Egs 041

JOURNAL OF PETROLOGY VOLUME 53 NUMBER 10 PAGES 1999^2026 2012 doi:10.
1093/petrology/egs041
Melt Segregation in Deep Crustal Hot Zones:

a Mechanism for Chemical Differentiation,
Crustal Assimilation and the Formation
of Evolved Magmas
Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

J. M. S. SOLANO1*, M. D. JACKSON1y, R. S. J. SPARKS2, J. D. BLUNDY2
AND C. ANNEN2
1
DEPARTMENT OF EARTH SCIENCE AND ENGINEERING, IMPERIAL COLLEGE, LONDON SW7 2AZ, UK
2
SCHOOL OF EARTH SCIENCES, UNIVERSITY OF BRISTOL, WILLS MEMORIAL BUILDING, QUEENS ROAD,
BRISTOL BS8 1RJ, UK
RECEIVED JANUARY 28, 2011; ACCEPTED MAY 18, 2012

ADVANCE ACCESS PUBLICATION SEPTEMBER 9, 2012
Mantle-derived basaltic sills emplaced in the lower crust provide a broadly granitic in composition. Chemical differentiation is therefore
mechanism for the generation of evolved magmas in deep crustal hot driven by melt migration, because the melt migrates through, and
zones (DCHZ).This study uses numerical modelling to characterize chemically equilibrates with, partially molten rock at progressively
the time required for evolved magma formation, the depth and tem- lower temperatures. Crustal assimilation occurs during partial melt-
perature at which magma formation occurs, and the composition of ing, and mixing of crustal and residual melt occurs when residual
the magma. The lower crust is assumed to comprise amphibolite. In melt migrates into the partially molten crust, yielding chemical sig-
an extension of previous DCHZ models, the new model couples heat natures indicative of a mixed crustal and mantle origin. However, re-
transfer during the repetitive emplacement of sills with mass transfer sidual melt is volumetrically more significant than crustal melt,
via buoyancy-driven melt segregation along grain boundaries.The re- except at the highest emplacement rates. Contamination of crustal
sults shed light on the dynamics of DCHZ development and evolu- melt by residual melt from basalt crystallization appears to be an in-
tion.The DCHZ comprises a mush of crystals plus interstitial melt, evitable consequence of melt segregation in DCHZ, and can explain
except when a new influx of basaltic magma yields a short-lived the mixed crust^mantle origin of many granites.
(20^200 years) reservoir of melt plus suspended crystals (magma).
Melt segregation and accumulation within the mush yields two con-
trasting modes of evolved magma formation, which operate over time- KEY WORDS: intrusions; mush; compaction; anatexis; assimilation
scales of c. 10 kyr^1 Myr, depending upon emplacement rate and and fractional crystallization
style. In one, favoured by emplacement via over-accretion, or emplace-
ment at high rates, evolved magma forms in the crust overlying the
intruded basalt sills, and is composed of crustal partial melt, and re- I N T RO D U C T I O N
sidual melt that has migrated upwards out of the crystallizing Continental and mature island arc crust has an estimated
basalt. In the other, favoured by emplacement via under- or andesite to dacite composition and is vertically stratified
intra-accretion, or by emplacement at lower rates, evolved magma from a less evolved, mafic lower crust to a more evolved,
forms in the intruded basalt, and the resulting magma is composed felsic upper crust (Rudnick & Fountain, 1995). A key ques-
primarily of residual melt. In all cases, the upward migration of tion, central to understanding the evolution of continental
buoyant melt yields cooler and more evolved magmas, which are and mature arc crust, concerns the origin of the more
y
Corresponding author. E-mail: [email protected] ß The Author 2012. Published by Oxford University Press. All
*Present address: Galson Sciences Limited, 5 Grosvenor House, rights reserved. For Permissions, please e-mail: journals.permissions@
Melton Road, Oakham LE15 6AX, UK. oup.com
JOURNAL OF PETROLOGY VOLUME 53 NUMBER 10 OCTOBER 2012
evolved, intermediate to silicic magmatic rocks that dom- extraction of the evolved partial melt leaves behind mafic
inate the upper crust, including the voluminous granite residues (restites), which may also contribute to mafic
batholiths (diorites, tonalites, granodiorites and granites) lower crust. Cumulates and restites in the lower crust can
and evolved volcanic rocks (andesites, dacites and rhyo- eventually delaminate into the underlying mantle (e.g.
lites). Most researchers now agree that evolved magmas Kay & Mahlburg-Kay, 1991; Seber et al., 1996; Ducea &
(granitic, sensu lato) form either by differentiation of pri- Saleeby, 1998b; Jull & Kelemen, 2001; Muntener et al.,
mary, mantle-derived magmas during cooling and crys- 2001; Zandt et al., 2004). Distinctive trace element signa-
tallization (e.g. Gill, 1981; Grove & Kinzler, 1986; tures (e.g. high Sr/Y) in evolved magmatic rocks are con-
Musselwhite et al., 1989; Rogers & Hawkesworth, 1989; sistent with high-pressure differentiation in which garnet
Muntener et al., 2001) or by partial melting of older crustal is a crystallizing phase (e.g. Smith & Leeman, 1987;
rocks (e.g. Rushmer, 1991; Atherton & Petford, 1993; Feeley & Davidson, 1994; Feeley & Hacker, 1995; Garrison

Tepper et al., 1993; Wolf & Wyllie, 1994; Rapp & Watson, & Davidson, 2003; Alonso-Perez et al., 2009).
1995; Petford & Atherton, 1996; Chappell & White, 2001). The crust is not normally partially molten and thermal
These processes can occur together to generate magmas models suggest that heat must be supplied to produce par-
of hybrid origin through assimilation and mixing (e.g. De tial melting. The intrusion of hot, mantle-derived primary
Paolo, 1981), with heat and volatiles liberated from magma into the lower crust is one likely mechanism (e.g.
mantle-derived magmas during crystallization triggering Hodge, 1974; Huppert & Sparks, 1988; Bergantz, 1989;
partial melting of the surrounding crust (Petford & Bergantz & Dawes, 1994; Petford & Gallagher, 2001;
Gallagher, 2001; Annen & Sparks, 2002). Assimilated or Annen & Sparks, 2002; Dufek & Bergantz, 2005; Annen
partially melted crust may be unrelated petrogenetically et al., 2006); other mechanisms that have been proposed in-
to the primary magmas and possess a distinct trace elem- clude tectonic burial of radiogenic sediments (England &
ent and isotope geochemistry; conversely, partial melting Thompson, 1984) and frictional heating associated with
can occur in crust formed from earlier intrusions of pri- lithospheric faults (Deves et al., 2010). Cooling and crystal-
mary magma, in which case the partial melt and parental lization of mantle-derived magmas transfers volatiles
basalt magma may have strong geochemical affinities (e.g. (principally H2O) into the surrounding crust, which af-
Heath et al., 1998; Dungan & Davidson, 2004). Evidence fects the electrical conductivity structure, consistent with
for crustal assimilation and mixing of melts and crystals geophysical data from the Cascades (Stanley et al., 1987)
from different sources is common in a wide range of tec- and the central Andes (Brasse & Soyer, 2001; Lezaeta &
tonic settings (Grove et al., 1988, 1997; Musselwhite et al., Brasse, 2001), and low-velocity zones observed in seismic
1989). data from the central Andes (Chmielowski et al., 1999;
A key question is the depth at which chemical differenti- Zandt et al., 2003) and Australia (Drummond & Collins,
ation occurs. The existence of shallow, sub-volcanic 1986) are interpreted to represent regions of partial melt.
magma chambers, within which differentiation occurs These observations have motivated the development of a
during cooling and crystallization, is well documented. model for the origin and evolution of intermediate to silicic
However, there is geological, geochemical and geophysical magmas, in which chemical differentiation begins in the
evidence to suggest that differentiation also occurs at lower crust and results from the intrusion of hot,
depth in the crust and, indeed, that the lower crust may mantle-derived basaltic magma (Fig. 1). Cooling and crys-
be the dominant source of evolved magmas in regions tallization of the basaltic magma yields evolved residual
termed deep crustal hot zones (DCHZ; Annen et al. 2006) melt, and heat and volatiles released by the cooling bas-
or meltingâssimilation^storage^hybridization (MASH) altic magma cause partial melting of the surrounding
zones (Hildreth & Moorbath, 1988). Seismic velocity pro- crust, yielding crustal melt. Throughout the paper we use
files (Holbrook et al., 1999; Krawczyck et al., 1999; the term melt to denote crystal-free liquid, and magma to
Stratford & Stern, 2006; Majdanski et al., 2007; Voss & describe melt plus suspended crystals. The residual and
Jokat, 2007; Sandrin & Thybo, 2008; Xia et al., 2010), gravi- crustal melts migrate and collect to form evolved
metric data (Kimbell & Richards, 2008) and xenolith stu- magmas, which can then migrate upwards out of the
dies (Cull et al., 1991; Ducea & Saleeby, 1998a; Klugel source region, via dykes, faults and fractures, to shallower
et al., 2005) indicate the presence of large volumes of mafic crustal levels. The magmas may move directly to their
lower crustal rocks in many areas worldwide and in vari- final emplacement level and be intruded to form plutonic
ous tectonic contexts. Differentiation of primary basaltic bodies, or form shallow magma chambers and differentiate
magma to yield batholithic quantities of silicic magma further prior to eruption. Conversely, the magmas may
must form large volumes of dense mafic and ultramafic cu- stall and differentiate in one or more intermediate magma
mulates and, as these are not observed in the upper crust, chambers. Further differentiation at shallower crustal
they are a strong candidate to explain the geophysical ob- levels can yield polybaric geochemical signatures in both
servations. Likewise, partial melting of crustal rocks and intruded and erupted magmas, although evidence of
2000
SOLANO et al. MELT SEGREGATION IN DCHZ

