of a lay- Sato, H., Sacks, I. S., Murase, T., Muncill, G" and Fukuyama.
ered igneous complex in the Ivrea zone of the Western H. (l989h). Qr-melting temperature relation in peridotite
Alps, 1. Petrol. 22, 127-153. at high pressure and temperature: Attenuation mechanism
Roberts. P. M.. Aki, K.. and Fehler, M. C. (1991). A low- and implications for the mechanical properties of the upper
velocity zone in the basement beneath the Valles caldera, mantle, J. Geophys. Res. 94, 10647-10661.
New Mexico, 1. Geophys. Res. 96, 21583-21596. Scarfe, C. M., and Takahashi, E. (1986). Melting of garnet pe-
Robertson, E. C. (1972). "The Nature of the Solid Earth," Mc- ridotite to 13 GPa and the early history of the upper
Graw-Hill, New York. mantle, Nature 322,354-356.
Roy, R. E, Blackwell, D. D., and Decker, E. R. (1972). Conti- Schatz, 1. E, and Simmons, G. (1972). Thermal conductivity
nental heat flow, in "The Nature of the Solid Earth" (E. C. of earth materials at high temperatures, 1. Geophys. Res.
Robertson, ed.), pp. 506-543, Mcflraw-Hill, New York. 77,6966-6983.
Rubin, A. M., (1993). Dikes vs. Diapirs in viscoelastic rock. Sclater, 1. G., and Francheteau, 1. (1970). The implication of
Earth and Planet. Sci. Lett. 119,641-659. terrestrial heat flow ohservations on current tectonic and
Ryall, E, and Ryall, A. (1981). Attenuation of P and S waves geochemical models of the crust and upper mantle of the
in a magma chamber in Long Valley caldera, California, earth, Geophys. 1.R. Astron. Soc. 20, 509-542.
Geophys. Res. Lett. 8, 557-560. Sclater, J. G., Jaupart, C.. and Galson, D. (1980). The heat flow
Ryan, M. P. (1980). Mechanical behavior of magma reservoir through oceanic and continental crust and the heat loss of
envelopes: Elasticity of the olivine tholeiite solidus, Bull. the earth, Rev. Geophys. Space Phys. 18, 269-311.
Volcanol. 43, 743-772. Secor, D. T., and Pollard, D. D. (1975). On the stability of open
Ryan, M. P. (1993). Neutral buoyancy and the structure of hydraulic fractures in the Earth's crust, Geophys. Res. Lett.
mid-ocean ridge magma reservoirs, J. Geophys. Res. 98, 2,510-513.
22321-322338. Shankland, T. 1., and Waff, H. S. (1977). Partial melting and
290 Hiraki Sata and Michael P. Ryan

electrical conductivity anomalies in the upper mantle, Waff. H. S., and Bulau, 1. R, (1979). Equilibrium fluid distri-
J. Geophys. Res. 82, 5409-5417. bution in an ultramatic partial melt under hydrostatic stress
Shervais, 1. W. (1979). Thermal emplacement model for the conditions, 1. Geophys. Res. 84, 6109-6114.
Alpine lherzolite massif at Balmuccia, Italy, 1. Petrol. 20, Waff, H. s., and Bulau, 1. R, (1982). Experimental determina-
795-R20. tion of near-equilibrium textures in partially molten sili-
Sleep, N. H. (1988). Tapping of melt by veins and dikes, cates at high pressures, in "High Pressure Research in
J. Geophys. Res. 93, 10255-10272. Geophysics, Vol. 12" (S. Akimoto, and M. H. Manghnani,
Smith, R.L., and Luedke, R.G. (l9R4). Potentially active vol- eds.), pp. 229-236, Advances in Earth and Planetary Sci-
canic lineaments and loci in western conterminous United ences, Center for Academic Publications, Tokyo.
States, in "Explosive Volcanism, Inception, Evolution and Walsh, 1. B. (1969). New analysis of attenuation in partially
Hazards" (F. R. Boyd, ed.), pp, 47 -66. National Academy melted rock, J. Geophys. Res. 74, 4333-4337.
Press, Washington, DC. Wannamaker, P. E., Booker, 1. R., Jones, A. G., Chave, A. D"
Soller, D. R" Ray, R. D., and Brown, R. D. (1982). A new Filloux,1. H., Waff, H. S., and Law, L. K. (1989). Resis-
global thickness map, Tectonics 1, 125-149. tivity cross section through the Juan de Fuca subduction
Stanley, W. D.. Mooney, W. D., and Fuis, G. S. (1990). Deep system and its tectonic implications, 1. Geophys. Res. 94,
crustal structure of the Cascade Range and surrounding re- 14127-14144.
gions from seismic refraction and magnetotellurgic data. Weertman. J. (197Ia). Theory of water-filled crevasses in gla-
J. Geophys. Res. 95,19419-19438. ciers applied to vertical magma transport beneath oceanic
Takada, A. (1990). Experimental study on propagation of ridges, J. Geophys. Res. 76,1171-1183,
liquid-filled crack in gelatin: Shape and velocity in hydro- Weertman.T, (l97Ib). Velocity at which liquid-filled cracks
static stress condition, J. Geophys. Res. 95, 8471-8481. move in the Earth's crust or in glaciers, J. Geophys. Res.
Takada, A. (1994). Accumulation of magma in space and time 76,8544-8553.
by crack interaction, in "Magmatic Systems" (M. P. Ryan, Wilshire, H. G. (1990). Lithology and evolution of the crust-
ed.), Academic Press, San Diego. mantle boundary region in the southwestern Basin and
Takahashi. E, (1986). Melting of a dry peridotite KLB-I up to Range province, J. Geophys. Res. 95, 649-665.
14 GPa: Implications on the origin of peri dati tic upper Wyllie, P, 1. (1988). Solidus curves, mantle plumes, and
mantle, J. Geophys. Res. 91, 9367 -9382, magma generation beneath Hawaii, J. Geophys. Res. 93,
Takahashi, E., and Kushiro, 1. (1983). Melting of a dry perido- 4171-4181.
tite at high pressures and basalt magma genesis, Am. Min- Young, C. T., and Kitchen, M. R, (1989). A magnetotelluric
eral. 68, 859-879. transect in the Oregon coast Range, 1. Geophys. Res. 94,
Thelin, G. P, and Pike, R. 1. (1991). "Landforms of the Con- 14185-14193.
terminous United States," digital shaded relief image, Al- Young, c.-Y., and Ward, R. W. (1980), Three-dimensional
bers equal area conic projection, scale: I: 3,500,000, MI- Q- . model of the Coso hot springs known geothermal re-
MAP 1-2206, U.S. Geological Survey. source area, J. Geophys. Res. 85, 2459-2470.
von Bargen, N., and Waff, H. S. (1988). Wetting of enstatite Zandt, G., and Furlong, K. P. (1982). Evolution and thickness
by basaltic melt at 1350°C and 1.0-2.5 GP, pressure, of the lithosphere beneath coastal California, Geology 10,
J. Geophys. Res. 93,1153-1158. 376-381.
 Chapter 13 Aspects of Magma Generation and Ascent
in Continental Lithosphere

George W. Bergantz and Ralph Dawes

Overview Units

Ste Stefan number dimensionless

The variety of melt-producing reactions and rheological
T Temperature K,oC
conditions in the crust make it difficult to generalize
many aspects of crustal magmatism. Basalt provides TCR Initial country rock temperature K,oC
both the material and the heat to initiate and sustain TM Initial basalt temperature K, °C
magmatism as demonstrated by unequivocal geophysi- W Width of underplatingmodel m
cal and geological evidence. Following intrusion of ba-
X Thickness of melt region m
saltic magma in the deep crust, the quantity of melt
b Grain diameter m,cm
produced and its composition are controlled by the
presence of hydrous phases; however, it is inappropri- cp Specificheat at constant pressure J. kg-
'
ate to parameterize the amounts of melt produced by g Scalar acceleration of gravity m. S-2

consideration of the modal percentage of the hydrous k Thermal conductivity J·s-'·m-'·K-'

phases alone. Laboratory and numerical experiments Time
on the heat transfer following hypothetical basaltic un-
1/ Horizontal component of the m· S-I
derplating indicate that the thermal exchange is largely velocity
conductive, in agreement with geplogical examples.
w Verticalcomponent of the velocity m· s
Aggregrate viscosity is the most important physical pa-
rameter in controlling melt homogenization and ascent, x Horizontal independent variable m
although there is some ambiguity as to the utility of ex- Verticalindependentvariable m
isting models for magma rheology. Compaction, diapir- f3 Thermal expansion coefficient K-'
ism, and diking do not appear to be viable ascent mech- Power-lawcoefficientin viscosity dimensionless
'Y
anisms when considered alone. The observations that relation
midcrustal plutons often occupy crustal scale shear rn 2
K Molecular thermal diffusivity • S-I
zones and that melt extraction occurs at a variety of
JL Dynamic viscosity kg' m- I • S-I
scales suggest that models of magmatism require an ap-
proach that is not simply cast in terms of simple end JL, Reduced dynamic viscosity dimensionless
member mechanical models. Crustal magmatism might 1J Kinematic viscosity m 2 • S-l
be usefully considered in a geodynamical context where Po Reference density kg m v ·3

magma segregation is explicitly keyed to tectonic con- Volumefraction solids dimensionless

<P
ditions and petrologic diversity is generated and prop-
<PM Limiting volume fraction solids dimensionless
erly understood on a crustal scale.
X Porosity dimensionless

Notation
Units

H Height of underplating model m Introduction

I Thickness of basalt layer m
L Specific latent heat J. kg I. K-' A number of models for melt migration and com-
Nu Nusselt number dimensionless positional diversity have been constructed for the
P Pressure kg m I. s 2
i
mantle (McKenzie and Bickle, 1988). Less well
Magmatic Systems Copyright © 1994 by Academic Press, Inc.
Edited by M. P. Ryan 291 All rights of reproduction in any form reserved.
292 George W. Bergantz and Ralph Dawes

understood are the processes that drive petrologic of chemical and physical change IS appropriate
diversity in the crust. The differences lie in that when trying to compare magmatic systems or
the crustal composition is not readily generalized, generalize the conditions of petrologic diversity.
nor are the changes in chemical potential that Clearly, broad patterns emerge when the style of
drive melt production well understood. The move- intrusion and eruption is indexed to tectonic re-
ment of melt is also more complex, due to the gime, but within individual centers a diversity of
extreme changes in temperature and rheological processes is evident.
state in the crust. Thus, the chemical and dynamic Petrologic diversity is usually considered to
generalizations that have permitted some degree involve crustal melting with variable removal of
of progress in the modeling of processes in the refractory phases, or crystal fractionation of a
mantle are not widely applicable. Despite the ac- mantle-derived melt that may mix with crustally
cessibility of crustal geologic features, even first- derived components. The importance of crustal
order answers to classical questions such as the sources and the potential contribution of the par-
origin and meaning of migmatites or the "room tial retention of restite in the variations in grani-
problem" in granite magmatism are not yet re- toids are indicated by the presence of complexly
solved. The crustal sections that have been de- zoned zircons (Miller et al., 1992; Paterson et al.,
scribed by metamorphic petrologists and struc- 1992), the geochemistry of some enclaves (Chen
tural geologists are often not like those that would et al., 1990), and rather remarkable agreement
seem to be required from the mass balance and between calculated magma temperatures and pla-
transport calculations of igneous petrologists. The gioclase compositions (Burnham, 1992). These
goal of this study is to inventory many aspects of observations indicate that the restite unmixing re-
crustal magmatism around a central theme of ba- lations proposed by Chappell and White (1992)
saltic injection and underplating. The geological and Chappell et al. (1987) have a place in petro-
expression of magmatism and transport is empha- genetic schemes. However, major and trace ele-
sized. The purpose is to present and discuss the ment systematics often require a more complex
kinds of data that are potentially relevant to the interpretation for the origin of many plutons and
construction of transport models. volcanic suites, involving the presence of a mantle
The face of magmatism can be complex, and component that yields petrologic diversity by frac-
temporal and spatial variability is evident: at Ar- tionation and by providing a mixing component
enal volcano, significant chemical variability is for anatectic melts (DePaolo ct al., 1992). The
manifested on time scales of 101 to 103 yr (Reagan geochemical evidence for a mantle component is
et al., 1987), which is in contrast to composition- discussed in more detail below.
ally monotonous systems such as Mt. Hood, and Both scenarios of magma generation require a
to Katmai where spatially overlapping vents erupt change in thermal conditions in the crust. One po-
a variety of magma types with diverse plumbing tentially unifying element in the characterization
systems and apparently unrelated eruptive styles of magmatism comes from the suggestion of Hil-
(Hildreth, 1987). There is also some question as dreth (1981) that crustal magmatism is fundamen-
to the characteristic rates of intrusion and the life- tally basaltic. In keeping with the broad interpre-
times of chemical processes in magmatic systems. tation suggested by Hildreth (1981), we do not
Halliday et al. (1989), for example, estimate that mean simply a Bowen-style fractional crystalliza-
the silicic magma at Glass Mountain may have tion relation. Instead basalt acts as a heat source
had a residence time of 0.7 Myr (also see Mahood, for crustal melting and as a chemical reservoir for
1990), whereas the models of Reagan et al. (1991) magma mingling/mixing. It also provides the nec-
and Gill and Condomines (1992) posit transit essary changes in rheological state to permit ex-
times from 8,000 to 50,000 yr for melt generation, tension and large tectonic strains. We explore
transport, fractionation, and eruption. Intrusive to some elements of the premise that much of the
extrusive ratios vary as well: from 5: 1 in oceanic petrologic diversity in magmas originates in the
settings to 10: 1 for continents (Crisp, 1984). The mid- to deep crust where the thermal conditions
physical controls on the variability and evolution enhance the likelihood of melt generation and
of magmatic systems remain poorly character- mingling between the basalt and crustal melts.
ized: it is not clear what generic feature or index The term "underplating" has been used to de-
 13. Magma Generation in Continental Lithosphere 293
scribe the accumulation of basaltic magma in the Hildreth and Moorbath (1988) (melting-assimi-
mid- to deep crust. According to the underplating lation-storage-homogenization), which holds that
scenario magma is, initially at least, overlain by the "base level" chemical signature in some mag-
upward propagating regions of partial melt and matic systems may he due to intrusion, partial
also subject to continued intrusion from below. melting, and hybridization of the deep crust by
The Moho is often suggested as the site of basaltic mantle-derived melts.
magma accumulation and indeed may be defined (h) Taken together, the geophysical data such
by such a process. Underplating has also been as the high heat flow, uplift, the seismic reflectiv-
used to describe the tectonic accretion of material ity of the lower crust and a well-defined Moho are
from a subducting slab, or even the attachment of consistent with an origin by the accumulation of
the slab itself onto the bottom of the overriding basaltic magma in the mid- and lower crust. The
plate. Admittedly, this "underplating" nomencla- rapid rise time of the very high heat flow in the Ba-
ture is potentially confusing; in the rest of our dis- sin and Range can be attributed only to a combi-
cussion we use underplating only to refer to the nation of tectonic extension and intrusion of basalt
interaction of basaltic magma and the crust. (Lachenbruch and Sass, 1978; Mareschal and Ber-
The geological and geophysical evidence for gantz, 1990). The presence of well-defined sub-
the presence of basaltic magma is as follows: horizontal reflections and downward-increasing
(a) The petrologic diversity of crustal igneous seismic velocities, from 6.8 to 7.8 km- s I, in the
rocks requires basaltic magma as a material source, lower crust of the Basin and Range is consistent
and just as importantly, as a means of providing with magmatically underplated rocks (Klemperer
the thermal input to generate partial melting and et 01., 1986; Valasek et 01.• 1987). Seismic velocity
subsequent hybridization of the adjacent crust. In models for the deep-crustal structure in southern
this model basalt is not simply a parent magma, Alaska suggest that magmatically emplaced rocks
but rather just one of several possible geochemical may form part of the "crustal" root (Fuis and
reservoirs. Numerous petrologic studies of both Plafker, 1991). Variations in seismically deter-
intrusive and extrusive rocks (Davidson et 01., mined crustal thickness in Australia can be attrib-
1990; Feeley and Grunder, 1991; Hildreth et 01., uted to crustal underplating by mantle-derived
1986; Manduca et 01., 1992) have documented the magmas (Drummond and Collins, 1986). A num-
presence of basalt and other geochemical reser- ber of other deep-reflection profiles have features
voirs that have interacted with the basalt. Repeated that would permit an interpretation involving ba-
intrusion of basalt may lead to partial melting of saltic magmatism in the deep crust (Mereu et 01.•
previous underplates, which would comprise a 1989). In one sense, the Moho is a "Stefan"
young and mafic lower crust. Repeated melting of boundary in that its spatial variation and definition
this material could give rise to evolved suites that represents a degree of freedom in the response of
would be difficult to geochemically distinguish the crust to thermal throughput on a planetary
from simple basalt fractionation (Kay et 01.• 1990; scale (Nelson, 1991).
Tepper et 01., 1993) if the lower crust is isotopi- (c) Exposures of the deep crust provide direct
cally young. Among the most dramatic examples examples of magmatic additions to the lower crust
of the type of interaction of basaltic magma and with the subsequent generation of zones of partial
crust is the basalt-rhyolite association at the Yel- melting and granulite formation. The best studied
lowstone Plateau volcanic field (6000 km' rhyo- of these is the Ivrea-Verbano zone of northern
lite, 100 km" basalt). The isotopic character of the Italy where a crustal column containing mantle
eruptive rocks and the absence of intermediate peridotite, overlain by layered and homogeneous
compositions require a large-scale, deep-crustal gabbro and amphibolite facies metasediments, is
hybridization between basalt and Archaean crust exposed (Handy and Zingg, 199I; Voshage et 01.,
(Hildreth et 01.• 199I). The origin of the rhyolites 1990; Zingg, 1990). The isotopic similarity be-
of the Snake River Plain-Yellowstone Plateau tween the magmatic assemblage and the overlying
province is also consistent with an origin by deep- granitoids, and the positive Eu anomaly in the
crustal melting following the intrusion of basalt mafic portions point to a genetic relationship be-
(Leeman, 1982). Observations of this type pro- tween the two and may be an example of the lower
vide the key elements of the MASH hypothesis of crustal regions of partial melting and mixing.
294 George W. Bergantz and Ralph Dawes

Quick et al. (1992) have mapped pervasive defor- xenoliths requires an origin by basaltic intrusion
mation features in the Ivrea-Verbano zone that of the lower crust, followed by simultaneous par-
must have formed under near solidus conditions, tial melting of overlying crust, which may include
yielding fabrics that indicate substantial lateral the residue of previous episodes of underplating,
mass transfer while the melt was in a rheologi- fractionation of the mafic melt, and the production
cally "mushy" state, much like textures observed of melts representing varying degrees of hybrid-
in ophiolite complexes. The compositional variety ization (Kempton and Harmon, 1992; Rudnick,
and volumes of mafic magma suggest that the 1992). Continued magmatic underplating leads to
Ivrea-Verbano zone may represent a very ther- the cratonization of the crust, yielding a crust that
mally mature system with sustained thermal input. has grown from below, perhaps with episodes of
The other end member, perhaps representing the delamination of hybridized and cumulate regions
incipient generation of a MASH zone, where the into the mantle (Arndt and Goldstein, 1989; Kay
melting of surroundings is minor, may be the Fi- and Kay, 1991).
ambala gabbronorite in Argentina (Grissom et al.,
1991). This sill-like body was emplaced at a depth
of 21- 24 km, synchronous with the local devel- Melt Generation
opment of granulite facies conditions. The amount
of crustal partial melting as a result of the thermal The primary factors that control melting of the
perturbation during intrusion is much less than lower crust are (a) bulk composition and mineral
that observed in the Ivrea-Verbano zone, and like mode, (h) temperature, (c) pressure, (d) volatiles
that estimated by Bergantz (1989b) for a single (primarily H2 0 and CO 2 ) , and (e) the physical dy-
episode of underplating. namics of the melting process. The first four fac-
(d) The occurrence of counterclockwise P-T tors control the chemical potential of the melt
paths and the coincidence of peak metamorphism phase, and thus directly influence the extent of
and peak temperatures in granulite terrains are con- melting and the melt composition. The fifth factor
sistent with a metamorphic history where crustal includes kinetic factors and transport mechanisms
thickening occurs at least partly as a result of the that may overlap with the melt segregation pro-
addition of magma to the lower crust (Bohlen cess. This section outlines recent progress in un-
and Mezger, 1989). An in-depth inventory of the derstanding the phase equilibria that are likely to
granulite "controversy" is beyond the scope of be involved in lower crustal melting. We suggest
this paper (the interested reader is directed to the that what is currently known about lower crustal
comprehensive volume edited by Vielzeuf and melting is consistent with the following: (a) the
Vidal, 1990). We note however that some granu- generation of many common magmatic suites is
lites clearly have mineral phases consistent with initiated by intrusion of the crust by mantle-
an origin by the generation and removal of partial derived mafic magma; (h) heat and volatiles from
melt, although there are few examples where the the mafic magmas partially melt proximal lower
evidence is unequivocal on a regional basis. In- crust, yielding a major component of resulting
complete melt removal (Rudnick, 1992) and the magmatic suites.
uncertainty in the assumptions of trace element A good experimental understanding of the hap-
partitioning during melting (Bea, 1991; Sawyer, logranite system is fairly well in hand, but the ex-
1991) may render the standard geochemical argu- periments are only indirectly applicable to melt-
ments of little use in discriminating between a ing of natural rocks besides acid granites. As stated
restitic and nonrestitic granulite. Nonetheless, the by Tuttle and Bowen (1958), lower-crustal melting
abundance of mafic xenoliths recording pressures is likely to be water-undersaturated and involve
of 0.5-1.0 GPa and magmatic or near magmatic mica and amphibole in the phase equilibria that
temperatures and textures suggest that one com- control the melting reactions,the water activities in
ponent of crustal growth is the ponding of basic the melts, and melt productivity. It is now generally
magmas at or near the Moho (Cull, 1990; Rud- accepted that crustal melts that mobilize to form
nick, 1990; Rudnick and Taylor, 1987). The iso- voluminous granitoids are water-undersaturated
topic and trace element character of lower crustal (Burnham, 1967; Clemens, 1984; Clemens and
 13. Magma Generation in Continental Lithosphere 295
Vielzeuf, 1987; Powell, 1983; Thompson, 1983). ter and Wyllie (1988), Wolf and Wyllie (1989),
This has led to a focus on so-called "damp melt- Patifio-Douce and Johnston (1991), Beard and
ing," in which the only water in the system is that Lofgren (1991), and Rushmer (1991). The follow-
contained in hydrous metamorphic minerals pres- ing points derive from these studies. (0) Closed-
ent before melting began. system melting of metapelites produces peralumi-
Free water is unlikely to be present in pore nous melt compositions similiar to S-type granites
spaces at > 15- 20 km depths (Yardley, 1986). It (Patifio-Douce and Johnston, 1991; Vielzeuf and
may be possible for water-rich fluids to enter Holloway, 1988). The main melt producing reac-
the lower crust by exsolution from crystallizing, tion in both studies is biotite + plagioclase +
mantle-derived magmas (Lange and Carmichael, aluminosilicate + quartz = garnet + melt. At
1990; Wickham and Peters, 1992). Similarly, CO 2 higher temperatures garnet is also consumed to
or mixed H2 0 - C 02 fluids released from mafic form melt, with spinel and aluminosilicate the
magmas may playa role in lower-crustal melt- most refractory phases. (h) Amphibolite melt-
ing although the source of the CO 2-dominated ing at lower pressures (7 -10 kbar) occurs by
fluids remains to be resolved (Peterson and New- "damp" melting mainly via breakdown of horn-
ton, 1990). Although appinites and lamprophyres, blende, producing clinopyroxene ± orthopyrox-
common in calc-alkaline intrusions, suggest a ene (Rushmer, 1991; Wolf and Wyllie, 1989).
possible role for H20 or H20-CO z fluid-saturated Melts range from granite or trondhjemite to ton-
mafic magmas in the generation of calc-alkaline alite with increasing temperature and cross the
granitoids, little research has been devoted to this boundary from slightly peraluminous to metalu-
topic. The production of plausible granitoid com- minous. (c) The importance of hydrous minerals
positions in fluid-absent, lower-crustal melting in controlling melt productivity and composition
experiments, and the discovery by Beard and Lof- is consistently highlighted in all of these studies;
gren (1989) that water-saturated melts of arnphi- the linked parameters mineral mode and bulk rock
bo ite, generated at midcrustal pressures, are un- composition exert a strong control on the amount
like natural granitoids, are oblique evidence in of melt produced in fluid-absent melting.
support of fluid-absent melting as the dominant This third point, on the relation between min-
process in the lower crust. eral mode and melt production, is illustrated by
Recent chemographic and experimental work the very different melt fraction vs temperature re-
on lower-crustal melting has focused on four types sults for melting of metapelite obtained by Viel-
of systems: model systems (i.e., specific reactions zeuf and Holloway (1988) and Patifio-Douce and
involving a few components and phases), amphi- Johnston (1991). This difference, and the more
bolites (of more or less basaltic composition), potassic and felsic nature of the melts produced in
common granitoids (e.g., tonalites), and metape- the experiments of Vielzeuf and Holloway, is as-
lites. Theoretical phase equilibria of model sys- cribed primarily to the different bulk compositions
tems provide a basis for understanding the melting and mineral modes of the two metapelites, and how
behavior of rocks containing hydrous minerals well or poorly the initial mineral mode corresponds
(Brown and Fyfe, 1970; Burnham, 1979; Thomp- to the stoichiometry of the major melt-producing
son, 1982; Thompson and Algor, 1977; Thomp- reaction, which in this case involves biotite dehy-
son and Tracy, 1979; Wyllie, 1977). Reactions in dration (Patifio-Douce and Johnston, 1991).
natural rocks generally have greater thermody- The results of a variety of partial melting ex-
namic variance than those in model systems. Melt- periments (Fig. 1) suggest that melt production vs
ing is likely to take place through both continuous temperature of natural, hydrous mineral-bearing
and discontinuous reactions (Thompson, 1988), metamorphic rocks covers a spectrum. At one end
making it difficult to generalize melt production. of the spectrum are rocks in which the relation be-
Melting experiments that used natural rocks as tween modes of hydrous minerals, quartz, and
starting materials, evaluated melt fraction through feldspar, and bulk minimum-melt components,
the melting interval, and applied pressures equiva- fits well enough with reaction stoichiometry that
lent to the middle to deep crust, have been per- most melting occurs at the reaction that consumes
formed by Vielzeuf and Holloway (1988), Rut- the hydrous mineral and releases HzO fluid to dis-
296 George W. Bergantz and Ralph Dawes

to those found in granulites (Rudnick, 19tJ2). Me-

tapelites and mica-rich quartzofeldspathic rocks
80 (e.g., some types of metagraywackes) produce the
largest volume proportions of melt. What rock
types are present in the crust is uncertain, but it
.... 60
Qj is not necessarily limited to direct products of
~ mantle-derived magmas. Metasupracrustal rocks
~ 40 may also be emplaced in the lower crust, particu-
larly in convergent margin orogens (Kempton and
20 Harmon, 1992).
In generic terms, the experimental evidence ac-
OL.-. L.",::;..J""o:;",a;.. . . .I. . ."",.I.-. . L. . . .L.~_"--..l-. .l-I cords with the geological evidence, in that both
700 800 900 1000 1100 1200 are consistent with the initiation of crustal-scale
Temperature, °C magmatism by the intrusion of basaltic magma
at the base of the crust, and the involvement of
Figure 1 Percentage melt as a function of temperature for a
number of lithologies. Pressures range between 7 and 10kbar, lower-crustal melting as an important part of the
Compositions are metapelites (1,2), silicic plutons (4,5), and process. Major episodes of lower-crustal melting
amphibolite (3,6). All are damp melting: no free water is pres- are associated with temperatures achievable only if
ent at the start of the experiment. Note rapid and nonmono- basalt intrudes the lower crust, based on the tem-
tonic changes in percentage melt as temperature increases.
peratures required for the generation of interme-
Curve 7 is the crystallization of a basalt. References for indi-
vidual curves: (I) Vielzeuf and Holloway (1988); (2) Patino- diate magmas. Similarly high temperatures are in-
Douce and Johnston (1991); (3) Rushmer (1991): (4) Wyllie dicated for many incompatible-element-depleted
(1977); (5) Rutter and Wyllie (1988); (6) Beard and Lofgren granulites (Bohlen and Mezger, 1989; Clemens,
(1989); (7) Marsh (1981). 1990). The compositions of magmatic rocks that
may have been generated in the lower crust are very
solve in the melt. Such rocks will produce melt in similar to the experimentally produced melts from
step-like fashion, with large volumes produced vapor-absent runs, with an important exception.
over a narrow temperature interval. At the other The major point of disagreement between ex-
end of the spectrum are rocks that do not contain perimentally produced melts and natural mag-
the components in proportions suitable to main- matic rocks is that the Mg numbers of experimen-
taining quartz- and feldspar-saturated melting via tal melts are consistently lower than those of most
dehydration reactions. These rocks will not pro- natural rocks interpreted as prohable lower-crustal
duce major volumes of melt over narrow tempera- melts. This is particularly true at intermediate
ture intervals within the melting range. Because Si0 2 values. Assuming that crustal melting was
the total volume of rock undergoing partial melt- involved, this discrepancy may be accounted for
ing will probably vary over some range of hetero- by partial mingling or mixing of basaltic magmas
geneity, it seems likely that in many real situations with lower-crustal melts that were generated by
the intermediate pattern of the temperature vs melt the intrusion of the basaltic magmas, or by restite
fraction spectrum holds, with a dampened corre- retention, which assumes that Ca- and Mg-rich re-
lation between breakdown reactions of hydrous fractory minerals from the crustal melting source
minerals and pulses of increased melt production. are entrained in the melt when it leaves the source
In sum, recent work on lower crustal melting area. Even if restite unmixing (Chappell and Ste-
suggests a dominant role for fluid-absent, "damp" phens, 1988) is an important process, intrusion of
melting. Dehydration of biotite occurs at 800- the lower crust by basalts must still be invoked
900°C, producing melt fractions of >35% with (Burnham, 1992). The question is then whether
granite or felsic granodiorite composition. De- magmas are mobilized en masse as melt plus re-
hydration of hornblende occurs at 900-1000°C, sidual solids or whether the melt segregates effi-
producing melt fractions of >35% with tonalite ciently from the residual solids, although the an-
or intermediate granodiorite composition. The re- swer is undoubtedly somewhere in between and
sidual, anhydrous mineral assemblages are similar will vary on a case-by-case basis.
 13. Magma Generation in Continental Lithosphere 297
Modeling the Physical Interaction shape and induce a feedback that is particular to a
of Basalt and Crust given geometry, shear rate, etc. Despite these dif-
ficulties, various empirical relationships can be
The effectiveness of melt generation, segregation, taken from the engineering literature, although it
and homogenization will depend on the rate at is important that these results be applied to geo-
which melt is generated, the differences between logic conditions with caution.
the viscosity of the crystal-liquid mixture, the To the viscosity calculated for the crystal-free
melt-only viscosity, and the bulk viscosity of the liquid, a "correction" for the influence of crys-
solid matrix. Hence. one of the most important tals can be added. One such expression often cited
elements in characterizing the dynamics is the is a form of the Krieger-Dougherty expression
rheological state of the partial melt zone (Ber- given by Wildemuth and Williams (1984),
gantz, 1990; Wickham, 1987). In addition, there
1
is a growing body of geological evidence that (1)
there is a feedback between the stress field in the
crust and the style and timing of ascent. Before where u, is the reduced viscosity, which is the
we consider models of underplating and segrega- actual viscosity divided by the viscosity of the
tion mechanisms, we will consider the rheological crystal-free liquid at the same temperature, pres-
characterization of crystal-melt mixtures. sure, and composition; ¢> is the volume particle
fraction; and ¢>'v1 is the volume fraction of particles
Rheology of Suspensions at which the three-dimensional contact between
the particles causes strain to cease at a given shear
Models for the viscosity of magmatic liquids are stress. It should be appreciated that ¢>M is shear
given by Bottinga and Weill (1972) and Shaw stress dependent and using the absolute maximum
(1972) and a comprehensive summary ofthe avail- packing fraction for ¢>M may overestimate ¢>\1 for
able data on the viscosity of melts of geologic in- many magmatic conditions where strain rates as-
terest can be found in Ryan and Blevins (1987). sociated with convection may be low. ¢>M ranges
Obviously, estimates of viscosity based on con- from about 0.5 at low shear stress to a value of
sideration of only the changing liquid composi- 0.75 for infinite shear stress, and "y is a coefficient
tion have little application to dynamical models of with experimentally determined values that vary
crystal-melt systems; the viscosity can range from from about 1.3 to almost 3 with a value of 2 being
nearly infinite under near solidus conditions to typical of many systems (Barnes et al., 1989; Sen-
I Pa . s under near liquidus conditions. Develop- gun and Probstein, 1989; Wildemuth and Wil-
ing robust quantitative models for the influence liams, 1984). Although weakly shear stress de-
of increasing crystallinity on the viscosity is dif- pendent, "y is nearly invariant with variations in
ficult; some scaling relations are reviewed by stress over two orders of magnitude for a wide
Jomha et al. (1991). Even making reliable mea- variety of particle sizes and mixture properties
surements of the viscosity of suspensions has (Wildemuth and Williams, 1984).
proven challenging: the same crystal-liquid sus- The MELTS program (a substantial refinement
pensions yield different viscosities when mea- of the SILMIN algorithm of Ghiorso (1985» was
sured in different viscometers and the measure- used to calculate the viscosity as cooling proceeds
ments are often difficult to reproduce even when for a given initial composition and pressure using
repeated evaluations are made on the same vis- the method of Shaw (1972). The MELTS program
cometer. These results cast doubt on the notion is robust in that it allows for the crystallization of
that a single expression can adequately character- practically all the relevant phases: pyroxenes, com-
ize the viscosity of dense suspensions (Cheng, mon feldspars, quartz, biotite, olivine, the Fe-Ti
1984; Frith et al., 1987). The origin of the irre- oxides, and others. In addition, the oxygen fu-
producibility may reside in variations of the solid gacity can be controlled. Using MELTS, changes
packing structure, an organization of the crystals in melt composition and volume percent crystal-
that is an inevitable consequence of flow. This will linity can be calculated as a function of tempera-
vary depending on the flow regime and crystal ture, including the effects of water in the melt.
298 George W. Bergantz and Ralph Dawes

GRANODIORITE BASALT
12 ...................................I""T'"............,,.......,,...,..,...,....,..., 4
til 11 til
\
\
.
~
Q.. ~
Q..
\ \
, \

-
10

->;
'iii
0
u
9
,
,,
,, \
>; 3
'iii
0
u
\
,
,
,
,
\
\
til 8 ,
" \, ,
-
';: til
.;: \
0 7 ,'-' . e Ql M
Ql M .... 2

," ,
eJI '~
0 6 OIl .3
...:l .3 0 .S
.S
...:l
5 .7
.7
4 ......... 1
500 600 700 800 900 1000 1100 1180 1220 1260 1300 1340
Temperature, °C Temperature, °C
Figure 2 Calculated variation of viscosity with temperature Figure 3 Calculated variation of viscosity with temperature
for the Red Lake granodiorite (Noyes et al.. 1983) at two dif- for the mafic inclusions (basaltic composition) in the Burnt
ferent initial water contents, I and 3 wt% H2 0 . The calcula- lava flow of Grove et al. (1988). The calculations were done
tions were done with the MELTS algorithm assuming the oxy- with the MELTS algorithm assuming the oxygen was buffered
gen was buffered at Ni-NiO and a total pressure of 2 kbar, at Q-F-M and a total pressure of 8 kbar. These curves are
Water content will increase as crystallization of anhydrous typical of a variety of basalts. Adding small amounts of water
phases continues. Very high water contents yield low solidus to the system (1-2 wt%) had a negligible effect on the calcu-
temperatures; given the absence of experimental data under lated viscosity.
these conditions, viscosity values for temperatures under 600°C
should be used with caution.

ture can be very different depending on whether

temperature is falling, such as in the case of a
Thus at any temperature, the calculated melt com- magma, or temperature is rising, which is the case
position, water content, and volume fraction crys- in partial melting. Volume fraction curves tend to
tals can be used in Eq. (1) to yield the viscosity, be smoother for solidification as the nucleation is
Two examples of the calculated viscosity are typically heterogeneous and the undercoolings are
given here, a granodiorite in Fig. 2 and a basalt in demonstrably small (Cashman, 1993). In compo-
Fig. 3. Three curves are shown in each of the fig- sitionally similar, but "dirtier" rocks, retrograde
ures; each curve corresponds to a different as- assemblages can profoundly affect the melt frac-
sumed value of the critical volume fraction, cPM' tion distribution during partial melting; see Fig. 1.
There is little unambiguous geological evidence Hence, the disposition of the partial melt may not
of the value of the critical volume fraction as dis- be simply related to temperature as discussed in
cussed by Bergantz (1990) but a value around 0.5 Bergantz (1992). This has been demonstrated by
is suggested from observations on extrusive and Wolf and Wyllie (1991) where it was observed
plutonic rocks (Marsh, 1981; Miller et al., 1988). that melt distribution in anisotropic partial melts
For each curve in Figs. 2 and 3, all values of melt was not controlled by dihedral angles. This raises
viscosity and crystallinity are the same at a given the question as to the degree to which equilibrium
temperature, the only difference among the curves is maintained during partial melting. The agree-
is the value of cPM. The curves for a critical vol- ment between experimentally produced and cal-
ume fraction of 0.3 and 0.7 delimit an envelope of culated major element melt compositions often
reasonable values. It is apparent that increasing agrees well with those seen in the rock record
crystallinity dominates the variation in viscosity, (Ashworth and Brown, 1990; Burnham, 1992).
Two initial water contents are given for the gran- Direct natural evidence is uncommon; however, in
odiorite to demonstrate the effect that increas- a remarkable study of the partial melting of a
ing water content in the melt will have on the magma chamber's walls that was interrupted by
viscosity. eruption, Bacon (1992) documents that equilib-
This is important when one considers that the rium was not achieved between plagioclase and
volume fraction of solids as a function of tempera- melt on time scales of 102_10 4 yr. The implica-
 13. Magma Generation in Continental Lithosphere 299
appropriate melt fraction. However, as the thermal
Equilibrium melts gradient extends outward, the melt fraction may
not be simply related to temperature as the rates
Disequilibrium melts of heating and cooling will be less. Unfortunately,
no complete kinetic model for melt-forming reac-
tions exists. This is due in part to the experimental
melt residuum difficulty in establishing the reaction stoichiome-
segregation curve
try and developing the appropriate kinetic model;
race element diffusion curve the most recent summary of metamorphic kinetics
can be found in Kerrick ( 1991). Hence, estimating
Temperature
the rheological state during partial melting is sub-
Figure 4 Schematic of relationship between melt segrega- ject to considerable uncertainty. The available ex-
tion and rate of trace element equilibration between melt and periments are few and reveal a strong sensitivity
residuum. After Sawyer (1991).
of deformation style to melt fraction, particularly
near the solidus (Dell' Angelo and Tullis, 1988).
tions of disequilibrium melting during melt gen-
eration and the geochemical interpretation of mig- Numerical Model of Basaltic Underplating
matites have been addressed by Bea (1991) and
Sawyer (1991). The approach to equilibrium can We begin our discussion by invoking the simplest
be considered a competition between diffusion imaginable process, as illustrated in Fig. 5. Ba-
and segregation and is shown schematically in saltic magma at the liquidus intrudes some re-
Fig. 4. Rubie and Brearley (1990) explore a num- gion of the crust that is at ambient temperature
ber of simple models of disequilibrium melting to TCR ' In response to the intrusion, melting begins
examine the sensitivity of the melt fraction distri- in the country rock and, subsequently, a region of
bution to the overstepping of the solidus tempera- solid and liquid will form in the region once oc-
ture and the effects of variable latent heats of melt- cupied by the magma. This simple picture has
ing, the volume fraction of the melting species, been examined using both numerical and labora-
grain sizes, and the diffusion activation energy. For tory experiments by a number of workers (Ber-
the conditions of magmatic underplating, the ini- gantz, 1989a, 1989b; Campbell and Turner, 1987;
tially large thermal overstepping rapidly yields the Fountain et al.. 1989; Hodge, 1974; Huppert and

T T

Rock at constant solid rock

initial temperature:
TCR
z

Contac

solidus, Ts--.>r-----l
Melt at the partially
liquidus solidified:
temperature: T mr crystals & melt
convective liquidUS, To"',...,---1
melt
o x
A B
Figure 5 The initial conditions at the time of basaltic underplating, with subsequent evolution. Well-mixed basalt at the liquidus
temperature forms a planar interface contact against country rock, which is at some ambient temperature Te R• (A) The tempera-
ture profile initially has a step profile that decays, yielding (B) regions of crystal-melt "mush."
300 George W. Bergantz and Ralph Dawes

Sparks, 1988; Wells, 1980; Yoder, 1990; Younker are that we are using a set of "best guess" calcu-
and Vogel, 1976). Each of these studies has em- lated physical property variations: viscosity, den-
phasized different physical aspects of basaltic sity, and all the thermophysical properties vary
underplating and yielded results that are not all temporally and spatially as crystallization and con-
in agreement. The laboratory experiments often comitant partial melting proceed. In particular, the
demonstrate a wide variety of convective phe- model attempts to retain the appropriate relation-
nomena; however, the solid fraction-enthalpy, ships for enthalpy content and transport properties
and hence transport property, relationships be- in the combined basalt-country rock system. Con-
tween the model basalt and country rock do not vection in the basalt occurs, and the time depen-
match those found in natural systems. The ana- dence, vigor, and form of the convection can be
lytical models usually assume that the enthalpy directly related to the crystallization history, which
transfer is only by conduction, and so there is al- itself is constrained by the phase relations deter-
ways some question as to the role of convection in mined under geologically relevant conditions and
the magma in enhancing the heat transfer and par- the progress of melting in the country rock. Thus,
tial melting of the country rock. the questions of interest are: How might magmatic
The style and vigor of convection in magma convection influence the timing of partial melting
chambers has been subject to uncertainty, and even in the overlying country rock and what is the style
controversy (Huppert and Sparks, 1991; Marsh, of this convection? How does variable viscosity
1991). Much of the uncertainty lies in the difficulty generated by the changing crystallinity and com-
in forming testable hypotheses. The simple fluids position of the melt influence the convection?
available for table-top experiments rarely demon- What might the characteristic convective veloci-
strate the strong rheological contrasts and buoy- ties be, and what is the form of the instabilities?
ancy relationships present in magmatic systems. What are the overall heat transfer rates from the
On the other hand, the products of magmatism system relative to those where the transfer is by
rarely provide unambiguous evidence of the pro- conduction only?
cesses involved in their generation and transport. Oldenburg and Spera (1992) have developed a
Buoyancy in magmas is also generated in com- clever hybrid model for the numerical simulation
plex ways, with thermal effects being important of crystallizing systems. Their model is a combi-
near the liquidus and compositional ones impor- nation of two end-member states: the relative mo-
tant near the solidus. There is no question that tion condition where interdendritic melt can flow
magmas convect, although time dependence and through the crystal pile, and the no relative motion
the presence of multiple length scales often make condition where the fluid that is moving through
the fluid structures and thermal history difficult any control volume includes the crystals and equi-
to generalize with the conventional parameteriza- librium is assumed. The model considered here
tions developed for simple fluids. The emphasis of uses the no relative motion assumption. In prac-
the model presented here is convection and partial tice this means that at any point in the flow field,
melting under one set of arguably geologically the crystallinity, composition, density, viscosity,
relevant conditions. We note that our model as- and all thermophysical properties are determined
sumptions do preclude some aspects of the heat from the phase constraints directly. This is essen-
and mass transfer in magmas, such as volatile re- tially just application of the phase rule to the con-
lease and the capacity to model individual crystal trol volume over which the local average tempera-
growth and settling. However, it is hoped that cal- ture is determined.
culations presented here, which emphasize some The advantage of this approach is that all the
of the known rate-limiting features present in geo- transport property variations can be indexed to
logical systems, will encourage a discussion of temperature. What is required is a means of cal-
magmatic convection that directly addresses geo- culating how density, specific heat, crystallinity,
logical conditions. viscosity, and conductivity vary as a function of
One model of the conjugate enthalpy transfer temperature, including the effects of melt compo-
between crystallizing basalt and simultaneous par- sition as crystallization proceeds. The MELTS al-
tial melting of the overlying country rock is pre- gorithm of Ghiorso (discussed previously) pro-
sented. The important new features of this model vides these quantities. The disadvantage of this
 13. Magma Generation in Continental Lithosphere 301
approach is that it does not allow for nonequili- conservation of momentum,
brium crystal-liquid segregation. For example,
if crystals begin to settle out individually or in po[au + uau + W al.l] = _ap +.i. (fJ,au)
the form of a plume, it is possible that interstitial at ax a:: ax ax ax
melt could be removed as the plume falls. In our
model, any time magma convects to a new region + -a( fJ--
au) -r, -afJ- -au + -afJ- -al1' (3)
a:: az ax ax a:: ax
where temperature is different, it is assumed that
the crystallinity changes accordingly; temperature
Po [-aw + uaw aw]
- + 11'- = - -
ap
and crystal fraction are linked. Thus, convection at ax a:: a::
driven strictly by crystals falling cannot be ac-
commodated. We do not consider this a restrictive + pg + ;- (fJ,aw) + ~(fJ- aw) (4)
ax ax a:: a::
condition as it is not clear what role settling crys-
afJ, au afJ, aw
tals may have in driving hulk convection. +-- +--
The combined basalt-country rock system ax az a:: az'
shown in Fig. 5 was represented on a two- and conservation of energy.
dimensional 75 X 75 grid of total width W of
15 m and total height H of 25 m. The depth of the PoC [aT + u aT + 11' aT] - a ( k aT)
initial basalt layer is 1 and was 20 m thick. The
p
at ax a:: - ax ax (5)
origin was placed at the lower left corner and Car-
tesian coordinate axes were introduced: x being + i!..
a::
(k aT),
az
the horizontal coordinate axis and :: the vertical
axis, where z is taken as positive in the upward with initial and boundary conditions on region
direction. Fifty-five grid points in the vertical di- and boundary temperatures,
rection were assigned to the fluid layer and twenty T(x.::,O) o ::; z ::; I;
to the overlying country rock. Variable grid spac- (6)
T(x,::,O) 1 ::; z ::; T('R.
ing was used in both the vertical and horizontal
aT
directions to ensure that the boundary layers near
-«r I
the intrusive contact could he adequately char-
acterized. Using fewer grid points (40 X 40)
-
ax n I, = 0,
ax \ II
= o.
(7)
aT
yielded the same Nusselt number relationships - = 0, T(xH,t) = T('R,
discussed here. but some details of the thermal
az I 0

structure of the plumes were lost so we elected to and on mass flux (motion components) at the
use the finer grid. houndaries,
A comprehensive discussion of the continuum
equations describing the heat and mass transfer -aul =0 a~a.. I.. = 0,
in systems undergoing crystallization. with pos- az .r II ' \ 1\

sible double-diffusive convection is available u(x,OJ) = 0, u(x.!,t) = 0 (8)