Fig. 1. Schematic illustration of a deep crustal hot zone (DCHZ), feeding magma chambers and intrusions at shallower levels in the crust. We
focus on melt generation and segregation processes that occur in the DCHZ, when melt is first formed following the intrusion of hot,
mantle-derived basaltic sills. These are shown injected at 30 km depth. We do not attempt to describe the later evolution of magma composition
during ascent. The rectangle shows the location of Figure 2. Modified from Annen et al. (2006).
high-pressure differentiation within DCHZ is often pre- M E LT G E N E R AT I O N A N D

served (e.g. Ariskin et al., 1995; Macdonald et al., 2000;
Reubi & Nicholls, 2004; Grove et al., 2005; Rodriguez
S E G R E G AT I O N I N D E E P C RU S TA L
et al., 2007; Ulmer, 2007). The conceptual model shown in HOT ZON E S
Fig. 1 has been proposed to explain the origin of evolved Models of heat transfer during the emplacement of
(granitic, sensu lato) magmas by numerous workers (e.g. mantle-derived basaltic sills or dykes in the lower crust
Hodge, 1974; Pitcher, 1979, 1993; Chappell, 1984; Hildreth demonstrate that large volumes of crustal and residual
& Moorbath, 1988; Miller et al., 1988; Leake, 1990; Clarke, melt may be generated in DCHZ over timescales of
1992; Atherton, 1993; Tepper et al., 1993; Brown, 1994; c. 0·1^10 Myr, depending upon the emplacement rate and
Petford, 1995; Petford et al., 2000; Annen & Sparks, 2002; composition of the intruded basalt, and the composition
Jackson et al., 2003; Dufek & Bergantz, 2005). of the crust (Hodge, 1974; Huppert & Sparks, 1988;
The physical processes in DCHZ that govern the gener- Bergantz, 1989; Bergantz & Dawes, 1994; Petford &
ation, segregation and accumulation of melt to produce Gallagher, 2001; Annen & Sparks, 2002; Dufek &
evolved, silicic magmas (broadly granitic in composition) Bergantz, 2005). These heat transfer models consider the
that can leave the source and ascend to shallower crustal thermal evolution of a DCHZ, but neglect the migration
levels are investigated in this study, which focuses on pro- of melt once it has formed. Nevertheless, the formation of
cesses within the DCHZ and does not consider the subse- evolved magma in DCHZ requires melt migration to sep-
quent ascent of the magmas, or their further chemical arate crustal and residual melts from their solid residues,
differentiation prior to emplacement or eruption. and produce evolved magmas with bulk compositions that
2001
are different from the parent crustal protolith or primary (including the models presented here) and also consider-
basaltic magma. These evolved magmas then migrate to ations of crustal strength, which suggest that large liquid
higher structural levels. Models that consider heat transfer, bodies cannot be maintained in the crust over a long time-
but not melt migration, only partially describe the physical scales (e.g. Grosfils, 2007).
and chemical processes responsible for the formation of Because the DCHZ is in a mush state, residual and crus-
evolved magmas in a DCHZ. Our objective here is to tal melts located along grain boundaries must migrate
extend the DCHZ concept by linking their thermal evolu- relative to the solid matrix, and collect until the melt frac-
tion to the dynamic processes of melt migration and tion locally exceeds the SLT and an evolved magma forms
segregation. that can leave the source region. This process is termed
During melting and crystallization, the crystal^melt segregation. Melt segregation requires that the melt is con-
mixture undergoes profound rheological changes as the nected along grain boundaries, that there is a fluid poten-

melt fraction changes. At high melt fraction, the rheology tial gradient to drive melt flow relative to the matrix, and
of the mixture is that of a liquid containing suspended that there is space available to accommodate the melt
crystals (often termed a ‘slurry’); at low melt fraction, the (Jackson et al., 2003, 2005). The available evidence suggests
crystals form an interconnected matrix and the rheology that melt forms an interconnected network at melt frac-
of the mixture is that of a solid containing liquid along tions greater than 0·01 when the matrix grains are in tex-
grain boundaries (often termed a ‘mush’). The switch be- tural equilibrium (von Bargen & Waff, 1981) and as low as
tween these two rheological domains occurs when the 0·02 when they are not in textural equilibrium (Wolf &
melt fraction reaches the ‘solid-to-liquid transition’ (SLT) Wyllie, 1991; Lupulescu & Watson, 1999). This melt fraction
of Rosenberg & Handy (2005) or the ‘critical melt fraction’ is the ‘melt connectivity transition’ (MCT) of Rosenburg
(CMF) of van der Molen & Paterson (1979); a similar con- & Handy (2005) and the ‘liquid percolation threshold’
cept is encapsulated by the ‘rheological limit of eruptabil- (LPT) of Petford (2003). If the source rock is layered, and
ity’ (Marsh, 1981). The value of the SLT or CMF is the layers have different rheological properties, then tec-
uncertain and depends on several factors, such the crystal tonically driven deformation can provide both a potential
shape and size distribution, and the applied strain rate gradient and accommodation space. Because the layers re-
(Barnes, 1999; Petford, 2003; Costa, 2005; Rosenberg & spond differently to deformation, potential gradients form
Handy, 2005; Costa et al., 2009). At low strain rate, the between them and space is created at dilatant sites such as
available evidence suggests that the SLT most probably boudins and fractures (e.g. Brown, 1994; Sawyer, 1994;
lies in the range 0·4^0·7 (Rosenberg & Handy, 2005; Brown et al., 1995; Brown & Rushmer, 1997; Rabinowicz &
Castrucchio et al., 2010). The range of 0·4^0·6 modelled Vigneresse, 2004). Melt migration is also driven by buoy-
here is taken from Rosenberg & Handy (2005). ancy, either along grain boundaries or through a network
In this study, we address uncertainty in the value of the of fractures (e.g. McKenzie, 1984; Wickham, 1987; Petford,
SLT by modelling a range of plausible values. Shortly 1995; Petford & Koenders, 1998; Rushmer, 2001). If flow
after intrusion, the melt fraction in the basaltic magma is occurs along grain boundaries, then creep processes pro-
above the SLT; however, as has been shown in numerous vide a mechanism for changing the morphology of the
previous studies (e.g. Annen & Sparks, 2002; Jackson grains (Pharr & Ashby, 1983; Cooper & Kohlstedt, 1984;
et al., 2003, 2005; Dufek & Bergantz, 2005; Annen et al., Hollister & Crawford, 1986; Kohlstedt & Chopra, 1994),
2006; Annen, 2009, 2011) and as our results also demon- so the partially molten source rock can compact in re-
strate, cooling rapidly reduces the melt fraction to a value sponse to melt flow (McKenzie, 1984). Compaction pro-
below the SLT. Cooling is especially rapid early in the life vides space for the melt to accumulate in the source
of the DCHZ, when the temperature of the surrounding region (Jackson et al., 2003, 2005).
crust is low relative to the intrusion temperature. During Geochemical models of continental crust evolution gen-
cooling of the intruded magma, heat and volatile release erally neglect melt segregation, assuming instead one of
can cause partial melting of the surrounding crust; how- two end-member cases. In the case of fractional melting
ever, unless the solidus of the crust is significantly lower or crystallization, melt or solid is instantaneously removed
than that of the intruding basalt, bulk melting of the crust from the residual matrix or magma as it forms and the
is not possible; rather, crustal partial melt fractions are ini- bulk composition changes in response. In the case of equi-
tially below the SLT. Consequently, both residual and librium (batch) melting or crystallization, melt remains in
crustal melts lie along grain boundaries, and the DCHZ is contact with the residual crystals and the bulk composition
in a mush state except when a new influx of hot basaltic stays constant until the melt is instantaneously removed.
material generates a transient magma chamber. The The generation of mid-ocean ridge basalt (MORB) in the
DCHZ therefore comprises a crystalline mush for most of mantle has been described by polybaric partial melting
its life, rather than a largely liquid magma chamber. This models in which single batches of melt are generated over
view is supported by the results of thermal modelling a range of depths during adiabatic upwelling of the
2002
mantle (e.g. Klein & Langmuir, 1987; McKenzie & Bickle, cooler regions of the DCHZ, so cooling and crystallization
1988). These batches of melt then segregate and ascend results in the composition of the melt within the layers
through the mantle to the base of the crust, where they becoming more evolved. The depth, and hence tempera-
mix to yield MORB compositions. In this scenario, the ture, at which the melt layers form therefore plays a key
batches of melt apparently do not interact chemically with role in controlling melt composition (Rushmer & Jackson,
the mantle during ascent, which suggests that they must 2006; Getsinger et al., 2009). Furthermore, melt segregation
migrate through melt channels, pipes or fractures, rather can lead to the accumulation of large volumes of evolved
than along grain boundaries (Spiegelman & Kenyon, melt over timescales of c. 10 kyr^10 Myr, which are much
1992). The chemical compositions of the batches of melt shorter than the timescales predicted by purely thermal
are typically predicted from equilibrium melting experi- models that omit melt migration (Jackson et al., 2005).
ments, although it has recently been recognized that these However, the melt segregation models of Jackson and