(Beckermann and Viskanta, 19S9, 1993; Bennon
and Incropera, 1987; Ni and Beckermann, 1991; awl
- = 0, -awl =0
Oldenburg and Spera, 1991. 1992). We will not ax 0 .r ax .r II" '

u(x,O,t) = 0, w(x.!,t) = 0 (9)

revisit these derivations. but direct the reader to
these sources for clarification of the continuum where 1 is the thickness of the fluid layer. The
expression of the model. initial conditions are constant temperature in the
The equations required to describe the veloc- country rock Tc R , and the basalt is assumed to
ity and temperature fields following underplat- have intruded at its liquidus, TM' The horizontal
ing, invoking the Boussinesq approximation, are boundary conditions on temperature and velocity
given for are reflecting conditions where it is assumed that
the magma layer is infinite in the horizontal di-
conservation of mass (continuity),
mensions. Thus, effects due to the presence of side
au aw walls as sources of drag or heat are not consid-
- +- = 0, (2)
ax a:: ered. The vertical conditions are that the magma
302 George W. Bergantz and Ralph Dawes

layer is effectively insulated at the bottom. This 9000

means that the convection is driven only by cool- 8000

ing from above and is fundamentally different
7000 latent & sensible heat
from systems that are heated from below and
'-;'
cooled from above, and buoyancy is generated at :::.:: 6000
'-;'
both boundaries. This seems reasonable as it is QJl
.:.: 5000
difficult to imagine what heat source would be ~

hotter than the basalt itself. U e, 4000

The numerical algorithm used to solve Eqs.
(2-5) is a modified version of the TEMPEST 3000

code developed at Pacific Northwest Laboratories 2000

(Trent and Eyler, 1991). It uses a semi-implicit,
finite volume method. We have established code 1000
1065 1115 1165 1215 1265 1315
accuracy and reliability by doing benchmark runs
Temperature.v'C
on a number of classical natural convection prob-
lems, and we have found excellent agreement with Figure 6 Calculated effective specific heat capacity for the
published numerical and experimental results, in- model basalt. This effective specific heat includes both the ef-
fects of phase change and an overall drop in system tempera-
cluding conditions of variable transport properties.
ture. It also includes the temperature-dependent heat capacities
We consider only one set of calculations here, of the minerals and melt. The peak in effective heat capacity
but one that reveals the important features ob- at about 1275°C is due to the onset of feldspar crystallization;
served in a variety of simulations. The model ba- the peak at about 1085°C is due to the appearance of spinel.
salt composition used was that of the Burnt Lava
flow (Grove et al., 1988). The crystallization was
assumed to occur under oxygen-buffered condi- havior of the effective specific heat as a function
tions of quartz-fayalite-magnetite. It was also as- of temperature is the result of the magma becom-
sumed that crystallization occurred in the frac- ing saturated with new phases as cooling pro-
tionation mode. This does not mean that the ceeds. Once the melt has become saturated with
crystals were physically removed as temperature a phase, with a concomitant "burst" of crystal
dropped, but rather that as crystals grow, they do growth, the effective specific heat will diminish as
not back-react and the melt composition evolves the increase in crystallinity per degree of tempera-
accordingly. This is really a statement that thermal ture drop diminishes. Although the liquidus phase
equilibrium might be maintained, but the crystals was olivine, the increase in crystallinity with tem-
will remain zoned. If resorption is required by perature at the liquidus is modest; the two domi-
advection of crystals to a region of higher tem- nant peaks in Fig. 6 are due to the saturation of
perature, it will initially involve the rim of the the magma with plagioclase at about l275°C and
crystal. This decision was based on the fact that with spinel at I085°C. Pyroxene appears at about
plagioclase was the dominant near liquidus phase. l260°C. Sensitivity analysis of the model results
Very little difference in the convective state or indicated that geologically reasonable variations
the progress of melting of the country rock was in the absolute values, or reasonable variations in
found in simulations where the physical proper- the functional form of the effective specific heat
ties were generated assuming complete equilib- function, had little impact on the results. Calcu-
rium between melt and solid phases. lated density variations are shown in Fig. 7. The
The MELTS algorithm was used to calculate system density was used in the calculations, in
the required physical property data. The numeri- keeping with our no relative motion postulate. This
cal prescription of the enthalpy changes associ- system density reflects both changes in the melt
ated with phase changes was done by defining a density as cooling and composition of the melt
temperature-dependent effective specific heat ca- changes during crystallization and increases in
pacity that includes changes in both sensible and density associated with the appearance of the solid
latent heats. Figure 6 is a plot of the calculated phases. The viscosity-crystallinity-temperature
variation in effective specific heat as a function of relationships for the crystallizing basalt are shown
temperature for the basalt. The nonmonotonic be- in Fig. 3.
 13. Magma Generation in Continental Lithosphere 303
rock. The maximum Nusselt number under these
conditions is 2 and would occur if the magma
"J 2900 were perfectly well mixed by some unspecified
'a stirring agent. To obtain the conduction results for
ell
.:.: calculation of the Nusselt number, we did a nu-
.s 2800 merical simulation identical to that with convec-
.~ System: Melt & Crystals
<l>
tion, except that no fluid motion was permitted.
Q 2700 The Nusselt number as a function of the dimen-
sionless conduction time is shown in Fig. 8; four
2600 ........................................L..L......................L..I........................................
curves are shown and we consider each in turn.
1065 ms 1165 1215 1265 1315 The first is the case of constant viscosity. In this
Temperature.v'C case, crystallization is occurring, but the viscosity
is held constant at the value the magma had at the
Figure 7 Calculated density of the model basalt. This den- time of intrusion; hence, there is no effect on the
sity includes all the phases present as both crystals and liquid.
viscosity due to increasing crystallinity or chang-
ing composition and temperature. Although this is
The melt fraction-temperature relationship clearly unrealistic for magmas, it provides a useful
used for the model country rock was that of Rutter end-member result. The Nusselt number rises rap-
and Wyllie (1988) who melted a tonalite under idly to a maximum of about 1.77 and then de-
vapor-absent conditions (Fig. 1). They do not re- creases as the magma reservoir temperature de-
port enthalpy changes associated with the melt- creases and the country rock heats up. The other
ing and so a constant value of 1.5 X 10' J . kg 1 three curves correspond to cases where the vis-
was used. This value is consistent with those cal- cosity varied as a function of critical solid frac-
culated by Hon and Weill (1982). Temperature- tion, <PM' The curve for a critical volume fraction
dependent thermal conductivity curves for both of 0.7 is queried as there appeared to he abrupt
the basalt and the tonalite protolith were taken and unreproducible oscillation in the thermal flux,
from Touloukian et al. (1981).
We initialize the model by assuming a pressure
of 8 kbar with an ambient temperature of TCR
0e
equal to 450 and a basalt liquidus, T'M equal to
1315°C. After a short period following intrusion 1.8

to allow the cold boundary layer at the roof to de-

velop, the initial instability was generated by in-
...
<l> 1.6
..c
troducing a random thermal perturbation of 1 to C
3°e at each grid point near the basalt-country
=
c 1.4
oW
?
0:;
rock boundary. These rather high perturbation val- '"
ues were required as the presence of variable vis-
'"= 1.2
Z
cosity rendered the cooling boundary layer very
stable. Once convection started, the form and dis-
position of the instabilities were independent of the 0.8 ......._-..L.~~ ........ ~ ........._ _......._ - - . I

location and magnitude of the initial perturbations. o 0.05 0.1 0.15 0.2 0.25
Many of the questions relatedto the nature of 2-Jii
the coupled crystallizing basalt and melting of the I
country rock can be resolved by considering the Figure 8 Nusxelt number as a function of the dimensionless
time-dependent Nusselt number for the system. In conductive cooling time. The thickness of the melt layer is J.
this instance the Nusselt number is defined to be the thermal diffusivity is K, and time is I. Four curves are
the ratio of the total heat transfer in the presence shown: one for the case where viscosity is held constant and
three that correspond to the viscosity-temperature relations
of convection to that which would occur under
given in Figure 3. The curves for critical volume fractions, <PM'
conditions of conduction only. This ratio was cal- of 0.3 and 0.5 are practically indistinguishable from each
culated using the thermal flux determined at the other. See text for a discussion of the query on the curve for
contact between the magma layer and the country <P" equal to 0.7.
 304 George W. Bergantz and Ralph Dawes

perhaps indicating the abrupt destruction of the The partitioning of the fluid into a region with
stagnant layer at the roof (see discussion follow- a rigid lid where high viscosity inhibits convec-
ing); because this was not reproducible, that curve tion, and an adjacent region where convection can
should be interpreted with caution. The cause of occur, has been noted in other studies of variable-
this behavior is still under investigation; we will viscosity convection (Chen and Pearlstein, 1988;
not discuss it further. Chu and Hickox, 1990; Ogawa et al., 1991; Olson
The curves for critical volume fractions of 0.3 et al., 1988; Richter et al., 1983). Of particular
and 0.5 fall practically on top of each other in relevance are the studies of Jaupart and Parsons
Fig. 8. Both have Nusselt number values so close (1985) who examine the case where a variable-
to 1 that it is not apparent from Fig. 8 that there is viscosity fluid layer is cooled only from above,
any difference between them. Under these condi- and Brandeis and Jaupart (1986) and Smith (1988)
tions, even though convection is occurring, the who also add the complication of simultaneous
heat transfer to the country rock, and hence the crystallization. Smith (1988) uses linear stability
progress of melting. is as if the magma were cool- analysis to determine the critical Rayleigh number
ing by conduction only. This is despite the fact and demonstrates that, in the ncar-critical regime,
that the initial temperature difference between the high aspect ratio cells that may not penetrate to the
magma and the country rock was 865°C. For these bottom of the chamber form and that the appropri-
model runs the region of the magma in which con- ate length scale represents a balance between ther-
vection is initiated is where the viscosity is vary- mal diffusion and propagation rate of the solidifi-
ing by no more than an order of magnitude, and cation front. For supercritical conditions where the
more typically by no more than a factor of 2, and critical Rayleigh number is substantially exceeded,
where the crystallinity is quite low. This indicates it is estimated that the convection will still be gen-
that convection is occurring in a region where the tle, as the convective vigor is modulated by vari-
temperature differs by no more than a few (one to able viscosity and the rate-limiting steps involved
five) degrees. Another important observation was with heat transfer out of the roof by conduction.
that the Nusselt number decayed monotonically; The results obtained from our modeling vali-
there was no melting back or resorption of the date the suggestions of Smith (1988) and are in
high-viscosity sublayer. agreement with the laboratory ex periments of Ber-
The disposition of the isotherms and plume gantz (1989a). The convection is never vigorous,
morphology was similiar to that seen in the study although it certainly does occur. The high aspect
of variable viscosity convection by Olson et al. ratio cells predicted by Smith (1988) were ob-
(1988); see their Figs. 17, 18, and 21. For all the served, but they appear on an intermittent basis,
cases considered in this study, the convection which is to be expected in the supercritical re-
formed plumes that fell from the leading edge of gime. This type of behavior has been seen in geo-
the crystallizing front. There was no regularity in physical models of lithospheric convection below
the positions from which the plumes originated or the oceanic crust where the mantle has been mod-
in the numbers of plumes that might exist in a eled as a variable-viscosity fluid (Buck and Par-
fully developed state at any time; the plume mor- mentier, 1986). These results are also consistent
phology was broadly like that of classic thermals. with the models of magma chamber convection
The flow field was always irregular but not tur- proposed by Marsh (1989), who addresses some
bulent; the maximum vertical velocity associated of the petrologic implications and details of the
with this convection was of order 10 4m . s - I, parameterization of convection.
and typically much smaller, about 10 5 m . s I. The calculated progress of melting of the coun-
Despite these low velocities the fluid motion was try rock due to the presence of a convecting basalt
able to keep the interior largely isothermal, prob- layer follows directly from the low values of the
ably due to the small temperature contrast associ- Nusselt number. For the cases where the critical
ated with convection. Although the flow would on melt fraction cP\1 was 0.3 a~d 0.5, the amount of
occasion be organized into cells with high aspect melt generated at the intrusive contact and the way
ratios (length/width), which did not penetrate to it changed as melting proceeded were indistin-
the floor, it was more common for the return flow guishable from that calculated in the conduction-
to be dispersed and irregular. only models (Bergantz, 1989b). It thus appears
 13. Magma Generation in Continental Lithosphere 305
that within the framework of the assumptions in- aid melt production in the country rock, which we
voked in this model, calculating the progress of have not considered. Other geometries, such as
melting following a single episode of underplat- the multiple dikes depicted by Grove et al. (1988),
ing can be facilitated by assuming a simple, but may yield regions of overlapping thermal gradi-
still nonlinear, conductive cooling model. These ents and enhanced melting. It was also found in
results clearly indicate that multiple intrusions are the course of a sensitivity analysis that some vari-
required to thermally mature a deep crustal sec- able viscosity-temperature relationships did not
tion to incite melt generation on a regional scale. yield a steady-state viscous lid and whole-layer
However, it should be remembered that most geo- convection occurred. Although these viscosity-
chemical models of partial melting (see preceding temperature relationships are not like those calcu-
discussion) require melting in the range 10-35%, lated by the MELTS code for use in this example,
which is like that estimated here and in Bergantz it is not unreasonable to speculate that there is more
(I 989b ). The compositions of melts produced by to be learned about variable-viscosity convection.
wholesale melting of the lower crust deviate from Similar results, although on a system that was
naturally occurring compositions (Beard and Lof- heated from below, were found by Ogawa et al.
gren, 1991) and models that would predict melt- (1991). A complete parametric treatment of our
ing in excess of 50% should be invoked with cau- work on variable-viscosity convection, crystalli-
tion. Thus, there is no justification for models that zation, and partial melting will appear elsewhere.
suggest that the incipient stages of partial melting The results of our numerical experiments differ
involve bulk melting of the lower crust. from the laboratory experiments of Huppert and
Our results are also in keeping with obser- Sparks (1988) and Campbell and Turner (1987).
vations made during large-scale melting experi- These deservedly influential papers evaluate a
ments. An artificial magma chamber, or melt layer, number of the fluid dynamical and petrological
3 X 1.5 m was generated by melting soil at the consequences of basaltic underplating. We agree
Oak Ridge National Laboratory (Jacobs et al., with many of the geological implications for melt
1992). The melting was initiated and sustained by production and the origin of petrologic diversity
applying electrical power to the ground through discussed in these works. However, our contin-
four graphite electrodes. The melt pool convected uum models yield estimates of time and length
vigorously as it was thermally "pumped" and su- scales of magma generation and cooling that are
perheated conditions were attained: melt tempera- quite different from theirs. The important differ-
tures reached 1500°C. However, once the power ences are that Huppert and Sparks (1988) predict
was shut off, convection could be driven only by very turbulent conditions in the magma, leading
heat losses to the environment under conditions of to rapid cooling and concomitant bulk melting of
thermal decay; convection ceased once the liqui- the overlying country rock. The differences can be
dus of the melt was attained (Dunbar, 1993; oral ascribed to the following: (a) we explicitly model
communication). This is an important demonstra- strongly variable-viscosity flow, and as impor-
tion of the fact that the rheological conditions as- tantly, (b) the melting and solidification relations
sociated with the onset of crystallization control between the country rock and the model basalt in
the subsequent dynamic evolution of the body. our work are like those in natural systems. We
It should be noted that a number of conditions preface our discussion by noting that magmas
that could alter the broad conclusions of our model are rarely superheated and that the melt fraction-
might exist. If the magma chamber experienced temperature curves used in models must retain the
reintrusion and subsequent disruption of the up- relationships between melting intervals between
per viscous boundary layer, then a pulse of melt- basalt and country rock as shown in Fig. 1: the
ing might follow. If the roof were not infinitely underplated material must have a higher liquidus
rigid, as assumed here, it could buckle under its temperature and a narrower and higher solidus-
own weight and delamination of the roof could to-liquidus temperature range than the country
occur. This would abruptly bring together par- rock.
tially melted country rock with near liquidus tem- The laboratory experiments of underplating of
perature magma, yielding a condition that could Huppert and Sparks (1988) use a superheated
lead to magma mixing. Any volatile release will aqueous solution of NaN0 3 , underplated beneath
306 George W. Bergantz and Ralph Dawes

a wax (PEG 1000). We were unable to find any Melt Collection and Ascent
reference to the the actual weight percent of
NaNO, for the experiments of Huppert and Once partial melting has been initiated, a gradient
Sparks (1988); however, the liquidus slope of the in melt fraction will exist away from the under-
NaNO, solution varies between 0 and - 18.1°C plating interface. Wickham (1987) has considered
(Kirk-Othmer, 1978). The NaNO:, was substan- a number of the controls on the segregation and
tial1y superheated to 70°C before underplating. transport of magma during partial melting; we
The PEG wax used as the country rock has a melt- will focus largely on refinements and new para-
ing interval between 37 and 40°C and at the start digms. A number of processes of homogenization,
of the experiment was isothermal with a tempera- segregation, and transport can occur during partial
ture of 20°e. Thus, in no circumstances could melting: convective overturn in the style of porous
the underplated "basalt" solution crystal1ize in medium convection; magma mixing involving un-
the models, and hence no rheological penalty for derlying partial1y molten basalt; bulk convection
crystallization is possible. This is unlike the geo- of the partial1y molten region; compaction of the
logical case in many important respects, where matrix, yielding discrete melt bodies; and, finally,
basalt crystal1izes at temperatures above the am- melt (and matrix) transfer by virtue of diking or
bient temperatures of the country rock. The labo- diapirism.
ratory roof melting experiments of Campbel1 and
Turner ( 19R7)show more complex behavior as si- Melt Convection in a Porous Medium
multaneous crystallization and melting occur in
some of their experiments. However, the model Once a partly melted region is generated. a variety
basalt in their experiments is also superheated, of dynamical states are possible: simple convec-
and hence the same cautions detailed above ap- tion of the melt in a rigid matrix, movement of
ply. We hasten to add that the experiments of Hup- melt and matrix due to compaction, or bulk con-
pert and Sparks (19R8) and Campbel1 and Turner vective motion of the region of partial melting.
(1987) are wel1 motivated and that it should be The first is the most simple: it is assumed that the
appreciated that getting the right combination of unmelted residuum forms a rigid matrix. This
material in any analog model of underplating is may have application to magma chambers that
very difficult. are composed of regions of largely crystal-liquid
In summary. numerical experiments using geo- mush, such as that now proposed for mid-ocean
logically relevant thermophysical properties indi- ridges (Nicolas et al., 1993). This is appropriate
cate that partial melting from underplated basalt for conditions where the melt fraction is below
wil1 be largely conductive. in the early stages of about SOIk , although this parameter is very poorly
underplating. This is in agreement with many fea- constrained (Bergantz, 1990). Assuming that an
tures from the rock record, including measured interconnected network forms, which can be for
amounts of partial melting at plutonic contacts as little as 2 wto/c·melt for a mixture of olivine and
and the compositions of natural1y occurring gran- basalt powder (Daines and Richter, 19RR), the mo-
itoids. We concur with the important conclusions tion of the melt may be considered flow in a po-
from Huppert and Sparks (1988) regarding the rous medium, and the formalism for convection in
thermal and compositional evolution of regions of a porous medium may be applied as noted by
underplating: if further underplating occurs, the Lowel1 (1982). For the underplating conditions
region wil1 thermally mature. and reach a condi- described previously, where the region undergo-
tion where large amounts of melt can be generated ing partial melting is being heated from below and
by a modest increase in temperature caused by in- growing upward, the permeability will be aniso-
cremental addition of magma to the region of par- tropic and heterogeneous and may wel1 produce
tial melting. Neither the model presented here nor local reversals in density. due to the presence of
those in Huppert and Sparks (1988) address these water, which will also affect the density. There is
conditions in a rigorous way. Without some geo- no theoretical formalism to predict the onset of
logical evidence, constructing a model that could convection for these geological1y relevant condi-
be generalized would be difficult. tions. However, if one assumes that permeability
 13. Magma Generation in Continental Lithosphere 307
is homogeneous and isotropic, the model relation- able porosity was considered by David et al.
ships of Lowell (1982) can be used. These are (1991), and variable viscosity of the melt by
summarized in the following expression, which Blythe and Simpkins (1981). Given the nonlinear
specifies the dependence of the required thick- nature of these effects, it is not possible to com-
ness of the melt region, X(t), on other system bine the results of these different studies and to
properties, assess their importance for convection in partial
melts. None of the current studies considers the
2887T3K IJ C
X(t) 2:: p , (10) conditions for convection in a growing melt layer
f3gX3b2L . Ste where the permeability and melt viscosity are
where K is the thermal diffusivity, IJ the kinematic changing as a function of temperature. In sum-
viscosity, c, the specific heat, f3 the thermal ex- mary, it is unclear whether a distributed melt will
pansion coefficient, g the scalar acceleration of be homogenized on anything but small scales.
gravity, b the average grain diameter, L the latent Whether melt homogenization occurs depends
heat of melting, and Ste the Stefan number modi- most importantly on melt viscosity, which will
fied by multiplying the latent heat term by the vary dramatically as a function of volatile content
melt fraction. For typical values of these quanti- and hence melt progress.
ties and a constant melt fraction of 0.3, Eq. (10)
is plotted in Fig. 9. Low-viscosity melts may be- Compaction
come homogenized over scales of a few meters to
tens of meters, which is the length scale over If the unmelted matrix is not infinitely rigid, com-
which the model assumptions may be relevant. paction and subsequent reorganization of the par-
For more viscous melts, the layer thickness will tial melt region can occur. Compaction refers to
exceed any reasonable length scale of lithologic the change in melt fraction that occurs as buoyant
homogeneity and convection cannot be simply interstitial fluid moves upward and solid material
characterized by an expression like (10). is displaced downward. The governing equations
Expression (10) is based on model assumptions that describe compaction are developed in the
that severely limit its application to crustal melting context of mixture theory or averaging techniques
except as an end-member estimate. A number of and hence the calculated values of the dependent
additional, potentially relevant complexities have variables represent averages at scales that are
been addressed in the engineering literature: Ka- large relative to individual grains but small rela-
viany (1984) considered the case where the lower tive to the scale of the gradients in the dependent
boundary temperature is increased linearly. vari- variables. This is the usual approach in the con-
tinuum description of porous media and mixed
phase processes and does not limit the utility of
10
4 the model for most problems of geologic interest.
E
The physical processes and petrologic implica-
....
~ tions of this form of melt segregation have been
~ considered in the context of melt migration in the
'"
'"
OJ
mantle (Fowler, 1985, 1990a, 1990b; McKenzie,
c
..:.: 1984, 1985; Ribe, 1985; Scott and Stevenson,
<:.J
;S 2
1986; Sleep, 1974), in magma chambers (Shirley,
..c 10 1986, 1987), and in the crust (Fountain et al.,
'"
::l
1989; Lowell and Bergantz, 1987; McKenzie,
:;
1985; Wickham, 1987).
10
10. 3 10. 2 1O,! 1 A succinct summary of the equations describ-
2 ing compaction is given by Ribe j 1987). Of inter-
Kinematic viscosity, m s·1
est here is the case where compaction is occurring
Figure 9 Critical mush thickness for porous media convec-
tion to begin as a function of kinematic viscosity. For these
in a region of variable partial melt and with gra-
curves the grain diameter, b. is in centimeters. Ste is the Stefan dients where the partial melt fraction goes to zero.
number. After Lowell (1982). Under these conditions, melt can collect into "sol-
 308 George W. Bergantz and Ralph Dawes

itary waves," dubbed solitons or magmons (Scott tures not explicitly considered in the model of
and Stevenson, 1984; Stevenson and Scott, 1987), Fountain et al. (1989) are the specifics of thermal
where under certain conditions the melt fraction or compositional magmatic convection; e.g., the
can go to 100% and fully liquid magma chambers model is spatially one-dimensional. These authors
can ostensibly form. The model of Fountain et al. argue that the density will be dominated by the
(1989) demonstrates this in the context of melt influence of crystals falling off the roof, an ad hoc
migration during crustal anatexis following un- but not unreasonable assumption once the grow-
derplating by mafic magma. Provided the melt ing melt layer is itself largely crystal free, Lowell
fraction at the contact exceeds 25%, it is demon- and Bergantz (1987) also consider compaction of
strated that compaction can yield melt bodies with a growing partial melt layer and conclude that the
dimensions on the order of kilometers extending compositional effects on density may yield insta-
from the contact to a few kilometers above the bilities in the growing melt layer (see discussion
contact. These length scales are in rough agree- on porous media convection above) and hence
ment with some geological examples: however, yield conditions in the region of compaction that
the upward migration is limited to regions that would be hard to generalize and would require a
have undergone partial melting. Figure 10 shows multidimensional model to adequately represent
the calculated distribution of partial melt. Fea- the structure of the fluid and solid flow.

iOO,OOO yrs 300,000 yrs Diapirism and Diking

u The crustal scale transport of magma is often con-

-Eo 4 sidered in terms of two end members: diapirism
U and diking. This is somewhat surprising as there
3
s
Q)

.0
is little theoretical or unambiguous geological evi-
dence to suggest that either of these mechanisms
<1
alone is dominant in the crustal transfer of silicic
melts. Diapirism is a type of Rayleigh-Taylor in-
stability and has traditionally been invoked to ex-
plain the growth and disposition of salt domes, the
~=""-~--'--~-,JOO oo!-----'"-~--'--~
50 100 spacing of volcanoes (Marsh, 1979), and the as-
cent of granitic magma in the crust (Miller et al.,
% Liquid
1988). The conditions leading to the onset of the
instability are developed by Turcotte and Schu-
bert (1982) and more recently by Lister and Kerr
500,000yrs 700,000yrs
(1989) who caution that variable physical proper-
-u
.E
c:
4
ties, geometries, and the presence of a deformable
lower layer render many of the more classical es-
o
U timates of the wavelength and rise time of the fast-
Q)
est growing instability inappropriate. These con-
s
.0 ditions may be present in partial melt regions that
<I:
...
en
Q)
have vertical lithological and rheological varia-
tions where a region of greater partial melt is
Q3
E sandwiched between regions with smaller amounts
..Q
::>:: of partial melt and the viscosity contrasts are not
00·'-----'--~--'---1-,J00
50
0 0L---'--~--'---~
50 100
large between the layers.
The important rate-limiting element in mag-
% Liquid
matic diapirism lies in the thermal coupling of the
Figure 10 Distribution of granitic melt in the country rock at
diapir and the surrounding rock. It is thought that
four different times, 100,000.300.000.500,000, and 700,000 yr
after intrusion. The stippled area is defined by the curve of
heat loss from the magma to the surrounding rock
percentage liquid versus distance from the intrusion. From lowers the viscosity of the country rock allowing
Fountain et al. (1 \lX9). it to "flow" around the magma body. Implicit in
 13. Magma Generation in Continental Lithosphere 309
this model is that there is a finite distance that the bly isotropic fabric internally. An element not in-
magma can rise before the enthalpy difference be- cluded in this or other models is the inevitable
tween the diapir and the surroundings diminishes crystallization that must occur at the boundary as
to the point where the density and viscosity con- diapirism proceeds. This crystallization will effec-
trasts are no longer sufficient to permit continued tively act to move the isotherm associated with in-
ascent. This is the basis of what has become ternal slip inward, yielding a telescoping system
known as the hot Stokes model of diapirism. The where fractionation can occur by virtue of ascent,
quantitative description of the model has been de- and the time history of fabrics will be even more
veloped by Marsh (1982), Morris (1982). Daly complex than that given by Cruden (1990). These
and Raefsky (1985), and Mahon et al. (1988). It considerations cast some doubt on the notion that
should be noted that when the hot Stokes model kinematic indicators taken from intrusions and the
of diapirism was first proposed by Marsh and oth- surrounding rock can discriminate between dia-
ers. the intended application was melt transfer in pirism, as understood in the classic sense, and
the mantle. magma chamber development following initial
The hot Stokes model of diapiric transfer of transport in a dike.
magma contains features that. in principle, should Mechanical and thermal models of magma
be testable by geologic observation (Bateman, transport in dikes are considered in this text. It is
1984; England, 1990; Paterson etal., 1991). How- important to note, however, that some models of
ever, the ambiguity in uniquely identifying strain melt transfer in dikes suggest that it is a very ef-
fields that record diapiric ascent as opposed to the ficient means of magma transfer. Clemens and
fabric generated during (final) emplacement pre- Mawer (1992) present a model where it is esti-
cludes a direct interpretation in many cases. Pat- mated that a 2000-km3 batholith can be filled in
erson et al. (1991) have considered a wide variety less than 900 yr. Although this model requires
of natural examples and conclude that the width some end-member assumptions, it demonstrates
of aureoles and the intensity of wall rock strain the kinds of time scales that may be possible if
around granitoids are generally not as great as magma is transported in dikes.
those required by laboratory or numerical models
for diapirism. England (1990; 1992) argues that Tectonic Regimes and Magma Ascent
the fabrics generated during an emplacement pro-
cess, such as "ballooning," may make it difficult The previous discussions considered simple sys-
to recognize the signature of diapirism. This con- tems where the material surrounding the melt was
dition is exacerbated if the diapir is composition- considered to be a mechanical continuum, and the
ally nonuniform, because a number of mixing role of tectonism and the rheological partitioning
conditions can arise if there are multiple mag- of the crust were not explicitly included. Conflict-
matic elements in the diapir, leading to strain lo- ing and ambiguous geological evidence indicate
calization and nonuniformity in the thermal and that this approach, while amenahle to modeling, is
compositional fields. Weinberg (1992) notes that not sufficient to account for the variety of scales
reverse zoning would be the most common pluton of melt extraction and ascent that are observed in
geometry if the zoning were controlled only by the rock record. For example, on the basis of the
internal circulation. However, most zoned systems compositions of migmatites and partial melts of
have a normal zonation pattern that progresses the midcrust, Sawyer (1991) provides evidence
from more mafic to felsic, and it appears that an that the rates of melting and melt extraction often
internal circulation driven by internal shear cou- are greater than the rates at which chemical equi-
pling may be of limited importance in generating librium can be achieved in the melt. Thus, it may
zonation patterns. Isothermal experiments of dia- be inappropriate to model many important melt-
pirism by Cruden (1990) reveal some of the fab- forming reactions with an assumption of thermo-
rics that can occur within an upward moving melt dynamic equilibrium. There are numerous studies
body by virtue of the viscous coupling to the that demonstrate the broadly syntectonic nature of
country rock. The deformation is strongly time magma ascent. Some studies even infer that a
dependent and a variety of fabrics can form with feedback exists between magma generation and
penetrative foliations at the margins and a possi- tectonism. It is apparent that the material transfer
310 George W. Bergantz and Ralph Dawes

required to accommodate the ascent of magma oc- axis of the strike-slip faults, and the occurrence of
curs simultaneously at a variety of scales; there- syntectonic fabrics whose orientation is coherent
fore, models that attempt to describe the ascent on a pluton scale and consistent with an origin by
and magma chamber assembly process only on dextral shearing.
the basis of the near field deformation will neces- Mineral fabrics provide the only direct means
sarily be incomplete. for the determination of the kinematic conditions
The possible role of crustal scale structures during ascent, emplacement, and subsequent de-
controlling the ascent and emplacement of mag- formation. These fabrics are defined by mineral
mas is typically determined by the use of kine- orientations and their mechanical condition. In
matic indicators measured in and at the margins of general, a plutonic rock can have a fabric that is
plutons and the disposition of magmatic centers the result of a sequence of stress fields, from those
relative to the positions of structural features. A generated during flow in the magmatic state, to
hypothesis linking the ascent of magmas and the near solidus and subsolidus brittle fabrics due to
development of crustal scale structural features regional stresses. The use of kinematics to deter-
where magmas can possibly follow up and into mine the flow in magmatic rocks has been re-
any zones of weakness created by faulting is at- viewed by Nicolas (1992). The criteria for the de-
tractive in that it appears, on the face of it at least, termination of those elements of a rock fabric that
to provide a solution to the "room problem" in the have an origin by magmatic flow, as opposed to
near field. We are not suggesting that fault bends solid-state deformation, are reviewed by Paterson
produce "holes" in the crust that magmas then et al. (1989). The criteria involve observations at
pour into, but rather that magmas can intrude along the outcrop and thin section scale: alignment of
shear zones while they are active. With this type of near liquidus phases indicates flow where crystals
intrusion mechanism the need for a crustal scale are dispersed in a fluid medium. Tiling of crystals
ductile halo as required for diapiric rise is obvi- can also indicate magmatic flow with a high solid
ated. Gravity studies indicate that the form of gra- fraction. The presence of plastic strain such as
nitic plutons can be correlated with the tectonic kinking, undulatory extinction, and crystal re-
style and occurrence of shear zones: intrusion dur- growth indicates deformation in the solid state.
ing extension yields thin plutons composed of a When deformation is taking place near the critical
number of subunits and intrusion during transcur- melt fraction, the textures can be ambiguous as a
rent shear yields a few feeder zones that lie off the transition from magmatic to solid-state flow may
shear zone (Vigneresse, 1994). occur. Establishing the sense of shear under these
The structural control of the intrusion of plu- conditions, which are of particular interest in mid-
tons is further suggested by the frequent occur- to deep-crustal exposures where magma genera-
rence of plutons in linear belts as noted by Pitcher tion and extraction may be occurring, is complex
and Bussell (1977) in the Andes. This tectonic and the interested reader is referred to the study of
control on the character of magmatism at conver- Blumenfeld and Bouchez (1988) who consider
gent margins has been considered in the broadest deformation in migmatites and partial melts. The
sense by Glazner (1991) who suggests that plu- magnetic susceptibility anisotropy of plutonic
tonism is favored when the tangential component rocks has also been used to determine flow direc-
of convergence is large, and volcanism is favored tion when other means of determining the fabric
when the normal component of the convergence are inadequate or ambiguous. General agreement
vector is dominant. However, no specific mecha- is found between the magmatic mineral fabric and
nism for this correlation is proposed. Considering the orientation of the magnetic susceptibility an-
tectonic elements at a similar scale, Tikoff and isotropy (Bouchez et al., 1990),
Teyssier (1992) provide a model where pluton- Syntectonic mineral fabrics, that is, mineral
ism and batholith assembly occurs in zones of fabrics with a component that can be ascribed to
dilatation that accompany the development of en both magmatic and regional shear during solidifi-
echelon P-shear arrays, which result from a trans- cation, provide compelling evidence that magma
pressional strain regime created by oblique con- is, in some instances, intruding during tectonic
vergence. The evidence used to support this is the movement. Crustal "openings" and bends, which
linear distribution of intrusions that lie just off the are the inevitable consequence of shearing, pro-
 13. Magma Generation in Continental Lithosphere 311
vide access for magmas that reside at deeper crustal noted in the discussion on diapirism, they may
levels. Hutton (1988, 1992) and Hutton et al. record mechanical conditions quite different than
(1990) review a few case histories where magma those that existed at the time of melt extraction
has intruded along shear zones both vertically and ascent. Nonetheless, these fabrics are the only
and nearly horizontally. Antonellini and Cambray structural information directly available.
(1992) document magma transport by the process The tectonic control of the distribution of vol-
of stepping up along bedding-parallel shear zones canic vents at both regional and local scales is evi-
in rift systems. Although these models of magma dent, although complex, perhaps due to the redi-
transfer along shear zones appear to provide a rection of regional stress fields in the near surface.
resolution to the space problem in the near field, This is well documented in the study of Bacon
the space must be accommodated elsewhere, os- (1985) who also notes the correlation between
tensibly by thrusting or extension in the far field. earthquakes and the timing of eruption. More com-
Thus, the rate-limiting steps in magmatic ascent plex relationships between the clustering of cinder
will involve some combination of regional stresses cones and the presence of fault zones are recog-
and local buoyancy forces related to the magma nized in Mexico (Conner, 1990). Vent clustering
density contrast. was found to be pervasive and the azimuths of the
Among the more provocative corollaries of clusters appear to be dominantly controlled by the
such models are the suggestions that shearing it- convergence direction, and less so by the orienta-
self may act as an agent of melt production, or tion of faults. The long-standing conundrum in
conversely, the presence of melt may act to local- this regard is the spacing and linear arrangements
ize shear. Karlstrom et al. (1993) document melt of volcanic centers in arcs (Jarrard, 1986; Marsh,
extraction in conjunction with thrusting in a small 1979; Sherrod and Smith, 199.0). The positions of
midcrustal pluton: tectonic movement was instru- these centers often persist for millions of years,
mental in driving melt segregation within the erupting a variety of magma types with no clear
pluton. Hutton and Reavy (1992) propose that association with faulting, suggesting instead a
strike-slip transpression thickens the crust, yield- control due to processes in the magma supply re-
ing anatectic granitic magmas that will have the gion. Or it may be that the crust is sufficiently
usual geochemical signature of crustal melts. This fractured that magmas can at any time exploit an
thickening may also yield an undulating Moho, available shear zone, and so the connection be-
which itself may undergo melting and mixing with tween magmatism and shear zones potentially
mantle diapirs, yielding magmas with geochemi- tells us little about the rate-limiting elements in
cal attributes that indicate interaction with mantle- crustal magmatism.
derived materials. Thus, thrusting and transpres- Thus, the distribution and disposition of plu-
sion provide a means to generate magmas whose tons often indicate some form of structural con-
geochemical indicators suggest diverse origins. trol, while volcanism reflects structural control in
although additional geologic tests need to be ar- some cases and not in others. One feature that
ticulated to fully test this hypothesis. The pres- both plutonic and volcanic systems do share is
ence of melt in shear zones may also act to en- some degree of open system behavior. This is ex-
hance and/or localize deformation as argued by pressed in plutons as repeated intrusion and plu-
Davidson ct al. (1992), who combine a thermal ton assembly in an incremental fashion (Harry
model with kinematic indicators of shearing to and Richey, 1963) which can yield a wide variety
demonstrate that large amounts of crustal strain of zoning patterns (Bergantz, 1990) as well as
can be accommodated by regions of melt. One chemical evidence of magma mixing. In the vol-
key element in this mechanism is that the crust is canic record it is expressed by the diversity of
at near solidus temperatures and hence the melt magmas erupting from a single center: as indi-
layer has a long thermal lifetime. vidual units often unrelated geochemically in any
It should be appreciated that kinematic indica- simple way, or mingling with adjacent magma
tors interpreted from the mineral fabric record the bodies as demonstrated in the spectacularly ex-
last episode of the crystal-melt organization in posed Quaternary arc volcano, Tatara-San Pedro
magma. They record the end stages of the process, in the Chilean Andes (Dungan, 1992).
whether one of ascent or emplacement, and as It appears then that the time-integrated geo-
312 George W. Bergantz and Ralph Dawes

dynamic and petrologic expression of magmatism propagates upward in the crust, the conditions
might be that of a crustal scale conduit rather than and style of magmatic ascent will change as the
a single chamber (Singer et al., 1989). Implicit in temperature field changes both spatially and tem-
this conduit model is that magmatic systems are porally. The success of the methods taken from
assembled from the bottom up, which yields a structural geology in illustrating some of the kine-
downward and outward flaring thermal anomaly. matic aspects of magma ascent and the connection
The region from which magmatic contributions between rheological conditions and temperature
can be generated by partial melting is potentially suggest that a geodynamic approach to the quanti-
much larger at the bottom than at the top and tative description of magmatic processes may pro-
petrologic diversity is generated in this case by vide the missing link in developing scientifically
crystal-liquid separation by virtue of ascent and sound models of crustal scale magma genesis.
open system processes. The style of ascent will
vary from processes that are ductile in the deep
crust to diking and intrusion along fault-generated
Acknowledgments
bends in the upper crust as rheological conditions
will migrate upward with the thermal anomalies.
This work was supported by National Science Foundation
As the thermal anomaly that is driving magma- Grants OCE-90009993 and EAR-9019217 to G. W. B. We are
tism in the deep crust migrates upward, it will grateful for helpful reviews by C. Wayne Burnham. Juliet
"consume" the geological evidence of the early Mckenna, Peter Reiners, Donna Whitney. and in particular.
stages of magma generation, ascent, and assem- Steve Wickham. These reviews sharpened our discussion con-
siderably, and we acknowledge that our interpretations are not
bly. The end result may be a thermal anomaly on
shared by all the reviewers. The patient and detailed editorial
a crustal scale like that interpreted from compres- assistance of Michael P. Ryan is gratefully acknowledged.
sional wave velocity studies at Long Valley, Cali-
fornia (Dawson et al., 1990). Geological evidence
of the early stages of this process would have to
be found in regions where this process has failed, References
for example, an immature episode of underplating
Antonellini, M. A.. and Cam bray. f. W. (1992). Relations
as discussed previously. If melt is generated by
between sill intrusions and bedding-parallel extensional
anatexis during transpressive orogenesis, a similar shear zones in the Mid-continent Rift System of the Lake
style of melt generation and ascent may result. Superior region, Tertonophysics 212, 3.11-349.
D'Lemos et al. (1992) document these transi- Arndt. N. 1'.. and Goldstein, S. L. (19X9). An open boundary
tions in intrusive style with rheological state on a between lower continental crust and mantle: Its role in
crust formation and crustal recycling, Tectonophysics 161,
crustal scale where the important physical ele-
201-212.
ments linking regions of melt generation and final Ashworth, J. R.. and Brown, "1. (I 99()). An overview of diverse
pluton formation are "rnegadikes." responses to diverse procev-es at high crustal tempera-
Given the temporal complexity in the spectrum tures, ill "High-Temperature Metamorphism and Crustal
of chemical compositions within single centers, Anatexis" (1. R. Ashworth and M. Brown, eds.). pp. i-IX.
Lnwin Hyman. London.
and the presence of intermediate magmas in the
Bacon. C. R. (] 9R5). Implications of silicic vent patterns for
midcrust, it is clear that we need to move beyond the presence of large crustal magma chambers, J. Geo-
paradigms of chemical petrology that rely solely phvs. RES. 90,11243-11252.
on the concept of parental magma chambers as Bacon, C. R. (1992). Partially melted granodiorite and related
sites for the generation of petrologic diversity. rocks ejected from Crater Lake caldera, Oregon. Trans.
Roy. Soc. Edinburgh 83, 27-47,
Nor is it clear that simple mechanical models ade-
Barnes, H. A., Hutton, 1. E, and Walters. K. (19X91. "An Intro-
quately express the many scales at which magma duction to Rheology," Elsevier, New York.
generation, extraction, and ascent occur simulta- Bateman, R. (1984). On the role of diapirism in the segrega-
neously. As discussed previously, basaltic mag- tion, ascent and final emplacement of granitoid magma,
matism is a ubiquitous and seemingly necessary Tectonophvsics 110,211-231.
Bea, F. (Jl)l) 1). Geochemical modeling of low melt-fraction
element in the generation of crustal melts in some
anatexis in a peraluminous system: The Pena Negra Com-
tectonic regimes and represents the first of the plex (central Spain), Geochem. Cosmochim. ALta 55,
likely connections between the mantle and crustal IX59-IX74.
magma chambers. As this "master" perturbation Beard, 1. S., and Lofgren, G. E. (19X9). Effect> of water on the
 13. Magma Generation in Continental Lithosphere 313
composition of partial melts of greenstone and amphibo- Burnham, C W. (1979). Magmas and hydrothermal fluids, in
lite, Scienc e 244, 195 -197. "Geochemistry of Hydrothermal Ore Deposits" (H. L.
Beard, J. S., and Lofgren, G, E, (1991), Dehydration melting Barnes, ed.), 2nd ed., pp. 71-136, Wiley, New York,
and water-saturated melting of basaltic and andesitic green- Burnham, C W. (1992), Calculated melt and restite composi-
stones and amphibolites at I, 3, and 6.9 kb, 1. Petrol. 32, tions of some Australian granites, Trans. Roy. Soc, Edin-
365-401. burgh 83, 387-397.
Beckennann, C, and Viskanta, R. (1988). Double-diffusive Campbell, L H.. and Turner, J. S. (1987). A laboratory inves-
convection during dendritic solidification of a binary mix- tigation of assimilation at the top of a basaltic magma
ture, Physicochem. Hydrodyn. 10, 195-213, chamber. J, Geol. 95,155-172,
Beckermann, C, and Viskanta, R. (1993). Mathematical mod- Cashman, K. V. (1993). Relationship between plagioclase
eling of transport phenomena during alloy solidification, crystallization and cooling rate in basaltic melts, Contrib.
Appl, Mech. Rev, 46, 1-27. Mineral. Petrol. 113, 126-142,
Bennon, W. D., and Incropera, E P, (l9k7). A continuum Chappell, B. w., and Stephens, W. E. (1988). Origin of infra-
model for momentum, heat and species transport in binary crustal (I-type) granite magmas, Trans. Roy. Soc. Edin-
solid-liquid phase change systems. L Model formulation, burgh 79, 71-86.
Int. J. Heat Man Trans. 30, 2161-2170. Chappell, B. w., and White, A. J. R. (1992), 1- and S-type
Bergantz, G. W. (l989a). Thermal and dynamical state of the granites of the Lachlan Fold Belt. Trans. Roy. Soc. Edin-
crust following underplating: Implications for melt genera- burgh 83, 1-26.
tion and petrologic diversity, EOS Trans. Am. Geophys. Chappell, B. w., White, A.1 R.. and Wyborn, D. (1987). The
Union 70, 1320. importance of residual source material (restite) in granite
Bergantz, G, W. (I 989b). Underplating and partial melting: petrogenesis, J. Petrol. 28, 1111-1138,
Implications for melt generation and extraction, Science Chen, Y.-M" and Pearlstein, A1 (1988). Onset of convection
245,1093-1095. in variable viscosity fluids: Assessment of approximate
Bergantz. G. W. (1990). Melt fraction diagrams: The link be- viscosity-temperature relations, Phys. Fluids 31, 1380-
tween chemical and transport models, in "Modern Meth- 1385.
ods of Igneous Petrology: Understanding Magmatic Pro- Chen. Y. D.. Price, R. C, and White, A. J. R. (1990). Mafic
cesses" (1 Nicholls and 1 K. Russell, eds.), pp. 240-257, inclusions from the Glenbog and Bluegum granite suites,
Mineralogical Society of America. southeastern Australia,l. Geophvs. Res. 95,17757 -17785.
Bergantz. G. W. (1992). Conjugate solidification and melting Cheng, D. C.-H. (1984). Further observations on the rheo-
in multicomponent open and closed systems, Int. J. Heat logical behavior of dense suspensions, Powder Tech, 37,
Mass Tram. 35, 533-543. 255-273.
Blumenfeld, P.. and Bouchez, J. 1" (1988), Shear criteria in Cbu, T. Y., and Hickox. C E. (1990). Thermal convection with
granite and migmatite deformed in the magmatic and solid large viscosity variation in an enclosure with localized
stages, J. Struct, Geol. 10,361-371. heating, J. Heat Trans. 112,388-395.
Blythe. P. A, and Simpkins, P. G. (19k I), Convection in a po- Clemens. 1 D. (1984). Water contents of silicic to intermediate
rous layer for a temperature dependent viscosity, Int. J. magmas, Lithos 17,273-287.
Heat Mass Trans. 24, 497-506, Clemens, 1 D. (1990). The granulite-granite connection, in
Bohlen, S. R., and Mezger, K. (1989). Origin of granulite ter- "Granulites and Crustal Evolution" (D. Vielzeuf and P,
ranes and the formation of the lowermost continental crust, Vidal. eds.), pp. 25-36, Kluwer, Dordrecht.
Science 244, 326 -329. Clemens, 1 D., and Mawer, C K, (1992), Granitic magma
Bottinga, Y., and WeilL D. F. (1972). The viscosity of mag- transport by fracture propagation, Tectonophysics 204,
matic silicate liquids: A model for calculation, Am. 1. Sci. 339-360,
272,438-475. Clemens, J. D" and Vielzeuf, D, (1987). Constraints on melt-
Bouchcz, 1-1", Gleizes, G .. Djouadi, L and Rochette, P. ing and magma production in the crust, Earth Planet. Sci.
( 19(0). Microstructure and magnetic susceptibility applied Lett. 86, 287 - 306.
to emplacement kinematics of granites: The example of Conner, C B. (1990). Cinder cone clustering in the Trans-
the Foix pluton (French Pyrenees), Tertonophvsics 184, Mexican volcanic belt: Implications for structural and pet-
157-171. rologic models. 1. Geophys. Res. 95, 19395-19405.
Brandeis, G .. and Jaupart, C (1986). On the interaction be- Crisp, 1 A. (1984). Rates of magma emplacement and vol-
tween convection and crystallization in cooling magma canic output, J. Vole. Geotherm. Res. 20,177-21 L
chambers, Earth Planet. Sci. Lett. 77, 345-301. Cruden, A R. (1990). Flow and fabric development during the
Brown, G. C, and Fyfe, W. S, (l97(J), The production of gra- diapiric rise of magma, J. Geol. 98, 681-698.
nitic melts during ultrametamorphisrn, Contrib. Mineral. Cull, J. P. (I 990l. Underplating of the crust and xenolith
Petrol. 28, 310-318, geotherms in Australia, Geophys. Res. Lett. 17, 1133-
Buck, W. R., and Parmentier, E. M. (l9k6). Convection be- 1136.
neath young oceanic lithosphere: implications for thermal D'Lernos, R. S., Brown, M., and Strachan, R. A (1992). Gran-
structure and gravity. 1. Gcophys. Res. 91, 1961-1974. ite magma generation, ascent and emplacement within a
Burnham, C. W. (1967 J. Hydrothermal fluids at the magmatic transpressional orogen, J. Geol. Soc. London 149, 487-
stage, in "Geochemistry of Hydrothermal Ore Deposits" 490.
(H, L. Barnes, ed.), pp, 34-76, Holt, Rinehart & Winston, Daines, M. 1., and Richter, F. M. (1988). An experimental
New York. method for directly determining the interconnectivity of
314 George W. Bergantz and Ralph Dawes