experiments do not account for the depletion of the solid co-workers consider only crustal melts generated by the
phase as melting proceeds. Thus experiments have been heat from a single basaltic intrusion. The thermal model-
designed that attempt to capture the close to fractional ling results of Annen & Sparks (2002) and Annen et al.
nature of polybaric melting (e.g. Kinzler & Grove, 1992; (2006) demonstrate that the generation of large volumes
Hirose & Kawamura, 1994; Kinzler, 1997; Kushiro, 2001). of silicic and intermediate magma in a DCHZ requires re-
In the continental crust, numerous thermal models sug- peated intrusion of basaltic magma, and involves the con-
gest that the steep thermal gradient through the source tribution of both crustal and residual melt. To date, no
region will dominate the compositional evolution of melt models have been developed that describe heat and mass
and matrix; pressure variations are small because of the re- transfer in a DCHZ formed from repeated intrusion of
stricted vertical extent over which melting occurs (e.g. basalt. Such models are essential to investigate the relative
Hodge, 1974; Bergantz, 1989; Bergantz & Dawes, 1994; contribution of crustal and residual melts, and the degree
Jackson et al., 2003). If melt segregates via fractures, which to which these may interact and mix during the formation
efficiently drain melt before it has interacted with the of evolved magma. Such models are also essential to inves-
matrix, then the segregated melt compositions may ap- tigate whether the rate and/or geometry of basalt intrusion
proach the limit of fractional melting. Batches of melt have an impact on the volume and composition of evolved
from different depths may mix, yielding something akin magma, and on the timescale of magma formation. These
to a polythermal partial melting model, in which single issues have not yet been addressed by quantitative models
batches of melt are generated over a range of depths (and of melting and melt segregation in the continental crust.
therefore temperatures) during heating from below. The aim of this study is to characterize the impact of
However, if melt flows along grain boundaries, then it mi- melt segregation processes on the formation of intermedi-
grates through a temperature field that varies spatially ate and silicic magmas in a DCHZ during the repeated in-
and temporally, whilst interacting thermally and chem- trusion of hot, mantle-derived basaltic sills into
ically with the surrounding matrix. The local bulk com- amphibolite crust. We recognize that further modification
position changes as melt migrates into, and out of, of magma chemistry may occur after the magma leaves
different regions (Rushmer & Jackson, 2006; Getsinger the DCHZ (e.g. Fig. 1); however, melt must originate some-
et al., 2009); moreover, melt and matrix reach local where and we argue (consistent with numerous previous
thermodynamic equilibrium during segregation, except at workers, e.g. Hodge, 1974; Pitcher, 1979, 1993; Chappell,
very high melt flow rates or low component diffusivity in 1984; Hildreth & Moorbath, 1988; Miller et al., 1988;
the solid phase (Jackson et al., 2005). Consequently, there Leake, 1990; Rushmer, 1991; Clarke, 1992; Atherton, 1993;
is a continuous exchange of components between melt and Tepper et al., 1993; Brown, 1994; Petford, 1995; Petford &
matrix as each migrates and equilibrates at new local con- Gallagher, 2001; Annen & Sparks, 2002; Dufek &
ditions of temperature and bulk composition. It is difficult Bergantz, 2005; Annen et al., 2006; and many others) that
to see how equilibrium or fractional melting or crystalliza- most of the melt originates in the deep crust. Hence we
tion models can fully capture the evolution of melt and focus our attention on melt generation and segregation
matrix chemistry during segregation in this dynamic melt- within a DCHZ.
ing environment. We describe the generation and migration of both crus-
Models of melt segregation developed by Jackson & tal and residual melt within a DCHZ using a quantitative
Cheadle (1998) and Jackson et al. (2003, 2005) describe the model that combines the repetitive intrusion model of
migration of melt through a DCHZ via buoyancy-driven Annen & Sparks (2002) with the heat and mass transfer
flow along grain boundaries. Their results show that com- model of Jackson et al. (2003). We use the model to investi-
paction leads to the formation of high-porosity melt gate the impact of emplacement rate and style on the gen-
layers, which migrate upwards through the DCHZ. As the eration of evolved magmas, focusing in particular on the
layers migrate upwards, they move into progressively relative contribution of crustal and residual melt, the
2003
timescale of magma formation, and the composition of the modelled amphibolite crust is taken from Petford &
magma. We recognize that our model does not capture Gallagher (2001) and is based on melting experiments on
the full complexity of melt generation and segregation pro- amphibolite samples with initial, structurally bound water
cesses in the continental crust, where additional pressure contents generally 51·5 wt % at lower crustal pressures
gradients may be imposed by tectonic forces and melt (0·7^1·6 GPa) (Beard & Lofgren, 1991; Rushmer, 1991; Sen
transport may occur through fractures (e.g. Brown, 1994; & Dunn, 1994; Wolf & Wyllie, 1994; Rapp & Watson, 1995;
Sawyer, 1994; Brown et al., 1995; Petford, 1995; Petford & see Fig. 3). The melting relationship is similar to that used
Koenders, 1998; Rushmer, 2001; Rushmer & Miller, 2006). by Dufek & Bergantz (2005). We also follow Annen et al.
However, segregation always requires at least some (2006) in assuming that the intruded, mantle-derived
grain-boundary flow and buoyancy is always present to basalt contains 2·5% water and is initially at a tempera-
drive this; moreover, there is abundant evidence that com- ture of c. 1560 K (12858C). The melting relationship for the

paction occurs in a variety of settings within the continen- modelled basalt is taken from Annen et al. (2006) and is
tal crust (e.g. Wager, 1961; Irvine, 1974; Wilson et al., 1981; based on experimental data from Mu«ntener et al. (2001)
Shirley, 1986; Boudreau & Philpotts, 2002; Bachmann & and Sisson et al. (2005) (Fig. 3). Other crustal lithologies
Bergantz, 2004; Hersum et al., 2007; Tegner et al., 2009). No could be used in the model in place of amphibolite, but
other models have been developed that predict tempera- the main principles can be elucidated with amphibolite.
ture, melt fraction and melt transport rates during melt Partial melting of amphibole-bearing mafic rocks in the
generation and segregation in the crust. We argue that our deep crust has been invoked as the source of evolved
approach is a reasonable first step, and that a model that magmas in numerous studies (e.g. Pitcher, 1979, 1993;
captures at least some of the physics of melt segregation Chappell, 1984; Hildreth & Moorbath, 1988; Miller et al.,
provides useful insights that cannot be obtained from 1988; Leake, 1990; Rushmer, 1991; Clarke, 1992; Atherton,
models that omit segregation entirely. 1993; Atherton & Petford, 1993; Tepper et al., 1993; Brown,
1994; Petford & Atherton, 1996; Dufek & Bergantz, 2005).
Annen & Sparks (2002) and Annen et al. (2008) presented
M O D E L F O R M U L AT I O N the results of purely thermal models for different crustal
Repetitive sill intrusion into the lithologies.
lower crust The modelling of repetitive sill intrusions follows Annen
We model the repeated emplacement of mantle-derived et al. (2006). The initial temperature at the emplacement
basaltic sills into the lower crust following the approach of depth is c. 870 K (6008C), which is below the solidus of the
Annen & Sparks (2002) and Annen et al. (2006). We follow modelled amphibolite, so the crust is initially solid with
Annen & Sparks (2002) in defining a linear initial geother- no melt present. At t ¼ 0, the first basaltic sill is emplaced
mal gradient of 20 K km1 and initial emplacement depth (Fig. 2a and e). Further sills, each 50 m thick, are then
of 30 km (Fig. 2). We assume that each sill is emplaced in- emplaced at regular intervals to yield the chosen
stantaneously at its liquidus temperature and consider time-averaged emplacement rate. We term the region be-
only the vertical transport of heat and mass (i.e. the tween the top of the uppermost sill and the base of the
model is one-dimensional; 1-D). These assumptions have lowermost sill the basalt layer, to distinguish it from the
been justified in numerous previous models of basaltic in- over- and underlying older amphibolite crust (Fig. 2).
trusion into the lower crust; the former is reasonable The modelled sills are emplaced in one of three styles: (1)
given that the time to emplace a single sill is short com- over-accretion, in which each sill is intruded above the pre-
pared with the time over which the DCHZ thermally vious sill (Fig. 2a^d); (2) under-accretion, in which each
evolves; the latter is reasonable given the large aspect new sill is emplaced below the previous sill (Fig. 2e^h);
ratio (several tens of kilometres in plan-view; a few kilo- (3) intra-accretion, in which each sill is emplaced at the
metres in thickness) of a DCHZ (Hodge, 1974; Huppert & centre of the basalt layer. Regardless of emplacement
Sparks, 1988; Bergantz, 1989; Bergantz & Dawes, 1994; style, the material below each new sill is displaced down-
Petford & Gallagher, 2001; Annen & Sparks, 2002; wards, which approximates isostatic equilibrium.
Jackson et al., 2003, 2005; Dufek & Bergantz, 2005; Annen Estimates of the rate of basalt emplacement vary over
et al., 2006). orders of magnitude; in this study, modelled rates range
We follow Annen et al. (2006) in assuming that the lower from 0·5 to 50 mm a1, which is greater than the range of
crust is homogeneous and composed of amphibolite, 2^10 mm a1 suggested by Annen et al. (2006). The geomet-
which represents an end-member proxy for arc systems ric mean value of 5 mm a1 is typical of time-averaged
and is consistent with earlier studies of basalt intrusion magma productivity in arc settings, simplified in a 1-D
into the lower continental crust (e.g. Petford & Gallagher, geometry (Crisp, 1984). Values greater than 10^20 mm a1
2001; Jackson et al., 2003, 2005; Dufek & Bergantz, 2005; are more characteristic of transient rates associated
Annen & Sparks, 2006). The melting relationship for the with formation of large crustal magma chambers
2004
SOLANO et al.
2005
MELT SEGREGATION IN DCHZ
Fig. 2. Schematic illustration showing the behaviour of a DCHZ during the repeated emplacement of mantle-derived basaltic sills into amphibolite crust. Plots (a) and (e) show a 50 m sill that
has been emplaced at 30 km depth before cooling has taken place. The downwards displacement of the crust is indicated. The corresponding temperature (T) and melt fraction (f) profile is
shown to the right of each plot. Plots (a)^(d) show emplacement by over-accretion whereas (e)^(h) show emplacement by under-accretion. The order in which the sills are emplaced is indicated
by the numbers 1^8 to the right of each plot, with 1 being the first sill to be emplaced. Melt migrates upwards through the crystalline mush, as shown by the grey arrows, and accumulates at
the top of the emplacement region. The upper and lower boundaries of the deep crustal hot zone are indicated by the solidus temperature (Ts).

1
using the model of Jackson et al. (2003). The melt and solid
Basalt phases are assumed to be in local thermodynamic equilib-
Amphibolite rium, which is reasonable for the range of melt flow veloci-
0.8 ties considered (Jackson et al., 2005). The conservation
equations for mass, energy and momentum are the same
as those presented by Jackson et al. (2003), except that we
0.6 solve for enthalpy, rather than temperature, in the energy
Melt Fraction
conservation equation (Katz, 2008). This allows the model

Basaltic Andesite to capture melting and crystallization at the eutectic,
0.4 which occurs at constant temperature. A further improve-
Andesite ment on the model of Jackson et al. (2003) is the inclusion
of a porosity-dependent matrix viscosity (see Table 1 for

0.2 Dacite the nomenclature)
4
Rhyolite Z þ x ¼ Z ¼ Z0 fb ð1Þ
3
0 where the value of the exponent b is chosen to be 0·5
900 1000 1100 1200 1300 1400 1500 1600
Temperature (K) (Connolly & Podladchikov, 1998). It should be noted that
Fig. 3. Static melt fraction (X) vs temperature (T) relationships for the matrix viscosity is Newtonian (i.e. is independent of
basalt and amphibolite at 1GPa (after Annen et al., 2006). The basalt strain rate) because we assume the matrix deforms by
contains 2·5 wt % H2O. Grey dashed lines divide very approximate melt-enhanced diffusional creep (Cooper & Kohlstedt,
fields in which crustal melts sourced from the amphibolite, and re-
sidual melts sourced from the basalt are, broadly speaking, rhyolite
1984; Dell’Angelo et al., 1987; Ranalli, 1987). Maximum
to basaltic andesite in composition. strain rates predicted by the model are 5^7 orders of mag-
nitude smaller than the estimated maximum for which
compaction will occur by dislocation creep, rather than
(Annen, 2009). It should be noted that the time between melt-enhanced diffusion creep. Conservation of mass is
each sill emplacement and the thickness of a sill do not given by
affect the thermal evolution of the DCHZ, so long as they
@f @ @X 1 @X
are much smaller than the total duration and thickness of ¼ ½ð1 fÞws þ þ 2 ws ð2Þ
@t @z @t f @z
emplacement (Annen & Sparks, 2002).
where the final term on the right-hand side models phase
Heat transfer and melt segregation change in a reference frame which accounts for the migra-
Each sill intruded into the crust acts as a source of heat, tion of melt and matrix relative to the mixture reference
which causes partial melting in the over- and underlying frame. The contribution of this term is negligibly small
rock (which may be crust or earlier intruded sills), and over the parameter range used here. Conservation of
also as a source of residual melt as it cools and crystallizes. energy, defined in terms of enthalpy, is given by
We assume that the melt accumulates along grain bound- @h k @2 T @
aries, forming an interconnected network. The melt is ¼ þ Lf ½ð1 fÞws : ð3Þ
@t r @z2 @z
buoyant with respect to the matrix, which provides a pres-
sure gradient to drive flow (McKenzie, 1984); moreover, Conservation of momentum, including the porosity-
the matrix is able to deform in response through dependent matrix viscosity, is given by
melt-enhanced diffusional creep (Pharr & Ashby, 1983;
@ @ mf
Cooper & Kohlstedt, 1984; Dell’Angelo et al., 1987; Z0 fb ws ¼ ð1 fÞðrs rl Þg ws ð4Þ
Ranalli, 1987; Kohlstedt & Chopra, 1994). Compaction @z @z kf
causes melt to accumulate until the melt fraction reaches Permeability is given by
the SLT, whereupon a buoyant, mobile magma forms.
Such a high melt fraction cannot reside in the crust over kf ¼ ab2 fa : ð5Þ
long timescales and we assume that the magma leaves the
source region shortly after reaching the SLT, ascending to Equations (1)^(5) are strictly valid only below the SLT
shallower crustal levels through dykes, fractures or faults (i.e. where the DCHZ comprises a mush of crystals and
(e.g. Fig. 1). We explore the impact of uncertainty in the interstitial melt), but for convenience they are invoked for
SLTon our results by considering a range of values. all melt fractions. After a fully molten sill is emplaced, the
We describe the coupled processes of phase change and melt fraction is greater than the SLT for a short period,
buoyancy-driven melt migration along grain boundaries but this transient behaviour lasts only for a few time-steps
2006
Table 1: Nomenclature In ‘static’ melting (or crystallization) models and experi-

ments, there is no relative migration of melt and matrix,
so melt fraction and composition vary only with tempera-
Symbol Description Unit
ture and are directly related to each other for a given ini-
tial (bulk) rock (or magma) composition. Indeed, melt
a grain radius m fraction is often used as a proxy for melt composition in
b permeability constant none static models; for example, Fig. 3 (which was derived from
cp specific heat capacity J kg1 K1 static melting experiments) may be used to suggest that a
g acceleration due to gravity m s2 rhyolite melt in an amphibolite protolith corresponds to a
h enthalpy J 20% melt fraction. In ‘dynamic’ models, such as the model
k thermal conductivity W K1 m1 presented here, melt fraction varies with temperature, and
also in response to melt segregation; for example, a rhyolite