melt in a partially molten system, Geophys. Res. Lett. 15, Frith, W. J., Mewis, J., and Strivens, T, A. (1987). Rheology
1459-1462. of concentrated suspensions: Experimental investigations,
Daly, S. E, and Raefsky, A. (1985). On the penetration of a hot Powder Tech. 51, 27-34.
diapir through a strongly temperature dependent medium, Fuis, G. S., and Plafker, G. (1991), Evolution of deep structure
Geol. J. Roy. Astron. Soc. 83, 657 -682. along the trans-Alaska crustal transect, Chugach Moun-
David, E., Lauriat, G., and Cheng, P. (1991). A numerical so- tains and Copper River basin, Southern Alaska, J. Geo-
lution of variable porosity effects on natural convection in phys. Res. 96, 4229-4253.
a packed-sphere cavity, J. Heat Trans. 113,391-399. Ghiorso, M. S. (1985). Chemical mass transfer in magmatic
Davidson, C, Hollister, L., and Schmid, S. M. (1992). Role of systems. I. Thermodynamic relations and numerical algo-
melt in the formation of a deep crustal compressive shear rithms, Contrib. Mineral. Petrol. 90,107-120.
zone: The Maclaren Glacier metamorphic belt, south cen- Gill, 1., and Condomines, M. (1992). Short-lived radioactivity
tral Alaska, Tectonics 11, 348-359. and magma genesis. Science 257,1368-1376.
Davidson, 1. P., McMillian, x J., Moorbath, S., Worner, G., Glazner, A, E (1991). Plutonism, oblique subduction, and con-
Harmon, R. S., and Lopez-Escobar, L. (1990). The Neva- tinental growth: An example from the Mesozoic of Cali-
dos de Payachata volcanic region. II. Evidence for wide- fornia, Geology 19, 784-786.
spread crustal involvement in Andean magmatism, Con- Grissom, G. C, DeBari, S. M., Page, S. P., Page, R. EN.,
trib. Mineral. Petrol. 105,412-432. Villar, L. M., Coleman, R. C, and Ramirez, M. V. (1991).
Dawson, P. B., Evans, 1. R., and Iyer, H. M. (1990). Teleseimic The deep crust of an early Paleozoic arc: The Sierra de
tomography of the compressional wave velocity structure Fimbala, northwestern Argentina, in "Andean Magmatism
beneath the Long Valley region, California, J. Geophys. and Its Tectonic Setting" (R. S. Harmon and C. W Rapela,
Res. 95, 11021-11050. eds.), Special Paper 265, pp. 189-200, Geol. Soc. Arn.,
Dell'Angelo, L. N., and Tullis, 1. (1988). Experimental defor- Boulder, CO.
mation of partially melted granitic aggregates, J. Meta. Grove, T. L., Kinzler, R. 1., Baker, M. B., Donnelly-Nolan,
Geol. 6,495-515. 1. M., and Lesher, C. E. (1988). Assimilation of granite by
DePaolo, D. 1., Perry, E v.. and Baldridge, W S. (1992). basaltic magma at Burnt Lava flow, Medicine Lake vol-
Crustal versus mantle sources of granitic magmas: A two- cano, northern California: Decoupling of heat and mass
parameter model based on Nd isotopic studies, Trans. Roy. transfer, Contrib. Mineral. Petrol. 99, 320-343.
Soc. Edinburgh 83, 439-446. Halliday, A. N., Mahood, G. A., Holden, P., Metz, J. M.,
Drummond, B. 1., and Collins, C. D. N. (1986). Seismic evi- Dempster, T. J., and Davidson, 1. P. (1989). Evidence for
dence for underplating of the lower continental crust in long residence times of rhyolitic magma in the Long Val-
Australia, Earth Planet. Sci. Lett. 79, 361-372. ley magmatic system: The isotope record in precaldera
Dunbar, N. W, Riciputi, L. R., Jacobs, G. K., and Christie, W. lavas of Glass Mountain, Earth Planet. Sci. Lett. 94,
(1993). Generation of rhyolitic melt in an artificial magma: 274-290.
Implications for fractional crystallization processes in nat- Handy, M. R., and Zingg, A. (1991). The tectonic and rheo-
ural magmas, J. Volc. Geotherm. Res., 57, 157-166. logical evolution of an attenuated section of the continental
Dungan, M. (1992). The life history of an Andean volcano, crust: Ivrea crustal section, southern Alps, northwestern It-
EOS Trans. Am. Geophys. Union 73, 406-407. aly and southern Switzerland, Geol. Soc, Am. Bull. 103,
England, R. W. (1990). The identification of granitic diapirs, 236-253.
J. Geo!. Soc. London 147, 931-933. Harry, W T., and Richey, 1. E. (1963). Magmatic pulses in the
England, R. W. (1992). The genesis, ascent, and emplacement emplacement of plutons, Liverpool Manchester Geol. J. 3,
of the Northern Arran Granite, Scotland: Implications for 254-268.
granite diapirism, Geo!' Soc. Am. Bull. 104,606-614. Hildreth, W. (1981). Gradients in silicic magma chambers: Im-
Feeley, T. C, and Grunder, A. L. (1991). Mantle contribution plications for lithospheric magmatism, J. Gcophys. Res.
to the evolution of middle tertiary silicic magmatism dur- 86,10153-10192.
ing early stages of extension: The Egan Range volcanic Hildreth, W (1987). New perspectives on the eruption of 1912
complex, east-central Nevada, Contrib. Mineral. Petrol. in the Valley of Ten Thousand Smokes, Katrnai National
106,154-169. Park, Alaska. Bull. Volc. 49, 680-693.
Fountain.J, C, Hodge, D. S., and Shaw, R. P. (1989). Melt seg- Hildreth. W, Grove, T. L., and Dungan, M. A. (1986). Intro-
regation in anatectic granites: A thermomechanical model, duction to the special section on open magmatic systems,
J. Volc. Geotherm. Res. 39, 279-296. 1. Geophys. Res. 91, 5887 -5889.
Fowler, A. C. (1985). A mathematical model of magma trans- Hildreth. W, Halliday, A. N., and Christiansen, R. L. (1991).
port in the asthenosphere, Geophys. Astrophys. Fluid Dyn. Isotopic and chemical evidence concerning the 'genesis
33,63-96. and contamination of basaltic and rhyolitic magma be-
Fowler, A. C. (1990a). A compaction model for melt transport neath the Yellowstone Plateau volcanic field, J. Petrol. 32,
in the Earth's asthenosphere. Part I. The basic model, in 63-138.
"Magma Transport and Storage" (M. P. Ryan, ed.), pp. 3- Hildreth. W, and Moorbath, S. (1988). Crustal contributions
14, Wiley, Chichester/Sussex. England. to arc magmatism in the Andes of Central Chile, Contrib.
Fowler, A. C (l990b). A compaction model for melt transport Mineral. Petrol. 98, 455-489.
in the Earth's asthenosphere. Part II. Applications, Hodge, D. S. (1974). Thermal model for the origin of granitic
in "Magma Transport and Storage" (M. P. Ryan, ed.), batholiths, Nature 251, 297-299.
pp. 15-32, Wiley, Chichester/Sussex, England. Hon, R., and Weill, D. (1982). Heat balance of basaltic intru-
 13. Magma Generation in Continental Lithosphere 315
sion vs granitic fusion in the lower crust, EOS Trans. Am. tending Ii thosphere and heat flow in the Basin and Range
Geophys. Union 63,470. province, Geol. Soc. Am. Bull. Mem. 152,209-250.
Huppert, H., and Sparks, R. S. J. (1991). Comments on "On Lange, R. A., and Carmichael, I. S. E. (1990). Hydrous basal-
convective style and vigor in sheet-like magma chambers" tic andesites associated with minette and related lavas in
by Bruce D. Marsh, J. Petrol. 32, 851-854. western Mexico, J. Petrol. 31,1225-1259.
Huppert, H. E., and Sparks, R. S. J. (1988). The generation of Leeman, W. P. (1982). Rhyolites of the Snake River Plain-
granitic magmas by intrusion of basalt into continental Yellowstone Plateau province, Idaho and Wyoming: A
crust, J. Petrol. 29, 599-624. summary of petrogenetic models, in "Cenozic Geology of
Hutton. D. H. W. (1988). Granite emplacement mechanisms Idaho" (B. Bonnichsen and R. M. Breckenridge, eds.),
and tectonic controls: Inferences from deformation stud- pp. 203-212, Idaho Geological Survey.
ies, Trans. Roy. Soc. Edinburgh 79, 245-255. Lister, J. R., and Kerr, R. C. (1989). The effect of geometry on
Hutton, D. II. W. (1992). Granite sheeted complexes: Evi- the gravitational instability of a buoyant region of viscous
dence for the dyking ascent mechanism, Trans. Roy. SO(. fluid, 1. Fluid Mech. 202, 577-594.
Edinburgh 83, 377-382. Lowell, R. P. (1982). Thermal convection in magmas gener-
Hutton, D. H. w., Dempster, T. J., Brown, P. E., and Becker, ated by hot-plate heating, Nature 300, 253-254.
S. D. (1990). A new mechanism of granite emplacement: Lowell, R. P., and Bergantz, G. W. (1987). Melt stability and
Intrusion in active extensional shear zones, Nature 343, compaction in a partially molten silicate layer heated from
452-455. below, in "Structure and Dynamics of Partially Solidified
Hutton, D. II. w., and Reavy, R. J. (1992). Strike-slip tectonics Systems" (D. E. Loper, ed.), pp. 383-400, Martinus Nij-
and granite petrogenesis, Tectonics 11, 960--967. hoff, Dordrecht.
Jacobs, G. K., Dunbar, N, w., Naney, M. T., and Williams, Mahon, K. I., Harrison, T. M., and Drew. D. A. (1988). Ascent
R. T. (1992). Petrologic and geophysical studies of an of a granitoid diapir in a temperature varying medium, J.
artificial magma, EOS Trans. Am. Geophys. Union 73, Geophys. Res. 93, 1174-1188.
411-412. Mahood, G. A. (1990). Second reply to comment of R. S. J.
Jarrard, R. D. (1986). Relation" among subduction parameters, Sparks, II. E. Huppert, and C. J. N. Wilson on "Evidence
Rei'. Geophys. 24, 217-284. for long residence times of rhyolitic magma in the Long
Jaupart, C., and Parsons, B. (1985). Convective instabilities in Valley magmatic system: The isotopic record in the pre-
a variable viscosity fluid cooled from above, Phys. Earth caldera lavas of Glass Mountain," Earth Planet. Sci. Lett.
Planet. Inter. 39, 14-32. 99,395-399.
Jomha, A. I.. Merrington. A., Woodcock, L. v.. Barnes, H. A.. Manduca, C. A" Silver, L. T., and Taylor, H. P. (1992). "Sri
and Lips, A. (1991). Recent developments in dense sus- "Sr and \'0/"0 isotopic systematics and geochemistry of
pension rheology, Powder Tech. 65, 343-370. granitoid plutons across a steeply-dipping boundary be-
Karlstrom, K. E., Miller, C. F.. Kingsbury. J. A., and Wooden, tween contrasting lithospheric blocks in western Idaho,
J. L. (1993). Pluton emplacement along an active ductile Contrib. Mineral. Petrol. 109,355-372.
thrust zone. Piute Mountains, southeastern California: In- Mareschal, J.-C., and Bergantz. G. B. (1990). Constraints on
teraction between deformational and solidification pro- thermal models of the Basin and Range province, Tecton-
cesses, Geol. Soc. Am. Bull. 105,213-230. ophysics 174,137-146.
Kaviany, M. (1984). Onset of thermal convection in a satu- Marsh, B. D. (1979). Island arc development: Some observa-
rated porous medium: Experiment and analysis, Int. J. tions, experiments and speculations, J. Geol. 87, 687 - 713.
Heat Mass Trans. 27, 2101-2110. Marsh, B.D. (1981). On the crystallinity, probability of occur-
Kay, R. w., and Kay, S. M. (1991). Creation and destruction rence, and rheology of lava and magma, Contrib. Mineral.
of lower continental crust, Geo!' Rund. 80, 259-278. Petrol. 78, 85-98.
Kay, S. M., Kay, R. w., Citron, G. P., and Perfit, M. R. (1990). Marsh, B. D. (1982). On the mechanics of igneous diapirism,
Calc-alkaline plutonism in the intra-oceanic Aleutian are, stoping and zone melting, Am. J. Sci. 282, 808-855.
Alaska, in "Plutonism from Antarctica to Alaska" (S. M. Marsh, B. D. (1989). On convective style and vigor in sheet-
Kay and C. W. Rapella, eds.), Special Paper 241, pp. 233- like magma bodies, 1. Petrol. 30, 479-530.
255, Geol. Soc. Am., Boulder, CO. Marsh, B. D. (1991). Reply to comments of Huppert and
Kempton, P. D., and Harmon, R. S. (1992). Oxygen isotope evi- Sparks, J. Petrol. 32, 855-860.
dence for large-scale hybridization of the lower crust dur- McKenzie, D. (1984). The generation and compaction of par-
ing magmatic underplating, Geochem. Cosmochim. Acta tially molten rock, J. Petrol. 25, 713-765.
56, 971-986. McKenzie, D. (1985). The extraction of magma from crust and
Kerrick, D. M. (1991). Contact metamorphism, Rei'. Mineral. mantle, Earth Planet. Sci. Lett. 74, 81-91.
26,847. McKenzie, D., and Bickle, M. J. (1988). The volume and com-
Kirk-Othmer (1978). "Encyclopedia of Chemical Technol- position of melt generated by extension of the lithosphere,
ogy," Vol. 21, 3rd ed., Wiley, New York. 1. Petrol. 29, 625-679.
Klemperer, S. L., Hauge, T. A., Hauser, E. C, Oliver, J. E., Mereu, R. E, Mueller, S., and Fountain, D. M. (1989). "Prop-
and Potter, C. J. (1986). The Moho in the northern Basin erties and Processes of Earth's Lower Crust," Geophysical
and Range province, Nevada, along the COCORP 40o~ Monograph 51, Amer. Geophys. Union.
seismic-reflection transect, Geol. Soc. Am. Bull. 97, 603- Miller, C. E, Hanchar, J. M., Wooden, J. L., Bennett, V. C,
618. Harrison, T. M., Wark, D. A., and Foster. D. A. (1992).
Lachenbruch, A. H., and Sass, J. II. (1978). Models of an ex- Source region of a granite batholith: Evidence from lower
316 George W. Bergantz and Ralph Dawes

crustal xenoliths and inherited accessory minerals, Trans. Powell, R. (1983). Processes in granulite-facies metamor-
Roy. Soc. Edinburgh 83, 49-62. phism, ill "Migmatites, Melting, and Metamorphism"
Miller, C. E, Watson, E. B., and Harrison, T. M. (1988). Per- (M. P. Atherton and C. D. Gribbs, eds.), pp. 127-139,
spectives on the source, segregation and transport of gran- Shiva, Cheshire.
itoid magmas, Trans. Roy. Soc. Edinburgh 79, 135-156. Quick, 1. E., Sinigoi, S., Negrini, L., Demarchi, G., and Mayer,
Morris, S. (1982). The effects of strongly temperature depen- A. (1992). Synmagmatic deformation in the underplated
dent viscosity on slow flow past a hot sphere, J. Fluid igneous complex of the Ivrea-Verbano zone, Geology 20,
Mech. 124, 1-26. 613-616.
Nelson, K. D. (1991). A unified view of craton evolution mo- Reagan, M. K., Gill, J. B., .vlalavassl. E., and Garcia, M. O.
tivated by recent deep seismic reflection and refaction re- (1987). Changes in magma composition at Arenal volcano,
sults, Geophys. J. lilt. 105,25-35. Costa Rica, 1968-1985: Real time monitoring of open
Ni, 1., and Beckermann, C. (l99\). A volume-averaged two- system differentiation, Bull. vote. 49, 415-434.
phase model for transport phenomena during solidifica- Reagan, M. K., Herrstrom, E. A., and Murrell, M. T. (1991).
tion, Metall. Trails. B 22, 349-361. The time scale of magma generation in Nicaragua trom
Nicolas, A. (1992). Kinematics in magmatic rocks with special "'U-series nuclide abundances, Gcol. Soc. Am. Abst, A 114.
reference to gabbros,.I. Petrol. 33, 891-9/5. Ribe, N. M. (1985). The generation and composition of partial
Nicolas, A., Freydier, C.. Godard, M., and Vauchez, A. (1993). melts in the Earth's mantle, Earth Planet. Sci. Lett. 73,
Magma chambers at oceanic ridges: How large? Geology 361-376.
21,53-56. Ribe, N. M. (1987). Theory of melt segregation-A review, J.
Noyes. H. 1., Wones, D. R., and Frey, F. (1983). A tale of two \'olc. Geotherm. Res. 33, 241-253.
plutons: Petrographic and mineralogic constraints on the Richter, EM., Nataf, H.-C., and Daly, S. F. (1983). Heat
petrogenesis of the Red Lake and Eagle Peak plutons, Cen- transfer and horizontally averaged temperature of convec-
tral Sierra Nevada, California, J. Ceol. 91, 353-379. tion with large viscosity variation. J. Fluid Mech. 129,
Ogawa, M., Schubert, G., and Zebib, A. (1991). Numerical 173-192.
simulations of three-dimensional thermal convection with Rubie, D. C, and Brearley, A. 1. (1990). A model for rates of
strongly temperature-dependent viscosity, J. Fluid Mech. disequilibrium melting during metamorphism, ill "High-
233,299-328. Temperature Metamorphism and Anatexis" (1. R. Ash-
Oldenburg, C. M., and Spera, F. J. (1991). Numerical model- worth and M. Brown. eds.), pp. 57-86, Lnwin Hyman,
ing of solidification and convection in a viscous pure bi- London.
nary eutectic system, lilt. J. Heat Mass Trans. 34, 2107- Rudnick, R. L. (1990). Growing from below, Nature 347,
2121. 711-712.
Oldenburg, C. M., and Spera, E 1. (1992). Hybrid model for Rudnick, R. L. (1992). Restites, Eu anomalies, and the lower
solidification and convection, Numer. Heat Trans. B 21, continental crust, Geochem. Cosmo. 56, 963-970.
217-229. Rudnick, R. L, and Taylor, S. R. (1987). The composition and
Olson, P., Schubert, G.. Anderson, C, and Goldman, P. (1988). petrogenesis of the lower crust: A xenolith study, J. Geo-
Plume formation and lithosphere erosion: A comparison of phys. ReI. 92,13981-14005.
laboratory and numerical experiments, J. Geophys. Res. Rushmer, T. (1991). Partial melting of two arnphibolites: Con-
93, 15065-15084. trasting experimental results under fluid-absent conditions,
Paterson, B. A., Stephens, W. E., Rogers, G., Williams, 1. S., Contrib. Mineral. Petrol. 107,41-59.
Hinton, R. W, and Herd, D. A. (1992). The nature of zir- Rutter, M. 1., and Wyllie. P. J. ( 1(88). Melting of vapor-absent
con inheritance in two granite plutons, Trans. Roy. SOt . tonalite at 10 kbar to simulate dehydration-melting in the
Edinburgh 83, 459-472. deep crust, Nature 331, 159-160.
Paterson, S. R., Vernon, R. H.. and Fowler, T. K. (1991). Au- Ryan. M. P., and Blevins. 1. Y. K. (1987). The Viscosity of
reole tectonics, ill "Contact Metamorphism" (D. M. Ker- Synthetic and Natural Silicate Melts and Glasses at High
rick, ed.), Reviews in Mineralogy, Vol. 26, pp. 673-722. Temperatures and I Bar (10' Pascals) Pressure and Higher
Paterson, S. R.. Vernon, R. H., and Tobisch, O. T. (1989). A Pressures," U.S. Geological Survey Bulletin 1764.
review of the criteria for the identification of magmatic Sawyer, E. W (I YY I). Disequilibrium melting and the rate of
and tectonic foliations in granitoids, 1. Strurt. Ceol. 11, melt-residuum separation during migmatization of malic
349-363. rocks from the Grenville Front Quebec, J. Petrol. 32,
Patifio-Douce. A. E., and Johnston, A. D. 099\). Phase equi- 701-738.
libria and melt productivity in the pelitic system: Implica- Scott, D. R., and Stevenson, D. J. (1984). Magma solitons,
tions for the origin of peraluminous granitoids and alumi- Geophys.Res.Lett.II,1161-11M.
nous granulites, Contrib. Mineral. Petrol. 107,202-218. Scott, D. R., and Stevenson, D. 1. (1986), Magma ascent by
Peterson. 1. W, and Newton, R. C. (1990). Experimental porous flow, 1. Geophys. Res. 91, 2973-29XX.
biotite-quartz melting in KMASH-CO, system and the role Sengun, M. Z., and Prohstein, R. F. (lY89). High-shear-limit
of CO, in the petrogenesis of granites and related rocks, viscosity and the maximum packing fraction in concen-
Am. Mill. 75, 1029-1042. trated monomodal suspensions, Physicochem, Hydrodvn.
Pitcher, W S., and Bussell, M. A. (1977). Structural control of 11,229-241.
batholith emplacement in Peru: A review, 1. Ceol. Soc. Shaw, H. R. (1972). Viscosities of magmatic liquids: An em-
LOlldon133,249-256. pirical method of prediction, Am. J. Sci. 272, 870-893.
 13. Magma Generation in Continental Lithosphere 317
Sherrod, D. R., and Smith. J. G. (1990). Quaternary extru- light of experimental studies in the system NaAISi,O,-
sion rates of the Cascade Range, Northwestern United KAISi,O,-SiO,-H,O. Geol. Soc. Am. Memoir 74.
States and Southern British Columbia. .f. Geophys. Res. Valasek, P. A., Hawrnan, R. B., Johnson. R. A, and Smithson,
95,19465- 19474. S. B. (1987). Nature of the lower crust and moho in eastern
Shirley, D. N. (1986). Compaction of igneous cumulates. .f. Nevada from" wide angle" reflection measurements, Geo-
Geol. 94,795-809. phys. Res. Lett. 14, 1111-1114.
Shirley, D. N. (1987). Differentiation and compaction in the Vielzeuf, D.• and Holloway. 1. R. (1988). Experimental deter-
Palisades Sill, "lew Jersey, J. Petrol. 28, 835-865. mination of the fluid-absent melting relations in the pelitic
Singer, B. S., Myers, J. Do, Linneman, S. R., and Angevine, system, Contrib. Mineral. Petrol. 98, 257-276.
C. L. ( 1(89). The thermal history of ascending magma dia- Vielzeuf, D., and Vidal, P.(1990). "Granulites and Crustal Evo-
pirs and the thermal and physical evolution of magmatic lution," ""ATO ASI, Series C, Vol. 311, Kluwer, Dordrecht.
conduits, J. v'oic . Geotherm.Res. 37, 273-289. Vigneresse, J.-L. (! (94). Granite emplacement and the re-
Sleep, 1\1. H. (1974). Segregation of magma from a mostly gional deformation field. Tectonophysics, in press.
crystalline mush, Geol. 50('. Am. son. 85, 1225-1232. Voshage, H., Hofmann, A. Wo, Mazzucchelli, M., Rivalcnti,
Smith. M. K. (1988). Thermal convection during the direc- G., Sinigoi, S., Raczek. I., and Demarchi. G. (199m. Iso-
tional solidification of a pure liquid with variable viscosity, topic evidence from the Ivrea Zone for a hybrid lower crust
J. Fillid Mech. 188,547-570. formed by magmatic underplating. Nature 347, 731-736.
Stevenson, D. J., and Scott, D. R. (1987). Melt migration in Weinberg, R. E (1992). Internal circulation in a buoyant two-
deformable media, in "Structure and Dynamics of Par- fluid Newtonian sphere: Implications for composed mag-
tially Solidified Systems" (D. E. Loper, ed.), pp. 401-416, matic diapirs, Tectonophysics 110, 77-94.
Martinus Nijhoff, Dordrecht. Wells, P. R. A (1980). Thermal models for the magmatic
Tepper, J. H., Nelson, B. K., Bergantz, G. W., and Irving, A 1. accretion and subsequent metamorphism of continental
(1993). Petrology of the Chilliwack Batholith, North Cas- crust, Earth Planet. Sci. Lett. 46, 253-265,
cades, Washington: Generation of calc-alkaline granitoids Wickham, S. M. (1987). The segregation and emplacement of
by melting of mafic lower crust with variable water fu- granitic magmas, J. Geol. Soc, London 144,281-297.
gacity, Contrib. Mineral. Petrol. 113,333-351. Wickham, S. M., and Peters, M. T. (1992). Oxygen and car-
Thompson, A B. (1982). Dehydration melting of pelitic rocks bon isotope profiles from Lizzies Basin, East Humboldt
and the generation of H,O undersaturated granitic liquids. Range. Nevada: Constraints on mid-crustal metamorphic
Am. J. Sci. 282,1567-1595. and magmatic volatile fluxes, Contrib. Mineral. Petrol.
Thompson, A B. (1983). Fluid-absent metamorphism, J. Geol. 112,46-65,
Soc. London 140,533-547. Wildemuth, C. R., and Williams, M. C. (l9R4). Viscosity of
Thompson, A. B. (1988). Dehydration melting of crustal rocks, suspensions modeled with a shear-dependent maximum
Rend. 501. Ital. Mineral. Petrol. 43, 41-60. packing fraction, Rheol. Acta 23, 627-63.~.
Thompson, A B., and Algor, 1. R. (1977). Model systems for Wolf, M. B., and Wyllie, P. 1. (1989). The formation of tona-
anatexis of pelitic rocks. I. Theory of melting reactions in litic liquids during the vapor-absent partial melting of am-
the system KAIO,-NaAIO,-SiO,-H,O, Contrib. Mineral. phibolite at 10 kbar, EOS Trans. Am. Geophys. Union 70,
Pnrol. 63, 247-269. 506.
Thompson, A. B., and Tracy, R. 1. (1979). Model systems for Wolf, M. B., and Wyllie, P. 1. (1991). Dehydration-melting of
the anatexis of pelitic rocks. II. Facies series melting re- solid amphibolite at 10 kbar: Textural development, liquid
actions in the system CaO-KAIO,-NaAIO:-AI,O.-H,O, interconnectivity and applications to the segregation of
Contrib. Mineral. Petrol. 70, 429-438. magmas, Mineral. Petrol. 44,151-179.
Tikoff, B., and Teyssier, C. (1992). Crustal-scale, en echelon Wyllie. P. 1. (1977). Crustal anatexis: An experimental review.
"P-shear" tensional bridges: A possible solution to the Tectonophysics 43, 41- 71,
batholithic room prohlem, Geology 20, 927-930. Yardley, B. W. D. (1986). Is there water in the deep continental
Touloukian, Y. S., Judd, W. R., and Roy. R. E (1981). "Physi- crust? Nature 323, Ill.
cal Properties of Rocks and Minerals," McGraw-Hill, Yoder. H. S. (1990). Heat transfer during partial melting: An
New York. experimental study of a simple hinary silicate system, 1.
Trent, D. S., and Eyler. E. E. (1991). "TEMPEST: A Computer ~ ole. Geotherm. Res. 43, 1-36.
Program for Three-Dimensional Time-Dependent Hydro- Younker, L. w., and Vogel, T. A (1976). Plutonism and plate
thermal Analysis. Vol. I. Lxers Manual." Battelle. Pa- tectonics: The origin of circum-Pacific batholiths. Can.
cific Northwest Laboratory publication' PNL-4348. Vol. I, Mineral. 14,238-244.
Rev. 3. Zingg, A (1990). The Ivrea crustal cross-section (northern
Turcotte, D. L., and Schuhert, G. (1982). "Geodynamics: Ap- Italy and southern Switzerland). in "Exposed Cross Sec-
plications of Continuum Physics to Geologic Problems," tions of the Continental Crust" (M. H. Salisbury and D. M.
Wiley, New York. Fountain. eds.), pp. 1-20. Kluwer, Dordrecht.
Tuttle, 0. E, and Bowen, N. L. (1958). Origin of granite in the
This page intentionally left blank
 Chapter 14 Two-Component Magma Transport and the Origin
of Composite Intrusions and Lava Flows

Charles R. Carrigan

Overview through a dike is proposed as an alternative to the often

used box and tube withdrawal model. This new model
Observations of layering and other evidence for chemi- has the advantage that two or more layers can be tapped
cal variations in intrusions and in lava flows are pre- simultaneously without concern for the draw-up-depth
sented to evaluate several fluid dynamical models for parameter that is conventionally associated with lay-
two-component effusive volcanism. The newest model ered withdrawal models. The model is then used to de-
involves the hydrodynamic process of encapsulation or, velop a scenario for the contemporaneous effusion of
more generally, viscous segregation. Viscous segrega- three composite domes in the Inyo volcanic chain of
tion is described as the migration of the low viscosity Long Valley, California. Finally, self-lubrication is con-
component of a two-component magma to the high sidered as a means for enhancing the probability that
shear regions of a flow that are normally adjacent to the highly silicic magmas in the presence of lower vis-
walls of a dike or conduit. The process is explained in cosity components will reach Earth's surface. The bias
terms of the tendency for a two-component flow to have of magma transport toward lubricated silicic magmas
its energy losses to viscous dissipation minimized. In reaching the Earth's surface over unlubricated silicic
effect, the lower viscosity component becomes a lubri- magmas suggests that the frequency of association of
cant for the passage of the higher viscosity component. two end-member magmas, such as rhyolite and basalt
Both laboratory studies of the encapsulation of a higher in crustal reservoirs, may be somewhat lower than
viscosity layer by a lower viscosity layer during flow in would be indicated by their occurrence in zoned lava
a pipe or slot and numerical simulations show that the flows and in near surface intrusions.
viscosity ratio of the two components is the major fac-
Notation
tor in determining whether a flow evolves to become a
Units
zoned, lubricated flow. The lubrication equations for
two-component flow in a rectangular dike geometry are constants, Eq. (36) W·m- 2 •
derived assuming a power-law rheology. In addition to W·m
the substantial reduction in equivalent viscosity result- constant, Eq. (10) Pa,MPa
ing from the presence of thin, lower viscosity wall lay-
ers, it is found that, for certain power-law magmas, the
C, constant, Eq. (10) Pa,MPa
D constant, Eq. (23), or diameter of Pa-I/n. s 1.
equivalent viscosity is less for two-component flow m{l+2n)/nor m
extrusion die
than for either of the components flowing singly in the
E constant, Eq. (23) Pa lIn. S-I.
dike. With fracture propagation models, it is shown that m(l+2nl/n
lubrication can substantially increase propagation rates
F constant, Eq. (23) Pa lIn. S-I.
for viscous magmas and decrease the probability that
m(I+2n)/n
solidification will occur in transit. Viscous heating re-
sulting from the concentration of shear into the narrow G negative of driving pressure gra- Pa . m I

lubricating layers may explain how such layers of basalt dient. Eq. (9)
can remain fluid in the presence of both cooler rhyolitic G,.o" value of G at which viscosities of Pa m- I
magmas and wall rock. The effect of inlet conditions two different power-law magmas
involving the distribution of the two magmatic compo- are equal at interface between
nents at the entry of the dike is also considered. In this magmas
context, a generalized and considerably more realistic J constant coefficient, Eq. (36) Pa . S' 1•

m-(I+n)/n
model for the simultaneous withdrawal of two magmas

Magmatic Systems Copyright © 1994 by Academic Press, Inc.

Edited by M. P. Ryan 319 All rights of reproduction in any form reserved.
 320 Charles R. Carrigan

Units Unih

L length of tube or extrusion die m I absolute value

M elastic shear modulus, Pa \' del or nabla, the gradient operator m I
Eqs. (27)-(30)
Q volumetric flux of magma per unit m1 . s ·m
horizontal length of dike
Re Reynolds' number (lI'dp/fl.) dimensionless
Introduction
S volumetric viscous heating rate W·m-'
T temperature of magma °C
Petrologic studies of lava flows and magmatic in-
T, temperature at which magma is °C
trusions tend to stress chemical variations and
rigid
their spatial relationships, since it is believed that
a width of lubricating layer at dike m
wall
such information can be used to constrain par-
J. kg-I. °C-I tially the physical and chemical state of the mag-
(r specific heat at constant pressure
matic source region as well as the petrogenic
d width of dike m
nature of the local processes involved in magma-
unit vector across dike dimensionless
tism. Thus, a full analysis of how multicomponent
k thermal conductivity W·m- I.
°C-I
magmas flow should involve more than just treat-
ing the efficacy of transport between the magma
k unit vector along dike (vertical) dimensionless
reservoir and the surface. It is also necessary
I fracture or dike length. Eq. (27) m
to understand how transport along a magmatic
m power-law coefficient Pa s-v

pathway can affect and/or produce the observed

11 power-law exponent dimensionless chemical and physical variations in flows and in
p fluid pressure Pa shallow intrusions. This is a requirement if one is
p. fluid pressure (chamber) Pa to distinguish between the characteristics of the
p, fluid pressure (outlet) Pa flow that are contributed by the source region and
q heat flux W·m- 2 those contributions imposed by the transport pro-
time cess itself.
It horizontal velocity component m. S-l
At one end of the range of possible volcanic
activity, large and violent eruptions, such as are
v flow velocity vector ill' S-l

m.
responsible for the thick ash flow deposits of Long
II' vertical velocity component S-l
Valley Caldera, California, and Yellowstone, Wy-
x horizontal coordinate m
oming, probably are the result of significant over-
vertical coordinate m
pressuring of magmatic reservoirs that are rich in
cP,· dissipation function volatiles. In these eruptions, flow velocities may
E strain rate in magma be great (300-500 m . S-I; Wohletz arid Valen-
E""r'lss strain rate at which two power-law tine, 1990) and the volume of erupted products
magmas have same viscosity may be 10% of the total magma reservoir volume
1) generalized Newtonian viscosity Pa· s (Smith, 1979). Because of the large magmatic
fl. molecular viscosity Pa· s driving pressures and the presence of large quan-
j.Llx local viscosity, Eq. (19) Pa· s tities of gas associated with the exsolution of vol-
fl.", equivalent viscosity, Eq. (20) Pa· s atiles that ultimately lead to the complete disrup-
/I Poisson's ratio dimensionless tion of the continuous molten phase, the magmatic
p magma density kg -rn
, viscosity is likely to play, at most, a secondary
T viscous stress tensor Pa,MPa role in the dynamics of this class of eruptions. In
ijth component of viscous stress Pa,MPa
particular, magmatic viscosity probably will not
r;
tensor significantly affect the ordering of the layered
c as sub/superscript, indicates core eruption products and the layering on the surface
layer will tend to be a simple inversion of the layering
w as sub/superscript, indicates wall of the contents in the magmatic reservoir (Hil-
layer dreth, 1981). Toward the other end of the range,
 14. Two-Component Magma Transport 321

eruptions of an effusive nature probably involve the common characteristic that the two different
much lower magmatic driving pressures and pos- magmas involved erupt sequentially from a vent.
sibly lower magmatic volatile contents than those A more rational model that involves encapsulating
that are typical of large pyroclastic eruptions. Mag- and self-lubricating two-component flow is pre-
matic viscosity may now exert a much stronger sented and describes how different magmas can
influence in determining what is observed at the erupt simultaneously from a vent to form a com-
surface. In this chapter, the focus is on effusive posite flow in a single-step process. As a reason
eruptions involving two different magmatic com- for the occurrence of viscous segregation, the hy-
ponents. A general model is synthesized for the drodynamic implications of the minimization of
fluid dynamics of two-component magmas that dissipative shear forces in two-component flows
can explain why the effusive eruption of very vis- are considered. To support the argument that en-
cous magmas is often associated with the occur- capsulation is an important mechanism for the
rence of discrete chemical zoning and why this formation of composite intrusions and flows, a
zoning at the surface is often a reversal of that collection of results is presented from laboratory-
found in much larger and more energetic eruptions. scale experiments and from numerical models of
It is presumed in the model that silicic magmas that the two-component mode of transport in pipes and
are associated with a component having a signifi- channels. Then, a general derivation of the lubri-
cantly lower viscosity can be "self-lubricating." cation equations is used to investigate the degree
In certain cases, the low viscosity component to which encapsulation can mitigate the effects of
(usually the more mafic component) will encap- high viscosity on driving pressure and dike propa-
sulate the higher viscosity component during flow gation. In addition, the concentration of shear into
in a dike or conduit. giving rise to the common lubricating layers coupled with viscous dissipa-
mafic-to-silicic zoning. It is further argued that si- tion is investigated as a means of offsetting the
licic magma lubricated by a less viscous compo- heat losses that can lead to solidification. On the
nent has a greater probability of ascending to basis of our understanding of the flow of multi-
the Earth's surface than does the silicic magma component fluids in pipes and channels, we out-
by itself. The model also considers those two- line the general types of inlet conditions that
component magmas flowing in a dike that do not would be required to produce the kinds of zoning
as readily produce encapsulation of the higher vis- that occur at the surface. We also suggest how dif-
cosity component by the lower viscosity one. In ferent magmatic compositions can be sampled si-
such flows little zoning may be present. Rather, multaneously by a single dike that breaches the
the mixture, at the surface, takes the form of a nonhorizontal roof of a magma reservoir to pro-
magma "emulsion" containing blobs of one com- duce a variety of layering in near-surface intru-
ponent dispersed throughout the continuous "ma- sions and in surface flows. Having now developed
trix" component. a preferred model for two-component magma
To provide an observational basis for the fluid transport, we apply it to the formation' of three
dynamical models discussed here, two of the contemporaneously extruded composite flows that
many extensive field investigations of the varia- are part of the much studied Inyo volcanic chain
tions in chemical zoning that are manifested in of Long Valley, California.
surface flows and intrusions are summarized. In
one example of two-component flows and intru-
sions, the ranges of chemical and physical dif- Observations of Composite Flows,
ferences between the two magmatic components Dikes, and Conduits
(rhyolite and basalt) are large, whereas in the
other example (rhyolite and rhyodacite), chemical Basalt and Rhyolite Systems
analyses are required for the differences to be ap-
parent. The discussion is limited to features that An enduring mystery is how basaltic and rhyolitic
are thought to be archetypal of the petrology of magmas, with their strongly contrasting chemical
such systems. Some of the earlier models devel- and physical properties, must be spatially distrib-
oped to explain the origin of the discrete zoning uted at depth to produce the contemporaneous ex-
of rock types are then reviewed. Such models have trusion of layered lavas. Linking the observed dis-
 322 Charles R. Carrigan

tribution of the two components at the surface to part of the newly accreting crust (Gudmundsson,
their initial distribution in a magma chamber is 1990; Ryan, 1990). Possibly by the melting of the
one goal of the modeling of two-component trans- hydrated crust or by extreme crystal fractionation
port that may help resolve a mystery that has pe- of mafic magma, or a combination of both these
trogenic implications. Another goal of the model- processes (Yoder, 1973; Vogel & Wilband, 1978;
ing is to better understand how the thermal and Huppert and Sparks, 1980; Vogel, 1982; Furman
flow regimes may interact to permit the more re- and Spera, 1985), significant quantities of silicic
fractory basalt to flow large distances in a thin magmas have been produced and have flowed
layer that loses heat to the wall rock on one side within the crust and to the surface.
and to the somewhat cooler rhyolite on the other. In a study of the Tertiary volcanics of eastern
Some of the best examples of end-member Iceland by Gibson and Walker (1963), five com-
rocks in composite intrusions and rarer composite posite lava flows with associated intrusions were
lava flows are found in Iceland. A plausible model mapped (Fig. 1). The intrusions are dikes of the
to explain both the thermal and mafic magma in- classic or normally zoned type in which a silicic
put into the diverging Icelandic crust via dike core is sandwiched between layers of more mafic
swarms and sheets requires the emplacement of rock (Walker and Skelhorn, 1966). In such intru-
shallow mafic magma chambers residing several sions, the mafic margins tend to be thinner than
kilometers beneath the surface in the thinnest the more silicic core. According to Gibson and

Vertical scale:
(=2x horiz. scale)
Horizontal scale:
4a.~.•. .; . 4~ o 1 mile 1000 flI300m
- - - _ ;.;;:.:.:;:;:::;W ---- I

(Not composite) o 1 km o 0

ta, 1b. ,...... 2.

~""""""~
~ .:.._IIIIIIiIiiii;m·;·;::i/"·:::: .::.;-..:;~-.: -- •.....
......... ~..
~
.::,.

5.