kf permeability m2
Lf latent heat J kg1 melt composition may accumulate to have a melt fraction
t time s much greater than 20%. The local melt composition in dy-
tm accumulation time with melt migration Myr namic models depends on temperature, but only indirectly
tnm accumulation time without melt migration Myr on melt fraction, through the influence of the latter on the
T temperature K
local bulk composition. Melt segregation changes the bulk
Ta amphibolite crystallization temperature K
composition, and hence may change the local phase behav-
Tliq liquidus temperature K
iour [see Getsinger et al. (2009) for a discussion of melting
experiments designed to capture this].
Tsol solidus temperature K
Excepting the short-lived, transient behaviour of the
ws matrix velocity m s1
intruded sill immediately after emplacement, the model is
wl melt velocity m s1
allowed to evolve until the melt fraction somewhere
X static melt fraction (degree of melting) none
within the DCHZ reaches the SLT, at which time it is
Z distance m
assumed that a mobile magma has formed that rapidly
a permeability exponent none
leaves the source region. The model is then stopped and
b viscosity exponent none the resulting temperature and melt fraction profiles are re-
Z matrix bulk viscosity Pa s corded. If the melt fraction fails to reach the SLT within
Z* effective viscosity Pa s 25 Myr, the model is stopped and it is assumed that a
Z0 effective matrix viscosity Pa s mobile magma cannot form. After this time, at least
m melt shear viscosity Pa s 12·5 km of basalt has been emplaced, which represents
x matrix shear viscosity Pa s almost half of the initial thickness of the modelled crust
rl liquid density kg m3 and is similar to the maximum thickness assumed in previ-
rs solid density kg m3 ous models (Annen et al., 2006). In all models that include
j solid–liquid transition (SLT) none segregation, a mobile magma forms long before this upper
f porosity (melt fraction) none limit is reached.
Material properties
The thermal properties of the amphibolite crust and
intruded basalt are taken from Petford & Gallagher
(2001) (summarized here in Table 2) and Annen et al.
as the sill cools (typically corresponding to a period of (2006) (summarized here in Table 3), respectively, and are
20^200 years). Equations (1)^(5) are non-dimensionalized reasonably well constrained. The parameters governing
and solved numerically using finite-difference methods, in buoyancy-driven melt segregation and compaction of the
a code developed by the first two authors. Equation (2) is matrix are taken from Jackson et al. (2003) (summarized
approximated using a forward-time centred-space scheme, here in Table 4). In contrast to the thermal properties,
equation (3) using a second-order Lax^Wendroff scheme, many of the melt segregation parameters are poorly con-
and equation (4) using a centred scheme (Morton & strained; for example, estimates of the matrix bulk viscos-
Mayers, 2005). ity, deforming by melt-enhanced diffusion creep, vary
We use the term melt fraction to denote the volume frac- over several orders of magnitude. The value of the SLT is
tion of melt present at a given point in space and time also uncertain, and we vary the values of all these param-
(i.e. it is directly equivalent to porosity). However, it is im- eters to investigate the impact of uncertainty on the pre-
portant to note that melt fraction in this model is not syn- dicted behaviour of the DCHZ. The melt viscosity and
onymous with melt composition, owing to the relative the density contrast between melt and matrix are assumed
migration of melt and matrix during melt segregation. constant in each model run, and the same values are used
2007
in both the melting amphibolite crust and the crystallizing measured values for rhyolite varying by seven orders of
basalt. The viscosity of a silicate melt depends primarily magnitude over the temperature range investigated
on its composition (including volatiles, such as water) and (e.g. Shaw, 1963; Murase & McBirney, 1973; Giordano &
temperature (Bottinga & Weill, 1972). The variation of vis- Dingwell, 2003; Giordano et al., 2008). However, during
cosity with temperature for evolved silicic melts of constant partial melting or crystallization in a mush, the compos-
composition has been investigated experimentally, with ition of the melt continually changes and this must be ac-
counted for.
Viscosities calculated for melts produced during partial
Table 2: Material properties of amphibolite (Petford & melting experiments on amphibolite (consistent with
those used to develop the melt fraction vs temperature rela-
Gallagher, 2001)
tionship shown in Fig. 3) vary by approximately one order

of magnitude over the melt fraction and temperature
Symbol Description Values used Units range of interest (Fig. 4). The comparatively small vari-
ation in melt viscosity arises because of the competing ef-
cp specific heat capacity 1200 J kg1 K1 fects of silica and water content. As temperature and melt
k thermal conductivity 3 W K1 m1 fraction increase, the composition of the melt evolves to
Lf latent heat 293000 J kg1 have lower silica and water content; high silica content in-
Tliq liquidus temperature 1473 K creases melt viscosity, but high water content lowers viscos-
Tsol solidus temperature 1095 K ity. For dry melts, no such counterbalancing of silica and
r density 2900 kg m3 water content is possible, so melt viscosity increases dra-
matically with decreasing temperature (and melt fraction)
(Fig. 4). As the present study concerns only hydrous basalt
or amphibolite, this complicating effect is not considered.
The values of constant melt viscosity investigated in this
Table 3: Material properties of basalt (Annen et al., study are based on those shown in Fig. 4, and the range
2006) given in Table 4 reflects the uncertainty in predicting melt
viscosity, rather than a range of melt compositions.
The phase relations of the basalt assume that the water
Symbol Description Values used Units
content is 2·5 wt % at 1GPa (Annen et al., 2006). Above
and below the amphibole crystallization temperature (Ta)
cp specific heat capacity 1480 J kg1 K1
at 1350 K, the melting behaviour is linear (Fig. 3). The
k thermal conductivity 2·6 W K1 m1
melting relationship is described by
Lf latent heat 400000 J kg1
Tliq liquidus temperature 1558 K T Tsol
X ¼ Xa , Tsol T Ta ð6aÞ
Tsol solidus temperature 993 K Ta Tsol
r density 2830 kg m3
X ¼ 3 25 103 ðT Tliq Þ þ 1, Ta T Tliq ð6bÞ
Table 4: Physical parameters describing melt migration (Jackson et al., 2003)
Symbol Description Min.–max. values Values used Units
a matrix grain radius 5 104–5 103 2·5 103 m

b permeability constant 1/2500–1/50 1/500 none
a permeability exponent 3 3 none
b viscosity exponent 0·5 0·5 none
Z matrix shear viscosity 1014–1018 1014–1018 Pa s
m melt shear viscosity 103–106 103–105 Pa s
x matrix bulk viscosity 1014–1018 1014–1018 Pa s
rs–ri liquid–solid density contrast 400–700 500 kg m3
j solid–liquid transition 0·4–0·6 0·4–0·6 none
2008
Evolution of melt fraction through time

within the DCHZ
We begin by investigating the evolution of melt fraction
within the DCHZ, as heat is supplied by repeated basalt
intrusions and buoyancy forces cause melt to migrate up-
wards. Figure 5 shows melt fraction (porosity) as a func-
tion of depth through the DCHZ during basalt
emplacement by over-accretion at a rate of 10 mm a1.
Each line shows melt fraction at a different snapshot in
time, separated by 5000 years in Fig. 5a, and by 100 years
in Fig. 5b. The snapshots in Fig. 5a are taken immediately

before the next intrusion of a mantle-derived basalt sill;
the snapshots in Fig. 5b show how the melt fraction evolves
before, during and after the emplacement of a sill. The
effect of varying emplacement style and rate is investigated
in the following sections.
Early in the evolution of the DCHZ, the melt fraction
Fig. 4. Melt viscosity as a function of temperature, calculated using lies below the SLT (i.e. the DCHZ comprises a mush),
compositional data from dehydration melting experiments on am- except when a new intrusion of basalt yields a transient,
phibolite at 0·7^1·5 GPa (Beard & Lofgren, 1991; Sen & Dunn, 1994; short-lived basaltic magma chamber, as discussed above.
Rapp & Watson, 1995). Viscosity calculated using the model of
Giordano et al. (2008). ‘Wet’ results use reported melt water contents The melt fraction in the basaltic magma chamber rapidly
where available, or are calculated using the model of Clemens & falls below the SLT as the basalt cools and crystallizes; in
Vielzeuf (1987).‘Dry’ results assume zero melt water content. Fig. 5, which shows typical results from the spectrum of
cases investigated, the melt fraction within the initially
liquid basaltic sill has fallen below the SLT within 100
Xa ¼ 3 25 103 ðTa Tliq Þ þ 1 ð6cÞ
years of intrusion. Hence we model the DCHZ as a mush,
where Tsol and Tliq denote the solidus and liquidus tem- with melt present along grain boundaries, rather than as
peratures respectively. The amphibolite melting relation- a large liquid body containing suspended crystals.
ship at 1GPa is described by a third-order polynomial However, the upwards migration of buoyant melt results
equation (Petford & Gallagher, 2001; Annen et al., 2006) in the formation of one or more melt layers (zones of high
porosity), similar to those observed in previous numerical
T Tsol
X ¼ ð1 a0 þ 0 5a1 Þ experiments (e.g. Jackson et al., 2003, 2005; Fig. 5a). These
Tliq Tsol layers form when the melt flux decreases upwards in re-
ð7Þ
T Tsol 2 T Tsol 3 sponse to the upwards decrease in melt fraction and per-
þ ða0 þ 1 5a1 Þ þa1
Tliq Tsol Tliq Tsol meability towards the solidus isotherm, which acts as a
barrier to melt flow at the top of the DCHZ. Melt locally
with a0 ¼ 0·4577 and a1 ¼ ^0·771. accumulates when the melt flux decreases upwards, and
the matrix dilates to accommodate it, resulting in the for-
mation of a melt layer. The melt fraction in each layer in-
R E S U LT S creases until the SLT is exceeded and a mobile magma
In describing the formation of evolved magmas in DCHZ, forms. The time required for this to occur is the accumula-
we are interested in the distribution of melt within the tion time. Melt segregation clearly has a profound impact
DCHZ, especially the relative contribution of crustal and on melt distribution within the DCHZ; in the next section,
residual melt, and the timescales of melt segregation and we compare our model predictions against the purely ther-
magma formation (which we term the accumulation mal (static) model of Annen & Sparks (2002) and Annen
time). We are also interested in the depth at which magma et al. (2006), for a range of basalt emplacement rates and
formation occurs (the accumulation depth) and the tem- styles. For simplicity, we refer to models that include melt
perature of the mobilized magma (i.e. the temperature at migration as ‘dynamic’ and those without melt migration
the accumulation depth, which we term the accumulation as ‘static’.
temperature). These latter parameters control the compos-
ition of the magma that leaves the DCHZ. Each of the Impact of melt segregation on melt
above may be affected by basalt emplacement rate and distribution within the DCHZ
style (over-, under- or intra-accretion). No previous Figures 6^8 show snapshots of temperature and melt frac-
models have investigated all these issues. tion through the DCHZ for different emplacement styles
2009
Buoyant crustal
(a) and residual melt
migrates upwards
and accumulates
at the top of the
DCHZ