[TI Non-xenolithic rhyolite

~:: : .: : j Xenolithic rhyolite

_ Basalt ICELAND
1. Upper Gilsardalur flow:
a) N. of Gilsardalur

WEST b) Kjappeyrarmuli

2. Ornolfsfjall flow

3. Kdttur flow
acid tu
4. Lower Gilsardalur flow:

a) Gilsardalur

b) N. of Ornolfsfjall
Upper GlIsa'rdalur Composite lava 5. Berufjordur flow

Figure 1 Mapped profiles of five composite rhyolite/basalt lava flows and intrusions located in the Tertiary volcanic region of
eastern Iceland. As discussed in the text, several of the profiles clearly indicate that basalt and rhyolite lava effused simultane-
ously from feeder dikes to form these flows. The inset shows a detailed profile of a composite flow that filled a basin-like
depression, The thin layer of basalt following the contours of the basin could only have been produced by the simultaneous
effusion of the two lavas. Adapted from Gibson and Walker (1963. Fig. 6) with permission of the Geologists' Association,
London.
 14. Two-Component Magma Transport 323
Walker, all the field evidence suggests that the ba- the basaltic layers of the flow and that the billowy
saltic and rhyolitic magmas flowed at the same nature of these inclusions or enclaves suggests that
time in feeder dikes and effused simultaneously they were molten at the time of inclusion in the
from these dikes to form layered lava flows. The rhyolitic component. Numerical models of two-
form and distribution of the basalt and rhyolite, component flow in a channel (Stockman et al..
such as the wrapping of the basalt margin around 1990) show that parcels of one fluid can be ripped
the toe (left side) of the Upper Gilsardalur flow off by the shear and entrained in the flow of the
(Fig. la and inset), require the simultaneous flow other component. Another example of this appar-
of these two components. If the two components ent tendency for mafic xenoliths to detach from
had effused sequentially, with the basalt coming mafic margins and be captured by the rhyolitic
from the vent first, basalt would have had to com- core flow is well documented by Skelhorn et al.
pletely fill the depression on the west side of the (1969) in a map (see their Fig. 12a) of a composite
feeding dike before rising any further along the intrusion at Scallastle Bay, Isle of Mull, Scotland.
toe. Instead, the basalt forms a thin layer that fol-
lows the contours of the basin, a feature that can Rhyodacite and Rhyolite Systems
be explained only by the simultaneous filling of
the basin from a composite layered feeder dike. More subtle chemical variations characterize the
Furthermore, from examination of Fig. 1, the rela- zoning in a volcanic conduit that has effused a lay-
tively close correspondence in the horizontal di- ered flow, which, along with its feeder conduit,
mension of the two components of the Kottur and has been the subject of a recent scientific drill-
Berufjordur flows (3 and 5) and the complemen- ing investigation (Younker et al., 1987). Obsid-
tary nature of variations in the thickness of the ba- ian Dome is one of seven rhyolitic flows making
salt and rhyolite layers (1b, 2 and 5) indicate that up the 12-km-long Inyo chain that also includes
both components were in the fluid state at the phreatic explosion craters and normal faults that
same time and that the effusions of both the upper cut across the northwestern boundary of the Long
and lower layers occurred simultaneously. If the Valley caldera, California. The fact that three of
magmas had flowed at separate times, there is no the domes including Obsidian Dome and Inyo
reason other than chance to explain the correspon- Craters, and a collection of explosion craters at the
dence of the horizontal dimensions of the upper south end of the chain, erupted along a line about
and lower flows or the complementary nature of 600 yr ago (Miller, 1985; Vogel et al.. 1989, and
thickness variations in the two layers. More de- references therein) suggests that these eruptions
tailed observations also support the simultaneous were the surface manifestations of a dike em-
effusion of the flows. For example, the rhyolite placed near the surface over a 10-km horizontal
veins basalt in the Orn6lfsfjall flow while the ba- distance. In fact, a dike having a thickness of 7 m
salt is chilled against the rhyolite. In the Upper was intersected during another drilling study at a
Gilsardalur flow (Fig. 4 of Gibson and Walker, depth of 600 m just south of Obsidian Dome at
1963), rhyolite is chilled to a pitchstone at the the northern end of the Inyo chain (Eichelberger
contact with the host rock only where the thin, in- et al.. 1985; Younker et al., 1987). Although the
sulating basaltic layer is missing. At other loca- dike itself does not break the surface, the three
tions where the 0.3- to 0.5-m insulating basaltic contemporaneously erupted domes are associated
layer exists, hot basalt has protected the rhyolite with conduits that evidently "budded" from the
from chilling. These observations are consistent dike sheet (Delaney and Pollard, 1981; Bruce and
with basalt and rhyolite both being molten in each Huppert, 1989, 1990), causing a significant vol-
other's presence. ume of magma to reach the surface. Since cy-
The simultaneous flow of basalt and rhyolite lindrical feeder conduits are more efficient for
might also be inferred from observations that the magma transport than dikes because the pressure
rhyolite of composite flows tends to be rich in gradient required to drive a given magma flux is
xenoliths of basalt that are invariably similar to lower and also because the rate of heat loss for a
the basaltic component of the flow (Gibson and given flow rate is lower, it is reasonable to expect
Walker, 1963). Furthermore, it was found that that widened parts of the dike, which eventu-
xenoliths of basalt tend to be more abundant nearer ally become conduits, are more likely to permit
324 Charles R. Carrigan

. : Depth (Ill)
200

.» 400
.. '
.c.>

..'

Dike N

Figure 2 Cross section of the Obsidian Dome composite flow located in the Inyo volcanic chain, Long Valley, California. One
core hole sampled the flow near its margin (1) and another slant hole (2) sampled the composite flow near its vent, a basal zone
of breccia shed before the advancing margin of the flow and also the composite feeder conduit. The data on compositional zoning
presented in Fig. 3 were obtained from hole 2. To test the hypothesis that a subsurface dike connects Obsidian Dome with two
others, a third hole was drilled between Obsidian Dome and Glass Creek Dome (see Fig. 22). The hole intersected a shallow,
unzoned rhyolitic dike, Adapted from original figure provided courtesy of J. C. Eichelberger.

magma to reach the surface and are favored for port a model for the contemporaneous effusion of
long-term ascent. rhyolitic and less viscous rhyodacitic lavas that
Obsidian Dome is a kidney-shaped flow with a compose Obsidian Dome (Carrigan and Eichel-
horizontal dimension that varies from I to 2 km berger, 1990). The arrangement of a more viscous
and a thickness that varies from 100 m over the magma overlying one of lower viscosity is shared
more or less central conduit to about 50 m at the with the two other domes erupted at about the
edge of the flow (see Vogel et al., 1989, and ac- same time in the Inyo Chain (Sampson and Cam-
companying references). Figure 2 illustrates the eron, 1987) and with a third lava flow on the
relationships between the dome, its feeder con- Medicine Lake Highland in California (Eichelber-
duit, and the associated dike. This diagram at- ger, 1975).
tempts to synthesize the surface geology and ob- The connection between zoning in the Ob-
servations obtained from three core holes. One sidian Dome lava flow and zoning in the feeder
hole was drilled about 150 m from the dome mar- conduit is shown in the chemical profiles (Fig. 3)
gin to sample the distal part. Another hole was of the drill cores obtained from the flow/conduit
slanted to penetrate both a central part of the dome hole. The cores clearly show that the more silicic
and the feeder conduit. As already mentioned, the rhyolite flowing up the center of the conduit
third hole was drilled a short distance away from formed the overlying layer and the lower viscosity
the dome and slanted to intersect the hypothesized rhyodacite flowed along the margin of the conduit
subsurface dike that was encountered. and vented to form the layer at the bottom of the
The slant hole through the central dome and flow. Another dome near Obsidian Dome that was
conduit (2 in Fig. 2) provided cores of the flow, exposed in cross section by an explosion exhib-
basal breccia, and the feeder that strongly sup- its flowlines in its texture that support both this
 14. Two-Component Magma Transport 325
Si0 2 (wt . %)
70 72

lava dome

40 basal breccia

-- - -- .... _- - ......... -
....

conduit ~ ...... _-- .... - .....

..... - - _ L.J._~=~:...:.:.='--....L...J low

--------",,--
74

... ...
... ....
-
's
't:l
r::::
-
"5
't:l
... "- o r::::

-
o o
... .... o'" '0 o
-.. "- in
o

'-"-.. 70 ~
...
"
o 20 40
Horizontal distance(m)
Figure 3 The silica content of cores from hole 2 of Obsidian Dome. Long Valley, California, as a function of location in the
flow and in the feeder conduit. The less silicic and presumably the less viscous component occupies regions at the walls of the
conduit and at the base of the intact flow that rests on a zone of basal breccia. The more silicic and more viscous component
composes the center of the conduit and the top layer of the flow. This arrangement indicates that distinct mafic and silicic layers
of magma flowed simultaneously up the conduit and effused onto the surface, with the marginal mafic layer forming the base of
the intact flow and the central silicic layer forming the topmost layer of the flow. Adapted from Carrigan and Eichelberger (1990)
with permission from Nature and Macmillan Magazines Limited.

model of effusion and the simultaneous venting tween the two lavas is physically continuous in the
of the two chemically distinct components of the flow and in the conduit with fine interbanding
lava (Fig. 4). The simultaneous transport of the present at the boundaries of the two components.
two components is also consistent with the obser- In addition, both rhyolite and rhyodacite are pres-
vation that both layers of the flow have margins ent in the basal breccia of the flow on which the
that are everywhere nearly coincident (Carrigan lower layer consisting of rhyodacite lies. This
et al., 1992). This would not be expected if both brecciated layer, which underlies the flow, has
flows arrived at the surface separately. Several been created from the slaggy crust falling off the
finer-scale observations also support the conclu- margin of the flow just ahead of the advancing
sion that both lavas vented together (Carrigan and flow front. If the lower rhyodacitic layer had ef-
Eichelberger, 1990). The chemical boundary be- fused from the vent alone and was followed later
 326 Charles R. Carrigan

Figure 4 Photomosaic-based illustration of explosively sectioned flow near Obsidian Dome. Flow lines are observable as
variations in texture and support the interpretation of drill cores (hole 2) that the dome was emplaced by the simultaneous flow
of a marginal zone of magma that flowed out horizontally to form a basal layer, and a central zone that flowed out on top of the
marginal layer to form the top of the flow. Adapted from Carrigan and Eichelberger (1990) with permission from Nature and
Macmillan Magazines Limited.

by the rhyolitic layer, the underlying breccia would ately following the rise of the mafic component,
not contain remnants of the crust of the upper the silicic component penetrates and flows along
rhyolitic layer. the center of the molten mafic dike or conduit,
which gives rise to the composite nature of the in-
trusion. At the surface, the silicic lava flows out
Models of the Origin of upon the more mafic layer to form the composite
Mafic-to-Silicic Zoning flow. Blake et al. (1965) suggest a version of this
model for the origin of the composite basalt and
Sequential Magma Transport rhyolite flows illustrated in Fig. 1. Such a sequen-
tial scenario may aid the rhyolite in rising to the
A traditional view of the origin of composite in- surface in two ways. The initial flow of basalt into
trusions and flows with mafic-to-silicic zoning the dike insulates the silicic magma from the cool
might be referred to as the sequential emplace- wall rock. The hotter basalt may also heat the
ment model. In the basic scenario, the mafic and rhyolite by a few tens of degrees and decrease its
silicic magmas may flow contemporaneously but viscosity, allowing it to flow in the dike more
also sequentially (one behind the other in the same easily.
dike). Initially, the mafic magma somehow fills Basalt and rhyolite lavas appear to have ex-
the dike or conduit first and may vent at the sur- truded sequentially to produce the Kjappeyrar-
face, forming the lowest layer of a flow. Immedi- rmili or Orn6lfsfjall flows of Iceland that exhibit
 14. Two-Component Magma Transport 327
mainly horizontal variations in lava type (Fig. I, draw-up-depth is a function of the withdrawal rate
flows 1band 2). According to the sequential through the tube. Layers shallower than the draw-
model, the basalt would have preceded the rhyo- up-depth will be tapped whereas the contents of
lite to the surface, forming a layer that was pushed deeper layers will not be entrained as magma
outward from the feeder upon the arrival of the flows into the tube. In principle, the withdrawal
rhyolitic magma. How the denser basalt, which process can explain how magmas from different
tends to underlie rhyolite in the typical model of a layers in a stratified reservoir can be tapped si-
stratified chamber, gets ahead of the rhyolite in the multaneously through a conduit even if the deeper
dike poses an interesting problem that will be con- layers are separated from the outlet by inter-
sidered further. An alternative model proposed vening magma layers. Such a draw-up process
here is that the basalt and rhyolite were simul- may be particularly applicable to the very large
taneously withdrawn and simultaneously flowed eruptions that are characterized by exceedingly
in the dike or conduit-side-by-side-in the nor- large magma withdrawal rates through the vol-
mally zoned arrangement, which is shown here to canic feeder dike/conduit. The mass flow rates
be dynamically stable. Because of its lower vis- that are thought to be typical of such eruptions fall
cosity, the basalt could, in certain circumstances, between 105 and 10"kg . S - I (Trial et al., 1992).
flow out of the vent from its margins at a higher For the conventional model of a stratified
rate, covering the surface ahead of the more magma chamber with a small diameter cylindrical
slowly advancing rhyolite flow effusing from the conduit centered on top, the draw-up effect is
central part of the vent. critical to explaining the origin of composite lay-
The sequential model cannot explain flows that ering in the intrusions and in the flows of effusive
are characterized by predominantly vertical vari- volcanism even though the effect is likely to be
ations in lava type involving more or less co- much weaker owing to the much lower magma
incidental margins and finer-scale features, indi- withdrawal rates of effusive eruptions (lO - 1-
cating simultaneous flow such as the Berufjordur, 102 kg . S - I ; Trial et al., 1992). In a later section,
Kottur, and Upper Gilsardalur flows (Fig. I, flows we examine the implications of the conventional
5, 3, and la). To obtain such layered flows re- model for transporting different magmas and we
quires both lava types to issue from the vent si- present a more general model for a magma reser-
multaneously. The sequential model is also inap- voir + transport system as an alternative to the
propriate in the case of Obsidian Dome, where symmetrical reservoir and pipe arrangement. The
both lava types must have arrived simultaneously new model is more realistic in terms of our under-
at the surface in a core-annular arrangement with standing of both how eruptions are initiated and
a more mafic rhyodacite in the annular part of the how magmatic pathways evolve, and it permits
flow adjacent to the conduit wall and a more si- withdrawal of two or more layers simultaneously
licic rhyolite occupying the core region of the flow. without regard to magma withdrawal rates and the
draw-up effect. This new model offers a better fit
Simultaneous Transport of Different to the extensive observations of the Inyo Domes
Magma Types system than do conventional models.
The withdrawal of two layers simultaneously
Various numerical and laboratory experiments from a stratified magma chamber will tend to pro-
have shown that withdrawing the contents of a duce a flow regime in which the lower (more
gravitationally stratified magma reservoir through dense) layer occupies the central region of a con-
a tube or slot centered in the lid of the reservoir duit while the upper (less dense) layer forms a sur-
can extract magma from different layers simul- rounding annular regime (Fig. 5). Since the mafic
taneously (Spera, 1984; Blake and Fink, 1987; layer in a two-component system is usually the
Blake & Ivey, 1986; Spera et al., 1986; Trial more dense, mafic magma will initially tend to
et al., 1992). A parameter that determines whether occupy the central region of flow during with-
a particular layer will be tapped during withdrawal drawal through a cylindrical conduit and the si-
is the so-called draw-up-depth. For a given reser- licic magma will make up the outer annular re-
voir geometry and set of fluid properties, the gion. How the transport of two magmas might
328 Charles R. Carrigan

drawal from a reservoir with a silicic layer over-

lying a mafic layer, they produced a core-annular
flow at the inlet of the tube with a central lower
viscosity region and an annular higher viscosity
region next to the wall of the tube. This was
accomplished by metering lower viscosity fluid
+
draw up along the central axis of the flow of higher vis-
depth cosity fluid. By adjusting the rate at which the
t lower viscosity component was metered into the
flow of the higher viscosity component, they found
that the stability of the flow was affected. For the
highest rates of flow of the central low viscosity
mafic component relative to the flow rate of the higher
magma
viscosity component, the central zone became
wavy and appeared to shed streamers of the cen-
Figure 5 Cross section of a conventional model of with-
drawal from a stratified magma chamber consisting of a flat-
tral lower viscosity component into the higher
lidded box with a cylindrical tube centered in the top. If the viscosity flow. Blake and Campbell (1986) used
withdrawal rate is sufficient. i.e.• the draw-up-depth is large these observations to argue correctly that the ar-
enough. the lower mafic layer will be drawn up simultane- rangement of high viscosity magma in an annular
ously with the overlying silicic layer. In this case. the initial
zone and low viscosity magma in the core zone of
configuration at the inlet of the conduit will be a core-annular
flow with more viscous silicic magma adjacent to the walls and
a volcanic conduit, as an inlet condition of their
less viscous mafic magma occupying the core. Experiments hypothesized withdrawal model, would lead to
and theory indicate that such an arrangement of magmas with mixing (breakup of the low viscosity core/high
the higher viscosity magma adjacent to the wall cannot persist viscosity annular flow) of the two magmas at
in the conduit. Such a box and tube model for magma with-
Reynolds' numbers far less than the critical value
drawal is likely to be too restrictive regarding the simultaneous
withdrawal of both the upper and the lower layers since there
required for the onset of turbulence in pipe flow.
is no reason why the magma chamber top or lid must be per- However, they also argued that composite intru-
fectly parallel to the layering such that only the top layer is in sions such as those of Fig. I could not be treated
contact with the lid in the absence of withdrawal. Furthermore. as "snapshots" of an entirely liquid structure.
a good case can be made that withdrawal will usually take
It was suggested, using the results of Koyaguchi
place through a slot-like orifice attached to a tabular structure
such as a dike rather than through the localized entry of a cy-
(1985), that during the vertical propagation of a
lindrical conduit. This model is not geologically realistic and dike, the lower viscosity mafic magma could as-
conveys false illusions about how real magma reservoirs work. cend in the center of a propagating, silicic magma-
filled dike and overtake the more viscous compo-
nent at the propagating tip of the dike. As the dike
allow this initially reversed orientation of zoning propagated upward beyond this point, the mafic
to be switched to the more normally observed component would advance ahead of the silicic
mafic- (annular) to-silicic (core) zoning has been component, filling the dike and plating its walls.
the object of laboratory experiments (Koyaguchi, The silicic magma would then follow along, flush-
1985; Blake and Campbell, 1986). ing the dike of the mafic component except at the
walls where the more refractory mafic magma had
Model for Simultaneous Inlet Flow already chilled. According to Blake and Campbell
Transformed to Sequential Flow (1986), the result of this overtaking process is
at the Outlet to transform simultaneous, core-annular flow at
In an experiment intended to model the dynam- depth to the sequential emplacement of the two
ics of pressure-driven, two-component magma magmatic components at shallower levels in the
transport in a long cylindrical conduit, Blake and crust (see Fig. 9 of their paper). As such, this
Campbell (1986) used the gravity-induced flow of model cannot explain those observations requir-
two fluids having nearly the same density but dif- ing the simultaneous emplacement and venting of
fering viscosities in a vertical cylindrical tube. To lavas. In addition, the Blake- Campbell and Ko-
simulate the inlet arrangement expected for with- yaguchi models also require that the simultaneous
 14. Two-Component Magma Transport 329
withdrawal of the two components be coincident passage of the higher viscosity component (Fig. 6).
with the vertical propagation of the dike. Such a This lubrication enhances the probability that a
restriction on timing is not necessary for produc- viscous magma will reach the surface by reducing
ing composite intrusions and flows in the model the required pressure gradient. In the case of the
developed here and discussed later. Deadman Dome flow in the Inyo chain, the lubri-
cation of crystal-rich, high silica rhyolite by a less
Modelfor Encapsulating and viscous rhyolite reduced the required pressure
Self-Lubricating Flow gradient to I % of that needed by the crystal-rich
A contrasting view for the origin of most nor- magma to reach the surface alone (Carrigan et al.,
mally zoned intrusions is presented in this chap- 1992). Even if the withdrawal of magma layers dic-
ter. It is argued here and elsewhere (Carrigan and tates in some cases a silicic- (wall) to-mafic (cen-
Eichelberger, 1990; Carrigan et al., 1992) that in tral) reversed ordering at the inlet, we argue that
many, but possibly not all, cases, composite intru- such a state is unstable and will break down quickly
sions do represent snapshots in time of zoned into the simultaneous flow of the two magmas
structures that were formed by two magmas flow- with a mafic- (wall) to-silicic (central) ordering
ing together in adjacent layers. We argue here without invoking Koyaguchi's overtaking model.
that the mafic-to-silicic zoning observed in dikes In considering inlet conditions that give rise to lu-
and conduits is often a natural consequence of bricated, core-annular flow, we also propose a
the fluid dynamics of the simultaneous transport considerably more realistic alternative model for
of two magmas with differing viscosities. Pres- simulating magma pathways and chambers than
sure differences arising between the magmas will the conventional withdrawal model employing a
cause the encapsulation of the more viscous com- symmetric, boxlike geometry. In addition to pro-
ponent by the second, less viscous component. viding a basis for understanding the origin of si-
Thus, the lower viscosity component isolates the multaneous side-by-side, two-component flow, it
higher viscosity magma from zones of strong is further argued that this alternative model can be
shear at the walls and acts as a lubricant for the used to explain observations where the lower vis-

Figure 6 The flow of a very viscous, single-component, silicic magma will include zones of high shear and high dissipation
near the walls of a dike (left). Encapsulation of the silicic magma component by the mafic magma component during flow in a
dike leads to isolation of the higher viscosity component from the zones of high shear at the dike walls (right). Most of the shear
(velocity gradient) will be concentrated in the mafic layers while the central, higher viscosity, silicic magma will tend to be
characterized by little-to-vanishing shear; i.e., it will tend to be transported as a solid body carried between the lubricating layers.
Lubrication of the more viscous component can substantially decrease the pressure gradient required to maintain a given flow
rate in the dike. Alternatively, lubrication can greatly increase the flow rate of magma in the dike for a given pressure gradient.
This can be important for determining whether a dike can propagate to the surface before solidifying.
 330 Charles R. Carrigan

cosity magma could have flowed ahead of the lower energy dissipation for a given mass flux
higher viscosity component in a conduit to effuse, than either the opposite arrangement of higher
for at least the initial part of the eruption, sequen- viscosity at the wall and lower viscosity in the
tially from a vent, e.g., Fig. 1, flows 1band 2. core (Joseph et al., 1984a) or a side-by-side ar-
rangement in which two different viscosity layers
are fed together (one on top of or on the side of
Fluid Dynamics of the Simultaneous Flow the other) into a pipe or channel (Karagiannis
of Two Magmatic Components et al., 1988). Another way of determining the pre-
ferred arrangement of high and low viscosity lay-
When two fluid layers of differing viscosity are ers in a flow is to perform a stability analysis.
driven by a pressure gradient in a channel or tube, Generally, flows that are not in a preferred state
an interesting process is initiated. As the two lay- for a given mass flux will be unstable to pertur-
ers flow along the channel, the lower viscosity bations that grow with time. Ultimately, the flows
layer will gradually flow around the higher vis- evolve to a preferred state that is stable. The case
cosity component encapsulating it (Fig. 7). This of a lower viscosity core flow surrounded by a
encapsulation of the higher viscosity component higher viscosity annular flow in a circular pipe
also results in self-lubrication of the flow since was shown to always be unstable at any value
the lower viscosity component now occupies all of the Reynolds' number, i.e., at any flow rate
zones of strong shear near the walls and the higher (Hickox, 1971). On the other hand, a tube with a
viscosity component is relegated to a zone of higher viscosity liquid occupying the central zone
weaker to vanishing shear in the center of the flow. and a lower viscosity liquid flowing in the annular
This encapsulation/self-lubrication process is ro- zone was found to be stable to applied perturba-
bust and has been long used to practical advantage tions over a wide range of Reynolds' numbers
in industry for transporting viscous oil in pipe- (Joseph et al., 1984b). However, it was also found
lines by the addition of water (Charles and Red- that if the zone of lower viscosity at the wall of the
berger, 1962; Oliemans and Ooms, 1986) and for tube becomes too thick (>30% of the tube ra-
forming composite plastic extrusions (Southern dius), the core-annular flow regime becomes un-
and Ballman, 1973; Han, 1975, 1976, 1981; Mi- stable. The authors speculated that the layering
nagawa and White, 1975; White and Lee, 1975a; might break up to form an emulsion in such cases.
Karagiannis et al., 1988). It is interesting to note that all the intrusions illus-
Why encapsulation occurs has often been ex- trated in Fig. 1 have basaltic margins character-
plained in terms of the minimization of the energy ized by thicknesses that fall well within the sta-
dissipated by viscosity in the process of pumping bility criterion obtained by Joseph et al. (l984b);
a given mass flux of two different liquids in a i.e., none are too thick for the flow to be regarded
channel or pipe (MacLean, 1973; Everage, 1973; as unstable. Furthermore, the generalized descrip-
Williams, 1975). A higher viscosity core and a tion of a composite intrusion with its thinner mafic
lower viscosity annular (near wall) zone produce margins (Walker and Skelhorn, 1966) is also con-

direction of flow
Figure 7 A channel sectioned at different locations along the direction of flow of two liquid layers with different viscosities
that are injected side-by-side illustrates the progressive nature of the encapsulation or self-lubrication process. After entering the
channel the initially flat interface between the layers begins to be distorted. The interface between the two liquids increasingly
bends back on itself at each successive station downstream as the lower viscosity component (light shading) of the flow en-
croaches on the contact between the wall and the higher viscosity component (dark shading). Eventually, the higher viscosity
component is completely isolated from the high shear wall region by the lower viscosity liquid.
 14. Two-Component Magma Transport 331
sistent with emplacement by a core-annular re- The results of a coextrusion experiment (Fig. 8)
gime that is stable according to the criterion of performed by Han (1973) illustrate the effects of
Joseph et al. It is possible that magma "emul- viscosity differences on the side-by-side flow of
sions" consisting of a silicic component with two polymers having different viscosities. Using
blobs or pillows of mafic magma with little or no cylindrical dies of different aspect ratios, it was
mafic layering (see Gautneb, 1988, Fig. 5) may be found that the lower viscosity component (low
produced when there is so much mafic magma density polyethylene) gradually flowed around the
that the layers are too thick to be stable. higher viscosity component (polystyrene) to com-
pletely encapsulate it. Although the higher vis-
Polymer Coextrusion Experimental Analogues cosity component never became centered in the
lower viscosity encapsulant for dies with length-
Of all the encapsulation studies, coextrusion ex- to-diameter (LID) ratios of up to 18, it was fully
periments involving molten polymers have dy- encapsulated (separated from the die wall) within
namics that are probably most like those charac- only four diameters of the die inlet. In polymer
terizing the transport of magmas. The dynamics experiments the exact details of the encapsula-
of flowing magmas are dominated by their high tion/self-lubrication process, such as the interface
viscosities. Depending on parameters such as tem- shape, may vary in ways that depend on the rheol-
perature, volatile content, and crystal content, the ogy of a particular melt system. (In flowing poly-
rheology of magma is typically found to vary meric systems, interfacial effects between two
from purely Newtonian (vanishing dependence of components will in general be a function of the
viscosity on shear rate) to a shear thinning power- viscosity and elasticity differences in the two lay-
law behavior (Shaw, 1969; Ryan and Blevins, ers.) However, available experimental evidence
1987; Ryerson et al., 1988; Spera et al., 1988). supports the conclusion (Han, 1981) that encap-
The flow of molten polymers in capillary extru- sulation in two-polymer systems is determined by
sion dies is also dominated by viscosity. As in the the differences in viscosity just as it is in purely
case of magmas, polymers can exhibit both New- Newtonian systems.
tonian and power-law behavior (Han, 1981). In Another set of experiments by Minagawa and
both two-component polymers and magmas, the White (1975) have produced similar results for the
surface tension can be neglected compared with side-by-side extrusion mode in both a tubular die
the viscous effects. Polymers may also have elas- and a rectangular (dike-like) die. The coextrusion
tic behavior, but this additional aspect is not re- experiment using tubular geometry (Fig. 9a) illus-
sponsible for the encapsulation process (White trates the gradual bending around of the interface
and Lee, 1975b). Encapsulation is also not limited between the melts as the higher viscosity (dark)
to either purely Newtonian or power-law rheol- component is encapsulated by the lower viscos-
ogy. The process is observed for a variety of dif- ity component with increasing distance down the
ferent liquids having a wide range of material tube. As a result, the high shear, contact area
properties and seems to depend only on viscosity between the higher viscosity component and the
contrasts between the liquids involved. tube wall gradually decreases with distance down
Another feature of polymer coextrusion ex- the tube. Encapsulation would eventually isolate
periments common to magma flow in a dike or the higher viscosity layer from the wall, as in the
conduit is that any inertial forces in the flow can case of Han's (1973) experiment, if a longer tu-
be neglected compared with the viscous forces. bular die had been used. The rectangular slot ex-
Thus, the Reynolds' number, which represents the periments (Fig. 9b) varied both the viscosity ratio
ratio of the inertial to the viscous forces, is much of the layers and their initial orientation at the inlet
less than unity for both the coextrusive magmatic of the slot. In the cases where both components
and polymeric flows of interest. The similarities had the same viscosity (viscosity ratio equal to 1),
in the physical properties of the molten materi- no encapsulation occurred so that the interfaces
als and the gross dynamics of the respective flow remained essentially flat at the outlet for both
systems suggest strongly an excellent correspon- the horizontal and the vertical initial orientations.
dence between the coextrusion experiments and When both layers had different viscosities (vis-
the simultaneous flow of two magmas in a dike. cosity ratio different from I), the interface be-
332 Charles R. Carrigan

Figure 8 A coextrusion experiment involving two molten polymer layers injected side-by-side into tubular capillary dies hav-
ing different length to diameter (L! D) ratios illustrates the encapsulation of the higher viscosity component. The ratio of the
viscosities indicated for the two layers varies from 0.14 to 0.34 over the range of extrusion rates studied. Careful examination of
the extrusion at L! D = 4 shows that the lower viscosity component has already almost completely flowed around the higher
viscosity component. With greater distances downstream the higher viscosity component migrates away from the wall toward
the center of the flow. Adapted from C.O. Han (1973) with permission of John Wiley and Sons, Inc.

tween the higher and the lower viscosity compo- core-annular flow with the higher viscosity melt
nents curved around so as to isolate the higher adjacent to the wall and the lower viscosity melt
viscosity melt from the wall of the slot by in- in the center of the flow. This configuration, of
terposing an intervening layer of low viscosity course, is a reversal of what is preferred from en-
melt-by the time the layers reached the outlet ergy minimization considerations. Such an inlet ar-
15 slot thicknesses downstream from the inlet. rangement is anticipated if the simultaneous with-
The slot experiments suggest that only relatively drawal of an upper silicic and lower mafic layer
small deviations of the viscosity ratio from unity occurs from a stratified magma chamber (Blake
« 1.56) are required for encapsulation to occur and Campbell, 1986). In the engineering and poly-
over distances (L) measured in terms of slot thick- mer physics fields, this ordering of layering at
ness (D) that are small (LID < 100) compared to the inlet has been the subject of both theoretical
the typical vertical lengths of dikes (LID> 1000). (Hickox, 1971; Joseph et al., 1984b) and labora-
For comparison, the viscosity ratio is estimated tory investigations (Han, 1975). The stability anal-
to fall between 2 and 200 for the extruded lavas yses of Hickox and Joseph et al. show that this
of the Inyo Domes (Carrigan and Eichelberger, particular inlet arrangement is unstable. A co-
1990). extrusion experiment by Han employing this re-
An alternative inlet condition to side-by-side, versed core-annular inlet condition produced ad-
two-component flow into conduits or dikes is ditional details regarding the instability of reversed
 14. Two-Component Magma Transport 333
b
Horizontal interface IlHigh / IlLow

a 1.00
"

~
g,..,m.··.,.',•.,;
tit 1.56

3.35

Vertical interface

L 9 28
o 5

• High viscosity melt raJ High viscosity melt

D Low viscosity melt D Low viscosity melt
Figure 9 (a) Another coextrusion experiment with side-by-side initial conditions illustrates the bending of the interface and
the gradual isolation of the higher viscosity component from the wall with increasing distance downstream (normalized by the
pipe diameter D). The viscosity ratio is 1.56. (b) In a rectangular, dikelike channel. side-by-side injection is studied for two
different orientations of the interface between the two layers and for different viscosity ratios. The aspect ratio (downstream
length: thickness) is 15. For viscosity ratios of unity, significant interfacial distortion does not occur. The maximumencapsulation
for given length of the channel takes place when the interface is parallel to the shortest dimension (thickness) of the channel.
This is the anticipated orientation that will give rise to encapsulation in the generalized model for magma withdrawal through a
dike. Adapted from Minagawa and White (1975, Figs. 6-9) with permission from the Society of Plastics Engineers.

zoning. For perfectly concentric inlet layering, stable. Within a few tens of diameters downstream
Han found that the lower viscosity core tended from the inlet, the results suggest that even the
to drift toward the wall of the tube, producing concentrically fed reversed arrangement would
a slightly eccentric interface at a distance of have completely broken down and started to form
L = 18D from the inlet. (It is anticipated that the the energetically preferred, normally zoned core-
lower viscosity core eventually contacts the wall in annular arrangement.
a longer die given the results of the next experi- In their investigation of two-liquid transport in
ment.) Han also produced a slightly eccentric inlet a cylindrical tube, Blake and Campbell (1986)
layering for both normal and reversed layering by also found the reversed core-annular arrangement
injecting the core off center of the die inlet. At the to be unstable for higher flow rates. However, it is
end of the die (L = 18D), the normal layering curious that at the low flux rates of the inner,
(Fig. lOa) continued to exist with the higher vis- lower viscosity layer, they obtained a seemingly
cosity melt along the center and the lower viscos- stable flow in contrast to the analyses of Hickox
ity melt on the outside. For the reversed case and Joseph et al. and the experiments of Han. The
(Fig. lab), the concentric layering had broken experiment of Blake and Campbell differs from
down by the end of the die and the lower viscos- these analyses and experiments in its use of the
ity component that originally made up the core gravitational body force to help drive the flow.
now contacted the wall of the tube along more Considering that it is the unequal pressure gradi-
than 50% of its circumference. Both the concen- ents arising in the two different components that
tric and eccentric experiments indicate that the cause encapsulation (White and Lee, 1975b), it is
reversed core-annular inlet arrangement is un- possible that gravity-driven flow may be stabiliz-
 334 Charles R. Carrigan

duced at the inlet of a dike or conduit by the simul-

taneous withdrawal of two layers from a magma
chamber should be transformed to the normal
core-annular arrangement by the time it arrives at
·the surface. Such a model is in contrast with the
models of Blake and Campbell (1986) and Koya-
guchi (1985), in which different magmas will ar-
rive at the surface sequentially rather than simul-
taneously in a normal core-annular arrangement.

Numerical Models of Encapsulation and

Self-Lubrication

Recent numerical models of two-component flow

support arguments that encapsulation involves a
process that tends to rearrange a two-component
flow to minimize viscous dissipation and the driv-
ing pressure gradient. Using a finite element ap-
proach, Karagiannis et al. (1988) showed that the
encapsulation of viscous fluids in the absence of
surface tension is predicted by the minimum en-
ergy dissipation principle. Figure 7 is consistent
Figure 10 (a) Normal core-annular zoning (higher viscosity
melt in the core) is stable when eccentrically introduced into
with the results of their model for side-by-side,
long cylindrical die. (b) Reversed core-annular zoning (lower two-component flow in a channel. More recently,
viscosity melt in the core) is unstable and rapidly migrates to a novel numerical model based upon the lattice-
wall when eccentrically introduced into long cylindrical die. gas approach was used to simulate side-by-side and
This suggests that any reversed core-annular flow produced at
dispersed two-component flow in a pipe (Stock-
a conduit inlet will change into a normally zoned flow at the
surface. Adapted from C. D. Han (1975) with permission from
man et al., 1990). The lattice-gas method simu-
John Wiley and Sons. Inc. lates the equations of motion for a liquid in an av-
eraged sense using collections of discrete particles
that are constrained to move on fixed triangular
ing the reversed core-annular regime by minimiz- lattices (Frisch et al., 1986; Rothman and Keller,
ing pressure differences between the components. 1988). As the particles move along the links of the
On the basis of Han's experiment, it is expected lattice network during a time step, they collide
that any reversed core-annular flow of magma will with other particles conserving both momentum
rapidly become unstable over distances down- and particle number. By averaging 'a number of
stream from the inlet of a dike corresponding time steps, the continuum behavior of the many-
to tens of dike thicknesses. The lower viscosity particle system is obtained.
magma will increase its contact with the wall until Both the reversed core-annular and the side-
the higher viscosity component is isolated and the by-side, two-component inlet conditions were used
normal core-annular arrangement is established. in the pipe flow simulations (Figs. 11B and II C).
Thus, extrapolating the results of Han's experi- It is not surprising that in both cases, the flows
ment, any reversed core-annular arrangement pro- break up to eventually form the normal core-

Figure 11 (A) Two-dimensional lattice gas model of a 50% mixture of high (black) and low (white) viscosity fluid flowing in
a channel. Each frame from the top (initial conditions) downward represents a successively later time in the development of the
flow with the final frame representing the evolved two-component flow morphology. (Numbers on the frame sides indicate the
number of site updates or time-steps required to reach that point in the flow's evolution.) Flow is to the right and the viscosity
ratio is 8 in cases A-C. Model predicts that viscous segregation will be more or less complete several thousand channel thick-
nesses downstream. (B) Reverse zoned flow is shown to be unStable as it evolves from the top frame downward to a normally
zoned flow. Higher viscosity layers at the walls are exchanged for layers having lower viscosity. (e) As in the case of polymer
 14. Two-Component Magma Transport 335

experiments, side-by-side flow is shown to change to a normal core-annular distribution. Note the thin high viscosity layer along
the bottom wall of the channel. Such a layer does not destabilize the normal core-annular regime as demonstrated by Hu and
Joseph (1989). (D) Same as (A) except that a viscosity ratio of 2.5 is used. Reproduced from Stockman et al. (1990) with
permission from Nature and Macmillan Magazines Limited.
336 Charles R. Carrigan

annular configuration. In one case (Fig. II C), a nent to migrate to the wall would seem to form
thin margin of higher viscosity material (black) much less effectively in a well-dispersed system.
remains on the bottom boundary while the rest of There is also some evidence that the evolution of
the flow has changed to the normal core-annular a two-component dispersed system is rather sen-
arrangement. Hu and Joseph (1989) found that sitive to both the type of flow and the rheology of
such an arrangement involving a thin layer of the suspending component (Han, 1981). The pro-
higher viscosity liquid on the boundary was in cess of segregation in dispersed, two-component
fact stable as long as the layer remained thin. This systems is clearly more complex than in side-by-
has an important implication for magma trans- side flow and is a good candidate for further in-
port. Magma flowing in a dike or conduit loses vestigation using both laboratory and numerical
heat to the host rock. As heat diffuses across the methods.
boundary of the dike, a boundary layer or nar- One conclusion can be made about the appli-
row zone of cooler magma develops. This cooler cation of viscous segregation to systems where
layer also becomes a more viscous layer at the one component is dispersed in the other. If the
boundary, due to the temperature dependence of dispersed low viscosity component must remain
viscosity in silicate melts. Boundary roughness, warmer than the matrix component to retain its
rising vapor bubbles, and other effects responsible fluidity, e.g., a basalt in rhyolite, then the charac-
for deviations from laminar flow will tend to work teristic size of the dispersion must be large enough
against thermal diffusion to keep this layer thin that the diffusion time scale for cooling is large
and the boundary temperatures high (Carrigan et compared to the total time that the more fluid
al., 1992). The results of Hu and Joseph (1989) component exists in the dispersed mode. A crude
and Stockman et al. (1990) suggest that such a estimate based upon the thermal diffusion of heat
thin, thermal boundary layer of magma will not from a spherical body suggests that a total time of
impede the formation of the core-annular regime. 3 h in the dispersed mode before segregation is
Another inlet condition considered by Stock- complete would require the suspended compo-
man et al. (1990) consisted of a 50% mixture of nent to have a parcel size of 0.5 m. If the dis-
dispersed higher and lower viscosity components. persed mode in a two-component system has
Such a mixture might be anticipated with the for- greater fluidity because of chemical differences,
mation of mafic pillows or inclusions (enclaves) e.g., a higher water content, then the dispersion
in the chamber of a silicic system (Blake et al., size can be much smaller since chemical diffusiv-
1965; Eichelberger, 1980; Vogel, 1982; Huppert ities are much smaller than the thermal diffusivity
et al., 1982; Bacon, 1986). In the numerical ex- of magma.
periment of Stockman et al. (1990), the mixture
also segregated into wall and core layers with flow
along the tube for Newtonian viscosity ratios of
Analytical Modeling of Lubricated
8: 1 and 2.5: 1 (Figs. 11A and lID). However,
Magma Transport
the segregation was never complete and blobs of
lower viscosity fluid continued to remain sus- Derivation ofLubrication Equations
pended in the core flow for both the dispersed and
The vector forms of the equations governing
the side-by-side cases. This observation is prob-
the flow of magma in a dike are
ably explained by a large reduction in shear in the
central region of the flow as the core-annular re- av -1 1
- + v.Vv =- Vp - - V. T (1)
gime is set up. (In the absence of shear, viscous at p p
segregation should vanish.) Polymer blends of
higher and lower viscosity components also show (motion)
a tendency to segregate with flow along a tube V·v o (2)
(Lee and White, 1975), although the segregation
rate is not nearly as dramatic as for side-by-side (continuity)
coextrusion. It is not surprising that blends do not et k YJ
form a core annular regime as rapidly as side-by- - + v . VT = - VlT + -<jJv (3)
at pc; pc;
side, two-component flow since the pressure dif-
ferences needed for the lower viscosity compo- (energy),
 14. Two-Component Magma Transport 337
where the temperature, pressure and two-dimen- even weak deviations from laminar flow such
sional, local flow velocity are T. p, and v = ui + as caused by the buoyant rise of gas or vapor
wk, respectively, with unit vectors i and k ori- bubbles or boundary roughness will suppress the
ented across and vertically along the dike. The T rate of growth of a thermal boundary layer and
in the equation of motion is the viscous stress ten- produce near core-flow magmatic temperatures in
sor. The form of T is determined by the nature of the boundary layer 1- 2 km downstream from the
the magma rheology, e.g., Newtonian or power dike inlet. What cooling of magma at the wall
law. Since a magma may exhibit both rheologies does occur, with its attendant increase in viscosity,
over different ranges of temperature and/or vola- should not immediately affect the dynamics of lu-
tile contents, the viscous term is left in its most bricated flow according to Hu and Joseph (1989).
general form. Also, note that the pressure gradient Thus, in most of the derivations we do not con-
term involving '\Ip is assumed to be much larger sider the energy equation although we do consider
than the body force resulting from thermal buoy- the thermal effects of viscous dissipation on the
ancy so that the latter can be neglected. If a vol- lubricating layers in a later subsection.
canic conduit becomes plugged and the forced The velocity of flow in a dike is assumed to
flow of magma is reduced to a small or vanishing have the form v = w(x) k, which automatically
amount, this assumption is no longer valid and the satisfies Eq. (2). The magnitude of flow is slow
thermal buoyancy term may provide the dominant enough for a given dike and range of magma
driving force resulting in convective circulations vis7151 cosities that the Reynolds' number is ap-
(Carrigan, 1983). The quantity YJ in Eq. (3) is a propriate for the limit of creeping flow, i.e., Re =
generalized Newtonian viscosity and cP., is the dis- wdplp: « 1 where d, p and J-L are the character-
sipation function (Bird et al., 1960). Vector and istic values of the dike width, magma density, and
scalar terms may be expressed in rectangular co- magma dynamic viscosity, respectively. The
ordinates as Reynolds' number may be interpreted as the ratio
of inertial terms (left side of Eq. (1», to the viscous
v . '\I v (u auax + w aU)i
a:
term in Eq. (1). In the creeping flow limit, only
the pressure gradient term is left to balance the
(4)
+ (u awax + aW)k,
w
az
viscous term so that Eq. (1) in rectangular coor-
dinates reduces to

ar..+a'T.
'\I''T= ( - -x ) I, (7)
ax az
(5)
where the tensor component 'T" is the shear stress
+ (a'T x : + a'T ==)k
ax a: ' in the z direction acting on a surface whose nor-
mal is in the x direction. The magnitude of the
and vertical component of velocity can be related to
aw+ au)"
(-
the stress tensor component for both Newtonian
ax -az and power-law rheologies by
(6) n

T _
.ra
= _ml dW
dx
/ I dw.
d.x (8)

Steady solutions of these equations have been When 11 = 1, this equation results in Newtonian
obtained for the flow of a single-component rheology in which m = J-L. For 11 > 1, the equa-
magma with temperature-dependent, Newtonian tion describes a shear thickening or dilatant fluid,
viscosity in high aspect ratio (10 3 : I or greater) whereas for 11 < 1 a shear thinning or pseudo-
dikes that lose heat across their boundaries into plastic fluid is modeled.
the host regime (Carrigan et al., 1992). For typi- To investigate the lubricative effects of two-
cal magma properties and rates of flow, thermal component flow on the flux of magma in a dike
boundary layers of the order of 0.1 m in thickness for a given pressure gradient or, alternatively,
develop in the magma adjacent to the boundaries on the pressure gradient reduction for a given
of the dike. However, Carrigan et al. found that magma flux, we can easily solve Eq. (7) for two-
338 Charles R. Carrigan

component magma systems with sidewall lubri- dw

cation. We consider a dike of half-thickness d/2 or dw =s dx. (13)
dx
with a lubricating layer of thickness a on each
Eliminating the shear stresses T;z and T~z in
sidewall. Consistent with the lubrication approxi-
Eqs. (l2a) and (l2b) using Eqs. (lla) and (llb),
mation, we assume that fluid pressure variations
and then taking the derivative with respect to x
in the cross-stream direction are small to vanish-
produces
ing compared to variations along the vertical di-
rection of flow. In our model, the negative of the n m i:: n w - I

pressure gradient is given by

dx = - w ~ de, 0::5 x ::5 a (14a)

G = _ dp Pc - Po n.m.e»:' d
dz L
(9) dx = c ~ de, a::5 x ::5 2' (l4b)

where subscripts c and 0 denote the chamber and Using Eq. (l3), we can now eliminate dx in
outlet (atmospheric) pressures, respectively, and L Eqs. (l4a) and (14b) to obtain
is the length of the dike. Because the problem is
dw = - n m sn w
de
symmetrical, we need only to solve for the ve- w w (l5a)
G '
locity or volume flux in the layers falling between
a wall and the centerline of the dike. Integrating n m i:: nc - I
d
dw = - c c de a::5 x::5-. (l5b)
Eq. (7) with respect to x in the wall and central G ' 2
layers gives
These last two equations may be readily inte-
T~, = Gx + Cr, 0 ::5 X ::5 a (lOa) grated to obtain expressions for w(x) over the in-
T~, = Gx + C~, a ::5 X ::5 d/2, (lOb) tervals 0 ::5 X ::5 a and a ::5 x ::5 d/2. Integration of
Eq. (l5a) between 0 and x utilizes w(O) = 0 and
with the superscripts wand c denoting the wall i::(0) as given by Eq. (21) as boundary conditions.
and central layers, respectively. The fact that Integration of Eq. (I5b) between a and x requires
T;, = T ~z at x = a means that the integration con- the evaluation of w(a) from the expression for w
stants C f and C ~ are equal and will be referred in the interval 0 ::5 X ::5 a and i::(a) from an equa-
to as C 1 • Rewriting Eqs. (lOa) and (lOb) as tion for i::(x) obtained from Eq. (l4b). These inte-
grations finally yield the desired result for the
O::5x::5a (lla)
cross-stream dependence of two-layer flow in the
T~z = G(x a ::5 X ::5 d/2, (Ll b) dike over the half-width 0::5 X ::5 d/2:

permits the constant C 1 to be interpreted as that w(x) = ~ (Q)lln

n; + 1 m;
W

(l6a)
x value in the dike where the shear stress vanishes.
Since this can happen only at one location in
the dike, i.e., x = C 1 , and the dike is symmetric
about the centerline, C 1 must be equal to d/2. The
absolute value sign appearing in the generalized O::5x::5a
shear stress + rate-of-strain relationship (Eq. (8»
can be eliminated if only one-half of the dike
channel is considered, where dwldx is already
greater than or equal to zero. Thus, from Eq. (8)
we can write the shear stress as a function of strain
rate in the wall and central layers as
o ::5 X ::5 a (l2a)

T Cxz = a ::5 x ::5 d/2, (l2b)

where m.i, nw , m., and n; are the coefficients and

exponents needed to define the power-law vis-
cosity for the wall layer and for the central layer, d
a ::5 x::5-.
respectively, and where 2
 14. Two-Component Magma Transport 339
The total volume flux of magma (volume per Gd 3
unit horizontal length of dike per unit time), Q, in /-Legv = 12Q' (20)
the dike can be determined by summing the con-
tributions from the wall and core layers, This expression is particularly useful in a later
section for evaluating the effect of viscosity on
(a (d/2
dike propagation rates. For time- independent flow,
Q = Qw + Qe = Jo W dx + Ja W dx, (17)
the velocity gradient or rate of strain at the bound-
ary is given by
where the integration of the expressions for w(x)
in Eqs. (l6a) and (l6b) over the limits of each Gd )I/nw
layer yields teO) = (- (21)
Zm.;
w
a. = ~ (Q.)I/n [a(~)(nw+l)/nw (l8a) and the velocity gradient in the wall layer of
n; + 1 m; 2 magma at the interface between the wall and cen-
+ (d/2 - a)<2n w+I)/n w - (d/2)(2n w+l)/nw ] trallayers is

(2n w + l)/n w
tea) (22)
G ) I/nc (l8b)
Qc = ne : 1 ( me In the common case of two magmas having the
same power-law exponent, n; = n; = n, such
·(1 - (2n e ~ 1)/nJ (~ as when both magmas are treated as Newtonian
fluids, the pressure gradient - G required to pro-

+--- -
n; (G)I/n w duce a given total volume flux of magma Q is
readily obtained from (l8a) and (l8b),
n; + 1 m;

· [ (~rW+IJ/nw (~_ a rW+IJ/nW] G = (D + ~ + F) n, (23)

where the terms in the denominator are written as

· (~ - a)' D = _ n (.2..)I/n [a(~)(n+IJ/n
Other useful expressions for power-law, 'single- n + 1 m; 2
and two-component flow may be developed or in- + (d/2- a)<2n+ll/n - (d/2)<2n+l)/nJ
ferred from these equations. The local viscosity of (2n+ l)/n . ,
a power-law magma at an arbitrary point in the
flow can be written as
I/n
/-Lloe = m ev: 1 (19)
E=n:l (~e )
(2n+I)/n
according to Eq. (8) over the half-width, where
t ~ O. This viscosity is just a constant for a New- ·(1 - ~ 1)/n)(~ (2n - a) ,

tonian magma (n = 1), but it is a function of shear and

rate and therefore location in the dike for non- I/n
Newtonian magmas (n 7' 1). Another kind of vis- n 1
F--- -
cosity might be called the equivalent viscosity, n + 1( m; )
/-Legv, corresponding to the viscosity of a New-
tonian magma that would result in an equivalent · [ (~r+I)/n (~_ a f+l)/n]
flux of magma through a dike for the same values
of the pressure gradient, - G, and dike width, d.
Equating the volume flux Q of magma for any · (~ - a)'
given type of flow in a dike, e.g., single-layer non-
Newtonian, multilayer non-Newtonian, or multi- When the power-law exponents are different for
layer Newtonian, to the expression for the flux of the wall and core layers of magma, the value of
a single-component, Newtonian flow yields G required for a given total flow rate may be
340 Charles R. Carrigan

10

0.4
x(m)

Figure 12 Velocity profiles are plotted over the dike half-width for single-component basaltic flows having different values of
the power-law index 11. The zero shear viscosity is 52 Pa . s for the three cases 11 = 0.75, I, \.5. For comparison with the n =
0.75 case, a Newtonian profile that would yield the same volumetric flux is plotted. In terms of the volumetric flux, the 11 = 0.75
case behaves as a Newtonian magma (11 = I) of 2\.6 Pa . s. The calculations assume a dike width (d) of I m and a pressure
gradient (- G) of 2 MPa . km '.