Emplacement depth
= top of basalt layer
(b)
Melt fraction
rapidly falls below
New basaltic
the solid-to-liquid
intrusion (sill) is
transition (SLT) as
initially liquid
intrusion cools
(melt fraction=1)
Fig. 5. Melt fraction vs depth during basalt emplacement by over-accretion at a rate of 10 mm a1. Each line shows melt fraction at a different
snapshot in time, separated by (a) 5000 years and (b) 100 years. The snapshots in (a) are taken immediately before the next intrusion of a sill.
at the relevant accumulation time. The melt distribution is rate, the snapshots shown in Figs 6^8 are taken at different
primarily determined by the style of emplacement (over-, times in dynamic and static models, which is why the
under- or intra-accretion) and the rate at which basalt is thickness of emplaced basalt and the temperature profile
emplaced. For comparison, we also show static cases that differ.
omit melt migration; these results are identical to those In dynamic models, one or more melt layers are
presented by Annen & Sparks (2002) and Annen et al. observed for all emplacement styles (Figs 6^8). Each
(2006). The accumulation time is generally longer in static model run is stopped when any one melt layer attains the
models (see Fig. 9 and associated discussion in the next sec- SLT; subsequent evolution of the other layers is not con-
tion); consequently, for a given emplacement style and sidered. The highest melt fraction occurs within the
2010
Temperature (K)
1000 1200 1000 1200 1000 1200 1000 1200
-29 (a) (b) (c) (d) -29
residual melt residual and
accumulates crustal melt
at top of accumulates
basalt layer in crust
-30 -30
top of
basalt layer
-31 -31
Depth (km)

base of
basalt layer
-32 -32
-33 -33
accumulation
time =1.56 Ma 0.31 Ma 0.17 Ma 0.05 Ma
0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8
Melt Fraction
1 mm/a 5 mm/a 10 mm/a 25 mm/a

Increasing Emplacement Rate by Over Accretion
Temperature (K)
1000 1200 1000 1200 1000 1200
Amphibolitic -29 crustal melt (e) (f) (g) -29
crust at location
of highest
temperature
Basalt layer
-30 -30
Melt Fraction
Temperature
-31 -31
Depth (km)
-32 -32
-33 -33
1.72 Ma 0.42 Ma 0.08 Ma

0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8
Melt Fraction
Fig. 6. Melt fraction (continuous line) and temperature (dashed line) vs depth for dynamic models with melt migration (a^d) and static
models without melt migration (e^g) during over-accretion of basaltic sills. Melt fraction is read from the lower x-axis and temperature from
the upper x-axis. The gray shaded area denotes the basalt layer. All runs have an SLTof 0·5 and the snapshots shown are taken at the accumu-
lation time, when the SLT is reached. The accumulation time is shown in each plot. Emplacement rates: (a) 1mm a1; (b, e) 5 mm a1; (c, f)
10 mm a1; (d, g) 25 mm a1. Matrix viscosity is 1016 Pa s and melt viscosity is 104 Pa s. Melt and temperature profiles for the static model at
the lowest emplacement rate are not shown, as the accumulation time was greater than 25 Myr.
2011
Temperature (K)
1000 1200 1000 1200 1000 1200 1000 1200
-29 (a) (b) (c) (d) -29
-30 top of
-30
basalt layer
-31 base of -31

Depth (km)
basalt layer

-32 -32
-33 -33
accumulation
time =1.10 Ma 0.09 Ma 0.04 Ma 0.05 Ma
0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8
Melt Fraction

Increasing Emplacement Rate by Under Accretion
Temperature (K)
1000 1200 1000 1200 1000 1200
Amphibolitic -29 (e) (f) (g) -29
crust
Basalt layer
-30 -30
Melt Fraction
Temperature
-31 -31
Depth (km)
residual melt
-32 at location -32
of highest
temperature
-33 -33
1.45 Ma 0.35 Ma 0.07 Ma

0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8
Melt Fraction
Fig. 7. Melt fraction (continuous line) and temperature (dashed line) vs depth for dynamic models with melt migration (a^d) and static models
without melt migration (e^g) during under-accretion of basaltic sills. Melt fraction is read from the lower x-axis and temperature from the
upper x-axis. The gray shaded area denotes the basalt layer. All runs have an SLTof 0·5 and the snapshots shown are taken at the accumulation
time, when the SLT is reached. The accumulation time is shown in each plot. Emplacement rates are (a) 1mm a1, (b, e) 5 mm a1, (c, f)
10 mm a1, and (d, g) 25 mm a1. Matrix viscosity is 1016 Pa s and melt viscosity is 104 Pa s. Melt and temperature profiles for the static model
at the lowest emplacement rate are not shown, as the accumulation time was greater than 25 Myr.
2012
Temperature (K)
1000 1200 1000 1200 1000 1200 1000 1200
-29 (a) (b) (c) (d) -29
-30 top of
-30
basalt layer
base of
-31 basalt layer -31
Depth (km)

-32 -32
-33 -33
accumulation
time =1.10 Ma 0.08 Ma 0.03 Ma 0.06 Ma
0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8
Melt Fraction

Increasing Emplacement Rate by Intra Accretion
Temperature (K)
1000 1200 1000 1200 1000 1200
Amphibolitic -29 (e) (f) (g) -29
crust
Basalt layer
-30 -30
Melt Fraction
Temperature
-31 -31
Depth (km)
residual melt
at location
-32 of highest -32
temperature
-33 -33
12.02 Ma 0.51 Ma 0.09 Ma

0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8
Melt Fraction
Fig. 8. Melt fraction (continuous line) and temperature (dashed line) vs depth for dynamic models with melt migration (a^d) and static
models without melt migration (e^g) during intra-accretion of basaltic sills. Melt fraction is read from the lower x-axis and temperature from
the upper x-axis. The gray shaded area denotes the basalt layer. All runs have an SLTof 0·5 and the snapshots shown are taken at the accumu-
lation time, when the SLT is reached. The accumulation time is shown in each plot. Emplacement rates are (a) 1mm a1, (b, e) 5 mm a1, (c,
f) 10 mm a1, and (d, g) 25 mm a1. Matrix viscosity is 1016 Pa s and melt viscosity is 104 Pa s. Melt and temperature profiles for the static
model at the lowest emplacement rate are not shown, as the accumulation time was greater than 25 Myr.
2013
leading melt layer, which, at high emplacement rates, is magnitude. At low emplacement rates (52 mm a1) the
located in the overlying amphibolite crust and contains SLT is not reached within the maximum run time of 25
both crustal melt and residual melt that has migrated up- Myr in the static models, so no data are shown for these.
wards from the cooling sills (Figs 6d, 7d and 8d). At intermediate emplacement rates, typical of
Conversely, at low emplacement rates, the leading melt time-averaged rates in arcs (c. 2^20 mm a1), dynamic
layer is located at the top of the basalt layer and contains models predict segregation times of c. 10 kyr^1 Myr and
only residual melt (Figs 6a, 7a and 8a). At intermediate show a systematic difference in accumulation time depend-
rates, the location of the leading melt layer depends upon ing upon emplacement style; over-accretion leads to
emplacement style. During over-accretion, it lies within longer accumulation times than intra- or under-accretion,
the overlying crust and comprises both crustal and residual because over-accretion leads to melt accumulating in the
melt (Fig. 6b and c); during intra- and under-accretion, it overlying crust rather than at the top of the basalt layer
lies at the top of the basalt layer and comprises residual (compare Fig. 6c with Figs 7c and 8c). The time required

melt with small volumes of crustal melt sourced from the for melt to migrate upwards and accumulate in the overly-
underlying crust (Figs 7b,c and 8b,c). ing crust is longer than that required for melt to accumu-
In static models (see Annen et al., 2006), the resulting late at top of the basalt layer. At high emplacement rates
melt profiles are markedly different. During over- (420 mm a1), melt migration has a smaller effect on accu-
accretion, the highest melt fraction occurs just above the mulation times; although they are still lower in dynamic
basalt layer, within the overlying amphibolite crust models, the difference between dynamic and static models
(Fig. 6; note that at the lowest emplacement rate, the SLT is less significant. Accumulation times in both models are
is not reached so no curves are shown). This is a fundamen- always shorter than c. 200 kyr, because melt production is
tal feature of all static models during over-accretion be- very rapid. However, such high emplacement rates are
cause, unlike the dynamic models, melt fraction and characteristic of transient rates associated with formation
temperature are directly linked in both the basalt layer of large crustal magma chambers, rather than long-term,
and the crust. The temperature of the DCHZ is highest at time-averaged rates in arcs. Uncertainty in the value of
the location of the most recently intruded sill and, at tem- the SLTdoes not significantly affect the results.
peratures higher than 1200 K, the amphibolite crust is
more fertile than the intruded basalt, and so yields a
larger fraction of melt (Fig. 3). Although the majority of
Accumulation depth
the melt present in the DCHZ is residual melt from the The depth at which melt accumulates to form a mobile
cooling sills, the highest melt fraction lies within the crust magma is significant because it affects the composition of
and comprises crustal melt. During static under-accretion, the magma in two ways. First, if accumulation occurs in
the highest melt fraction occurs just below the base of the the overlying crust, then the resulting magma is composed
basalt layer and comprises crustal melt, because the tem- of both crustal melts and residual melts, which have
perature of the DCHZ is highest here at the location of migrated upwards out of the intruded basalt through the
the most recently intruded sill (Fig. 7; note that at the relatively refractory matrix of the crust. Conversely, if ac-
lowest emplacement rate, the SLT is not reached so no cumulation occurs in the basalt layer, the resulting
curves are shown). During static intra-accretion, the high- magma is composed primarily of residual melt, which
est melt fraction occurs at the centre of the basalt layer may be contaminated by small volumes of crustal melt
where the temperature is highest and comprises residual that have migrated upwards from the underlying crust.
melt (Fig. 8; note that at the lowest emplacement rate, the Second, shallower accumulation depths yield cooler
SLT is not reached so no curves are shown). magmas, because the temperature decreases upwards. As
the melt migrates upwards, it cools and crystallizes, so its
composition changes to become more evolved; similarly,
Accumulation timescales as the matrix migrates downwards during compaction, it
Figure 9a shows the accumulation time required to form a heats and melts, so its composition also evolves to become
mobile magma as a function of emplacement rate in both more refractory (Jackson et al., 2003, 2005). Consequently,
dynamic and static models. The uncertainty bars show the shallower depths of accumulation correspond to cooler
effect of varying the value of the SLT between 0·4 and temperatures, which yield more evolved silicic (granitic
0·6. The accumulation time ranges from a minimum of c. sensu lato) partial melt compositions (Fig. 3). We investigate
10 kyr to c. 10 Myr and increases with decreasing emplace- segregation temperature in the next section.
ment rate. For emplacement rates less than c. 20 mm a1, Figure 9b shows the accumulation depth as a function of
melt migration has a significant impact on accumulation emplacement rate in dynamic and static models. The un-
time; segregation of melt into melt layers yields mobile certainty bars show the effect of varying the SLT. The ac-
magma much more rapidly than is achieved through melt- cumulation depth is always shallower in dynamic models;
ing alone, in many cases by more than an order of at low emplacement rates (52 mm a1) the SLT is not
2014
25
(a)
10
A cc u m u l a t i o n T i m e ( M a )
1
0.1
Dynamic (melt migration)
Static (no melt migration)

0.01 Over-Accretion
Under-Accretion
Intra-Accretion
-26 (b)
-28 Amphibolite Crust
-30
Accumulation Depth (km)
-32 Basalt Layer

-34
-36
-38
-40
-42
-44
-46
-48
-50
(c)
1400
Accumulation Temperature (K)
1300
1200
1100
1000
0.5 1 5 10 50
Emplacement Rate (mm/a)
Fig. 9. Accumulation time (a), accumulation depth (b) and accumulation temperature (c) vs emplacement rate for a range of solid to liquid
transition (SLT) values. Uncertainty bars show the range of varying the SLT from 0·4 to 0·6. Filled bars show cases with melt migration;
dashed bars show cases without melt migration. Blue denotes over-accretion; red denotes under-accretion; green denotes intra-accretion. For
clarity of presentation, data for under- and over-accretion have been shifted higher and lower on the x-axis by 5%. Matrix viscosity is 1016 Pa s
and melt viscosity is 104 Pa s. Solid lines in (b) denote the base of the basalt layer in the static melting case with SLT ¼ 0.5.
2015
reached within the maximum run time of 25 Myr in static the SLT, regardless of emplacement rate and regime. At low
models, so no data are shown for these. In dynamic emplacement rates (52 mm a1) the SLT is not reached
models at high emplacement rates (420 mm a1), the accu- within the maximum run time of 25 Ma, so no data are
mulation depth lies in the crust overlying the top of the shown for these. In static models magma formation always
intruded sills, so magmas that leave the DCHZ will com- occurs in the hottest part of the DCHZ, and the melt fraction
prise both crustal and residual melts, and have geochem- is governed only by the phase behaviour of the amphibolite
ical signatures that indicate both crust and mantle crust and intruded basalt. Accumulation temperature is
contributions. Conversely, in dynamic models at low em- most significantly affected by uncertainty in the value of the
placement rates (52 mm a1) the accumulation depth lies SLT, ranging from c. 1340 K to c. 1440 K over the modelled
at or near the top of the basalt layer, so magmas that range regardless of emplacement rate or style. The partial
leave the DCHZ will comprise primarily residual melts, melts formed at these temperatures and 30 km depth will be
and have geochemical signatures that indicate a domin- broadly andesitic in composition (Fig. 3 and Annen et al.,

antly mantle contribution. At typical time-averaged em- 2006), regardless of whether they are crustal or residual in
placement rates in arcs (2^20 mm a1) the accumulation origin. Static models therefore predict systematically higher
depth depends upon emplacement style; for over-accretion, and less variable magma temperatures, and consequently
magma accumulation occurs in the overlying crust, less evolved magma composition, than dynamic models,
whereas for intra- and under-accretion, magma accumula- and fail to capture the impact of emplacement rate on
tion occurs near the top of the basalt layer. magma temperature. Uncertainty in the value of the SLT
In static models, the accumulation depth corresponds to does not significantly affect the results.
the depth at which the highest temperature is reached
within the DCHZ, in the crust immediately overlying the Impact of uncertainty in melt segregation
top of the basalt layer for over-accretion, the base of parameters
the basalt layer for under-accretion, and the centre of the Several of the parameters governing buoyancy-driven melt
basalt layer for intra-accretion (Fig. 9b; lines denote the segregation and compaction of the matrix are poorly con-
base of the basalt layer assuming an SLT of 0·5). For strained, particularly melt and matrix shear viscosities,
over-accretion, the accumulation depth is therefore fixed, and the matrix bulk viscosity, with estimates ranging over
whereas for under- and intra-accretion, the accumulation several orders of magnitude (Table 4). Here, we explore
depth decreases with increasing emplacement rate, be- the impact of this uncertainty on the predicted timing of
cause mobile magma formation occurs more rapidly, magma formation, and the depth and temperature at
hence less basalt has been emplaced. Uncertainty in the which magma formation occurs. The value of the SLT is
value of the SLT does not significantly affect the results, also uncertain, although our results demonstrate that this
except in static models at low emplacement rates. does not have a significant impact over the modelled
range (Fig. 9). Our aim is to demonstrate that the results
Accumulation temperature and its impact shown in the previous sections and, more significantly, the
on magma composition conclusions drawn from these, are not conditional on the
Figure 9c shows accumulation temperature as a function of choice of particular values of these material properties.
emplacement rate in dynamic and static models. The un- Figure 10 shows the effect on the accumulation time,
certainty bars show the effect of varying the SLT. In dy- depth and temperature of uncertainty in the melt viscosity,
namic models, the accumulation temperature generally with the modelled (constant) value varying over two
increases with increasing emplacement rate; this is because orders of magnitude (103^105 Pa s; Fig. 10a^c), and the
mobile magma formation occurs more rapidly (Fig. 9a), matrix bulk and shear viscosities, with the modelled value
before melt has had time to migrate upwards into cooler varying over four orders of magnitude (1014^1018 Pa s;
regions of the DCHZ. Thus emplacement rate has a signifi- Fig. 10d^f). In general, uncertainty in the melt viscosity
cant impact on the temperature at which mobile magmas has a larger impact on magma accumulation than uncer-
form. Regardless of emplacement style, accumulation tem- tainty in the matrix viscosity, even though the modelled
peratures range from c. 1250 K at the highest emplacement range of matrix viscosity is much greater. As the melt
rates to c. 1000 K at the lowest emplacement rates, and are and/or matrix viscosity increases, the rate of melt segrega-
generally lower than 1200 K at typical time-averaged em- tion decreases and, in the limit of very large viscosity
placement rates in arcs (2^20 mm a1). The partial melts values, the model results tend towards those obtained in
formed at these temperatures and 30 km depth during the the absence of melt segregation.
intrusion of basalt into amphibolite crust are broadly rhyo- Accumulation time ranges over several orders of magni-
litic in composition, regardless of whether they are crustal tude (c. 10 kyr^20 Myr) and, in dynamic models, is similar
or residual melts (Fig. 3). regardless of emplacement style or the modelled values of
In static models, significantly higher temperatures are viscosity at high emplacement rates (420 mm a1).
required to generate melt fractions large enough to exceed However, at lower emplacement rates, the impact of
2016
25
Accumulation Time (Ma) 10 (a) (d)
0.1
Over-Accretion