30 .-----,--.-----,--.-----,---,.-----,--.-----,----, 100
obtained from Eqs. (17) and (18) using iterative
techniques.
80
Effect ofLubrication by Newtonian and
Non-Newtonian Wall Layers 20
":'1Il

Given the occurrence of viscous segregation, 60

".
":'E
the derived equations permit the evaluation of the '"§. !?;.
pressure reduction and flow enhancement associ- 0 go>
:1.
40
ated with the presence of lower viscosity New- 10
tonian or power-law wall layers. For comparison,
single-component velocity profiles for a given 20
nominal pressure gradient (2 X 103 Pa . m - I) ob-
tained from Eqs. (l6a) and (l6b) are illustrated in
Fig. 12 for several different values of the power-
0.6 0.8 1.0 1.4
law exponent n. The same value of the power-law Power law exponent (n)
coefficient m is used for each case and corre-
Figure 13 The effect of varying the power-law exponent n
sponds at n = I to a viscosity of 52 Pa . s, which on the volumetric flux Q. defined by Eqs. (17) and (18), and
is characteristic of basaltic compositions (Shaw the equivalent viscosity J.L,q., defined by Eq. (20), is plotted for
et al., 1968; Ryan and Blevins, 1987; Ryerson a single-component magma, with a power-law coefficient m =
et al., 1988). Not surprisingly, a shear-thinning 52. flowing in a I-m dike and driven by a 2 MPa . km t pres-
sure gradient. At /l = I, the viscosity has no dependence on
magma (n < I) has lower viscosity behavior in
the rate of strain and the flow is Newtonian with molecular
regions of high shear where the majority of vis- viscosity equal to the equivalent viscosity. For n < I, the
cous dissipation occurs with the result that maxi- magma is shear thinning and is characterized by an equiva-
mum flow velocity and volumetric flux (Fig. 13) lent viscosity that is less than that for the Newtonian case
are greater than for the Newtonian (n = I) or (52 Pa . s). For /l > I, the magma is shear thickening and in
terms of the volumetric flux behaves as a Newtonian magma
shear-thickened cases (n > I). The shear-thinned
with viscosity exceeding 52 Pa . s. The quantity Q is the vol-
case considered here (n = 0.75) has an equiva- ume of magma flowing per unit time and per unit horizontal
lent viscosity, as defined by Eq. (20), of about length of the dike. (Arrows indicate appropriate axis for each
21.6 Pa . s according to Fig. 13 and the Newto- plot.)
 14. Two-Component Magma Transport 341
1000 4
Dike Dike
wall center-
a =0.3 m Iline
4 I - - - - - - - - -•. l - - - - - - i 800
3

0.2m
...
e,
>
1

o ........."'-'-"="-:..:..:..0.:..:..--'-_ _- ' - _ - - 1_ _- ' - _ - - ' Lubricating layer thickness a (m)

0.0 0.2 0.4 0.6
x(m) Figure 15 The volumetric flux Q and the equivalent vis-
cosity J.L,q. are plotted as a function of the lubricating layer
Figure 14 The effect of varying the width, a, of the lubricat-
thickness, a. assuming the same parameter values given in
ing layers on a two-component velocity profile is plotted for
Fig. 14. The equivalent viscosity rapidly decreases for 0 < a
three values of a. For a basaltic viscosity of 52 Pa . s in the
< 0.05 m. Beyond a width of a = 0.1 m, the equivalent vis-
lubricating layer and a rhyolitic viscosity of2.2 X 10' Pa . s
cosity approaches the molecular viscosity (52 Pa . s) of the
in the core layer, it is found that a lubricating layer of almost
lubricating layers much more slowly.
any thickness strongly affects the profile and, hence, the flow
rate of the viscous component through the dike. For compari-
son, the single-component velocity profile (multiplied by 10)
for rhyolite is shown as a dashed curve. As a approaches 30% layer with 0.10- to 0.15-m lubricating wall layers
of the dike width (I m), the ascent rate of the rhyolite core produces volumetric fluxes in the 2 m' . m - 1 . S-1
becomes comparable to a purely basaltic flow (n = I, Fig. 12)
range. In this case, lubrication produces a 250-
driven by the same pressure gradient (2 MPa . km -I).
fold increase in the volumetric flux for the same
driving pressure gradient, an effect that is particu-
larly important for the production of dikes by
nian profile for a magma with this viscosity is magma-driven fractures. The role oflubrication in
plotted in Fig. 12 for comparison with the shear- magma-driven fracturing is discussed in more de-
thinned case. tail later.
The effects of lubrication on the velocity pro- So far we have looked only at two-layer systems
file, volumetric flux, and equivalent viscosity are with purely Newtonian rheologies (n = 1). When
shown in Figs. 14 and 15 for Newtonian, rhyo- each viscosity is characterized by a different value
litic magmas with Newtonian, basaltic lubricat- of the power-law exponent, Eqs. (l8a) and (l8b)
ing layers. As is apparent from Fig. 14, the shear predict a curious result. Two-component flows
is Virtually eliminated in the central rhyolitic in a dike can produce volumetric fluxes for a
core (2.2 X 104 Pa . s; Carrigan and Eichel- given pressure gradient that exceed the fluxes pro-
berger, 1990), even with thin lubricating layers duced at the same pressure gradient by single-
(52 Pa . s). From Fig. 15, the equivalent viscosity component flows of either of the magma con-
falls from rhyolitic values to less than 200 Pa . s stituents. Alternatively expressed, the equivalent
when the lubricating layer on each wall is about viscosity, as given by Eq. (20), can be smaller for
0.05 m thick in a l-m-thick dike and to about two-component flow than the equivalent viscosity
75 Pa . s with 0.15 m-lubricating layers. Also of either component flowing in the dike by itself.
plotted in Fig. 15 is the volumetric flux through Models for the coextrusion of power-law materi-
the l-m-wide dike as a function of lubricat- als (Han and Chin, 1979) predict that this behav-
ing layer thickness. For a pressure gradient of ior can occur when both magmas have the same
2 MPa . km -I, a single-component rhyolitic local viscosity (Eq. (20» at the interface, x = G,
flow is characterized by a volume flux of only between the wall and core layers. Equating the
8 X 10--' rn" . m -) . s I whereas a rhyolitic core shear stresses in Eqs. (12a) and (l2b) and solving
342 Charles R. Carrigan

for the value of strain rate at which the viscosity Gem" as given by Eq. (26). The power-law expo-

er
relationships cross give nents (n l = 0.5 and n2 = 0.8) and coefficients
(m) = 100 and m 2 = 50) assumed in this example
.
Scros~ = • (24) appear to be characteristic of values that are ap-
mw propriate for a picritic magma at different tem-
The value of eemss is the value of shear strain rate peratures and crystal contents (Ryerson et al.,
where both magmas have the same viscosity. If 1988) with magma I representing the lower tem-
this value of the shear strain rate also occurs at the perature component. In the vicinity of Gem" =
interface, Eq. (22) can be used to write 794 Pa . m- I (=0.8 MPa . km r the maximum
'),

w value of Q is achieved for magma 1 in the lubri-

eeroS' = [n~w (~ a) fn (25)
cating wall layer and magma 2 in the core layer.
The minimum value results from the reversal of
The value of G where this occurs is obtained by this arrangement. In between fall the values of Q
eliminating eem" between Eqs. (24) and (25): for the two magmas flowing in the dike singly. In
1 W terms of the equivalent viscosity, the two-magma
o.: = (d/2m~ a) [(::)"w/ -II,}n (26)
example giving the highest flow rate at Gem" has a
/-Leqv of 30.6 Pa . s while the two components flow-

ing alone in the dike have values of /-Leqv of 32.3

Thus, for given magma viscosities, dike widths,
and 33.6 Pa . s. One implication of such a predic-
and wall layer thicknesses, we can now calculate
tion is that two-component, power-law magmatic
the pressure gradient at which a two-magma sys-
flows are potentially better for minimizing vis-
tem will have a lower equivalent viscosity than ei-
cous losses during dike propagation than either
ther of its components.
component alone propagating the dike with the
Figure 16 shows the dependence of the volu-
same pressure gradient. Whether this behavior can
metric flux Q on the parameter G for a I-m-wide
really occur for a particular two-component sys-
dike with O. I-m wall layers. Equations (17), (I Sa),
tem will depend critically on the validity of em-
and (I 8b) were evaluated for a range of G about
pirically determined power-law relationships of
each component in the vicinity of eemss'

Range of G where 2 comp flow

Effect ofLubricated Flows
exceeds flow rates of each on Dike Propagation
4 component separately

'en
... \
1 comp
Analyses indicate that viscous losses associ-
ated with magma flow in narrow fractures may be
(n=0.8, m=50.) far more important than rock fracture toughness
'e (often designated K ,) in determining if the pres-
(0)
g 2 sure in a magma chamber is adequate to propagate
0
\
2 comp-
a dike significant distances (e.g., Spence and Tur-
reversed cotte, 1985; Lister and Kerr, 1991). Thus, it is
_ 1 comp expected that lubrication may play an important
(n=0.5, m=100.)
role in determining whether a particularly viscous
OL...!S<'=L.-L..J'--l'--l---l.---l.---l.....J......J..--'---'---'---'---l
0.0 200 400 600 800 1000 1200 1400 1600 magma, such as rhyolite, is able to penetrate to
Driving pressure gradient G (Pa . rn") the surface through the crust overlying a magma
chamber or significant distances along the horizon
Figure 16 For two power-law magmas, the two-component
volumetric flux can exceed the single-component volumetric
of neutral buoyancy (Ryan, 1987). To evaluate the
flux for the same value of the pressure gradient. The power- effect of lubrication on dike propagation, any of
law exponents (n, = 0.5 and n: = 0.8) and coefficients (nI, ~ the analyses by Geertsma and Haafkens (1979),
100 and m, = 50) are within the parameter range observed by Spence and Turcotte (1985), or Nilson and Grif-
Ryerson et al. (1988). The maximum flux with component 1
fiths (1986) could be used. The differences are mi-
in the lubricating layer and component 2 in the core is obtained
in the range of G between the vertical dashed lines. Reversal
nor and we use the Geertsma and Haafkens (1979)
of this order reduces the flux to values less than would be ob- derivation for the propagation of parallel-sided
tained for either component flowing in the dike individually. fractures that is consistent with the plane-parallel
 14. Two-Component Magma Transport 343
models derived here. It should be noted that the Replacing ,u with ,ueqv allows the effect of lubri-
Geertsma and Haafkens analysis does not take cation to be evaluated for two-component flows
into account the decrease in the confining stress as that exhibit either Newtonian or power-law be-
the fracture tip propagates vertically upward. On havior. As an example, consider the injection of
the other hand, the magmatic pressure-head loss a rhyolitic magma into a fracture. Once the frac-
associated with the weight of the magma in the ture is established and magma is flowing in the
vertical fracture is also not considered. Both ef- dike, the required driving pressure (not necessar-
fects approximately cancel each other. (Depend- ily the actual chamber pressure) would fall off
ing on the initial crustal stress state, other, more from about 22 MPa to only a few MPa if the dike
detailed fracture propagation models may be re- were to reach 5 km in length (Fig. 17). During the
quired to adequately predict fracture propagation propagation of the dike out to a distance 5 km
rates.) The solutions used also assume that the en- from the chamber, its width would increase from
ergy associated with fracturing at the crack tip is about 0.3 m to about 1.5 m (Fig. 17). However,
negligible compared to viscous dissipation losses. the fracture tip velocity as given by Eq. (28) and
This is not a good assumption at the time just after plotted in Fig. 18 would fall from about 0.04 m .
a fracture is initiated if the crustal rock is com- s - I to about 0.01 m . s- I, which is probably much
petent at the point of initiation, but it becomes a too slow to prevent solidification over the 5-km
progressively better assumption as the magma- distance. Now consider the same driving pressure
filled fracture lengthens. However, for the pur- history of Fig. 17 for a rhyolitic magma lubricated
poses of estimating dike propagation rates between by 100 Pa . s basaltic wall layers that are 16% as
a magma reservoir and the surface, it probably is thick as the dike width. The distribution of dike
a good assumption even near the time of fracture width (Fig. 17) with fracture length is the same as
initiation at the magma reservoir boundary since predicted for the rhyolitic flow assuming that it
the wall rock there is likely to be at least partially somehow traversed the 5 km. But now, rather than
molten so that the fracture toughness will be sub- having a viscosity of 2.2 X 104 Pa . s as in the
stantially reduced. case of the pure rhyolitic flow, the equivalent vis-
The model assumes a constant flux Q (m? . S-I cosity of the lubricated flow is only 146 Pa . s,
. m - I) of fluid into the fracture with time so that
the driving pressure in the magma chamber can 2 ....--,...--,..---,..---,..---,..------, 30
therefore vary as the propagating fracture both
widens and lengthens with time. The fracture
m
length as a function of time t is written as 0.
~
ell
20 :;
I = 0.68 QI/2 ( M, ) 1/6 t2/1 (27) Is: Ul
Ul
J.l(l - v) , ell
~
'i
a.
Cl
where M, (2 X 10 Pa: Spence and Turcotte,
10
e "
:~

".
:::J
1985) and v (0.25) are the elastic shear modulus 10 -0
and Poisson's ratio, respectively. Taking the de- u: ¥
'5
rivative of Eq. (27) with respect to time yields the C'
ell
a:
instantaneous velocity of the fracture tip:
0'----'-----'-----'-----'-----'------'0
dl 0.453Q1/2 ( M, ) 1/6 t -1/3. (28) o 2000 4000 6000
dt ,u(l - v) . Fracture length (rn)

The width of the fracture d is given as Figure 17 In the formulation for fracture propagation pre-
sented here, a constant volumetric flux of magma is injected
d = 1.87QI/2 (29) into the fracture as it propagates vertically toward the surface.
As injection occurs, the fracture grows in length and it also wid-
ens. Since the required driving pressure is more sensitive to
and the pressure is changes in the fracture width than in its length, the required
3/4 driving pressure for a given rate of injection of magma falls off
M, with length. In this case the fracture grows from a width of about
P = 1.13(Q,u)1/4 I 1/2 (30)
( ,u(l - v) ) . 0.3 m to about 1.5 m as it lengthens from about 0.2 to 5 km.
344 Charles R. Carrigan

term involving the dissipation function 4>,. in

Eq. (3). According to the energy equation, this
term represents the amount of heat produced per
'Ill
unit volume per unit time by the dissipation of
flow energy. For an incompressible plane paral-
lel flow, the energy S dissipated per unit volume
.1
Rhyolite w/o lubrication per unit time assuming a general power-law rhe-
~4pas ology is
2
.01
dw
S Tl4>. = TI ( dx )
(31 )
.001 '--_--'----'---'---'-l...J....L..J.J.._ _'---...J--'-L-J....L..L.U
100 1000 10000 = m; dw '" I(dw )2
Fracture length (rn)
( dx ) dx

Figure 18 Comparison of fracture propagation velocities In terms of the gradient of the heat flux q, we have
produced by a rhyolite (2.2 X 104 Pa . s) without lubrication
and a rhyolite lubricated by a basalt (a = 0.16d. /-Lw ~ 100 Pa dq
s, /-Leq. = 146 Pa . s) subjected to the same magma chamber
- = S = m; t"w+I , (32)
dx
pressures (Fig. 17). The lubricated rhyolite drives the fracture
at about 150 times the rate of the unlubricated rhyolite for the where
same pressures.

t(x) = [~w (~ x) J'' ' (33)

which is about 150 times smaller. The resulting for 0 ::5 X ::5 a.
crack-tip velocity of the lubricated dike is more
than 100 times larger, varying between 6 and l m- Integration with respect to x of Eq. (32) gives
s- 1 as the fracture lengthens and widens (Fig. 18).
Thus, given the same pressure history as the q= -m
w
( nw )(.Q.)(".tl)/"w
single-component rhyolitic flow, the lubricated Zn; + I m; (34)
flow has a much greater chance of reaching the sur- d )12".II)/"W
face or extending significant distances in horizon- . ( '2 - x + C1 ,
tal (lateral) propagation modes. (Lateral rift zone
propagation (=40-60 km) can be some 10- 30
times the fracture length required to move magma where C, is a constant of integration. By Fourier's
from shallow crustal reservoirs to the surface.) law, we have

dT
Effect a/Viscous Heating in Lubricated Flows -k - = q. (35)
dr
In the two-component flow of a rhyolite and
basalt, virtually all the shear is concentrated in the To solve for temperature, the right side of Eq. (34)
thin wall layers (Fig. 14) and, as a result, virtu- is substituted for q in Eq. (35) and the equation is
ally all the viscous dissipation will occur in these integrated with respect to x. Another constant of
layers too. This presents yet another way that lu- integration is introduced. These are evaluated by
brication by thin wall layers can promote dike applying boundary conditions on the temperatures
propagation, since it has been shown already that at x = 0 and at x = a. In selecting temperatures
the distribution of viscous dissipation in single- for each endpoint, it is assumed that there is some
component flow may actually increase the amount temperature below which the lubricating magma
of heat in the magma as it flows along the dike but component is simply too viscous to flow. Because
still allow solidification of magma at the dike of cooling on both the wall and core layer sides of
walls (Carrigan et al., 1992). the lubricating layer, it is expected that thin, more
A model for the effect of viscous heating on the or less rigid zones of basalt will form when the
lubricating layers can be derived using the last temperature falls below that required to sustain
 14. Two-Component Magma Transport 345

Figure 19 The thermal effect of concentrating shear in the lubricating layer (a = 0.16 m) is examined assuming that the
boundaries of a basaltic layer are chilled against the cooler wall rock and rhyolite core and become too rigid to flow. The
temperature at which this occurs in the basalt is taken to be l l Hr'C, For the case with lubrication in Fig. 18, it is found that
within the range of predicted fracture velocities (6-1 m . s -1), substantial viscous heating ofthe lubricating layer would occur.
The center of the mobile basalt layer would reach a temperature of about I 180°C. Enough heat would be produced in this case
to heat both margins of the mobile zone at the rate of about 2 kW . m- 2 • By comparison, a single-component flow of rhyolite
could creep only at 0.04-0.0 l m- s -I and produce a negligible amount of heating for offsetting heat loss to the wall rock.

the ZOne of mobility (Fig. 19). In principle, such a which the transition from mobile to effectively
model would allow chilling of basalt at the wall rigid behavior occurs. After solving for the con-
rock contact and at the interface between the ba- stants of integration, one may then write a time-
salt and rhyolite with a zone lying in between that independent solution for the temperature in a
remains mobile. The stability, i.e., the transverse power-law mobile zone as
growth or shrinkage, of such a mobile lubrication
zone depends on the balance between the rate of
heat generation in the zone and the rate of heat
T = -HJ(~ - x rnw+ll/nw
loss across both faces of the lubricating layer into (36)
the wall rock and central rhyolite layer. In this par-
ticular model T, represents the temperature at + C1x+ C2J
346 Charles R. Carrigan

where A Rational Model for Magma

w+1)/nw, Chamber Withdrawal
J = mWCllwn~ 1) Cllwn~ 1) (~J (n
A cylindrical tube attached to the center of a lid
nw+1)/nw], on a density-stratified reservoir (Fig. 5) has be-
C = ~[ (~rnw+I)lnW (~_ ar come a frequently used model for magma cham-
I
ber geometry in considering the withdrawal and
and effusion of multiple layers of magma. It is likely
that such a model is so often employed because it
d) (3nw+IllnW]
C2 = - [ kT, + J ( - . is easier to analyze numerically or to construct for
2 laboratory experiments. Unfortunately, this stor-
age withdrawal system is likely to be very unreal-
Figure 19 illustrates the temperature distribu- istic in at least two important ways that are critical
tion in the mobile zone for the lubricated flow for an understanding of the transport of two mag-
considered in the previous section (d = 1 m, a = mas and the timing of their arrival at the surface.
0.16 m, J.t. = 100 Pa . s, u; = 2.2 X 104 Pa . s First, the lid of the conventional model is pre-
and k = 1.3 W . m -I. °C-I). Using the rheological cisely parallel to surfaces of constant density in its
data of Shaw et al. (1968) to estimate the tempera- stratified interior. This represents a singular ge-
ture below which a basaltic magma behaves rig- ometry since there are no physical constraints that
idly, T, has been set to I I 10°C, and it is readily require the top and constant density surfaces to be
apparent from Fig. 19 that the average tempera- perfectly parallel. When the lid is perfectly hori-
ture of this zone is well above this transition tem- zontal, it will be in contact only with the upper-
perature. The assumption of a viscosity of 100 Pa most layer in a density-stratified system. The only
. s for this mobile zone is reasonable and may even way that lower layers can reach the lid and, thus,
be a little large for the temperatures indicated. be tapped by the tube is to require a sufficiently
This model can be used to argue that localized vis- high withdrawal rate that the draw-up-depth is
cous heating can promote the transport of excep- great enough to sample these lower layers. Even if
tionally viscous magmas even in thermally unfa- the lid is domed or pitched, the symmetric place-
vorable environments. ment of the tube on the lid still requires a suf-
The temperature gradient in the mobile zone is ficient withdrawal rate for lower layers to be
given by sampled. Within the context of the conventional
dT
dx
-- CjI[
k
- J (3n n;+ I)
w
model, it is difficult to understand, for example,
how the relatively low velocities of transport re-
(37) quired for the effusion of a lava dome can produce
d )(2nw+I)lnW] the simultaneous withdrawal of layers, particu-
. (- - x
2 ' larly when the magmas may have significant den-
sity differences as in the case of simultaneous ba-
which can be used to evaluate the rate of heat loss salt and rhyolite venting in Iceland.
into the wall rock and the core layer. For the ex- Second, cylindrical conduits, as used in the
ample considered, viscous dissipation supplies standard model, represent, in general, an unreal-
well over 2 kw . m -2 to the boundaries of the mo- istic pathway for the flow of magma from a res-
bile zone. This is a significant rate of heating that ervoir. Models of fracture growth near magma
may more than balance losses to the rhyolite and bodies produce sheetlike channels (Spence and
the wall rock (Carrigan et al., 1992). Such a mo- Turcotte, 1985) and not conduits. Surface studies
bile zone may actually increase in mean tempera- suggest that conduits are only secondary struc-
ture and grow in width with time. We have mod- tures that "bud" from sheetlike dikes and gradu-
eled several important aspects of two-component ally evolve during eruption by mechanical erosion
flow in a dike. It is now worth looking in greater (Delaney and Pollard, 1981) or by melting of the
detail at the models for supplying two-component adjacent wall rock (Bruce and Huppert, 1989,
magmas to a dike. 1990) as magma is channeled from the dike into
 14. Two-Component Magma Transport 347
the conduit. Furthermore, a conduit-like structure of the magma reservoir and a more localized,
is likely to be a feature of only the top part of conduit-like vent at or near the surface.
a dike where the cooler host regime aids the Figure 20 illustrates a more general and con-
conduit-forming process by plugging narrower siderably more realistic alternative model for a
portions of the dike through solidification. How- reservoir + withdrawal system that avoids the two
ever, in the roof of the magma chamber, solidifi- concerns associated with the conventional model.
cation at the dike inlet during the flow of magma In this schematic model, the reservoir roof is in-
is unlikely-especially considering numerical es- clined at an arbitrary angle relative to the surfaces
timates of dike boundary temperatures near the in- of constant density, and the dike inlet for with-
let (Carrigan et al., 1992). Thus, it is argued that drawal of magma is fracture-like rather than
a linear or dike-like breach in the roof of a magma circular. The presence of a dike on the sloping
chamber is a much more realistic exit in both the boundary permits withdrawal from more than one
early and later stages of the flow of magma from layer in the system simultaneously-irrespective
its reservoir than is a circular hole that is a char- of the withdrawal rates or magma viscosities and
acteristic feature of many withdrawal models. densities. In this model, the magmatic compo-
Both thermal and mechanical considerations sug- nents can arrive at the inlet as two or more discrete
gest that a mature magmatic pathway should have layers. Such inlet flow conditions exactly corre-
a dike-like inlet at the end penetrating the roof spond to the so called two-layer, side-by-side flow

--
dike

encapsulation of silicic flow where

dike crosses zoning in magma
chamber

mafic
magma

deeper
layers

Figure 20 Generalized model of magma chamber and dike system in which the roofis not parallel to the compositional layering
in the chamber and withdrawal is through a dike positioned asymmetrically on the top. This geometry allows the dike to withdraw
more than one layer simultaneously at any flow velocity in contrast to the conventional model in which the draw-up-depth
parameter is a function of withdrawal rate and magma properties. Encapsulation, as illustrated by the inset. is a natural conse-
quence of the side-by-side withdrawal of the two magma layers.
348 Charles R. Carrigan

conditions used in polymer coextrusion experi- Another implication is that dikes propagating
ments to produce encapsulation (zoning) in slots normal to the dip of the magma chamber roof tend
with the low viscosity component at the outside of to "see" smaller compositional gradients since
the flow adjacent to the wall (see Fig. 9b, bottom). they will breach the roof parallel to surfaces of
Some potential implications of this model de- constant density and hence similar composition
serve further discussion. One is that a dike breach- (Fig. 21). Thus, one dike, propagating downward
ing a magma reservoir roof in the direction of along the slope of the chamber roof, may have sig-
the dip or tilt of the roof can sample-along its nificant compositional gradients occurring along
length-different layers of the magma body si- its length on the surface whereas a second dike,
multaneously. Such a dike will produce at the emplaced perpendicular to the first, may only
surface lavas that have a compositional gradient sample magma of one composition. A corollary
along horizontal segments of the dike length to this is that selecting dike orientations and the
(Fig. 20 inset). Such compositional variations are shape of the magma chamber roof in this model
observed in the lavas of the Inyo Domes chain, permits any layer in a reservoir to be tapped with-
which are thought to have contemporaneously as- out regard to a restricted consideration of the
cended in the same dike (Sampson and Cameron, draw-up-depth associated with a particular effu-
1987). Encapsulation of the higher viscosity com- sion rate as well as to the potentially restrictive
ponent by the lower viscosity fluid should result requirements on flow rates.
in chemical zoning of flows over locations where A third implication is that lower viscosity mag-
a horizontally propagating dike crosses between mas may determine both the order of reservoir
compositional zones in the magma as it breaches withdrawal and whether the higher viscosity mag-
the sloping roof of the magma chamber. It is sug- mas will ultimately reach the surface (or, in the
gested in the next section that at least three of the case of volcanic rift zones, the lateral extent of
Inyo Domes represent a manifestation of this en- dike propagation from the source reservoir). A
capsulation process as a common (master) dike breach in a magma chamber roof initiated where
intersected different compositional boundaries in a layer of very high viscosity magma resides may
the underlying magma chamber. not result in the propagation of a dike to the sur-

compositional gradient

1T===::::II
-:

/
mafic
magma

deeper
layers

Figure 21 An implication of the generalized model for magma withdrawal is that dike orientation and location alone can
determine what magma is tapped from the chamber. Here two differently oriented dikes produce lavas with either a compositional
gradient or a constant composition depending on how they breach the chamber roof.
 14. Two-Component Magma Transport 349
face from that location on the roof owing to the : t'1;.
excessive driving pressures required for magma t t ~t
t t t ~
t t .,.~
flow. In this case, magma pressure in the chamber t t t 1 III
is not relieved by vertical propagation and, there- / t t t t t., ()~.,"
"t"t r"t~
fore, the dike can propagate along a slope that ttttt'"
ttttt~
ttTttt~
breaches the roof until it begins to sample and tttttt//tl~1
withdraw another layer of magma having a lower ttttttt
tttttttt
viscosity. At this point encapsulation of the higher tttttttt
ttttttttt ..... Drill holes
ttttttttt
viscosity component by the lower viscosity com- ttttttttt
--L.. Nonnallault
tttttttt
ponent occurs. Lubrication of the higher viscosity tttttttt
ttttttt
tttttt • Phreatic crater
component in this newly forming composite in- t t .,. t t
t t t t t Deadman C> Rhyolite lava
trusion may then be adequate to permit magma to t t t
t t t Dome t +

r
reach the surface to form a composite flow. Such t t
t t- + t Pre caldera rocks
t t
a scenario is suggested by the relationship be- t t Deer Mtn*
r t
tween the rhyolitic dike that did not breach the t
t t
t
Inyo Craters
t t
surface and the composite feeder for Obsidian -r t
t t t
Dome consisting of a rhyolitic core lubricated by t t

a less viscous rhyodacitic annular zone.

t t t
't- t
t t t
t t t
t t t t
t
Inyotrend' __==:io_. 3 km

A Physical Model for the Contemporaneous Figure 22 Map view of Inyo domes with associated volcanic
Eruption of Three Domes in the Inyo Volcanic and geologic features. The numbered squares on and by Ob-
sidian Dome indicate the locations of scientific drill holes.
Chain, Long Valley, California
Adapted from Eichelberger et al. (1988) with permission from
the American Geophysical Union.
As mentioned previously Obsidian Dome is one
of three composite units (Fig. 22) erupted more
or less contemporaneously about 600 yr ago in and consists of a fine-grained, low crystal content
Long Valley along a north-south trending line rhyolite overlying a fine, low crystal content and
that crosses the northwestern boundary of the cal- low silica rhyolite or rhyodacite (Vogel et al..
dera (Miller, 1985; Sampson and Cameron, 1987). 1989; Carrigan and Eichelberger, 1990).
The hypothesis that a common dike connects all A model for dike propagation and composite
three domes has been partially validated by the flow formation that includes the breaching of a
drill hole intersection of a shallow, rhyolitic dike sloping magma chamber roof and the encapsula-
between Obsidian Dome and Glass Creek Dome. tion of different viscosity magmas can explain the
At least three physically distinct magmas have gross zoning of these three domes by the eruption
contributed to the formation of the composite of magma from a common, chemically zoned
domes according to Sampson and Cameron chamber. Figure 23 illustrates how the progressive
(1987). Deadman Dome, the southern-most flow breaching of the sloping roof of a zoned chamber
in the chain, consists of a coarse-grained, crystal- could produce the observed compositional varia-
rich rhyolite (70- 74% SiO l ) that overlies a finer- tions during a sustained period of increased pres-
grained rhyolite of lower crystal content, which sure in the chamber. The simplest, but not neces-
also tends to have a slightly lower silica content sarily the only, chronology for the formation of
(71.5%). It is possible that the latter lava is a mix- the domes is used. The formation of a common
ture of the coarser-grained, crystal-rich rhyolite dike may have started by a breach in that portion
and a finer-grained rhyodacite that is lower in of the roof in contact with a layer of coarse,
crystal content. The middle flow in the chain, the crystal-rich, high silica rhyolite. Owing to its very
Glass Creek Dome, lies on the topographic rim of high viscosity, the magma does not allow a dike
the caldera. Like Deadman Dome, it consists of to reach the surface. At the same time, the dike
a coarse, crystal-rich rhyolite overlying a finer- also propagates horizontally, breaching the cham-
grained rhyolite with lower crystal content. Ob- ber roof along the line indicated in Fig. 23. This
sidian Dome lies beyond the rim of the caldera horizontal breaching continues until the lower
350 Charles R. Carrigan

composite:
rhyodacite/
xtal poor
rhyolite xtal poor
rhyolite
dike
.,,/

\ \
\ \

\ xt~1 rich
\ rhyolite
\ \
\

rhyodacite

Figure 23 Using an encapsulation model for two-component magma transport and the generalized withdrawal model of
Fig. 20. a new model is suggested for the contemporaneous formation of three 600 yr-old composite lava flows in the Inyo
volcanic chain. If a breach in the magma chamber roof is initiated over the crystal-rich rhyolite, the high viscosity would tend to
prevent this lava from reaching the surface. As the breach progresses down slope, it crosses into a zone of lower, viscosity,
crystal-poor rhyolite. This magma lubricates the rise of the higher viscosity, crystal-rich component and results in the composite
Deadman Dome lava flow (X's indicating the points over which lava flows have formed). The continued breaching of the roof
causes the composite dike to finally "bud" a conduit that reaches the surface to form the Glass Creek Dome, which has a similar
composition to Deadman Dome. The breach continues its descent across the crystal-poor rhyolite zone and produces a rhyolite
dike that does not reach the surface. A rhyodacitic layer is finally sampled by the dike as it crosses a compositional boundary
and provides lubrication for the rhyolite to erupt to the surface to form Obsidian Dome as the magma chamber pressure wanes.

silica and lower crystal content rhyolitic layer ciently high to permit the continued breaching
begins to be withdrawn in addition. Because of of the chamber roof along the line. Because of
its lower viscosity, this magma encapsulates the the dike's proximity to both the crystal-poor and
crystal-rich rhyolite, successfully lubricating its crystal-rich rhyolite layers as determined by the
rise to the surface to form the composite Deadman path of breaching, both are withdrawn simultane-
Dome lava flow by the simultaneous effusion of ously and encapsulation ensues to form a compos-
both crystal-rich and crystal-poor lavas. Pressures ite dike that once again reaches the surface to
in the magma chamber evidently remain suffi- form the composite Glass Creek Dome consisting
 14. Two-Component Magma Transport 351
of the same types of lavas as Deadman Dome. Other, more conventional, models are also
Continuation of the breach takes it across the zone evaluated and critiqued on the basis of the evi-
of crystal-poor rhyolite away from the zone of dence presented. It was found that sequential-flow
crystal-rich rhyolite so that the dike now consists models, in which one magma follows another up
of a uniformly crystal-poor rhyolite. This is con- a dike or conduit to be extruded one at a time, can-
sistent with the cores obtained from drilling the not explain several important features of the com-
dike near Obsidian Dome. Finally, the breach in posite flows discussed here. It is possible, how-
the roof approaches and begins to sample the con- ever, that the sequential scenario provides an
tents of the less viscous rhyodacitic zone. Encap- explanation for that less common class of dikes
sulation and self-lubrication of the crystal-poor that are reverse-zoned with the mafic magma in
rhyolite by the rhyodacite results in the budding the center and the silicic magma at the walls
of a conduit to form the Obsidian Dome flow. (Gautneb, 1988, Fig. 4). In such a case, silicic
magma withdrawn from the top layer of a magma
chamber could have been followed by a more
Summary and Conclusions mafic, underlying component.
Given that encapsulation of a high viscosity
Voluminous and violent eruptions, such as the one component by a lower viscosity component oc-
that resulted in the formation of the Long Valley curs, we have derived time-independent equations
caldera and that vented silicic magma followed by for the lubrication of a magma of high viscosity
more mafic material, usually empty the underly- by a second one of lower viscosity. The derivation
ing magma chamber from the top layer downward is general enough to permit consideration of both
so that the sequence of layering on the surface is Newtonian and power-law magmas. It is shown
just an inversion of the chamber's layering prior to that only relatively thin lubricating layers « 10%
the eruption (Hildreth, 1981). In such cases the of the dike width) are needed to substantially re-
driving pressure is evidently so great that the vis- duce the pressure gradient required to drive a
cosity of magma, within a very wide range, plays magma of very high viscosity through a dike. Fur-
an insignificant role in determining whether an thermore, the addition of thin lubricating layers to
eruption is initiated. Whatever magma is in con- a flow of rhyolite can increase fracture propaga-
tact with the roof at the moment that it is breached tion rates several orders of magnitude for a given
by a dike is the first magma to arrive at the value of the chamber pressure. In a basalt-
surface. rhyolite system, the core layer of rhyolite will flow
In contrast, effusive eruptions appear to be as a solid body with virtually all the shear concen-
characterized by much smaller magmatic driv- trated in the basalt lubricating layers. Viscous dis-
ing pressures, and the effective viscosity of a sipation will then be concentrated in the lubrica-
flow may be the pivotal influence in determining tion layers. As a result, the thin basaltic layers can
whether a magma can reach the surface. As long become significant heat producers during the as-
as the dike taps only the most viscous top layer cent of a composite magma through a dike and
through the roof of the chamber, the propagation offset wall heat losses (>2 kW . m:") for reason-
velocity of the dike simply may be too low for it able dike propagation rates (2=1 m . S-I). If the
to reach the surface before cooling causes plug- two components of a flow are characterized by
ging. If the new dike also happens to sample a different power-law rheologies, a range of shear
lower viscosity component across a compositional rate can be shown to exist, over which the pressure
interface, then encapsulation and lubrication of gradient for two-component flow is less for a
the more viscous component by the less viscous given flux of magma than the pressure gradient
component may reduce pressure losses enough to required to drive the same flux of each component
allow the flow to reach the surface. This chapter flowing by itself in the dike.
has considered the surface observations, labora- The important effects of different dike inlet
tory experiments, numerical models, and analyti- conditions (in the magma reservoir roof or walls)
cal development needed to evaluate the processes on the encapsulation process have been discussed.
pertinent to this scenario. In this context, a new and considerably more gen-
 352 Charles R. Carrigan

eral model for the simultaneous withdrawal of two References

magmas through a dike has been presented. It is a
characteristic of this model that a generalized ge- Bacon, C. R. (1986). Magmatic inclusions in silicic and inter-
ometry and not special-and ad hoc-appeals to mediate volcanic rocks. J. Geophys. Res. 91,6091-6112.
magma withdrawal rates or to physical properties Bird, R. B.• Stewart, W. E., and Lightfoot, E. N. (1960).
permits the simultaneous flow of two magmatic "Transport Phenomena," Wiley. New York.
Blake. D. H" Elwell, R. W. D., Gibson. I. L., Skelhorn, R. R.,
components in a dike. This inlet model in con-
and Walker, G. P. L. (1965). Some relationships resulting
junction with the encapsulation process provides from the intimate association of acid and basic magmas,
the basis for explaining the contemporaneous ori- Quart. J. Geo!. Soc. London 121,31-49.
gin of three of the Inyo volcanic domes in Long Blake, S.• and Campbell, I. H. (1986). The dynamics of
Valley, California. magma-mixing during flow in volcanic conduits, Contrib.
Mineral. Petrol. 94, 72-81.
The process of self-lubrication of two-
Blake, S., and Fink, 1. H. (1987). The dynamics of magma
component magmas needs to be considered as a withdrawal from a density stratified dyke, Earth Planet.
potentially important factor in any explanation for Sci. Lett. 85,516-524.
the origin of intrusions and lava flows. In some Blake, S.• and Ivey, G. N. (1986). Magma mixing and the dy-
cases it seems likely that we observe at the surface namics of withdrawal from stratified reservoirs. J. Vol-
canol. Geotherm. Res. 27, 153-178.
a certain high silica and high crystal content lava,
Bruce, P. M" and Huppert, H. E. (1989). Thermal control of
such as the coarse-grained, high crystal and silica basaltic fissure eruptions. Nature 342, 665-667.
content rhyolite of Deadman Dome, because of Bruce. P. M., and Huppert, H. E. (1990). Solidification and
that lava's association with a more fluid compo- melting along dykes by the laminar flow of basaltic
nent that provided lubrication. In fact, the statis- magma, in "Magma Transport and Storage" (M. P. Ryan,
tics for the occurrence of extreme compositional ed.), Wiley, Chichester/Sussex, England.
Carrigan, C. R. (1983). A heat-pipe model for vertical,
contrasts in lava flows, e.g., rhyolites and basalts, magma-filled conduits, J. Volcanol. Geotherm. Res. 16,
may not remotely reflect the degree of coexistence 279-298.
of end-member magmas at depth because of a bias Carrigan. C. R., and Eichelberger, 1. C. (1990). Zoning of
of reservoir-to-surface transport toward such two- magmas by viscosity in volcanic conduits, Nature 343,
248-251.
component associations. In this regard, the ob-
Carrigan, C. R., Schubert, G., and Eichelberger, 1. C. (1992).
served products of a large eruption may be more Thermal and dynamical regimes of single- and two-phase
indicative of the original content of the source- magmatic flow in dikes, 1. Geophys. Res. 97, 17377-
magma chamber than are the composite lavas 17392.
from a somewhat later effusion driven by a much Charles, M. E.. and Redberger, P. 1. (1962). The reduction of
pressure gradients in oil pipelines by the addition of water:
reduced magmatic pressure.
Numerical analysis of stratified flow, Can. 1. Chern. Eng.
(Author's Note: A utility written in Fortran for 40,70-75. .
either Macintosh or MS-DOS operating systems Delaney P. T., and Pollard, D. D. (1981). "Deformation of
is available from the author for evaluating most of Host Rock and Flow of Mafic Magma during Growth of
the mathematical expressions and models devel- Minette Dikes and Breccia-Bearing Intrusions near Ship
Rock, New Mexico," U.S. Geological Surv. Prof. Paper
oped in this chapter.)
1202.
Eichelberger,1. C. (1975). Origin of andesite and dacite: Evi-
dence of mixing at Glass Mountain. California. and at
Acknowledgments other circum-Pacific volcanoes, Geol. Soc. Am. Bull. 86,
1381-1391.
This chapter is dedicated to the memory of my father, Charles Eichelberger. 1. C. (1980). Vesiculation of mafic magma dur-
F. Carrigan. I thank Charles Bacon, John Eichelberger. Wes ing replenishment of silicic magma reservoirs. Nature 288,
Hildreth. Richard Knapp, Bob Nilson. Michael Ryan, Rick 446-450.
Ryerson. Tom Vogel. Lee Younker, and George Zandt for both Eichelberger, 1. C.• Lysne, P. C.. Miller. C. D., and Younker,
enlightening comments and stimulating discussions that have L. W. (1985). Research drilling at Inyo Domes, California:
improved this manuscript. Many of the figures were redrawn 1984 results, EOS Trans. AGU 66, 186-187.
by Priscilla Proctor of the Technical Information Department. Eichelberger, 1. C., Vogel, T. A., Younker, L. W.. Miller, C. D.,
This work was performed under the auspices of the U.S. De- Heiken, G. H., and Wohletz, K. H. (1988). Structure and
partment of Energy by the Lawrence Livermore National stratigraphy beneath a young phreatic vent: South Inyo
Laboratory under Contract W-7405-ENG-48. Support was Crater, Long Valley Caldera, California, 1. Geophys. Res.
provided through a Lawrence Livermore LDRD grant and the 93, 13208--13220.
Institute of Geophysics and Planetary Physics. Everage, A.E. (1973). Theory of stratified bicomponent flow
 14. Two-Component Magma Transport 353
of polymer melts. I. Equilibrium Newtonian tube flow, Koyaguchi, T. (1985). Magma mixing in a conduit, J. Vol-
Trans. Soc. Rheol. 17,629-646. canol. Geotherm. Res. 25, 365-369.
Frisch, V., Hasslacher, B., and Pomeau, Y. (1986). Lattice-gas Lee, B.-L., and White, 1. L. (1975). Experimental studies of
automata for the Navier-Stokes equation, Phys. Rev. Lett. disperse two-phase flow of molten polymers through dies,
56, 1505-1508. Trans. Soc. Rheol. 19,481-492.
Furman, T., and Spera, F. 1. (1985). Co-mingling of acid and Lister, 1. R., and Kerr, R. C. (1991). Fluid-mechanical models
basic magma with implications for the origin of mafic 1- of crack propagation and their application to magma trans-
type xenoliths: Field and petrochemical relations of an port in dykes. J. Geophys. Res. 96, 10049-10077.
unusual dike complex at Eagle Lake, Sequoia National Macl.ean, D. L. (1973). A theoretical analysis of bicomponent
Park, California, U.S.A., J. Volcanol. Geotherm. Res. 24, flow and the problem of interface shape, Trans. Soc. Rheol.
151-178. 17,385-399.
Gautneb, H. (1988). Structure, age and formation of dykes on Miller, C. D. (1985). Holocene eruptions at the Inyo volcanic
the island of Smola, Central Norway, Norsk Geol. Tids- chain, California-Implications for possible eruptions in
skrift 68,275-288. the Long Valley Caldera. Geology 13, 14-17.
Geertsma, 1., and Haafkens, R. (1979). A comparison of theo- Minagawa, N., and White, 1. L. (1975). Co-extrusion of un-
ries for predicting width and extent of vertical hydrauli- filled and Ti0 2-filled polyethylene: Influence of viscosity
cally induced fractures, Trans. ASME J. Energy Res. Tech- and die cross-section on interface shape, Polym. Eng. Sci.
nol., 101, 8-19. 15, 825 - 830.
Gibson, I. L., and Walker, G. P. L. (1963). Some composite Nilson, R. H., and Griffiths, S. K (1986). "Self-Similar
rhyolite/basalt lavas and related composite dykes in east- Analysis of Planar Hydraulic Fractures Driven by a Con-
ern Iceland, Proc. Geol. Ass. 74, 301-318. stant Flow Rate," Sandia National Laboratories Report,
Gudmundsson, A. (1990). Mafic dykes and emplacement SAND86-8694.
mechanisms, in "Proceedings, 2nd Intern. Dyke Conf., Oliemans, R. V., and Ooms, G. (1986). Core-annular flow
Adelaide, 1990," pp. 47-62. of oil and water through a pipeline, in "Multiphase Sci-
Han, C. D. (1973). A study of bicomponent coextrusion of ence and Technology 2" (G. F. Hewitt, 1. M. Delhaye,
molten polymers, J. Appl. Polym. Sci. 17, 1289. and N. Zuber, eds.), pp. 427-472, Hemisphere, Wash-
Han, C. D. (1975). A study of coextrusion in a circular die, ington, DC.
1. Appl. Polym. Sci. 19, 1875-1883. Rothman, D. H., and Keller, 1. M. (1988). Immiscible cellular-
Han, C. D. (1976). "Rheology in Polymer Processing," Aca- automaton fluids, 1. Stat. Phys. 52,1119-1127.
demic Press, New York. Ryan, M. P. (1987) Neutral buoyancy and the mechanical evo-
Han, C. D. (1981). "Multiphase Flow in Polymer Processing," lution of magmatic systems, in "Magmatic Processes:
Academic Press, New York. Physicochemical Principles" (B. O. Mysen, ed.), pp. 259-
Han, C. D., and Chin, H. B. (1979). Theoretical prediction 287, The Geochemical Society, University Park, PA.
of pressure gradients in coextrusions of non-Newtonian Ryan, M. P. (1990). The physical nature of the Icelandic
fluids, Polym. Eng. Sci. 19, 1156. magma transport system, in "Magma Transport and Stor-
Hickox, C. E. (1971). Instability due to viscosity and density age" (M. P. Ryan, ed.), pp. 175-224, Wiley, Chichester!
stratification in axisymmetric pipe flow, Phys. Fluids 14, Sussex, England.
251-262. Ryan, M. P., and Blevins, 1. Y. K. (1987). "The Viscosity of
Hildreth, W. (1981). Gradients in silicic magma chambers: Im- Synthetic and Natural Silicate Melts and Glasses at High
plications for lithospheric magmatism, J. Geophys. Res. Temperatures and I Bar (10' Pascals) Pressure and at
86,10133-10192. Higher Pressures," V.S. Geo!. Surv. Bull. 1764.
Hu, H., and Joseph, D. D. (1989). Lubricated pipelining: Sta- Ryerson, F. 1., Weed, H. c. and Piwinskii, A. 1. (1988). Rhe-
bility of core-annular flow, Part 2, J. Fluid Mech. 205, ology of subliquidus magmas. 1. Picritic' compositions,
359-396. 1. Geophys. Res. 93, 3421-3436.
Huppert, H. E., and Sparks, R. S. 1. (1980). The fluid dynamics Sampson, D. E., and Cameron, K. L. (1987). The geochemis-
of a basaltic magma chamber replenished by influx of hot try of the Inyo volcanic chain: MUltiple magma systems in
dense ultrabasic magma, Contrib. Mineral. Petrol. 75, the Long Valley region, eastern California, J. Geophys.
279-289. Res. 92, 10403-10421.
Huppert, H. E., Sparks, R. S. 1., and Turner, 1. S. (19R2). Ef- Shaw, H. R., Peck, D. L., Wright, T. L., and Okamura, R.
fects of volatiles on mixing in calc-alkaline magma sys- (1968). The viscosity of basaltic magma: An analysis of
tems, Nature 297, 554-557. field measurements in Makaopuhi Lava Lake, Hawaii, Am.
Joseph, D. D., Nguyen, K, and Beavers, G. S. (I984a). Non- 1. Sci. 266, 225-264.
uniqueness and stability of the configuration of flow of im- Shaw, H. (1969). Rheology of basalt in the melting range,
miscible fluids with different viscosities, J. Fluid Mech. J. Petrol. 10,510-535.
141,319-345. Skelhorn, R. R., MacDougall, 1. D. S., and Longland, P. 1. N.
Joseph, D. D., Renardy, M., and Renardy, Y. (I 984b). In- (1969). "The Tertiary Igneous Geology of the Isle of
stability of the flow of two immiscible liquids with differ- Mull," Geo!. Ass. Guides, No. 20, Geologists' Associa-
ent viscosities, 1. Fluid Mech. 141,309-317. tion, London, Benham and Co., Ltd.
Karagiannis, A., Mavridis, H., Hrymak, A. N., and Vlacho- Smith, R. L. (1979). "Ash Flow Magmatism." Geo!. Soc.
poulos, 1. (1988). Interface determination in bicomponent Amer. Special Paper 180, pp. 5-27.
extrusion, Polym. Eng. Sci. 28,982-988. Southern,1. H., and Ballman, R. L. (1973). Stratified bicom-
354 Charles R. Carrigan

ponent flow of polymer melts in a tube, Appl. Polym. Sci. H. R. (1989). Petrology and emplacement dynamics of in-
20, 175-189. trusive and extrusive rhyolites of Obsidian Dome, Inyo
Spence, D. A., and Turcotte, D. L. (1985). Magma-driven craters volcanic chain, eastern California. J. Geophys. Res.
propagation of cracks, J. Geophys. Res. 90, 575-580. 94, 17937-17956.
Spera, F. 1. (1984). Some numerical experiments on the with- Walker. G. P. L.. and Skelhorn, R. R. (1966). Some associ-
drawal of magma from crustal reservoirs. 1. Geophys. Res. ations of acid and basic igneous rocks, Earth Sci. Rev. 2,
89,8222-8236. 93-109.
Spera, F. 1.. Yuen, D. A.. Greer. 1. c., and Sewell, G. (1986). White, 1. L.. and Lee, B.-L. (l975a). An experimental study of
Dynamics of magma withdrawal from stratified magma sandwich injection molding of two polymer melts using
chambers, Geology 14, 723-726. simultaneous injection, Polym. Eng. Sci. 15,481-485.
Spera, F.1.. Borgia. A.• Strimple, 1.. and Feigenson, M. (1988). White, 1. L., and Lee, B.-L. (l975b). Theory of interface dis-
Rheology of melts and magmatic suspensions. I. Design tortion in stratified two-phase flow, Trans. Soc. Rheol. 19,
and calibration of concentric cylinder viscometer with 457-479.
application to rhyolitic magma. 1. Geophys. Res. 93, Williams. M. C. (1975). Migration of two liquid phases in cap-
10273-10294. illary extrusion: An energy interpretation, AIChE 1. 21,
Stockman. H. w.. Stockman, C. T.• and Carrigan, C. R. (1990). 1204-1207.
Modelling viscous segregation in immiscible fluids using Wohletz, K. H., and Valentine. G. A. (1990). Computer simu-
lattice-gas automata. Nature 348, 523-525. lations of explosive volcanic eruptions, in "Magma Trans-
Trial. A. F.. Spera, F. 1.. Greer, J., and Yuen, D. A. (1992). port and Storage" (M. P. Ryan. ed.), pp. 113-136, Wiley,
Simulations of magma withdrawal from compositionally Chichester/Sussex. England.
zoned bodies. 1. Geophys. Res. 97,6713-6733. Yoder. H. S. (1973). Contemporaneous basaltic and rhyolitic
Vogel, T. A. (1982). Magma mixing in the acidic-basic com- magmas. Am. Mineral. 58,153-171.
plex of Ardnamurchan: Implications on the evolution of Younker. L. w.. Eichelberger. 1. C., Kasameyer, P. w.. New-
shallow magma chambers, Contrib. Mineral Petrol. 79, mark. R. L.. and Vogel. T. A. (1987). Results from shallow
411-423. research drilling at Inyo Domes, Long Valley caldera.
Vogel, T. A., and Wilband, 1. T. (1978). Coexisting acidic and California, and the Salton Sea geothermal field, Salton
basic melts: Geochemistry of a composite dike, J. Geo!. Trough, California. in "Proceedings 3rd Intern. Symp. on
86,353-371. Observation of the Continental Crust through Drilling.
Vogel, T. A., Eichelberger. 1. C, Younker. L. w.. Schuraytz, Mora and Orsa, Sweden, 1987." vol. 2, pp. 172-187.
B. c., Horkowitz, J. P., Stockman, H. W., and Westrich,
 Chapter 15 Fluid and Thermal Dissolution Instabilities
in Magmatic Systems

J. A. Whitehead and Peter Kelemen

Overview Units

A'(y,t) small area variation from pres- cm-, m'

Magma source regions at great depths in the mantle are
sure change
now thought to have melt dynamics much more poten-
D solid-state diffusion coefficient ern- . s - 1
tially complex than the common perception of laminar
grain boundary flows with a spatially smooth melt mi- E Coefficient of elasticity times dyn- crn r '

gration front. As the melt rises by porous flow along the geometric terms
edges of grains in partially melted material, it can de- F scaled force accumulated in elas- dimensionless
velop a hydrodynamic instability whereby fingers of tic region
faster moving melt will dissolve parts of the porous net- L length of slot em, m
work and lead to yet higher velocity flows. Criteria for thickness of gravity current m
this instability are discussed. Laboratory experiments (g'q'td//I)I/S
with water flowing through salt crystals demonstrate scaled pressure (pd 4/96/1HPKU) dimensionless
this instability. The results follow earlier studies of ther- Peelet number iwdhc) dimensionless
mal erosion that have been demonstrated in a laboratory
diffusive Peelet number (lld'/ dimensionless
experiment with syrup flowing through a chilled tube
DL)
(Whitehead and Helfrich, 1991). Flow is either steady
Q uniform volumetric flux per unit em- . S-I
or periodic depending on the temperature of the liquid
length
and the flow rate into the reservoir. An analytic theory
Q' scaled volume flux (QIW,d) dimensionless
indicates that the transition from steady to periodic
flows depends on the nonlinearities in the steady-state T temperature of the fluid in slot °C
relationship between pressure and flow rate. A general temperature at which viscosity °C
stability criterion that states that the Peelet number is VH
must be within a certain range for an instability to de- W scaled velocity (= wl W,) dimensionless
velop is then advanced. Parameters governing the oscil- velocity perturbation amplitude dimensionless
lation period are determined. The thermal theory has velocity scale (8KLld') cm . s 1
also been extended to flow through a conduit, and finger
d thickness of slot ern, m
development is predicted. The instability is very similar
crystal size em
to the corrosive instability of salt + water permeation
experiments. Qualitative laboratory experiments with (gravity) X (normalized density em . s-,
paraffin that spreads radially over a cold plate also re- difference) reduced gravity

veal the fingering and demonstrate features that mimic feW) scaled flow resistance in slot dimensionless
flow regimes in advancing lava flows. The instability is cross-section length scale of em, m
very similar to the corrosive instability of water flowing magma chamber
through salt. m measure of mass dimensionless
p pressure dyn . em -',
Notation MPa
Units volume flux per unit length (par- m' .
q S-I

A scale of viscosity variation dimensionless affin experiment)

constant magma chamber cross- crn-, rn- time

section area i' scaled time (I2/1Htld') dimensionless

Magmatic Systems Copyright © 1994 by Academic Press, Inc.

Edited by M. P. Ryan 355 All rights of reproduction in any form reserved.
 356 J. A. Whitehead and Peter Kelemen
Units tance to flow can either decrease due to the failure
of the surrounding rock, the thermal or chemical
(' scaled time for earthquake dimensionless
problem corrosion of the surrounding rock, or the forma-
fd duration time of paraffin
tion of gases (which have vastly lower viscosity),
experiment or instead increase due to the action of numerous
u' scaled earthquake sliding dimensionless processes that retard the flow upon cooling of the
velocity magma or upon ascent from great depth. Some
v velocity along the axis of magma em- S-I processes that increase the resistance of magma to
chamber flow are constrictions due to crystallization along
W vertical velocity leaving magma ern S-I conduit walls, or an increase in the fluid viscosity
chamber due to cooling. The increase in viscosity through
Wu steady scaled velocity dimensionless cooling (Hughes, 1982) can result from the in-
11" spatial and temporal perturbation dimensionless herent properties of the material's viscosity-tem-
toW perature relationship (Ryan and Blevins, 1987),
x direction across the slot cm bulk composition changes through crystal frac-
y direction along chamber axis cm tionation, or the addition of suspended crystals to
y' scaled direction along chamber dimensionless the fluid upon cooling, with a consequent dra-
axis y/L matic increase in bulk viscosity due to two-phase
z direction up the slot cm,m effects.
IH temperature difference between °C Changes in the aggregate flow resistance
top and bottom of slot through cooling lead to changes in the dynamic
r measure of change of resistance dimensionless pressure, which leads, in turn, to a number of in-
(4yiJj/ iJwo) teresting effects. In this chapter, we describe re-
v
.... paraffin surface roughness m- I cent studies on the dynamics of flows that develop
(o:JHhL) decreased resistances to flow due to chemical dis-
'I' ratio of cooling time to advection dimensionless solution, and we show that the results are similar
time scale t, to flows that develop an increased resistance as
a viscosity-temperature coefficient ern"- S--I .OC-l
they flow into cooler regions. In particular, we de-
f3 measure of friction coefficient dimensionless scribe the mechanics of a number of intriguing
8 measure of aspect ratio dimensionless fluid dynamic instabilities that may develop. We
(l2A,/;/d 3L) have concentrated our effort developing an ana-
y measure of pressure expansion dimensionless lytic theory of fluid flow instability in magmatic
[(l2lJ H/d')' (pLi Ed)l conduits and have combined this with revealing
a, standard deviation of thickness m laboratory experiments that illustrate some pos-
of paraffin
sible thermal, velocity, and pressure instabilities.
K thermal diffusivity cm 2 . S-I
In general, these fluid instabilities are charac-
p constant density of fluid g. cm " ' terized by the development of fingers of melt and/
lJ(T) kinematic fluid viscosity cm'· s 1
or time-dependent effects. Volcanic features such
lJ H kinematic fluid viscosity in cm 2 • s·\ as time-dependent surges, for example, compli-
magma chamber cated free-surface shapes such as pahoehoe tex-
lJp kinematic viscosity of paraffin em"- S-l ture development, and lava tubes and restricted
lava fountains are thought to be the result of this
change in flow resistance with temperature. The
Introduction intriguing possibility that these processes happen
at great depth also exists, since an increase of flow
Volcanism occurs where hot silicate liquids resistance with cooling is one of the most preva-
(magma) from deep in the Earth flow to the sur- lent processes in magmatic systems. Our approach
face, where they cool and solidify. Before solidi- has been to study simple problems that capture the
fication is complete, the flow resistance in both basic features that develop and then to suggest
deep and shallow conduits can change. This resis- possible applications. Duplication of the full geo-
 15. Fluid and Thermal Dissolution Instabilities 357
physical and geological complexities for a given that a Rayleigh-Taylor gravitational instability
application is not our present objective. Instead, might produce the roughly periodic spacing of
the aim is to capture and illustrate-in experiment island-arc volcanos. Whitehead et at. (1984) and
and in theory-the essential physical process. Whitehead and Helfrich (1990) followed this lead
Numerous geological systems-beyond those to suggest that Rayleigh-Taylor diapiric struc-
mentioned previously-also involve a fluid that tures may form under mid-ocean ridge spreading
develops a change in resistance upon cooling. centers. An analysis similar to the Saffman-
Many aquifers, for example, dissolve minerals Taylor (1958) instability but with more direct geo-
under high pore fluid pressures and elevated chemical applications was conducted by Chad-
temperatures. Such minerals in solution may be ham et at. (1986) and Ortoleva et at. (1987b). In
redeposited along fractures and veins in cooler their study an advancing front reacted with host
portions of the system. In both terrestrial and material to produce a scallop-shaped interface.
deep-sea hydrothermal systems, numerous in- They suggested that numerous geological fea-
stances occur in which the systems fluctuate in tures may be generated by this geochemical self-
their flow behavior, become restricted to a few lo- organization (Ortoleva et al., 1987a). Although
calized springs, and may ultimately have fluid they showed, in general, that such wavy interfacial
permeabilities that diminish due to the deposited disturbances may grow with time, no clear sta-
minerals. bility criteria that included a prediction of funda-
A well-known instability in fluid dynamics mental length scales or fundamental time scales,
that involves a low viscosity fluid flowing into for the predicted changes were developed. A sec-
regions containing a more viscous resident fluid ond physical process for the thermal control ofba-
is called the "Saffman-Taylor" (Saffman and saltic fissure eruptions was developed by Bruce
Taylor, 1958) instability. A fluid intrudes, for ex- and Huppert (1989, 1990). Melt flowing through
ample, into a porous region or a Hele-Shaw cell a dike is shown to be fundamentally unsteady and
composed of two plane walls separated by a small the flowing magma will either gradually solidify
gap that contains a second more viscous fluid. Un- on the conduit walls until it completely blocks the
der suitable conditions, the interface between the flow or melt back the walls of the conduit, eroding
two fluids will develop finger-like protrusions that the adjacent country rock. It was suggested that
contain the lower viscosity fluid and periodically these results show that such a process leads to
interpenetrate into the viscous fluid. The lower flow localization, as earlier suggested by Delaney
viscosity fluid possesses less hydraulic resistance and Pollard (1982). A recent third model by Ste-
to the large-scale pressure field and moves rapidly venson (1989) showed that a partial melt under-
into the finger. This forces the finger tip to ad- going deformation is unstable to the small-scale
vance yet further into the viscous fluid. The ex- redistribution of melt, so that melt migrates par-
amples of Saffman-Taylor instabilities that have allel to the minimum compressive stress and ac-
been studied to date, whether with mathematical cumulates in veins, where the bulk viscosity is
analysis or with laboratory experiments, are in- lowest and becomes even lower as more melt ac-
herently time dependent, and the tips continue to cumulates. Although only a very small length
move indefinitely. After a long time the region is scale was indicated by the instability analysis, the
filled with veins of the low viscosity fluid; each important mechanism in the accumulation of melt
vein is surrounded by islands of viscous fluid in slip planes has long been hypothesized by
that is slowly moving away from the source. The geologists.
final state is never truly steady or even periodic Here, a thermal approach that involves a gen-
with time. eral decrease in resistance with higher velocity
Although numerous thoughts about structure flow like that of Bruce and Huppert (1989, 1990)
generation in flow below volcanic systems have is combined with instability considerations to pro-
been expressed, actual attempts to analyze the me- duce a thermal and chemical equivalent to either
chanical behavior that lead to structure generation the Saffman-Taylor instability or the geochemical
or time-dependent flow are rare. One of the first self-organization cases of Chadham et at. (1986)
such studies was by Marsh (1979), who suggested or Ortoleva et at. (1987a, 1987b). It is completely
358 J. A. Whitehead and Peter Kelemen

analogous to a melt flowing through a matrix in melted paraffin. The rest of the paraffin stops ad-
which the matrix resistance to flow is altered by vancing and ultimately solidifies. At later times,
the rate of flow of the fluid. For simplicity, instead all but one of the tubes slow down and stop, and
of two fluids differing materially, we have one the melted paraffin flows in one final tube for as
fluid with a temperature- or chemically dependent long as the experiment continues. The paraffin ev-
viscosity. The fluid flows from the source as an erywhere else gradually solidifies. This experi-
initially hot fluid and is then cooled through ther- ment is simpler than experiments by Fink and
mal conduction to the the cold sidewalls of a Griffiths (1990) where paraffin was fed from a
Hele-Shaw cell or, by analogy, altered by chemi- central tube over a flat plate in water as a model of
cal interaction with the porous medium. Unlike viscous gravity currents with a solidifying crust.
previously studied problems, the final state may In this experiment the free surface of the paraffin
become truly steady (although periodic or chaotic varied in depth and the structure became consid-
states are also possibilities). erably more complicated. No axisymmetric flow
In this chapter, we lay the foundations for the was ever found. As in the slot experiments, the
chemical modulation of a flow problem by first structure of the solidified flow was more compli-
reviewing a laboratory experiment conducted by cated for Peelet numbers of order 1. Recent ex-
Whitehead and Helfrich (1991) illustrating one periments with paraffin flows into water with a
possible situation: hot fluid with a temperature- Cartesian geometry (Ito and Whitehead, in prog-
dependent viscosity flowing through a cold slot ress) find that sheet flows dominate when lava so-
from an elastic reservoir. If along-slot perturba- lidifies late in the flow sequence whereas pillow
tions are neglected, the flow is found to be either morphologies occur when solidification begins
steady or periodic, depending on the degree of earlier during the eruption.
viscosity increase produced in the relatively cool
region and the rate of reservoir replenishment. If
along-slot perturbations are ineluded, it is found Thermal Erosion
that small wavelength instabilities are the most
unstable and that instability is encountered when Transition of One-Dimensional Flows to
the Peelet number (ud/K) is of order Lid, where u Oscillations: An Experimental Study
is the velocity scale, K is the thermal diffusivity, L
is the length of the conduit and d is the width of Figure I shows the schematic of the experimen-
the conduit. This appears to be the first quantita- tal apparatus consisting of a vertical glass tube
tive criterion found for the onset of this elass of 3.8 em i.d. and 1 m long, located below a reservoir
instabilities. The flow in the limit of infinite wave- containing Karo brand corn syrup. A variable con-
length (zero wavenumber) reduces to the labora- trol valve leading from the reservoir allowed the
tory experiment described in the section Transi- syrup to flow into the glass tube at a controlled
tion of One-Dimensional Flows to Oscillations: rate. Projecting out from the bottom of the glass
An Experimental Study. tube was a stopper with a hole and a O.383-cm i.d.
Another laboratory experiment is then de- copper tube. Varying tube lengths were used. The
scribed in the section Stability of Thermally copper tube projected downward and was shaped
Eroded Flows, which illustrates the spatial in- like the letter "J." The lowest part of the J-tube
stability. To perform this experiment, hot paraffin was immersed in a refrigerated thermostatic bath
was made to spread radially through a slot from a with a plastic flexible tube extending from the
point source over a cold plate. When the Peelet copper tube to a point outside the bath over a
number is of the order of the conduit length di- beaker placed to catch outflow. In a typical ex-
vided by its width, the radially symmetric flow ex- periment, syrup flows from the reservoir into the
periences a transition to a fingering flow. At first a tube. Syrup in the glass tube builds up to a height
number of fingers are visible at the outer front of h that can be easily measured and flows out
the expanding circular pool of paraffin. In the in- through the bottom copper tube. The frictional re-
termediate stage, these fingers advance substan- sistance to flow takes place principally within the
tially. Each finger is fed by a tube of flowing copper tube because it is much smaller in internal
 15. Fluid and Thermal Dissolution Instabilities 359

Constant Flow Tank

Plexiglass

..
~
<1-- - h(t)

Ruler
T1m
T room EIillJ Corn Syrup

Rubber Stoppe

Clock
1 IllII Isothermal Bath

overflow
Antifreeze bath (Tcold )

Figure I Schematic of the apparatus for generating flow instability modulations from fluids with temperature-dependent vis-
cosity. Corn syrup is fed into a vertical tube at a constant rate. As the syrup accumulates, it increases pressure across a small
outlet "J" tube that is in contact with a cold bath. When the syrup flows rapidly, it remains hot with low viscosity, but when the
flow rate slows, it becomes relatively cool and viscous.

diameter than the glass tube. As the syrup flows Run-down experiments were first conducted to
out, it is cooled to some extent by thermal conduc- estimate the resistance as a function of flow rate.
tion through the copper tube in contact with the Theoretical considerations indicate that a time de-
liquid bath maintained below room temperature. pendence is not to be expected unless the fluid re-
The apparatus is a simple (and highly ideal- sistance is inversely proportional to the flow rate.
ized) upside-down model of portions of a mag- Figure 2 shows data from two runs. Experiments
matic system subjected to either thermal gradients in Figs. 2a, 2c, and 2e had a bath temperature set
and/or chemical corrosion. The glass tube repre- to aoe, a room temperature of 24.1"C, and a cop-
sents either a compressible magma chamber or a per tube 3a cm long. Experiments in Figs. 2b, 2d,
matrix that can hold some liquid. The height of and 2f had a bath temperature of - l l.O'C, a room
the free surface in the glass tube conceptually rep- temperature of 24.aoe, and a copper tube 14.5 ern
resents either pressure in the chamber or percent- long. Figures 2a and 2b show height versus time
age melt in the matrix. The copper tube in the re- from which velocity was found as a time deriva-
frigerated bath represents the tube through which tive to produce Figs. 2c and 2d. For Figs. 2a, 2c,
magma flows to the surface of the earth or the ma- and 2e the run-down is close to exponential,
trix through which melt flows upward. The cool which would be expected for constant viscosity
temperature of the bath represents either relatively conditions. In contrast, the run shown in Figs. 2b,
cool country rock or the fluid flow resistance of 2d, and 2f with the bath at - l l.O'C is similar to
the matrix in the presence of corrosion by the the first run only for the first thousand seconds or
magma. If the experimental temperature bath is so; then there is a transition to a much slower run-
sufficiently cool and if the syrup has a sufficiently down. Presumably at this latter stage the viscos-
great viscosity increase upon cooling, an unsteady ity of the syrup is very large due to cooling in
flow develops even if the flow from the reservoir the copper tube. The difference between the two
is steady. Some estimate of how much viscosity states is particularly clear in the height versus ve-
change was necessary was obtained from theoreti- locity plot and even more so in the log-log plot.
cal considerations. For the run on the right, the transition region from
 a b
120 - - 120
1
f!,o lr:o'(1 IT =-11'( I
0
o
100.,. 100
0
1
• 0

80 t-
80 0

h. (em) h. (em) "e

0
60 e
60 0
0
0

40

20
0 10

time (x 10 s) 2
20 3D
40

20
0
'---- 10

time (x 1 02 S )
20 3D

C
120
I IT=O'( I
d

'20 I IT =-11'( I

,001
0
0
"
0"
00
0
"'1 100
I
0

00 0

::j
#0
80t "
0
1, (em) ,##
h. (em) 0
c
m/ 60
0

/
0

40
@, %d"
I
I

::1
I

0.00
I
0.02 0.04
I
0.06
I I
0.08
20
0.0 0.1 0.2
J
0.3

v (em. S-1) v (em -s:'}

e
2.2
2.21
I

/l
IT = 0'( I IT =-11'(
2.1
0
2.0 0
0
c-
~
0
0
0

::1
•. 8

/
0

log h. log h. #t
1.8

1.7+
., 1.6
I~~
o ~
om oci'
0'"'

~
r!Jl
0
, .6
0 'A! 00

I
1.5 1.2 I I
-06 -0.4 -0.2 0.0 0.2 0.4 -3 -2 -1 0

log v log v
Figure 2 Results from two run-down experiments with different bath temperatures. In (a), (c), and (e), the bath is at O'C, and
in figures (b), (d), and (f), the bath is at -II.O°C. Height versus time is plotted in (a) and (b), height versus velocity is plotted
in (c) and (d), and log height versus log velocity is plotted for (e) and (0. In (a), (c), and (e), the run-down is very similar to
exponential decay. In (b), (d), and (f), the run-down is characterized first by a rapid decay when hot syrup fills the tube and then
by slower decay when cold syrup fills the tube. Transition from rapid to slow is characterized by a plateau in v, h space as shown
in (f).
 15. Fluid and Thermal Dissolution Instabilities 361
a
120 . - - - - - - - - - - - - - - - - - - - - . 1 0
. ..

-
100
",. ....._.,,/ .. ~.

o -10

h (em) 80 e

o
o
(:/1':
'"
[!] ~ [!]

60
1 e'"'"
~

'"'"
'"'"
40
0 10 20 30 40

time (X I 03S)

b C

120 2.1

2.0 '"
100 '"
e 1.9
'"
h (cm) 80 i.. ....... log h
'. '"
J ,.
.: c
1.8
"J
.
r·"
,. ""
60
1.7

40 1.6
0.00 0.01 0.02 0.03 0.04 0.05 -4 -3 -2 -1

v (cm rs ") log v

Figure 3 An experiment that first had a run-down and then had a steady source turned on at the time shown by the arrow in
(a). Plotted are (a) the height (squares) and temperature (dots) versus time. (b) the height versus velocity, and (c) log height
versus log velocity. In (b) and (c) run-down is shown by squares and steady source by dots.

fast to slow flow is characterized by a plateau in the top of the tube. The sequence illustrated in
the height-velocity logarithmic curve. Theoreti- Figs. 2b, 2d, and 2f with a rapid run-down fol-
cal considerations will illustrate the significance lowed by a slow run-down is clearly visible. After
of the plateau in the section Stability of Thermally this, the volume flux from the reservoir was
Eroded Flows. started at the time denoted by the arrow in Fig. 3a.
When the volume flux of the source was set Three complete and nearly identical oscillations
within a certain range of values, the fluid height os- were seen thereafter. The plot of h and velocity
cillated with time. Figure 3 shows fluid height ver- was made from the data in Fig. 3a by subtracting
sus time, fluid height versus velocity, and log fluid the constant fluid velocity from the source as mea-
height versus log time for one example. This run sured by the abrupt change in slope immediately
was started as a run-down with the glass tube before and after the arrow. The oscillations pro-
filled with syrup but with no inflow of syrup to duce a closed curve in fluid height-velocity space
 362 J. A. Whitehead and Peter Kelemen

that lies on the top of the plateau in the run-down.

The plots of the logarithms (Fig. 3c) more clearly
show the limit cycle oscillation on the plateau.
The experiments exhibited a transition from
steady flow to a more complicated flow pattern
when the velocity ranged between 0.003 and Conduit
0.025 em . S-I. Peclet numbers (ur!«, see next /
section), based on the above velocities, a tube in-
ner radius r of 0.2 em, and a thermal diffusivity
of K=I.4X 10- 3 ern- . s " range from 0.4 to
3.5. The parameter ur 2 /KL ranged from 0.005
to 0.048.

Magma Reservoir
Stability of Thermally Eroded Flows

General Equations
<>Q

A theory is developed for a magma chamber Figure 4 Sketch of the idealized system. A constant magma
flux Q enters at the bottom of the elastic-walled magma cham-
that is two-dimensional and in Cartesian coordi-
ber. Magma can flow up the slot with a local velocity wand
nates rather than one-dimensional and in cylindri- cool through the side walls. It can also flow along the chamber
cal coordinates as in the preceding experiment. In axis.
this way, spatial as well as temporal instabilities
are considered. The term magma chamber is used
in a very general sense. It really means the stor- of elasticity times some geometric terms that de-
age area within which and from which magma pend on the shape of the magma chamber.
can easily flow and which is cooled slowly if at The expression of conservation of volumetric
all. Pressure in the chamber changes as the vol- flux is then
ume of the chamber becomes larger or smaller.
The fluid leaves the chamber through a region ap av
where the flow resistance increases by sidewall £1 at =Q - dw(y,t) - Ac ay" (I)
cooling on the conduit. The fluid ultimately leaves
that cooling-resistance area and passes into the Velocity w is the velocity of the magma leav-
outside world, which is at constant ambient envi- ing the magma chamber in the vertical (+ z)direc-
ronmental pressure. tion and v is the (y component) velocity in the
A very simplified system with the minimal es- magma chamber along the axis of the magma
sential features is sketched in Fig. 4. A narrow slot chamber.
with, for example, hydrothermally cooled walls, is In the laboratory experiment described in the
fed from below by fluid from the magma chamber. previous section, there is no y direction, but oth-
The conduit width in the x direction is d and the erwise the preceding equation describes the ex-
depth of the slot in the vertical z direction is L. periment to a good first approximation. For such
The y direction is along the axis of the magma a description, Q in the equations represents the
chamber. The bottom of the magma chamber is volume flux, d represents the area of the exit tube
fed by a uniform volumetric flux per unit length "J," the term E 'represents the glass tube area,
Q. The magma chamber itself has a cross- and p represents fluid height in the glass tube H.
sectional area Ac + A', where Ac is considered Within the magma chamber, we also assume
constant and A'(y, t) is a small variation in area there is a balance between the viscous resistance
due to inflation or deflation. The chamber also has from the velocity v and the pressure drop, so that
linearly elastic walls so that the pressure p in the
magma chamber is related to A I by the formula (2)
p = EA I, where the coefficient E is a coefficient
 15. Fluid and Thermal Dissolution Instabilities 363
where II H is the viscosity of the hot fluid in the Z
magma chamber and Ie is a cross-sectional length L
scale of the magma chamber. For a circular
magma chamber, I~ will be of the order of the
cross-sectional area, A e , but other chamber cross
sections may have smaller values of I~ .
Combining (I) and (2),

ap
+ A_ e[2 -a
2p
E -1 - = Q _ wd (3) a
at II H P ai
We must now relate the vertical velocity w to the
Figure 5 Sketch of the temperature of the conduit wall and
pressure drop along the slot from the magma the temperature of the fluid centerline for this idealized prob-
chamber to the exit. Steady viscous flow is lem whose solution is Eq. (8).
assumed,

lap
paz
l211(T) -dw 2
+-
aw
at' (4) to TH • For wd 2 /4K < z < L, the temperature at the
centerline will be calculated using Eq, (8) evalu-
where the viscosity is given as ated at x = O. In that case, Eq. (4) assuming steady
flow is combined with Eqs, (5) and (8) and inte-
II = IIH + a(TH - T). (5)
grated as
Here T is the temperature of the fluid in the con-
duit, TH is the temperature at which the viscosity p
- l2W(fL
= -- II h dz + fL (
a TH
P d? 0 wd'/8K
(9)
is II H, and a is the viscosity-temperature coeffi-
cient. To find p as a function of w, it is necessary wtiTd 2

to determine the temperature of the fluid in the 8KL

conduit. Assume that the material is flowing with
uniform slab flow (w = constant across the slot) so
in the z direction, so that

w-
st aT
2
= K-. (6)
az ax
2
which simplifies to
Let the temperature of the boundaries decrease pd'
P = --'----
linearly in the z direction at a rate of tiT/L so that 96l1 HPKU
along the conduit walls

sr, d = -w
Wr
( 1 +- 1A(
2
(lla)
T = TH - - at x ± (7)
L 2
w < w,
A solution to (7) of the form
w
2 P w> W" (lIb)
T = T
H
_ tiTz + wtiT (d (8) W'r
L 2KL 4·
where
exists. In particular, the temperature of the center
plane (x = 0) is higher than the wall tempera- (12)
ture by an amount wtiTd 2 / 8KL. The temperature
of the sidewall and centerline from Eq. (8) is and
sketched in Fig. 5.
For z< wd 2 /4K, it will be assumed that a ther- atiT
A (13)
mal boundary layer is developing and T will be set
364 J. A. Whitehead and Peter Kelemen

a b

81 ,'\
,I \ IA=l00!
,I \ 'j
6 \,
\, o
\,
\
p 4
,
i \,
log P

,I \,
I, -1
\,
2 , \A=lOl \
\
j -'- ---
a ¥'O------t--~-_+--~-__1 -2 +------+-----t--~-__I
o 2 3 -2 -1 o
log W/W T

Figure 6 (a) Pressure drop across the slot calculated from Eq. (II) as a function of velocity of the fluid for the three values A
°
= (straight line), 10, and 100. (b) The same data plotted as jog-log.

For w > W" the fluid stays at TH during its entire The governing dimensionless numbers are
traverse of the slot.
The crucial result for possible hydrodynamic
instabilities is that the pressure can drop off with
A= '}' = (~~). (16)

an increase in fluid velocity within a certain range

of wlW, if A > 6. To illustrate this, the function P
is plotted as a function of wlW, for some fixed
The parameter'}' is a measure of the ability of
values of A in Fig. 6.
the magma chamber to expand with pressure and
It is useful to nondimensionalize the final equa-
8 is a measure of the aspect ratio of the slot.
tions using W = wlW, for the scaled velocity,
Q' = QIW,d for the scaled volume flux, t ' = Oscillatory Instability for
12v Htld 2 for the scaled time, P for the scaled pres- One-Dimensional Flow
sure, and y' = yll: for the scaled length so that the
If y variations are neglected, Eqs. (l4b), (l5a),
equations now read
and (l5b) resemble a simple set of earthquake
ap , equations (Whitehead and Gans, 1974) that have
'}' at' = Q (l4a) the dimensionless form
dF
and 1 - u' (17)
dt"
aw
-, + feW) = P; (l4b) du' u'
(18)
at m dt" + 132 + U'2 = F.
where the newly scaled flow resistance in the slot Here F is the scaled force accumulated in an elas-

feW) = W (1 + i (I - W)2), (l5a)

tic region, u' is the scaled sliding velocity, m and
13 are dimensionless measures of mass and friction
coefficient, respectively, and til is a scaled time.
W < 1
The friction law u'/(j32 + U ' 2) is similar to that in
feW) = W, W>l. (l5b) Eq. (l5a) in the sense that the sliding resistance
 15. Fluid and Thermal Dissolution Instabilities 365

a 6
b
I Pressure-Velocity History I 51
5 IT] I 4
I
r-
.-,

4 "j
,' ' ',,
," , ,,',,
.:1 1
',P +,'/ " " ,'/ I 3T ,
3
"
P
..
...
2Y ~.
,,
,
,',
..
,"

,~ A 'I 2t

'\U
I ', ,
,-J- '.-
I \ ,/ I :

o .to . ' ~.
0 10
,

20
I
30
\r:~
40
I
' .
50
0
2
0
TIME W

Figure 7 Dimensionless pressure and velocity history during flow in the conduit after exiting the magma reservoir.
(a) Numerical solution of Eqs. (l4a) and (l4b) with a/ay = 0 for P (dashed curve) and W (solid curve) versus time, The governing
parameters are (A. Q. y) = (50,0,0.7, 1.0). (b) P-W phase plane of the preceding solution. The steady-state relation P = fCW)
from Eqs, (15a) and (15b) is also shown as a dashed curve.

increases linearly as u' increases from zero, but it der Pohl equation with a region of negative fric-
then falls off as the velocity increases above a set tion bordered by a region of positive friction. For
value. The earthquake equations have a steady so- Q in the linearly unstable region (dfldW < 0) a
lution u' = I. When f3 > I this solution is stable. limit cycle oscillation will always develop. When
When f3 < I this solution is unstable and there are A is large and f( W) has a large region of negative
growing oscillations. The oscillations become in- slope, W will oscillate between long periods of
creasingly episodic when the inertia term m and nearly constant low flow and an eruptive phase in
the friction term f3 get very small. It is necessary which W increases rapidly and then returns to low
to add only one term to the earthquake equations flow-a behavior like that of the earthquake equa-
to make them similar to our present equations. tions. Behaviors for two sets of parameters in
Equation (18) possesses a friction term that falls these limits are shown in Figs. 7 and 8. When the
off to zero as u gets large, whereas the present parameter A has a value close to the critical value
problem has a friction term that stays proportional of 6, the oscillations are close to sinuosoidal, as
to W for large W where the constant of propor- shown in Fig. 9.
tionality corresponds to the viscosity of the hot
fluid, This feature could be incorporated into Stahility of Two-Dimensional Flows
the earthquake equations by adding a term Au. The equations are also unstable to spatial dis-
where A « I. Oscillations would then grow turbances. We now linearize (I4a) and (I4b)
when a 2 + A < I. around a steady flow so that W = W o + €W'. One
For no spatial dependence. fila)' = 0, and equation can be found,
Eqs. (I4a) and (I4b) can be combined to give
a2w' af aw' a'w'
y- + y-- - 8--
d 2W
+ af dW + (W Q') at2 awo at atay 2 (20)
O. (19)
dt'2 aw dt' y af a2 w'
- 8- -- + w' = O.
awo ayz
This is the form of an equation describing a mass-
spring system with nonlinear friction as given by Take w' = We eikY"T/, where We is a constant. The
aflaw. The topology is similar to the classic Van roots are
 366 J. A. Whitehead and Peter Kelemen
a b
5-r-------------~
IPressure-Velocity History I

W,p p

O-F:::::::t==-+---+-kt=:::;::;==+----1--4==-=-==+===l O+-----r----f----t----J
o 10 20 30 40 50 60 70 80 90 10C o 2 3 4

TIME IV

Figure 8 Dimensionless pressure and velocity history during flow in the conduit after exiting the magma reservoir.
Variables same as in Figs. 7a and 7b. except that governing parameters are (A. Q. y) ~ (50.0,0.7,5.0) so the magma chamber is
more easily inflated.

a b
3 4
IPressure-Velocity History I
3
2 IT]

'If)//
I
w,p p

0
0 10 20 30
:V, 0 1
I
2
I
3

TIME W
Figure 9 Dimensionless pressure and velocity history during flow in the conduit after exiting the magma reservoir. Variables
same as in Figs. 7a and 7b except that the governing parameters are (A, Q. y) = (10.0,0.7, 1.0) so the system is close to neutral
stability.

quired for instability. For k = 0 and when y » I,

the growth rate is aj/awo. When y « I the
growth rate is one-half as fast and the instability
is in the form of growing oscillations with fre-
quency of y-I/2. Figure 10 illustrates the behavior
of the real part of the growth rate. A number of
curves are shown corresponding to different val-
ues of r = 4y (aj/iJ wo) . For r < I, only steady-
growing modes are found. For r > I the steady
For k = 0 the result for the time-dependent so- root intersects an oscillating root that is shown as
lution is recovered, namely, that aj/awo < 0 is re- a straight downward-sloping line that intersects
 15. Fluid and Thermal Dissolution Instabilities 367
3

0.1
2
1.
4.0

-1

-2

-3
0 2 3 4 5 6 7 8 9 10

Wavenumber Squared
Figure 10 Plot of the growth rate from Eq. (21) as a function of wavenumber squared. The growth rate is the real part of (T
divided by (~) (aj/aw o ) and the wavenumber squared is k2 divided by (y/8j(aj/awo ). Four values of the parameter (y/4)(iJj/awo )
are shown. The curves asymptote to 2 for large wavenumber.

k = O. For some values of k the growth rate is high viscosity and no pressure can become suf-
negative so that the perturbations decay. However, ficiently large to make the fluid in the slot flow
for large enough k, all growth rates are positive rapidly enough to reach the limit where hot fluid
and all curves asymptotically approach the limit- fills the slot. In the latter case, fluid in the slot is
ing growth rate of 2. Therefore, the shortest wave- hot and nothing can make the fluid flow slowly
lengths grow most rapidly in this theory. enough to become cold. These first two features
These results lead to some physical insights were borne out by the laboratory results in the
into the nature of instabilities that are produced by preceding experimental section. These dynamical
an increased resistance to fluid flow in conduits features differ from the earthquake theory analog
upon cooling. First, for an instability of the uni- where an instability always happened above a
form flow, the parameter A must be above the critical slip rate.
value 6 so that the resistance must actually fall off Third, in this theory, the most rapid growth is
with increasing fluid velocity over some range of at large k, so small wavelengths are to be expected
w/Wr • As a dynamical comparison, this feature as the preferred mode of growth. This third pre-
was also found in the earthquake theory, and most diction was not verified in the preceding experi-
earth scientists probably would have agreed that ments because the experiments correspond to
this feature was necessary even before the theories k = O. Experiments with paraffin described in the
were developed. next section correspond to other values of k.
Second, w/Wr must be of order of some fixed Since the linear analysis predicts the fastest
number less than 1 and/or approaching I for a instability growth at infinite wavenumber, some
growing instability. The parameter group w/Wr modification of the theory is needed to produce
may be interpreted as Pe dll., where Pe is Peclet explicit predictions of the expected wavelength.
number = Wd/K. This is a measure of the ability Since fluid velocities in the conduit in the direc-
of a flow to advect heat in competition with the tion of the axis of the magma chamber were ne-
dissipative effects of diffusion. Thus, according to glected, the preceding theory can be thought of as
the theory, Pe must be between 8L/3d and 8L/d a long wavelength theory. A theory that includes
(for Q'~ T » 6) for an instability. For values of Pe flows in two directions in the slot has not yet been
outside this range (either smaller or larger), no in- developed. Secondly, no finite-amplitude theory
stability will occur. In the former case, the fluid has yet been produced. Such a theory might lead
will be so slow that all fluid in the slot will have a to long wavelength cutoffs as well.
 15. Fluid and Thermal Dissolution Instabilities 369
Experiments with Paraffin ing in the cell. Forty-eight seconds later, dye of
another color was injected, and by that time flow
Two-Dimensional Flow through a Slot was going out of only one finger through one tube
(Fig. 11F). For 40 more seconds, flow out through
Experiments with liquid paraffin demonstrate that finger continued in a clearly defined channel
a transition from uniform flow to fingering- with little apparent change.
instability-dominated flow as time progresses. The Distances were measured from the photo-
apparatus consisted of a 1.2-cm-thick square alu- graphs with dividers and then tabulated. For pho-
minum plate 61 em on a side in an ice water bath. tographs illustrating an almost circular intrusion,
A 1.1-cm-thick square Plexiglas plate 46 em on a the extremes of the radius were measured. For the
side was clamped over the aluminum with spacers finger cases, the radial distance to the tip of each
between the aluminum and the Plexiglas so that a finger from the feeding tube exit was measured
narrow gap of 0.25 em remained. A hole in the and one measurement was also taken of the dis-
center of the Plexiglas was connected to a reser- tance between each pair of fingers. The results are
voir containing melted paraffin. A camera was po- shown in Fig. 12. While the front was circular, it
sitioned above the apparatus to record the prog- advanced according to simple conservation of
ress of the experiments. mass laws; a line with the formula r = (Qtl1Th)l!2
As a run commenced, paraffin was delivered to is shown for comparison, where the volume flux
the hole at a rate of 5.5 cc . s -1. For approximately Q is 5.5 cc· S-I, t is time, and the gap width h is
the first 16 s, the paraffin spread out in a radially 0.241 em. As the fingers developed, the fronts of
growing pattern that was close to perfectly cir- the fingers accelerated and the front between fin-
cular (Fig. 11A). Small deviations from perfect gers decelerated, finally stopping. At this time
circles appeared to be produced from the surface the paraffin appeared to begin to solidify in the
tension effects arising from slight irregularities in quiescent regions. As time passed, many of the
the texture of the black painted aluminum, but smaller finger fronts also halted their advance, but
these deviations produced less than a 10% devia- those fingers remaining reaccelerated. The width
tion in the radius of the circle. After 16 s, the cir- of the final active channel as marked by the col-
cular front rapidly developed small radial undula- ored dye varied along the channel axis from 1.4
tions (Fig. lIB) that signified a sudden decrease to 1.7 em.
in velocity at select points on the circle. Between A simple explanation of why the instability
these undulations, there was a concomitant rapid must happen is that the paraffin would solidify if
growth of radial finger-like bulges (Fig. 11C) with it remained in a wholly circular flow pattern be-
round tips. Ten or twelve such fingers grew within cause then it would stay in the gap for more' than
4 s but many stopped growing during the next 4 s a thermal time constant. However, it was clear that
(Fig. lID). The only subsequent change in the the paraffin had never completely solidified any-
pattern was that four fingers reached the edge of where during this run (it appears to experience a
the tank, and the rest froze. Oil-soluble dye was dramatic increase in bulk viscosity as it cools).
then injected into the paraffin source, and it was Assuming that both the lid and the aluminum plate
observed that when the dye had arrived in the cell, cool the paraffin as it flows along the conduit, the
most of the dye fluid flux was moving into the two value of the thermal time constant for the paraffin
largest fingers (Fig. lIE). Melt tubes feeding the in a gap of 0.241 cm can be estimated from the
other fingers seem to have stopped and solidified. formula h2/4K. Using the above value of hand
The dye had begun to intrude onto a third finger k = 0.0004 em- . s -1, the time constant is 36 s.
tube but it apparently stopped shortly after arriv- We saw not only that the interface became un-

Figure 11 Photographs of the evolution of finger flow instabilities in paraffin flowing with constant volume flux through a
cooled slot from a point source. (A) A circular intrusion 12 s after start of the experiment. (B) Two small notches have grown at
16 s. (C) Numerous fingers break out by 20 s. (D) The fingers have grown considerably by 24 s. (E) At 52 s, dye reveals there is
flow through only two channels. There is also evidence that a third channel stopped just as the dye entered the tank. (F) At 92 s,
darker dye reveals that there is flow in only one channel. The scale bar shows alternating white and black rectangles I cm long.
370 J. A. Whitehead and Peter Kelemen
.30 -,-------------~_r_------__,

Radial
Smooth
20 Perimeter
Flow ~
8

10 0

Finger
Flow

0
0 10 20 .30

Time (sec)
Figure 12 Transition from radial smooth perimeter flow regime to finger flow. Hatchured band is approximate transition region.
Data (dots) for the advance of the front of the paraffin with time. As the unstable front breaks into fingers, the fingers advance
much more rapidly than the neighboring regions. In other places the front may actually stop. The smooth curve is for a solution
to the simple conservation of volume for a circular front, r = 2.7(1(2. Diverging lines are a hand-fit envelope for the rapidly
expanding radial distance spread after finger development.

stable after 16 s but also that the final channel of Delaney and Pollard (1982) and further developed
approximately 1.5 cm width admitted a flow from by Bruce and Huppert (1989, 1990) that flow
the hole to the edge of the plate of around 15 cm . through a dike is fundamentally unsteady and ei-
s- '. Therefore, fluid leaves the region after being ther is gradually blocked or melts back the coun-
in the slot for less than 2 s, a time that is short try wall rock could be combined with these con-
compared to the time constant estimated earlier. straints to produce yet more realistic stability
Of course, these numbers have considerable un- studies. Lowell (1990) suggests that the spatial
certainties and are meant to be only approximate focusing of black smokers may result from the
and suggestive, but the simple concept that a flow thermal contraction upon cooling of the vent
pattern is formed that would allow fluid to escape walls, although mineral precipitation along veins
before it cools is consistent with the experiment. and fractures is the primary cause of permeabi-
The parameter group wd/WrL = wd 2/8KL at lity reduction in hydrothermal systems. Kent
which the fingers were first seen can also be es- et al. (1992) have pointed out the presence of
timated. Using w = 3Q/41Trd, with r = L = lamprophyre intrusions in the Damodar Valley
13 em, we get wd/WrL = 0.6, which is in agree- of northeast India whose cylindrical melt tubes
ment with the notion that this number must be in have apparently been produced from low viscosity
the range of I for instability. melt penetration into adjacent carbonaceous
A detailed application to a specific geological sediments.
system has not been made. It would be premature
at this time and would have to be tailored spe- Two-Dimensional Flows with a Free Surface:
cifically to the geological case being considered. Lava Flow Modeling
Numerous magmatic and hydrothermal systems,
however, are known to possess an increase in fluid The paraffin experiments may be considered
flow resistance upon cooling. The hypothesis by the magmatic counterparts to experiments con-
 15. Fluid and Thermal Dissolution Instabilities 371
ducted on model lava flows by Fink and Griffiths ing ridges tend to erupt cooler lavas at lower erup-
(1990). They injected polyethylene glycol at the tion rates than fast spreading ridges.
base of a water bath and correlated the develop-
ing morphology of the solidified material with the
model temperature and flow rate. Their experi- Magma Flow Instabilities from
ments had the additional feature of possessing Chemical Corrosion
a large exposed free surface so that the result-
ing structures had additional degrees of freedom The final set of studies are attempts to understand
(such as the reduced gravity g') compared with the effects of chemical dissolution on melts as-
the preceding experiments. They observed that cending by porous flow in the Earth's mantle.
progressively cooler fluid experiments had sur- Porous flow accompanied by compaction is now
face morphologies that resembled features on a well-accepted mechanism for separating melt
cooling lava flows and lava lakes. For flows with from a mantle plume or diapir that is ascending
a crust that cooled very slowly, marginal levees and undergoing decompression melting (Sleep,
that contained and channeled the main portion of 1974; Turcotte and Ahem, 1978; Turcotte, 1981,
the current developed. Yet cooler flows with more 1982; Scott and Stevenson, 1984; McKenzie,
rapid crustal growth rates had regularly spaced 1984; Fowler, 1985, 1990a, 1990b) but must be
folds and multiarmed rift structures complete with understood as representing a single extraction me-
shear offsets and bulbous lobate forms similar to chansim in a large heirarchy of transport mecha-
pillow lavas. nisms that include ascent by magma fracture as
Ito and Whitehead (work in progress) are in- well as the rise of relatively viscous melt by melt-
vestigating the cooling of paraffin from a line dominated diapirism in island arcs and in the si-
source. By using paraffin, the final morphology of licic systems of contential interiors. However, the
the solid can be quantified with thickness mea- most successful models of compositional varia-
surements. Photographs of two typical samples tion assume "closed system melting" where a
are shown in Fig. 13. Preliminary results are that mantle source is melted and the liquid is removed
the parameter k = all: becomes progressively without further intimate chemical contact with the
smaller as the parameter 'It becomes larger, as matrix as it ascends to the surface. Such a rapid
shown in Fig. 14. In this case, a, is the standard removal of melt from its source seems necessary
deviation of the thickness of the flowing fluid, and in understanding some magmatic systems, for
L is the flow length. Here, L is volume of paraffin example, at mid-ocean ridges where liquids ap-
added divided by H; times tank width. H; is the parently record equilibrium with high pressure
thickness of a gravity current. In Huppert (1982), phases, most notably garnet (Salters and Hart,
H h = (g' q3 td /Vp ) I/5 , where g' is the reduced grav- 1989). However, perhaps chemical reaction dur-
ity = !1pg/p, q is the volume flux per unit length, ing ascent of liquid through the upper mantle does
vp is the viscosity of paraffin, and td is the duration playa role in melt extraction in other types of set-
of the flow (eruption). The parameter 'It is the ra- tings (Kelemen et al., 1990, 1992).
tio of time required to cool to the solidification It is not difficult to visualize that the same
temperature to the advection time scale t.. types of instabilities described previously could
The conclusion is that lava morphology de- be produced from chemical dissolution rather
pends on the lava solidification rate during the than from strictly thermal effects. We assume that
flow process. The primary controlling factors are melt is rising by compaction through a matrix of
the effusion rate and the lava temperature. More mantle residua and that as the liquid ascends to
rugged morphology results if the lava cools early lower pressures it departs from the melting point
during the flow process and smoother textures along an adiabatic decompression path. The ma-
form if solidification rates are slow. Bonatti and trix at the higher level must stay in equilibrium
Harrison (1988) found a larger abundance of pil- with this liquid, which is now slightly super-
low basalts at slower spreading ridges and a pro- heated. The result will be that the liquid, which is
portionally larger number of sheet flows at faster still in intimate contact with the matrix, dissolves
spreading ridges. This suggests that slow spread- some of the matrix such that the local region
372 J. A. Whitehead and Peter Kelemen

Figure 13 Photographs of representative wax flows labeled with run number. The flow direction is from top to bottom in each
frame. (A) Pillows and deep clefts. (B) Lobed flow fronts. Run numbers 14(a) and 19(b).

evolves to a new thermal equilibrium. This evo- meability (reduces the resistance to fluid flow) of
lution is sketched in Fig. 15. The dissolution the melt, since for a fixed geometry, permeability
therefore increases the melt fraction (the percent- is a function of melt fraction porosity (e.g., Tur-
age of melt) slightly and also increases the per- cotte and Schubert, 1982). Dissolution cannot
 15. Fluid and Thermal Dissolution Instabilities 373

Figure 13 Continued. Photographs of representative wax flows labeled with run number. The flow direction is from top to
bottom in each frame. (C) A rippled sheet; (D) Smooth sheet flow. Run numbers 18(c) and 16(d).

immediately return the melt/matrix system to slightly slower (on the order of 0.5 to <0.01 em .
thermodynamic equilibrium since the rate of dis- day- I at 1300 to 10500 C; Kutolin and Agafa-
solution is governed by the rate of diffusion across nov, 1978; Scarfe et al., 1980; Donaldson, 1985;
a boundary layer in the liquid. Dissolution will be Thornber and Heubner, 1985; Kuo and Kirk-
 x -Pillows present
o -Rifts present
'" - Lobes present
+ -Ripples present
o -Flat sheet flows

07
10'

Figure 14 Scaled roughness ~ versus '1'. Roughness decreases with increasing '1'. Morphology types are denoted by symbols.
There is a general progression from pillows to flows with deep clefts, to lobes, to folds and then finally to flat sheets.

10 patrick, 1985a, 1985b; Brearley and Scarfe, 1986;

Tsuchiyama, 1986a, 1986b) than the rate of po-
rous flow due to compaction (on the order of 2 to
0.1 ern . day - 1 at 1300 to 10500 C; Watson, 1982;
Riley and Kohlstedt, 1990, 1991; Riley et al.,
1990). This will be especially true as chemical
equilibrium is approached, since dissolution rates
for a given solid phase decrease markedly as satu-
ration of the liquid with that phase is approached
30 (Kuo and Kirkpatrick 1985a, 1985b; Brearley and
Scarfe, 1986).
Mantle This can be quantified by assuming a liquid
Adiabat diffusion constant of D = 10 -7 to 10 -10 ern- . s I
for components such as AI and Si at temperatures
1250 1450
of 1300 to 10500 C (Hofmann, 1980; Chekmir and
Temperature. ·C
Epel 'baum, 1991; Shimizu and Kushiro, 1991;
Figure 15 Sketch of the pressure-temperature relation for Watson and Baker, 1991) and an effective bound-
mantle material. The mantle material rises to the solidus. Liq-
ary layer thickness db ranging from 10 to 1000 ,urn
uid forms and flows upward by compaction. If the liquid were
rising without thermal contact with the residual mantle mate-
for the case of a well-stirred, moving liquid (Kuo
rial, it would rise at the adiabatic melt trajectory (right curve). and Kirkpatrick, 1985b). The e-folding time to re-
However, the melt and residual material are in thermal contact. turn to thermodynamic equilibrium dUD is about
so if they both rise at the same temperature without dissolu- 10' to 10M s. A "diffusive Peclet number" PeD is
tion, they rise at the adiabatic nonreacting melt and residual
generated by dividing this time by the time it takes
trajectory (middle curve). However, as the liquid rises to lower
pressures, the liquid is above its liquidus (left curve), so it is
melt to rise through a mantle region of interest
superheated. It would dissolve residual material and follow the Llw, so PeD = wdVDL. Recall that the thermal
dissolution trajectory over to the liquidus. studies discussed earlier in this paper showed that
 15. Fluid and Thermal Dissolution Instabilities 375
the parameter group wd 2 /KL must be of order 1
Water flow
for the instability to arise. It is probable that the
rise time for melt will be in the above range and
that instability will arise in regions with small L
(of order a few meters) where melt suddenly en-
2
counters a matrix it can dissolve.
Two exploratory laboratory experiments have
3
been conducted to attempt to produce this in-
stability. In the first, a thin tank of size 58.3 em 4
high X 63.5 em long X 1.0 cm wide was filled
with dry table salt. Water was slowly fed in from 5
above through a pipe with many small holes to
even the flow. The water level was increased to
10-20 em above the salt so that the flow evened
itself out through a natural stratification. The wa- 6
ter slowly descended through and dissolved the
salt. At the bottom, the water filtrate escaped
through a 3-cm layer of foam rubber that lay
above approximately 50 small holes in the base of 7
the tank.
For the 10 runs to date, flow rates have ranged
from 1 to 5 cc . S I, and in all cases the free sur-
face of the salt became unstable. Figure 16 shows
a typical evolution of the salt surface. At first, the
salt surface began to move downward uniformly Figure 16 Sketch of the evolution of the upper surface of
as the descending water dissolved the salt. How- salt produced by the downward percolation of water and exit
flow through a porous base. The numbers correspond to time
ever, within 5 min, undulations with a wavelength
progression.
of roughly 2 to 4 cm were visible on the surface
of the salt. As time progressed, the amplitude of
the undulations became greater and the wave- suits were complicated by the absence of a rigid
length appeared to increase. When the slope of the network in the porous media. The shape of the top
salt-free surface reached a fixed angle, probably salt surface was constantly modified by "salta-
the angle of repose, the salt would tumble down tion" of the salt grains, which rolled down from
the slope and the interface thereafter resembled a high points, where relatively little dissolution had
sawtooth pattern of peaks and valleys. As time occurred, into valleys where more dissolution had
progressed further, the distance between the peaks occurred. Thus the morphology of the dissolution
of the sawtooth became larger as small valleys channels was obscured.
were eliminated and large ones persisted. The A second experiment duplicated the first ge-
small valleys appeared to be worn down on one ometry but included glass spheres of 0.1 mm di-
side or the other by the larger valleys and by grain ameter mixed with roughly 30% salt (with grains
saltation down the slopes. 0.03 mm). The glass and salt grains were intended
The experiment, although suggestive, had some to duplicate relatively soluble (pyroxene) and
serious drawbacks. Air dissolved in the water was insoluble (olivine) components in the mantle ma-
driven out of solution as the salt dissolved and trix. The presence of glass balls provided a
the water became saturated. This happened even rigid network that impeded "saltation" of the salt
though the air had been degassed as much as was grains, allowing observation of the morphology of
practical by using very hot water over long time the dissolution channels. A thin layer of salt was
periods. Air bubbles would accumulate in the up- placed over the glass bead/salt mixture so that the
per few centimeters of the salt and then perco- descending water would be close to saturated be-
late upward. Additionally, the experimental re- fore it percolated down into the glass/salt mixture.
 15. Fluid and Thermal Dissolution Instabilities 377

Figure 18 The glass sphere and salt mixture experiment at an advanced stage of development. After about 17 min, one region
is open. Vertical fingers of dissolution are seen in other regions.

This experiment also exhibited the surface in- Acknowledgments

stability, but as the instability grew, the peaks and
valleys along the top surface were much more Particular thanks are due to Robert E. Frazel who assisted with
subdued since the glass ball network underwent numerous laboratory experiments and produced many of the
the photographs. We also thank Peter Olson and Frank Richter
relatively minor compaction due to salt dissolu-
for constructive suggestions. Supported by the Experimental
tion. Instead a series of finger-like vertical veins and Theoretical Geophysics Division, National Science Foun-
of a salt-free glass matrix were formed, alternat- dation, under Grants EAR 87 08033 and EAR 89-16857.
ing with regions still retaining a salt-glass mix-
ture. Figures 17 and 18 show photographs of the
tracing dye descending into the model matrix. The
water flow is channeled into the salt-free fingers References
where the permeability is higher. The qualitative
aspects of the instability, particularly channeliza- Bonatti, E., and Harrison. C. G. A. (1988). Eruption styles of
basalt in oceanic spreading ridges and seamounts: Effects
tion of the water flow into the salt-free zones with
of magma temperature and viscosity. J. Geophys. Res. 93,
a lower flow resistance, and the emergence of 2967-2980.
these channels as a hydrodynamic instability seem Brearley, M., and Scarfe, C. M. (1986). Dissolution rates of
clearly confirmed. upper mantle minerals in an alkali basalt melt at high pres-

Figure 17 Photographs of an experiment with a mixture of about 80% glass spheres and 20% salt. Water flows from the top to
bottom, (A) After about 2 min. injected blue dye reveals vertical fingers of flow. The dissolution front at the top is becoming
uneven. (B) After about 9 min, injected blue dye shows that water flows at higher velocity and penetrates preferentially in some
regions. (C) After about 13 min, dissolution fingers are clearly visible. (D) After about 13.5 min the injected blue dye shows that
almost all flow occurs within the largest fingers.
378 J. A. Whitehead and Peter Kelemen
sure: An experimental study and implications for ultra- Kutolin, V. A., and Agafanov, L. V. (1978). Composition of
mafic xenolith survival, J. Petrol. 27, 1157 -1182. the upper mantle in light of the relative stability of ultra-
Bruce, P. M., and Huppert, H. E. (1989). Thermal control of mafic nodules, Geol. Geofiz. 19,3-13.
basaltic fissure eruptions, Nature 342, 665-667. Lowell, R. P. (1990). Thermoelasticity and the formation of
Bruce, P. M., and Huppert, H. E. (1990). Solidification and black smokers, Geophys. Res. Lett. 17,709-712.
melting along dykes by the laminar flow of basaltic Marsh, B. D. (1979). Island arc development: Some observa-
magma, in "Magma Transport and Storage" (M. P. Ryan, tions, experiments and speculations,J. Geol. 87, 687 - 713.
ed.), Wiley, Chichester/Sussex, England. McKenzie, D. P. (1984). The generation and compaction of
Chadam, 1., Hoff, D., Ortoleva, P., and Sen., A. (1986). Reac- partial melts, J. Petrol. 25, 713-765.
tive-infiltration instability, J. Appl. Math. 36, 207 -238. Ortoleva, P., Merino, E., Moore, C; and Chadam, 1. (l987a).
Chekmir, A. S., and Epel'baum, M. B. (1991). Diffusion in Geochemical self-organization. I. Reaction-transport feed-
magmatic melts: New study, in "Physical Chemistry of backs and modeling approach, Am. J. Sci. 287, 979-1007.
Magmas" (L. L. Perchuk and I. Kushiro, eds.), Applica- Ortoleva, P., Chad am, 1., Merino, E., and Sen, A. (I 987b).
tions in Physical Chemistry, Vol. 9, pp. 99-119, Springer- Geochemical self-organization. II. The reactive-infiltration
Verlag, New York. instability, Am. J. Sci. 287, 1008-1040.
Delaney, P. T., and Pollard, D. D. (1982). Solidification of ba- Riley, G. N., Jr., and Kohlstedt, D. L. (1990). An experimental
saltic magma during flow in a dike, Am. J. Sci. 282, study of melt migration in an olivine-melt system, in
856-885. "Magma Transport and Storage" (M. P. Ryan, ed.),
Donaldson, C. H. (1985). The rates of dissolution of olivine. pp. 57 -92, Wiley, Chichester/Sussex, England.
plagioclase. and quartz in a basalt melt, Min. Mag. 49, Riley, G. N., Jr., and Kohlstedt, D. L. (1991). Kinetics of melt
683-693. migration in upper mantle type rocks, Earth Planet. Sci.
Fink, 1. H., and Griffiths, R. W (1990). Radial spreading of Lett. 105,500-521.
viscous-gravity currents with solidifying crust, J. Fluid Riley, G. N., Jr., Kohlstedt, D. L., and Richter, F. M. (1990).
Mech. 227,485-509. Melt migration in a silicate liquid-olivine system: An ex-
Fowler, A. C. (1985). A mathematical model of magma trans- perimental test of compaction theory, Geophys. Res. Lett.
port in the asthenosphere, Geophys. Astrophys. Fluid Dyn. 17,2101-2104.
33,63-96. Ryan, M. P., and Blevins, 1. Y. K. (1987). "The Viscosity of
Fowler, A. C. (I 990a). A compaction model for melt trans- Synthetic and Natural Silicate Melts and Glasses at High
port in the Earth's asthenosphere. Part I. The basic model, Temperatures at I Bar (10' Pascals) Pressure and at Higher
in "Magma Transport and Storage" (M. P. Ryan, ed.), Pressures," U.S. Geol. Surv. Bull. 764, p. 563.
pp. 3-14, Wiley, Chichester/Sussex, England. Saffman, P. G., and Taylor, G. I. (1958). The penetration of a
Fowler, A. C. (I 990b). A compaction model for melt trans- fluid into a porous medium or Hele-Shaw cell containing
port in the Earth's asthenosphere. Part II. Applications, a more viscous liquid, Proc. Roy. Soc. London A 245,
in "Magma Transport and Storage" (M. P. Ryan, ed.), 312-29. .
pp. 15-32, Wiley, Chichester/Sussex, England. Salters, V. 1. M., and Hart S. R. (1989). The Hf-paradox, and
Hofmann, A. W. (1980). Diffusion in natural silicate melts: the role of garnet in the MORB source, Nature 342,
A critical review, in "Physics of Magmatic Processes" 420-422.
(R. B. Hargraves, ed.), pp. 385-418, Princeton Univ. Scarfe, C. M., Takahashi, E., and Yoder, H. (1980). Rates of
Press, Princeton, N1. dissolution of upper mantle minerals in an alkali-olivine
Hughes, C. 1. (1982). "Igneous Petrology," Elsevier Scientific, basalt melt at high pressures, Carnegie Inst. Washington
New York. Yearbook 79,290-296.
Huppert, H. E. (1982). The propagation of two-dimensional Scott, D. T., and Stevenson, D. J. (1984). Magma solitons,
and axisymmetric viscous gravity currents over a rigid Geophys. Res Lett. 11, 1161-1164.
horizontal surface, J. Fluid Mech. 121,43-58. Shimizu, N., and Kushiro, I. (1991). The mobility of Mg, Ca
Kelemen, P. B., Johnson, K. T. M., Kinzler, R., and Irving, A. and Si in diopside-jadeite liquids at high pressures, in
1. (1990). High field strength depletions in arc basalts due "Physical Chemistry" (L. L. Perchuk and I. Kushiro, eds.),
to mantle-magma interaction, Nature 345, 521-524. Vol. 9, pp. 192-212, Springer-Verlag, New York.
Kelemen, P. B., Dick, H. 1. Boo and Quick, 1. E. (1992). Pro- Sleep, N. H. (1974). Segregation of magma from a mostly
duction of harzburgite by pervasive melt-rock reaction in crystalline mush, Geol. Soc. Am. Bull. 85, 1225-1232.
the upper mantle, Nature 358,635-641. Stevenson, D. 1. (1989). Spontaneous small-scale melt segre-
Kent, R. W, Ghose, N. C., Paul, P. R., Hassan, M. 1., and Saun- gation in partial melts undergoing deformation Geophys.
ders, A. D. (1992). Coal-magma interaction: An inte- Res. Lett. 16, 1067-1070.
grated model for the emplacement of cylindrical intru- Thornber, C; and Heubner, 1. S. (1985). Dissolution of olivine
sions, Geol. Mag. 129(6),753-762. in basaltic liquids: Experimental observations and appli-
Kuo, L. C., and Kirkpatrick, R. 1. (l985a). Dissolution of cations, Am. Mineral. 70, 934-945.
mafic minerals and its implications for the ascent veloci- Tsuchiyama, A. (1986a). Experimental study of olivine-melt
ties of peridotite-bearing basaltic magmas, J. Geol. 93, reaction and its petrological implications, J. Volcanol.
691-700. Geotherm. Res. 29, 245-264.
Kuo, L. C., and Kirkpatrick, R. 1. (I 985b). Kinetics of crystal Tsuchiyama, A. (I 986b). Melting and dissolution kinetics:
dissolution in the system diopside-forsterite-silica, Am. J. Application to partial melting and dissolution of xenoliths,
Sci. 285, 51-90. J. Geophys. Res. 91, 9395-9406.
 15. Fluid and Thermal Dissolution Instabilities 379
Turcotte, D. L. (1981). Some thermal problems associated mas" (L. L. Perchuk and 1. Kushiro, eds.), Applications in
with magma migration, 1. Volcanol. Geotherm. Res. 10, Physical Chemistry Vol. 9, pp. 120-151, Springer-Verlag,
267-278. New York.
Turcotte, D. L. (1982). Magma migration, Annu. Rev. Earth Whitehead, John A., and Gans, Roger F. (1974). A new, theo-
Planet. Sci. 10, 397 -408. retically tractable earthquake model. Geophys. J. Roy. As-
Turcotte, D. L.. and J. L. Ahern. (1978). A porous flow model tron. Soc. 39, 11-28.
for magma migration in the asthenosphere, J. Geophys. Whitehead, J. A., Jr., Dick, H. J. B., and Schouten, Hans.
Res. 83, 767 - 772. (1984). A mechanism for magmatic accretion under spread-
Turcotte, D. L., and G. Schubert, (1982). "Geodynamics, Ap- ing centers, Nature 312, 146-148.
plications of Continuum Physics to Geological Problems," Whitehead, J. A., and Helfrich, K. R. (1990). Magma waves
Wiley, New York. and diapiric dynamics in "Magma Transport and Storage"
Watson, E. B. (1982). Melt infiltration and magma evolution, (M. P. Ryan, ed.), pp. 53-76, Wiley, Chichester/Sussex,
Geology 10, 236-240. England.
Watson, E. B., and Baker, D. R. (1991). Chemical diffusion in Whitehead, J. A., and Helfrich, K. R. (1991). Instability of
magmas: An overview of experimental results and geo- flow with temperature-dependent viscosity: A model of
chemical applications, in "Physical Chemistry of Mag- magma dynamics, J. Geophys. Res. 96(B3), 4145-55.
This page intentionally left blank
 Author Index

Ahern, J. L., 57, 371 Beckermann, c. 301

Agafanov, L. v; 373 Beeman, M. L., 20, 51
Ake, 1. P., 189 Beere, w., 278
Aki,K .. 180-181,236 Bell, D. R., 213
Aleksandrov, K. S., 281 Bender, 1. F.. 113
Algor,1. R., 295 Bennon, W. D., 301
Al-Khatib, H. H., 260, 265-266, 268, 271, 276-'-277 Benz, H. M .. 181,278
Alt, J. c., 212 Berckhemer, H., 153,261
Anderson, A. T., 20 Bergantz, G. w., 280, 293-294, 297 -299,304-308,311
Anderson, D. L., 10,32,34 Berge, P., 115
Anderson, E. M., 253 Besse, 1., 6
Anderson, O. L., 211,224,237 Bickle, M, J" 7, 203, 217-218, 271, 273, 286
Anderson, R. N., 77,192,198,213 Bird, P., 192
Ando, M., 181 Bird, R. B., 337
Andrews, D. 1., 198 Bischoff, J. L., 162
Antonellini, M. A., 311 Bjomsson, A., 135
Argon, A. S., 51 Blackman. D, K., 69,169,172
Arndt, N. T., 7-8, 294 Blackwell, D. D., 260, 273-274, 277, 279, 285
Ashby, M. E, 40 Blake, S., 326-328, 332-334, 336
Ashworth, 1. R., 298 Blakely, R. 1.. 285
Atkinson, B. K., 211, 227 Blanpied, M. L., 211
Auzende,1. M., 77 Blevins, 1. v. K., 236, 297, 331, 340, 356
Aydin, A., 243, 248 Blumenfeld, P., 310
Blythe, P. A, 307
Bodri, B., 198,200
Bacon, C. R., 298, 311, 336 Bodri, L., 198,200
Bai, Q., 49-50 Bohlen, S. R., 294, 296
Baker, D. R., 374 Bonatti, E" 371
Ballman, R. L., 330 Bories, S. A., 161, 163
Barcilon, v.. 58 Bottinga, y" 113,297
Barenblatt, G. I., 230 Bouchez, 1.-L., 310
Barnes, H. A., 297 Boudier, E, 77 - 79,82,85
Barsh, G" 281 Bowen, N. L., 294
Barth, G. A.. 77, 146 Boyd, E R., 272
Barton, M. D., 211, 213 Bott, M. H. P" 141-142
Batchelor, G, K., 200 Braile, LoW., 277
Bateman, R., 309 Brandeis, G., 304
Batiza, R., 69, 173, 174 Brandsdottir, B., 112, 119-120
Bea, E, 294, 299 Bratt, S. R.. 143
Beard,1. S" 208, 295-296,305 Brearley, A. 1., 299
Bebout, G. E., 211, 213 Brearley, M., 278, 285, 374
Bechtel, T. D.. 260 Brenan,1. M., 211, 217
Becker, K., 144,163,212 Brodholt,1. P., 69

381
 \82 Author Index

3rooks, A., 203 Cooper, R. E, 20-21, 24-25, 30, 38,45,49-51,57,212

3rophy, J. G., 198 Cordery, M. J., 56
3rown, G. c, 295, 298 Cormier, M. H., 69
3rowning, P., 159-160 Courtillot, V., 6
3ruce, P. M., 238, 323, 346, 357, 370 Courtney, R. C., 286
3uck, W R., 56, 67, 304 Crandall, S. H., 21
3ulau,1. R, 20, 38, 46, 57, 212, 278 Creager, K C, 198
3unch, A. W H., 171 Crisp, J. A., 292
Burnett, M. S., 152 Cruden, A. R., 309
Burnham, C. W, 292, 294-296, 298 Cull, J. P., 294
Bussell, M. A., 310

Dabovski, c, 243-244, 248

Cambray, E W, 311 Daines, M. J., 20, 46, 56, 306
Cameron, K. L., 324, 349 Daly, S. E, 309
Campbell, I. H., 2, 7, 299, 305-306, 328, 332-334. David, E., 307
Canil, D., 278, 285 Davidson, C., 293
Cann, J. R, 141 Davidson, J. P., 311
Caress, D. W, 152-153 Davies, G. E, 4
Carlson, R. L., 134 Davies, J. H., 180, 185, 192, 198-199,202-203,205-211,
Carlson, R. W, 2 214-217
Carmichael, I. S. s., 295 Davis, P. M., 135,261,265,278-279
Carpenter, P. J., 261, 286 Dawes, R., 280
Carrigan, C. R., 324-326, 329, 332, 336-337, 341, 344, Dawson, P. B., 261, 285, 312
346-347,349 Decker, R, W, 135
Carter, N. L., 272-273 Delaney, P, T., 235, 238, 243, 245, 323, 346, 357, 370
Cashman, K v.. 298 DelJ'Ange10, L. N., 299
Cassard, D., 92 Denlinger, R. P., 142
Ceuleneer, G., 56, 78, 92 DePaolo, D. J., 292
Chadham, J., 357 Der, Z. A., 260
Chan, W W, 264 Detrick, R. S., 69, 71,102,105-106,108,115,135,
Chapman, D. S., 264, 275, 279 142-143,145-146,151,154-155,161,172-174
Chappell, B.w., 292, 296 Dewey, J. E, 141, 159
Charles, M. E., 330 Dickenson, W. R., 215, 274
Cheadle, M. J., 58, 210, 215, 278 Didwania, A. K, 278
Chekmir, A. S., 374 Dieter, G. E., 22
Chen, I. W, 51 Dieterich, J. H., 135
Chen, Y. J., 69, 132-133, 142, 147, 159-161, 164-165, Dixon, J. E., 49, 115
168, 171 D'Lemos, R. S., 312
Chen, Y. -M., 292, 304 Drew, D. A., 278
Cheng, D. C. -H., 297 Drummond, B. 1., 293
Chin, H. B., 341 Duffield, W. A., 135
Chopra, P. N., 39, 40-42, 50-51 Dunbar, N. W, 305
Christensen, D., 181 Duncan, R. A., 2
Christensen, N.I., 99-100,103, III, 115, 130, 133, 144, Dungan, M., 311
146,151,281 Dvorak, J. J., 131, 135
Christensen, U. R., 8, 224
Christeson, G. L., 115
Christie, D., 173 Eaton, J. P., 135
Chu, T. Y., 304 Egg1er, D. H., 216
Clarke, D. B., 112 Eichelberger, J. c. 323-326, 329, 332, 336, 341, 349
Clawson, S. R., 185-186, 285 Einarsson, P., 112, 119-120, 135
Clayton, R. W, 180, 183, 186, 277,279 E1thon, D., 112
Clemens, J. D., 294, 296, 309 Emslie, R. E, 23
Clowes, R. M., 143 England, R. w., 309
Coble, R. L., 50 Epel'baum, M, B., 374
Collier, J., 143 Erdelyi, A., 126
Collins, C. D. N., 293 Erdogan, E, 243
Combamous, M., 162-163 Evans, A. G., 227
Condomines, M., 292 Evans, J. R., 185-186,263
Conner, C. B., 311 Everage, A. E., 330
 Author Index 383
Ewart, J. A., 135 Griggs, D. T., 272-273
Ewing, J. I., 143 Grissom, G. C., 294
Eyler, E. E., 302 Grove, T. L., 69, 74, 173,298,302,305
Grunder, A. L., 293
Gudmundsson, A., 235, 322
Farentani, C. G., 7 Guffanti, M., 260, 274
Faul, U. H., 20, 23, 38, 47, 49
Fedatov, S. A., 235
Feeley, T. C., 293 Haafkens, R., 342
Fife, W. S., 295 Hagee, V L., 6
Findley, W. N., 25, 34 Hagen, J. T., 243
Fink, J. H., 235, 327, 358 Hager, B. H., 278
Finerty, A. A., 272 Halderman, T. P., 261, 265, 278-279
Fiske, R. S., 135,224,244 Hale, L. D., 103, 106, 108-109, 135, 143
Forsyth, D. W., 56, 69, 71, 74,159,168,170,172 Halliday, A. N., 292
Fountain, D. M., 281 Hamilton, W. B., 280
Fountain, J. C; 299, 307 - 308 Han, C. D., 330-332, 334, 336, 341
Fourney, W. L., 235 Handy, M. R., 293
Fowler, A. c, 7, 57-58, 60, 224, 307, 371 Hanson, G. N., 69
Fox, P. J., 153 Harding,A.J.,142-143,147,149,151-152
Francheteau, J., 144, 168,274 Harmon, R. S., 294, 296
Frank, E c, 57 Harris, R. A., 181,261,277
Freund, L. B., 229 Harrison, C. G. A., 371
Frisch, U., 334 Harry, W. T., 311
Frith, W. J., 297 Hart, S. R., 38, 51, 371
Frost, H. J., 40 Hasegawa, A., 180,182-183, 185-186,188-189,217-218
Fryer, P., 211, 213 Hasemi, A. H., 181
Fuis, G. S., 293 Hashida, T., 186
Fujii, N., 20, 208, 278 Hatherton, T., 215
Fujii, T., 99, 129,236 Hawkesworth, C. J., 217
Fujisawa, H., 203 Haymon, R. M., 143, 151
Fukao, Y., 181 Hearn, T., 260-261, 277
Fukuyama, H., 180,261-262 Heezen, B. C., 122-123
Furlong, K. P., 274 Heimpe1, M., 281
Furman, T., 322 Helffrich, G. R., 198
Furukawa, Y, 198 Helfrich, K. R., 131,243,355,357-358
Helgeson, H. c., 209
Henstock, T. J., 142, 146, 159, 161
Garmany, J., 77, 143, 154 Herrick, C. N., 134
Gautneb, H., 331 Herrin, E., 272-273
Geertsma, J., 342 Herron, T. J., 103, 135, 143
Gettrust, J. E, 144 Hickox, C. E., 304, 330, 332
Ghiorso, M. E, 297, 300 Hildreth, E. w., 292-293, 320, 351
Gibson, I. L., 322, 323 Hill, D. P., 101, 188
Giggenbach, W. E, 210 Hirahara, K., 180-181, 185
Gill, J. B., 198, 292 Hirth, J. G., 51, 61
Given,J. W.,32,34 Hodge, D. S., 299
Glazner, A. E, 310 Hofmann, A. w., 374
Goldstein, S. L., 294 Holbrook, W. S., 260, 277, 279
Gosnold, W. D., 274, 276, 285 Ho-Liu, P., 185
Goto, A., 214 Hollenberg, G. w., 22, 24
Goto, K.,191 Holloway, J. R., 295-296
Gough, D. I., 285 Homsy, G. M., 278
Gove, P. B., 102 Hon, R., 303
Graham, E. K., 281 Honda, S., 198
Green, D. H., 25, 30, 32, 206, 209, 213, 215-216, 273 Hooft, E. E., 115, 161, 172
Gregory, R., 163,213 Hopson, C. A., 141,166
Grew, P. c, 211, 224 Horai, K., 10
Gribb, T. T., 32 Hori, S., 189
Griffiths, R. w., 2, 7, 358 Horie, A., 181
Griffiths, S. K., 342 Horiuchi, S., 189
384 Author Index

Hsui, A. L 198, 204 Karson, 1. A., 144

Hu, H., 335-336 Kastner, M., 213
Huang, P. Y., 132, 155 Kay,R. W,293-294
Huang, W L., 210 Kay, S. M., 294
Hubral, P., 149 Kelemen, P. B., 85, 371
Huebner, 1. S., 373 Keller, J. M., 334
Hughes, C J., 356 Kempner, W C, 144
Hughes, T 1. R., 203 Kempton, P. D., 294, 296
Humphreys, E., 181 Kennett,8. L. N., 171
Humphreys. E. D., 261, 277-279 Kent, G. M., 147, 149, 154
Huppert, H. E., 99, 113-115,238, 299-300, 305-306, 322- Kent, R. W, 370
323,336,346,357,370 Kenyon, P., 61, 218
Hutton, D. H., 311 Kerr, R. C, 125, 127-129,224,236,238,243-244,252-
Hwang, C, 51 253,281,308,342
Kerrick, D. M., 299
Khiratov, N. I., 146
Ichikawa, M., 243 Kidd, W S. E, 141, 159
Ida, Y., 198, 208 King, S. D., 203
Ildefonse, 8., 78,92-93 Kinoshita, W T, 135
Inamori, T., 189 Kinzler, R. 1., 69, 74
Incropcra, E P., 301 Kirkpatrick, R. J., 374
Irvine, T. N., 113 Kissling, E., 260
Ishida, M., 181 Kitchen, M. R., 286
Ito, E., 272-273 Klein, E. M., 69, 173-174
Ito, E. I., 213 Klein, D. P.. 260, 285
Ivey, G. N., 327 K1emperer, S. L., 293
Iwase; R., 189 Kobayashi, T, 235
Iyer, H. M., 181, 185, 188,260,285 Koh1stedt, D. L., 20, 38-39,41-42,45,47,49-51,57,61,
212,374
Kong, L. S. L., 143, 155
Jackson, D. 8., 135 Kosuga, M., 191
Jackson, E. D., 225, 244 Kovach, R. L., 278
Jackson, H. R., 56, 69 Koyaguchi, T, 114,328,334
Jackson, I., 34 Koyanagi, R. Y., 135
Jackson, M., 78, 82, 92 Kuo, B.-Y., 69, 71,159,168
Jacobs, G. K., 305 Kuo, L. C, 373-374
Jacoby, W., 135 Kushiro, I., 8,10,99,129,180,192.236,261-262,271,
Jakus, K., 24 273-274
Jarrard, R. D., 311 Kutolin, V. A., 373
Jaupart, C, 304 Kuznir, ED., 141-142
Jeanloz, R., 10
Jha, K., 57, 67. 74
Jiracek, G. R., 285 Lachenbruch, A. H., 274, 276-277, 279, 285, 293
Johnsen, G.v.. 135 Lambeck, K., 10
Johnson, K. T. M., 51, 56 Langdon, T G., 227
Johnston, A. D., 295-296 Lange,EE,243-244
Jomha, A. I., 297 Lange, R. A., 295
Jordan, T. H., 198 Langmuir, C H., 69, 145, 173-174, 198
Joseph, D. D., 330, 332, 335-336 Larsen, G., 135
Julian, B. R., 237-238 Larson, R. L., 2, 4-6
Jurdy, D. M., 199 Lawn, B. R., 234-235, 243
Jurewicz, S. R, 20 Lee, B. -L., 330-331, 333, 336
Juteau, T, 110, 130, 134, 163 Lee, W H. K., 180
Leeman, W. P., 293
Levander, A. R., 278
Kamiya, S., 180, 185 Lewis, B. T, R., 143
Kampfmann, W, 153,261 Lin, 1., 69, 71, 73,155,158,168,170-172
Kappel, E, S., 153 Lippard, S. 1., III
Kappus, M. E., 153 Lister, CR. B., 142, 163
Karagiannis, A., 330, 334 Lister, 1.,125,127-129,224,243-244,231,236,238,252-
Karato, S. I., 40, 50, 51 253,281,308,342
Karlstrom, K. E., 311 ue, M., 6-7
 Author Index 385
Lofgren, G. E., 208, 295-296, 305 Morris, E.. 69, 71
Loper, D. C., 6 Morris, 1. D., 209, 217
Lowell, R. P., 306-308, 370 Morris, S., 10, 309
Luedke, R. G., 266-268 Morton. J. L., 105, 135, 142-143, 147, 162
Luetgert,1. H., 188 Muller, O. H., 224, 231, 235
Luther, D. S., 224 Murase, T., 146, 180,261-262
Muskhelishvilli, N. 1., 126,280
Mutter, 1.c, 7
Maaloe, S., 225, 244 Mysen, B. 0., 211, 261
Macdonald, K. c. 56, 107, 153, 167, 171-172
Macdonald, R., 235
Mackwell, S. 1., 40 Nagao, T., 198
Maclean, D. L., 330 Nakada, M.. 10
Madsen, J. A., 147, 153, 158 Nakamura, N., 254
Mahon, K. 1., 309 Nakanishi, 1.,181
Mahoney, 1. J., 2 Nakashima, Y., 281
Mahood, G. A., 292 Nathenson, M., 260, 274
Malinvemo, A., 74 Natland,1. H., 155
Mammerickx, J., 106 Nehlig, P., 85, 110, 134, 163
Manduca, C. A., 293 Nelson, K. D., 293
Manghnani, M. H., 146 Nesbitt, R. w., 192
Mareschal, rc., 293 Newmann, G. A., 170, 172
Mark, R., 225 Newman, S., 214, 217
Marquart, G., 135 Newton, R. c, 295
Marsh, B. D., 114, 146, 158, 192, 198,207,224,296,298, Ni, 1., 301
300,304,308,309,311,357 Nicolas, A., 77-79. 82, 85, 92,111-112,131,141,166,224,
Martin, D., 115 252,280-281,306,310
Matsumoto, S., 188 Niida, K., 213
Matsuzawa, T., 183 Nilson, R. H., 342
Mavko, G., 33, 278 Nishiwaki, M., 189
Mawer, C. K., 309 Niu, Y., 69,173-174
Maxey, M. R., 114 Nokes, R., 115
McBimey, A. R., 146, 192 Nolet, G., 183
McClain, 1. S., 143, 152 Noyes, H. 1., 298
McKenzie, D. P., 2, 7,10,20-21,57-58,161,165,192,198, Nye, C. 1., 218
203,218,224,271,273,278-279,286,307,371
Mercier, 1.-C., 272-273
Meredith, P. G., 227 Obara,K., 181, 192
Mereu, R. E, 293 Ogawa, M., 304-305
Meyer, D. w., 25 O'Hara, M. 1., 112
Meyer, R. P., 143 Okamura, A. T., 131, 135
Mezger, K., 294, 296 Olafsson, M., 216
Michael, P. 1., 49 Oldenburg, C. M., 300
Michaelson, C. A., 181 Oliemans, R. v.. 330
Mikhlin, S. G., 280 Olson, J., 243, 248-249
Mikumo, T., 180, 185 Olson, P., 6. 7, 8, 224, 281, 304
Miller, C. D., 323, 349 Ooms, G., 330
Miller, C. E, 292, 298, 308 Orcutt,J. A., 99-100, 103, 105-108,135, 143-144
Minagawa, N., 330-331, 333 Ortoleva, P., 357
Minear, 1. W., 198 Osamura, K., 278
Mitchell, B. 1., 260, 265-266, 268, 271,.276-277 Otsuki, K., 201, 204
Miyamachi, H., 181 Oxburgh, E. R., 56, 198
Mizoue, M., 189
Mogi, K., 135
Mooney, W. D., 188,277 Palister,1. S., 141. 166
Moorbath, S., 293 Palrnason, G., 101
Moore, 1. G., 135 Paige, C. c, 183
Moores, E. M., 144 Parmentier, E. M., 56-57, 59-60, 63, 65-66, 70, 158, 161,
Morgan, P., 272-274, 276, 285 171,304
Morgan, W. 1., 2, 164-165,168 Parsons, B., 304
Morita, N., 281-283 Paterson, M. S., 39, 41, 50-51, 234
Moriya, T., 181 Paterson, S. R., 292, 309-310
386 Author Index

Patino-Douce, A. E., 295-296 Roy, R. E, 274

Patton, H. J., 260, 265 Rubie, D.c.. 299
Peacock, S. M., 200, 203, 211-212 Rubin, A. M., 230, 232, 238, 284
Pearlstein, A. 1., 304 Rudnick, R. L., 294, 296
Peters, M. T, 295 Ruff, L., 213
Peterson, 1.w., 295 Rushmer, T, 295-296
Philippot, P., 210-211 Rutter, M. 1.. 295, 303
Phipps Morgan, 1., 56-58, 69, 74, 102, 132-133, 142, ISS, Ryall, A., 261, 285
158-162,164-165,168,170,172 Ryall, E, 261, 285
Pike, R. 1., 267, 269, 270-271, 275 Ryan, M. P., 4, 57,100,103,108,110,112,114-115,119-
Pitcher, W. S., 310 121,128,130-131,133,135,161,172, 234,236,238,
Plafker, G., 293 242-243,248,253,281-282,297,322,331,340,342,
Plank, T., 198 356
Pockalny, R. A., 74 Ryan, W. B. E, 153
Pollack, H. N., 264, 275, 279 Ryerson, R., 331, 340, 342
Pollard, D. D., 135,224,230-232,235,238,243,245,248- Ryzhova, T v.. 281
251,253,281,323,346,357,370
Powell, R., 295
Probstein, R. E, 297 Sacks, S., 146, 180, 185-186,218,261,263,264-265,
Purdy, G. M., 69, 143, 155 269,274
Saemundsson, K., 101
Saffman, P. G., 357
Quick, J. E., 142, 159, 162,294 Sakuyama, M.. 192
Salisbury, M. H., 130, 133
Salters, V. 1. M., 38, 51, 371
Rabinowicz, M., 56, 78, 93 Sammis, C. G., 237-238
Rafesky, A., 309 Sampson, D. E., 324, 349
Raj, R., 21, 33 Sanders, C. 0.,183,188
Raskin, G. S., 134 Sandwell, D. T.. 2, 167
Rasmussen, 1., 181 Sanford, A. R., 189,260,286
Ravaut, P., 78 Sass, 1.,260,273-274,276,279,285,293
Reagan, M. K, 292 Satake, K., 186-187
Reasenberg, P., 261 Sato, H., 146, 180,186, 192,218,261-265,269,272-274,
Reavy, R. 1., 311 279
Reches, Z., 235 Sawyer, E.w., 294, 299, 309
Redberger, P. 1., 330 Scarfe, C. M., 271, 373-374
Reddy, J. N., 164 Schatz.T, E, 275
Reid, I., 56, 69, 103, 106, 135, 143 Schilling, 1. G., 4, 14
Reid, M. R., 218 Schubert, G., 2, 20,163,198,217,308,372
Reiter, M., 273-274 Schwindinger, K. R., 20
Reymer, A., 217 Sclater, 1. G., 272 - 274
Ribe, N. M.. 8, 20-21, 33, 57-59, 307 Scott, D. R.. 7, 56, 59, 67, 224, 243, 307-308, 371
Richards, M. A., 2, 7, 14 Scott, D. S., 164
Richards, R., 225 Secor, D. T., 281
Richet, P., 31 Segall, P., 251
Richey, 1. E., 311 Sekiguchi, S., 186-187
Richter, EM., 20, 46, 56, 58, 304, 306 Selverstone, J., 210- 211
Riley, G. N., 20, 35, 38, 51, 374 Sempere,1. -C., 151
Rinehart, E. 1., 260 Sengun, M. Z., 297
Ringwood, A., 273 Shankland, T. J.. 285
Rivalenti, G., 280 Shaw, G. H., 115
Roberts, 1. L., 248 Shaw,H. R.. 57, 242,297,331,340,346
Roberts, P. M., 285 Shen, Y., 56, 74
Robertson, E. c, 260 Sherrod, D. R., 311
Roeder, P. L., 23, 113 w.,
Shervais,1. 280
Rohr, K M., 143 Shimazaki, K, 186
Rosenbauer, R. 1., 162 Shimizu, N., 374
Rosendahl, B. R., 135, 143 Shirley, D. N., 307
Rossman, G. R., 213 w.,
Shreyer, 214
Rothery, D. A., 112 Sigurdsson, S., 135
Rothman, D. H., 334 Sih, G. C., 243
 Author Index 387
Simmons, G., 10,275 Tharp, M., 122-123
Simpkins, P. G., 307 Thelin, G. P., 267, 269-271, 275
Singer, B. S.. 312 Thigpen, L., 32
Sinha, M., 143 Thompson, A. B., 214, 295
Sinton, 1. M., 102, 108, 144, 146, 155, 173 Thornber, c., 373
Skelhorn, R. R., 322-323, 330 Thurber, C. H., 182
Sleep, N. H., 4,14,57-58,103,105, 1000, 135, 139, 141- Tichelaar, B. w., 213
143,147,159,162,171,192,198,224-225,252,279- Tikoff, B., 310
280,307,371 Toksoz, M. N., 192, 198
Small, c., 167 Tolstoy, M., 170
Smewing,1. D., 103, 130, 141, 144, 166 Toomey, D. R., 74, 106-107, 110, 143, 152-155
Smith, M. K., 304, 311 Touloukian, Y. S., 303
Smith, R. L., 266-268, 320 Tracy, R. 1., 295
Smith, S. M.. 106 Trent, D. S., 302
Sneddon, I. N., 230 Trial, A. E, 327
Snyder, W. S., 274 Tritton, D. 1., 164,201
Soller, D. R., 260 Tryggvason, E., 135
Solomon, S. C., 74,132,143,154-155 Tsuchiyama, A., 374
Sotin, C., 56. 63, 66 Tullis, 1., 299
Southern. J. H., 330 Turcotte, D. L., 10,20,57, 163, 198,224,243,308,342-
Spakman, w., 183 343,346,371-372
Sparks, D. w., 57. 59-60, 63, 65, 70, 74,161 Tumer.J, S., 299, 305-306
Sparks, R. S. 1., 99, 113-115,300,305-306,322 Tuttle, O. E, 294
Spence, D. A., 224, 243, 342-343, 346
Spencer, K. 1., 2
Spera, E 1., 236, 300, 322, 327, 331 Urn, 1., 182
Spiegelman, M., 57-58, 61, 218 Umino, N., 185-186,218
Spudich, P., 144 Uyeda, S., 180, 198
Stanley, W. D., 285
Stebbins, 1. E, 10
Stefanick, M., 199 Valasek, P. A., 293
Stephens, W. E., 296 Valentine, G. A., 320
Stevenson, D. J., 7, 56-57, 59, 67,180,192,198-199,202- Van den Beukel, 1., 200
203,205-211,214-216,224, 243, ~07-308, 357, 371 VanderHilst,R., 180,183, 185
Stevenson,1. M., 153 Vaughan, P. 1., 20, 47
Stockman, H. w., 323, 334;,..336 Vera, E. E., 106-108, 115, 135, 142-143, 147, 152
Stolper, E., 99, 113-114,214,217 Vidal, P., 294
Stommel, H., 114 Vielzeuf, D., 294-296
Stoyanov, S., 243-244 Vigneresse,1. -L., 310
Su, W.,56,67 Viskanta, R., 301
Sudo,A.,213 Vogel, T. A., 300, 322-324, 336, 349
Sugimura, A., 198 Von Bargen, N., 38, 279
Suyehiro, K., 185 Voshage, H., 293
Suzuki, S., 191 Vrolijk, P., 211, 213
Swain, M. V., 243
Swanson, D. A., 135
Waff, H. S., 20-21, 23, 38, 47, 49, 57, 212, 278, 279, 285
Walck, M. C., 186
Tada,H.,248 Walker, D., 99, 113-114
Takada, A., 225,230,238,243-244,251-254,281 Walker, G. P. L., 322-323, 330
Takahashi, E., 8, 10,261,263-264,271,273,278 Wallace, M. E., 216
Takahashi, H., 251 Walsh,1. B., 135,281
Tanimoto, T., 3 Walther,1. v.. 209
Tapponier, R., 168 Wang, X., 153-154
Tarduno,1. A., 2, 3 Ward,R. W., 185,261,286
Tatsumi, Y., 192, 198,208,213-214,218 Wannamaker, P. E., 260, 286
Taylor, G. I., 357 Watson, E. B., 20, 57, 208, 210-211,217,374
Taylor, H. P., 163,213 Watson, S., 7
Taylor, S. R., 260, 265, 294 Weaver, C. S., 181
Tepper, 1. H., 293 Weertman, 1., 224, 231, 243, 251, 280-281
Teyssier, c., 310 Weiblen, P. w., 135
388 Author Index

Weill, D. E, 113,297 Wright, T. L., 135

Weinberg, R. E, 309 Wunder, B., 214
Weinstein, S. A., 6 Wyllie, P. 1., 206, 211, 216, 295-296, 298,303
Weissel,1. K., 105
Wells, P. R. A, 300
Westergaard, H. M., 248
Yamamoto, A, 189
White, A 1., 292
Yamauchi, K., 191
White, D. 1., 143
Yardley, B. W. D., 295
White, 1. L., 330-331, 333, 336
Yoder, H. S., 300, 322
White, R. S., 2, 69, 286
Yokobori, T., 243, 245
Whitehead, 1. A, 112, 131,224,243,355.357-358
Young, C. T., 286
Wickham, S. A, 295, 297, 306-307
Young, C. Y., 185,261,286
Wiederhom, S., 24
Wilband, 1. T.. 322
Younker, L.w., 300, 323
Yuen, D. A., 200
Wilcock, W. S. D., 153
Wildemuth, C. R., 297
Williams, M.c. 297, 330
Wilshaw, T. R.. 234-235, 243 Zant, G., 274
Wilshire, H., 280, 285-286 Zhang, Y. S., 3
Wilson, D. S., 56, 153-154, 162 Zhao, D., 180-183,217
Wohletz, K. H., 320 Zingg, A, 293
Wolf, M. B., 211, 295, 298 Zucca,J.J., 101, 128, 185-186, 188
WOrle!,R., 200, 211
 Geographical Index

Alaska, 181 Columbia River Plateau. U.S., 270, 272

Albuquerque volcanic center, New Mexico, 266-268 continental flood basalts on, 2
Amboy-Pisgah volcanic center, California, 266-268 Coso-Indian wells region, California, 186
Andes mountains, South America, 310 Coso volcanic center, California, 266-268, 277,286
Arizona, quaternary volcanism in, 266-268
Aspen volcanic center, Colorado, 266-268
Atlantis Fracture Zone, 73 Deadman Dome, California, 349-350
Azores, low velocity region, 3 Deccan Plateau, India
age relationships in, 5
continental flood basalts of, 2
Basin and Range province, U.S., 260, 270, 272-274, 276-
277,285,293
Bay of Islands. Newfoundland, Ophiolite complex, III Eagle Mountains volcanic center, California, 266- 268
Berufjordur flow, Iceland, 322, 327 East Ogorman Fracture Zone, 106
Big Pine volcanic center, California, 266-268 East Pacific Rise, 74, 77, 99-100, 123
Blackfoot volcanic center, Idaho, 266-268 compressional wave velocity structure of, 107-110, 144,
Black Rock Desert volcanic center, Utah, 266-268 146
Brazos volcanic center, New Mexico, 266-268 in situ density-depth structure of, 100-101, 114
northern section of, 143, 147
seismic structure of, 107-110, 144, 146, 149:-pO, 152
California sub-ridge magma storage in, 103-105,155
Deadman Dome, 349-350 topography and bathymetry of. 106, 123, 156-158
Glass Creek Dome, 349-350
Inyo Domes volcanic chain, 348-349
quaternary volcanism in, 266-268 Fish Creek volcanic center, Nevada, 266-268
California Coast Ranges, U.S.A., 274, 278
Cargo Muchacho Mountains volcanic center, California,
266-268 Garret Fracture Zone, at East Pacific Rise, 106
Carrizozo volcanic center, New Mexico, 266-268 Geysers-Clear Lake volcanic center, California, 266-268
Carson Desert volcanic center, Nevada, 266- 268 Gilsardalur flow, Iceland, 322
Cascade Mountains, U.S.A.. 181,266,270,273- 274,285- 286 Glass Creek Dome, California, 349-350
Cerro Verde volcanic center, New Mexico, 266-268 Glass Mountain, California, 292
Chile, Tatara-San Pedro volcano, 311 Gofar Fracture Zone, at East Pacific Rise, 106
Cima volcanic center, California, 266-268, 280 Goffs volcanic center, California, 266-268
Cinder Hill volcanic center, California, 266-268 Gondwana,S
Clarion Fracture Zone, 106 Gorda Plate, 260
Clear Lake volcanic center, California, 266-268 Great Plains. U.S., 270, 274
Clipperton Fracture Zone, on East Pacific Rise, 106
Cocos-Nazca Ridge
bathymetry of, 159 Hawaii
bouguer gravity anomaly of, 159 depths for magma storage, 104-105
Colorado, quaternary volcanism in, 266-268 in situ density-depth structure, 101, 114
Colorado Plateau, U.S., 270, 272, 274 Kilauea volcano, 10 I

389
 390 Geographical Index

Hawaii (continued) Mathematicians Seamounts, 106

low velocity region, 3 Mid-Atlantic Ridge
magma ascent by fracture flow, 242 bathymetry of, 159
Mauna Loa volcano, 101 bouguer gravity anomaly of, 159
Huchidake volcano, Japan, 190 Middle America Trench, 106
Hokkaido, Japan, 191 Mid-Pacific Mountains, 4
Mt. Lassen, California, 181
Mount Taylor volcanic center, New Mexico, 266-268
Iceland Mull complex, Scotland, 323
Berufjordur flow, 322, 327
depths for magma storage, 104-105
Gilsardalur flow, 322-323 Nakhl-Rustaq massif, Oman, 82-85, 91
in situ density-depth structure, 101, 114 Nevada, quaternary volcanism in, 266-268
Kottur composite flow, 327 New Mexico, quaternary volcanism in, 266-268
low velocity region, 3 New Zealand, North Island, 186-187
Orn61fsfjall flow, 322-323, 326 Nikko-Shirane volcano, Japan, 188-189
rhyolite + basalt mixed magmas, 322
three-dimensional plume structure, 4
Iceland Plateau, 263 Oak Ridge National Laboratory, Tennessee, 305
Idaho, quaternary volcanism in, 266-268 Obsidian Dome, California, 323-327, 349,351
Inyo Craters, California, 323-324, 329 Ocate volcanic center, New Mexico, 266-268
Inyo Domes volcanic chain, California, 348 Oman, ophiolite in, 78
Ivrea-Verbano zone, Northern Italy, 281,293-294 Ontong-Java oceanic flood basalts, 2
Izu-Bonin arc, 181 Orn61fsfjall flow, Iceland, 322-323, 326
Izu-Oshima volcano, Japan, 191 Orozco Fracture Zone, 106

Japan, 180, 184, 186, 189, 191-192

Pacific-Farallon plate boundary, 4
Jemez Mountains volcanic center, New Mexico, 266-268
Parana continental flood basalts, 2
Jornado del Muerto volcanic center, New Mexico, 266-268
Philippine Sea plate, 186
Juan de Fuca plate, 260
Pinnacate volcanic center, Arizona, 266-268
Portrillo volcanic center, New Mexico, 266-268
Prince Edward, 2
Kanab volcanic center, Utah, 266-268
Kanto-Tokai region, Japan, 186-187
Karoo continental flood basalts, 2
Kearn volcanic center, California, 266-268 Quebrada Fracture Zone, at East Pacific Rise, 106
Kerguelen oceanic plateau basalts, 2 Quemado volcanic center, New Mexico, 266-268
Kilauea volcano, Hawaii
intrusive dimensions within, 126, 129-130, 132
magma intrusion portrait, 121 Railroad Point volcanic center, Nevada, 266-268
Kjappeyrarrmili section, Gilsardalur flow, Iceland, 322-323, Raton volcanic center, New Mexico, 266-268
326 Reunion, 2
Kolob volcanic center, Utah, 266-268 Rio Grande Rift, New Mexico, 189,259,261,273-274,
Kottur composite flow, Iceland, 327 276-279,286
Krafla volcano, Iceland Rocky Mountains, U.S., 270, 272, 274
intrusive dimensions, 126, 129-130, 132
magma intrusion portrait, 119-120
Kyushu, Japan, 191 Salton Sea volcanic center, California, 266-268, 274-275
San Andreas fault, California, 274
San Bernardino volcanic center, Arizona, 266-268
Lanzo, Italy, 281 San Carlos, Arizona, 37-41,266-268
Lava Mountains volcanic center, California, 266-268 San Francisco volcanic center, Arizona, 266- 268
Leucite Hills volcanic center, Wyoming, 266-268 Scotland, Mull volcanic complex, 323
Long Valley volcanic center, California, 188,266-268,277, Sentinel-Arlington volcanic center, Arizona, 266-268
285-286,312,319-320,323-325 Shatsky oceanic flood basalts, 2
Lunar Crater volcanic center, Nevada, 266-268 Shimada seamount, 106
Siberia, continental flood basalts, 2
Sierra Nevada mountains, U.S., 270, 274
Manihiki oceanic plateau basalts, 2 Snake River Plain, U.S., 274, 285, 293
Maqsad massif, Oman, 78-81, 90 Sonora Pass volcanic center, California, 266-268
 Geographical Index 391
Springerville volcanic center, Arizona, 266-268 Valu Fa Ridge, and Lau Basin, 104-105, 123
Saint George volcanic center, Utah, 266- 268

Wadi Fayd, Oman, 88

Tatara-San Pedro volcano, Chile, 311 Wadi Tayin massif, Oman, 93
Tehuantepec Fracture Zone, at East Pacific Rise, 106 Wilkes Fracture Zone, at East Pacific Rise, 106
Tethys, plate boundary reconstruction, 5 Wyoming, quaternary volcanism in, 266-268
Timber Mountain volcanic center, Nevada, 266-268 Wyoming Basin, U.S., 270-274
Tohoku region, Japan, 184-185, 189, 192
Tokachi volcano, Japan, 191
Tristan da Cunha, 2
Yellowstone, Wyoming, 2, 320
low velocity region, 3
Troodos, Cyprus, ophiolite complex, III Yellowstone Plateau volcanic center, Wyoming, 266-268,
Truckee-Donner Pass volcanic center, California, 266-268 277, 285, 293

Ubehebe volcanic center, California, 266-268 Zabargad Island, Oman, 77

Utah, quaternary volcanism in, 266-268 Zuni-Bandera volcanic center, New Mexico, 266-268
This page intentionally left blank
 Subject Index

Ablation, of slab-wedge interface, 200-201 Compaction. of partial melt, 307 -308

Adiabat, in relation to U.S. geotherms, 271 Complex variables, in crack stress analysis. 248-251
Amphibole. formation in subduction zones, 214 Compliance, of a partial melt, 32
Amphibole, role in subduction zone magmatism, 206, 208, Compliance spectra, of partial melts. 32
213-215 Compression, and four-point beam bending, 22
Anelasticity structure, upper mantle of U.S.A., 265, 271 Computational grid, 3-D seismic structure, 182
Attenuation Conduction, and thermal boundary layers, 7
in partially molten mantle, 34-35 Conduit
and melt migration, 32 dissolution instabilities in, 355
Young's modulus determination of, 24-25, 30 in plume core, 7
Attenuation band, of partial melts, 32 wall temperatures of, 363
Conservation of energy, 164, 20 I, 30 I, 336·
Conservation of mass, 164, 20 I, 30 I, 336
Basalt, underplating of, 279-285, 293-294, 299-306 Conservation of momentum. 164, 20 I, 30 I, 336
Basalt + rhyolite mixed magmas, 321-326 Conservation of volumetric flux, 362
Basaltic underplating, and magma ascent, 279-285 Continental flood basalts
Bouguer gravity anomaly. of mantle at mid-ocean ridges, Columbia River Plateau, 2
71-74 Deccan Plateau, 2
Boundary conditions Karoo, 2
crack-normal stress differentials, 284-285 Parana, 2
decompression melting, 9 Siberia, 2
on momentum in subduction zones, 202 Contractancy, defined, 102
and parallel interacting cracks, 249 Convection
on thermal energy in subduction zones, 202 in Earth's mantle, 7-16
Boundary layer in mantle beneath ridges, 63-65
along dike walls, 337 in magma chambers, 300-306
cold thermal, 9, 165.363 of melt in porous media, 306-307
decompaction in, 58 and partial melting, 7-16
thermal, 363 Core-annular flow
Boussinesq approximation, 201-202 in lattice-gas simulations, 334
Boussinesq fluid, 8 in two-fluid mixtures, 334
Buoyant mantle flow, three-dimensional structure of, 63-65 Crack
Burgers solid, and partial melting, 25 . buoyancy-driven, 281-282
Burgers solid viscoelastic model, 31 propagation velocity in magma transport, 229-230
shape of in magma transport, 230-234, 281-282
Crack coalescence, conditions for, 251-252
Cobalt, in synthetic olivine, 19,23 Crack interactions
Capillarity, of melt phase, 20 and magma ascent, 241
Carbon dioxide, and mixed volatiles in subduction zones, 210 mechanics of, 243-244, 245-248
Chemical corrosion, 371-377 parallel offset, 248-251
Chemical dissolution, 371-377 Crack propagation, and magma transport, 223-225
Chlorite, stability of in subduction zones, 214 Creep, flexural, 25
Coextrusion. of 2 polymers, 331-334 Crustal production, peak rates, 12

393
394 Subject Index

Crustal strain, 165-166, 167 Emulsions, magmatic, 321

Crustal volume, oceanic, and plume melting, 12 Encapsulation, in multi-phase flow, 329-330
Continuous compliance distribution, 34 Energy
conservation of, 164,201,301,336
interfacial, 21
D'Arcy's Law, in partial melts, 33 solid-liquid interfacial, 21
Decompaction Energy loss, and attenuation in partial melts, 32
in boundary layers, 58, 61 Enthalpy, 8
length scale, 59 activation, for anelastic creep, 31
Density-depth gradients, and magma neutral buoyancy, activation, for creep in partial melts, 28
115-117 Enthalpy function
Density-depth relations at melting temperature, 8
East Pacific Rise, 99-101 and saturation melt fraction, 8
Hawaii,IOI Equilibrium depths, for ridge magma chambers, 104-105,
Iceland, 101 132-133
mid-ocean ridges, 99 -I 00 Equivalent viscosity, 339
Diapirs Expansion coefficient
partial melting within, 7 due to melting, 8
serpentine-rich, 211 thermal,8
solitary, 7
solitary axisymmetric, II
Diapirism
Finite difference methods, and decompression melting, 9
as crustal magma ascent mechanism, 308-309
Finite-element grid, in subduction zone modeling, 202
mantle, beneath ridges, 63-65, 78-79, 82
Flexure, four-point
Differentiation, and melt density, 112-118
and partial melting, 21-22
Diffusion, and Newtonian rheology, 25
and stress distribution, 22
Diffusive Peclet number, 374-375
Flow
Dihedral angle, in partial melt, 21
resistance to, 356
Dike flow
thermally-eroded,362-364
boundary layer along wall, 337
two-dimensional instabilities in, 365-368
flow within, 337 -344
two-dimensional through slot, 369-370
and magma chamber withdrawal, 347 -348
Flows, composite, 321-325
volume flux of magma within, 339
Fluid instability
cross-sectional aspect ratios, 127- 128
experimental simulations, 358-377
Dike formation
oscillatory, 364-365
and height variations, 129-130
Rayleigh-Taylor type, 357
height-width relationships, 126-128
Saffman-Taylortype, 357
and lateral intrusion, 118-122, 122-125
Fluid velocity-pressure drop relations, 363-366
and length variations, 131
Fourier transform, of compliance spectra, 32
and neutral buoyancy control, 118-122, 122-125
Fractional crystallization, and melt density, 112-118, 119
and width variations, 130-131
Fractional melting, 8
Dike tips, interacting, 250
Fractionation density, 113-114
Dike width, and lubrication layers, 341
Fracture propagation, and lubricated dike flow, 344
Diking, as crustal ascent mechanism, 308-309
Frictional heating, of slab + mantle wedge interface, 200
Dikes
composite, 321-325
diabase, in Oman, 78, 80, 81, 91
heights in ascent regions, 280-281 Geoid, I
microgabbro, in Oman, 85 height and decompression melting, II, 13, 15
pyroxenite, in Oman, 84, 90-91 Geomagnetic reversals, 4
widths in ascent regions, 280-281 Geomagnetic reversals, and oceanic magmatism, 6
Dikes and veins, in melt generation region, 279 Geotherm
Dilatational deformation, of melting matrix, 20 Basin and Range province, U.S., 276
Dimensional analysis, of magma fracture, 228-229 Rio Grande Rift, U.S., 276
Dissolution, instabilities in conduits, 355 Geotherms
Domes, composite, 321-325 Conductive, of western U.S., 271-273
derived from heat flow, 276
Global trend, in ridge geochemistry, 173-174
Elastic moduli, bimodulus interface, 282-285 Granitic melt, liquid fraction-distance-time relations, 308
Elastic modulus, Young's, 22 Granodiorite, viscosity-temperature behavior, 298
 Subject Index 395
Heat, sensible at melting temperature, 8 Magma reservoir structure, East Pacific Rise, 105-110, 139,
Heat capacity 141-143,145,147-149,160
at constant pressure, 8 Magnesium, in synthetic olivine, 19,23
basalt, at constant P, 302 Magma
basalt, variation with T, 302 mixed basalt + rhyolite, 321-326
Heat flow mixed rhyodacite + rhyolite, 323-326
and decompression melting, 12 volume flux in dike, 339
geotherms derived from, 276 Magma accumulation, by magma fracture, 252-253
time series due to partial melt, 14, 16 Magma ascent
of western conterminous U.S., 273-275 general aspects of, 161-162,223,234,236,279-285
Homologous temperature, 261 crack interactions during, 241
Hydrothermal cooling, 131-133, 162-164 by crack propagation, 223 - 225
Hydrothermal system, in relation to horizon of neutral buoy- discontinuous nature of, 280-281
ancy, 131- 133 into region of basaltic underplating, 279-285
Hydrous fluids, transport in subduction zones, 210 and tectonic regime, 309-312
Hydrous fluxing, into mantle wedge, 197-198,207 - 209 velocities by magma fracture, 236-238, 252
Hydrous minerals, and subduction zone magmatism, 206, Magma chamber
208,213-215 defined, 102
and a rational treatment of 2-Fluid magma flow, 346-349
stratified with mafic and silicic melts, 347 -348
Initial conditions withdrawal from, 346-349
and decompression melting, 9 Magma-induced fracture
one-dimensional thermal model of basaltic underplating, initial conditions for constant magma volume source, 225
299 boundary conditions for constant source magma flux, 225
Intrusion rates, crustal estimates, 292 crack-tip stress intensity factor, 226-227
crack velocity-width relations, 236
dike widths across a bimodulus interface, 282-283
J-tube, experimental fluid flow type, 358-359 dike width-magma flux relations, 237
dimensional analysis of, 228-229
experimental simulation of, 225-226, 244-248
geometry and source flux, 238
Krieger-Dougherty relation, 297
parallel offset cracks, 248-251
shape with constant fluid flux, 231-232, 233
shape with constant fluid volume, 230, 232
Latent heat, and melting, 8 velocity-stress intensity relations, 227, 234-235
Lattice-gas simulations, of 2-fluid flows, 334-336 velocity-volume relations, 226
Lava flows, experimental modeling, 370-373 Magma reservoir depths, as function of spreading rates,
Lava pillows, experimental, 372 170-171
Lava ripples, experimental, 373 Magma segregation, temperatures in subduction zones, 218
Lava sheet flows, experimental, 373 Magma storage, depths of
Lithosphere East Pacific Rise, 104-105, 155
continental, 271-273, 291 Hawaii, 104-105
thermal, 10,271-273 Iceland, 104-105
Local trend, in ridge geochemistry, 173-174 Mantle, and attenuation in partial melts, 34-35
Low velocity regions, in Mantle upwelling, buoyant flow beneath ridges, 61-63
Azores, 3 Mantle wedge
Hawaii, 3 hydrous fluxing of, 197, 198,207-209
Iceland, 3,4 induced flow in, 199
Japan, 179, 181, 191 water flux into, 212-213
Tristan da Cunha, 3 Marine bathymetry, East Pacific Rise, 106, 153, 154, 156-
Lubricated flows, and viscous heating, 344-346 157,159
Lubrication Mass, conservation of, 164,201,301,336
by Newtonian wall layers, 340 Matrix
by non-Newtonian wall layers, 340 velocity, 8
Lubrication equations, 336-337 viscosity of, 8
Lubrication layer development, and dike width, 341 Maxwell viscoelastic model, 32
Melt
convection of in porous media, 306-307
Magma extraction, and decompression melting, 10 hydrous, in subduction zones, 209-210
Magma reservoir, defined, 100-102 microporous topology of, 20-21
396 Subject Index

Melt tcontinued i production rates of,S, 154-158, 166

partial, in upper mantle, 259, 268 production of, 68-69, lSI. 168
retention in source regions, 278-279 predicted gravity anomaly of, 71-74
secondary generation in crust, 294- 296 topography of. 71-74, 168
segregation and trace elements, 299 Oceanic plateau basalts
variation with temperature, 296 Kerguelen, 2
Melt density Manihiki,2
and differentiation, 112-118 Mid-Pacific mountains, 2
and H,O content, 115 Ontong-Java plateau, 2
and Fe-Mg content, 112-114 Shatsky rise, 2
Melt distribution, 29-30 Olivine, 37
Melt fraction Olivine - basalt mixtures
critical. 8 deformation in, 41-42, 50-52
generalized, for western U,S.. 271-272 microstructure of. 42-49
and homologous temperature, 262 water-related species in, 49
ill situ, 7, 271-272 Ophiolite, Oman, 78
saturation, 8 Ophiolites, and East Pacific Rise structure, 144
Melt migration Oscillatory instability. one-dimensional flow. 364-365
and attenuation, 32
in flexural creep, 28
and four-point beam bending, 22 P-wave velocity structure, Tohoku region, Japan. 182
by porous flow, 57-58 Partial melt, in upper mantle, 259, 268
Melt squirt, and attenuation phenomena. 33 Partial melting
Melt extraction, efficiency of by porous flow, 69-71 general, 7, 259-268
Melting assumptions in theoretical treatments, 7
decompression generation of, 1,8-15,62 prior studies in outline, 7
experimental, in situ, 305 in thermal plumes, 7
progress of, 9 Peelet number
Melting function, depth derivative, 8 diffusive, 374-375
Melting law, univariant, 8 and fluid instabilities, 358
Melting region, sub-ridge structure of, 65-68,151-154 Peridotite
Melting temperature, 261 hydrous phase equilibria of, 206
Microcracking, and fluid-flow in subduction zones. 211 seismic velocities at P and T, 261
Micrographs, SEM and partial melting, 29 Permeability, of melting matrix, 20
Mid-ocean ridge Phase equilibria, of hydrous peridotite. 206
diapirs beneath, 79, 82 Plate boundary, reconstructions,S
reconstructions in Oman, 79, 82 Plate motion, effects on melting. 8
Mixed magmas, 321-326 Plume, mantle, 4
Mixed magma transport, at Long Valley, California, 349-351 Poiseuille flow. of magma in cracks, 236
Mohorovicic Discontinuity, 281 Polymeric flows, 330-334
Momentum, conservation of, 164, 20 I, 30 I, 336 Porosity, of melting matrix, 20
Multiphase magma flow Porous media melt migration, 57 -58
general,319-352 Positive huoyancy region, in mid-ocean ridges, 102,
and lattice-gas simulations, 334-336 107-110
Power law, and magma rheology. 339
Prandtl number, 8
Negative buoyancy region. in mid-ocean ridges, 97, 102, Pressure
107-110 applied, in partial melt. 21
Neutral buoyancy effective. in partial melt. 21
of basaltic magma, 97. 102-103 pore fluid, in partial melt, 21
horizon of, defined, 102-103 Pressure gradient, in melting matrix, 20
level of, defined, 102-103 Pressure solution, 30
Neutral buoyancy region Principal stress
evolution of, 133-134 distribution, around interacting cracks, 249-250
in mid-ocean ridges, 97 -98, 102-103, 106-110 maximum, around crack-tips, 249-250
Newtonian rheology, of partial melt, 22 maximum, trajectories of. 250
Nusselt number, variation with cooling time, 303-304 minimum, around crack-tips, 249-250
trajectories, variations at volcanoes, 253-254

Oceanic crust
and neutral buoyancy zone evolution. 133-134 Q, structure in upper mantle, 259
 Subject Index 397
Rayleigh melting, R Sheeted dike complex, and neutral buoyancy control,
Rayleigh-Taylor instability, 357 122-125
Reference frame Sheeted dikes, in ophiolites, 77, 80, R3
plastic flow in sub-ridge, 89 Simultaneous flow, of mixed magmas, 327 -330
ridge, 89 Sintering, and Co-Mg olivine + basalt sample preparation,
Region of negative buoyancy, in ophiolites, 110-111 23
Region of neutral buoyancy Solid-state flow, induced in mantle wedge, 199
defined, 102-103 Solidification, (freezing) length scale, 60
in ophiolites, 111-112 Stefan number, in basaltic underplating, 307
Region of positive buoyancy, in ophiolites, III Strain
Rheology anelastic, 27
of (Co-Mg) olivine + basalt mixtures, 19,23-32 recovery, 25
of magmatic suspensions, 297-299 and relaxation in partial melts, 27
of oceanic crust, 169-170 Strain rate
of polycrystalline olivine -i- basalt mixtures, 37, 41-42, creep, in four-point bend tests, 26
50-52 of power-law magma, 339
power law behavior for magma, 339 Stream function, 8
Rhyolite + basalt mixed magmas, 321-326 Stress components, and beam bending, 22
Stress-intensity factor, and magma fracture, 226-227, 234-
235,248
Saffman-Taylor instability, 357 Subduction zone
Scanning Electron Microscope, images of partial melt micro- Adiabatic upwelling within, 204-206
textures, 29 age of plate within, 204
Seismic anelasticity, in mantle peridotites, 263-264 dip of plate in, 204-207
Seismic attenuation slab velocity of, 202, 204, 205
general, 179, 180, 185-187, 191 thermal structure of, 203, 205, 207
of compressional waves, 264 uncertainties in modelling, 216-217
and homologous temperature, 264 volcanic front location in, 216
beneath Iceland Plateau, 264 Suspension rheology, Krieger-Dougherty relation in,
and upper mantle melting, 259 297
Seismic compressional waves, in melting peridotite, 262
Seismic ray path
general, IR2, 267-269 Tectonic regime, and magma ascent, 309-312
distribution in western U.S., 267-269 Tension, and four-point beam bending, 22
Seismic shear waves, in melting peridotite, 262 Thermal, solitary axisymmetric, Ii
Seismic Tomography, 3, 179 Thermal boundary layers, 7, 165, 363
Azores, 3 Thermal structure
Hawaii.B Basin and Range upper mantle, 259
Iceland, 3,4 generalized, of western U.S., 265-277
Tristan da Cunha, 3 Rio Grande Rift upper mantle, 259
resolution tests, IR3-184 Subduction zone, 203-205, 207
Seismic travel time residuals, 182-1 R3 Upper mantle, 259
Seismic velocity discontinuity, 182 Thermophysical properties, and decompression mantle melt-
Seismic velocity parameters, 183 ing, 10
Seismic velocity structure Time series, of geoid height, 12
East Pacific Rise, 103, 146, 149-150, 152 Topography, dynamic, above rising plume, II, 13, 15
Japanese island arc, 180-191 Trace elements, in subduction zones, 217
Seismic wave arrival time, 182-183 Trench migration, in subduction zones, 201
Seismic wave velocity Two-Fluid dynamics
in asthenosphere, 263 and magma transport, 330-352
in partial melts, 261 and volumetric flux, 342
Seismic waves, reflected, 187-189
Seismograms
Hiuchidake volcano, Japan, 190
Nikko-Shirane volcano, Japan, 188 Underplating, basaltic, 279-285,293-294,299-306
Self-lubricating flow, of mixed magmas, 329-330
Serpentine, role in subduction zones, 214
Shear deformation, of melting matrix, 20 Velocity
Shear stresses, during flow in dike, 338 of partially molten matrix, 20
Shear thickening, in power-law magmas, 340 of melt phase, 20
Shear thinning, in power-law magmas, 340 Velocity field, of subducting slab, 200
398 Subject Index

Viscoelastic models, linear, in partial melts, 31 and power-law exponents, 340

Viscoelasticity, in partial melting, 25 and two-component flow, 342
Viscosity
bulk,30
equivalent, 339 Wadati-Benioff zone (see subduction zone)
equivalent, and flux rate, 341 Water
shear, of partial melts, 30 flux, into mantle wedge, 212-213
Viscosity-Temperature coefficient, 363 free, in subduction zones, 215
Viscous heating, and lubricated flows, 344-346 Westergaard stress functions, in crack stress analysis, 248-251
Voigt-Kelvin viscoelastic model, 31
Volcanic centers, in western U.S., 266-268
Volcanic front, location in subduction zones, 216 Young's elastic modulus, 22
Volumetric flux
conservation of, 362
 International Geophysics Series

EDITED BY

RENATA DMOWSKA
Division ofApplied Sciences
Harvard University
Cambridge. Massachusetts

JAMES R. HOLTON
Department ofAtmospheric Sciences
University of Washington
Seattle, Washington

Volume 1 BENOGUTENBERG. Physics of the Earth's Interior. 1959*

Volume 2 JOSEPH W. CHAMBERLAIN. Physics of the Aurora and Airglow. 1961 *

Volume 3 S. K. RUNCORN (ed.). Continental Drift. 1962 *

Volume 4 C. E. JUNGE.Air Chemistry and Radioactivity. 1963 *

Volume 5 ROBERT G. FLEAGLE AND JOOST A. BUSINGER. An Introduction to Atmospheric Physics. 1963 *

Volume 6 L. DUFOUR AND R. DEFAY. Thermodynamics of Clouds. 1963 *

Volume 7 H. U. ROLL. Physics of the Marine Atmosphere. 1965 *

Volume 8 RICHARD A. CRAIG. The Upper Atmosphere: Meteorology and Physics. 1965 *

Volume 9 WILLIS L. WEBB. Structure of the Stratosphere and Mesosphere. 1966*

Volume IO MICHELE CAPUTO. The Gravity Field of the Earth from Classical and Modern Methods. 1967 *

Volume 11 S. MATSUSHITA AND WALLACE H. CAMPBELL (eds.). Physics of Geomagnetic Phenomena. (In two volumes.)
1967*

Volume 12 K. YA. KONDRATYEV. Radiation in the Atmosphere. 1969*

Volume 13 E. PALMEN AND C. W. NEWTON. Atmospheric Circulation Systems: Their Structure and Physical Interpretation.
1969*

Volume 14 HENRY RISHBETH AND OWEN K. GARRIOTT. Introduction to Ionospheric Physics. 1969*

Volume IS C. S. RAMAGE. Monsoon Meteorology. 1971 *

*Out of Print.

399
400 International Geophysics Series

Volume /6 JAMES R. HOLTON. An Introduction to Dynamic Meteorology. 1972*

Volume /7 K. C. YEH AND C. H. Ltv. Theory of Ionospheric Waves. 1972*

Volume /8 M. I. BUDYKO. Climate and Life. 1974*

Volume 19 MELVIN E. STERN. Ocean Circulation Physics. 1975

Volume 20 J. A. JACOBS. The Earth's Core. 1975*

Volume 21 DAVID H. MILLER. Water at the Surface of the Earth: An Introduction to Ecosystem Hydrodynamics. 1977

Volume 22 JOSEPH W. CHAMBERLAIN. Theory of Planetary Atmospheres: An Introduction to Their Physics and Chemistry.
1978*

I'olume 23 JAMES R. HOLTON. An Introduction to Dynamic Meteorology, Second Edition. 1979*

Volume 24 ARNETT S. DENNIS. Weather Modification by Cloud Seeding. 1980

Volume 25 ROBERT G. FLEAGLE AND JOOST A. BUSINGER. An Introduction to Atmospheric Physics, Second Edition. 1980

Volume 26 Kuo-NAN LIOL. An Introduction to Atmospheric Radiation. 1980

Volume 27 DAVID H. MILLER. Energy at the Surface of the Earth: An Introduction to the Energetics of Ecosystems. 1981

Volume 28 HELMUT E. LANDSBERG. The Urban Climate. 1981

Volume 29 M. I. BUDYKO. The Earth's Climate: Past and Future. 1982

Volume 30 ADRIAN E. GILL. Atmosphere-Ocean Dynamics. 1982

1'oil/me 3/ PAOLO LANZANO. Deformations of an Elastic Earth. 1982*

Volume 32 RONALD T. MERRILL AND MICHAEL W. McELHINNY. The Earth's Magnetic Field: Its History, Origin. and
Planetary Perspective. 1983

Volume 33 JOHN.S. LEWIS AND RONALD G. PRINN. Planets and Their Atmospheres: Origin and EvoluIion. 1983

Yolume 34 ROLE MEISSNER. The Continental Crust: A Geophysical Approach. 1986

Volume 35 M. U. SAGITOV, B. BODRt, V. S. NAZARENKO, AND KH. G. TADZHIDINOV. Lunar Gravimetry. 1986

Volume 36 JOSEPH W.CHAMBERLAIN AND DONALD M. HUNTEN. Theory of Planetary Atmospheres: An Introduction to Their
Physics and Chemistry, Second Edition. 1987

Volume 37 1. A. JACOBS. The Earth's Core, Second Edition. 1987

Volume 38 1. R. ApEL. Principles of Ocean Physics. 1987

Volume 39 MARTIN A. UMAN. The Lightning Discharge. 1987

Volume 40 DAVID G. ANDREWS, JAMES R. HOLTON, AND CONWAY B. LEOVY. Middle Atmosphere Dynamics. 1987

Volume 41 PETER WARNECK. Chemistry of the Natural Atmosphere. 1988

Volume 42 S. PAL ARYA. Introduction to Micrometeorology. 1988

Volume 43 MICHAEL C. KELLEY. The Earth's Ionosphere. 1989

Volume 44 WILLIAM R. COTTON AND RICHARD A. ANTHES. Storm and Cloud Dynamics. 1989

Volume 45 WILLIAM MENKE. Geophysical Data Analysis: Discrete Inverse Theory, Revised Edition. 1989

Volume 46 S. GEORGE PHILANDER. El Nino. La Nina, and the Southern Oscillation. 1990

Volume 47 ROBERT A. BROWN. Fluid Mechanics of the Atmosphere. 1991

 International Geophysics Series 401
Volume 48 JAMES R. HOLTON. An Introduction to Dynamic Meteorology, Third Edition. 1992

Volume 49 ALEXANDER A. KAl:FMAN. Geophysical Field Theory and Method. Part A: Gravitational, Electric, and Magnetic
Fields. 1992. Part B: Electromagnetic Fields 1. 1994. Part C: Electromagnetic Fields II. 1994

Volume 50 SAMUEL S. BCTCHER, GORDON H. ORIANS, ROBERT 1. CHARLSON, AND GORDON V. WOLFE. Global
Biogeochemical Cycles. 1992

Volume 5/ BRIAN EVANS AND TENG-FONG WONG. Fault Mechanics and Transport Properties in Rock. 1992

Volume 52 ROBERT E. HUFFMAN. Atmospheric Ultraviolet Remote Sensing. 1992

Volume 53 ROBERT A. HOUZE, JR. Cloud Dynamics, 1993

Volume 54 PETER V. HOBBS. Aerosol-Cloud-Climate Interactions. 1993

Volume 55 S. J. GIBOWICZ AND A. Kuco. An Introduction to Mining Seismology. 1993

Volume 56 DENNIS L. HARTMANN. Global Physical Climatology. 1994

This page intentionally left blank