0.01
Under-Accretion
Intra-Accretion
-26 Amphibolite Crust (b) (e)

-28
Accumulation Depth (km)
-30
-32 Basalt Layer
-34
-36
-38
-40
-42
-44
-46
-48
-50
(c) (f)
Accumulation Temperature (K)
1400
1300
1200
1100
1000
0.5 1 5 10 50 0.5 1 5 10 50
Emplacement Rate (mm/a) Emplacement Rate (mm/a)
Fig. 10. Accumulation parameters vs emplacement rate for a range of matrix and melt viscosities. In (a)^(c) the uncertainty bars show the
effect of varying the matrix viscosity from 1014 to 1018 Pa s; melt viscosity is 104 Pa s. In (d)^(f) the uncertainty bars show the effect of varying
the melt viscosity from 103 to 105 Pa s; the matrix viscosity is 1016 Pa s. Continuous lines denote static models with no melt migration. Colours
are the same as Fig. 9, with blue denoting over-accretion, red denoting under-accretion and green denoting intra-accretion. The SLT is 0·5.
2017
uncertainty in the matrix and, in particular, the melt vis- DISCUSSION

cosity is more significant, yielding order-of-magnitude
Importance of capturing melt migration
variations in accumulation time. Higher values of melt
in models of DCHZ
or matrix viscosity yield larger accumulation times in
dynamic models, because melt migration occurs more The dynamic model presented here demonstrates that
slowly as the melt or matrix viscosity increases. melting and melt segregation leads to the formation of
evolved magmas over timescales of 10 kyr^1 Myr, within
Nevertheless, the accumulation time is always shorter
a DCHZ formed by the intrusion of hot, mantle-derived
than that predicted by static models (denoted by continu-
basalt sills into amphibolite crust at typical arc magma
ous lines) and, at typical time-averaged emplacement
production rates (2^20 mm a1). Basalt emplacement rate
rates in arcs (2^20 mm a1), the difference is significant
has a significant impact on the timing of evolved magma
(2^20 times shorter). At the lowest emplacement rates
formation; the emplacement style and rate also affect the
(52 mm a1), the SLT is not reached within the maximum

depth and temperature at which evolved magma forma-
run time of 25 Myr in static models, so no data are shown
tion occurs, and hence the composition of the magma.
for these.
Two styles of magma generation are revealed. In one, fa-
The accumulation depth in dynamic models varies from
voured by over-accretion, or by emplacement at high rate,
c. 28 km to 38 km, with the impact of uncertainty in the
a melt layer forms in the crust overlying the intruded
melt and matrix viscosities generally decreasing with
basalt, and the resulting magma comprises both crustal
increasing emplacement rate. Uncertainty in the matrix
and residual melts that have migrated upwards through
viscosity has little effect at low emplacement rates
the partially molten rock matrix. In the other, favoured
(51mm a1), where magma accumulation always occurs
by under- or intra-accretion, or by emplacement at low
within the basalt layer. At higher rates (41mm a1),
rate, a melt layer forms within the basalt layer, and the re-
magma accumulation always occurs in the overlying crust sulting magma is dominated by residual melt. In both
during over-accretion, but may occur either in the overly- cases, the magmas consist of silicic melt that is broadly
ing crust or the basalt layer during intra- or rhyolitic in composition, which may explain why andesite
under-accretion, depending upon the modelled values of magmas in arc settings are commonly found to be mix-
melt and matrix viscosity. Higher values of melt viscosity tures of silicic and mafic magmas, with true andesite
yield deeper accumulation depths within the basalt layer, melts being rare (Reubi & Blundy, 2009). Our results sug-
whereas lower values of matrix viscosity yield shallower gest that the dominant product of differentiation within
accumulation depths within the overlying crust; this is be- DCHZ will be rhyolitic melts that can mix with basaltic
cause melt migration occurs more slowly as the melt or melts in shallow magma chambers. This is in marked con-
matrix viscosity increases, so the melt migrates upwards trast to static models, which predict large volumes of inter-
over shorter distances. Nevertheless, the accumulation mediate melt because temperatures across the DCHZ are
depth in dynamic models is always shallower than that higher when mobile magma forms.
predicted by static models. The results are sensitive to the modelled values of melt
Accumulation temperature, which is a proxy for melt and matrix viscosity. Magma accumulation is more rapid
composition, is the property most significantly affected by in models with low melt and matrix viscosity, and more
uncertainty in the matrix and, in particular, melt viscosity. often occurs in the overlying crust at lower temperature.
As the emplacement rate increases, the uncertainty in vis- During over-accretion, magmas that leave the DCHZ are
cosity has a larger impact on the range of modelled accu- predicted to comprise both crustal and residual melts at
mulation temperatures. At the highest modelled rates all but the slowest emplacement rates and the highest visc-
(420 mm a1), accumulation temperatures range from c. osities; during under- and intra-accretion, magmas that
1150 to 1400 K; at typical arc rates (2^20 mm a1), accumu- leave the DCHZ may comprise both crustal and residual
lation temperatures range from c. 1000 to 1400 K; at the melts, or primarily residual melts, depending upon the em-
lowest modelled rates (52 mm a1), accumulation tempera- placement rate and the melt and/or matrix viscosity. Melt
tures range from c. 1000 to 1150 K. In all cases, the range migration processes ensure that residual melt is the volu-
of accumulation temperatures depends upon the emplace- metrically and compositionally dominant component of
ment style. The accumulation temperature in dynamic the magma that leaves the DCHZ in all cases. Previous
models is always lower than, or the same as, that predicted models have largely neglected the contribution of residual
by static models. Lower accumulation temperatures in dy- melt in a DCHZ, focusing primarily on crustal melts
namic models are favoured by low melt and matrix viscos- formed by partial melting (Huppert & Sparks, 1988;
ity, because melt migration occurs more rapidly as the Bergantz, 1989; Bergantz & Dawes, 1994; Petford &
melt or matrix viscosity decreases. Consequently, the melt Gallagher, 2001; Jackson et al., 2003, 2005; Dufek &
accumulates more rapidly and at shallower depths (as dis- Bergantz, 2005). Moreover, previous models have failed to
cussed above) so the temperatures are lower. capture the importance of basalt emplacement style and
2018
its interaction with melt segregation processes. Annen & Emplacement Rate (mm/a)
Sparks (2002) and Annen et al. (2006) considered emplace- 5 10 25 50
ment style but neglected melt segregation, whereas
Over-Accretion (a)
Jackson et al. (2003, 2005) considered melt segregation but
did not consider emplacement style or the contribution of 25
residual melt.
Productivity (%)
Static (i.e. purely thermal) models, which neglect melt 20
segregation, predict that magma mobilization occurs over
significantly longer timescales than dynamic models that
15
include melt segregation. The difference in model predic-
tions is greatest at lower emplacement rates, for which ac-
10

cumulation times in static models can be more than an
order of magnitude larger than in dynamic models;
indeed, at the slowest emplacement rates, magma mobil- 5
Melt Migration
ization may never occur in static models (Fig. 9a). This is No Melt Migration
because accumulation times in static models correspond
to the time required for some part of the DCHZ to reach Under-Accretion (b)
the temperature that yields a melt fraction equivalent to
25
the SLT. In dynamic models, this melt fraction is reached
Productivity (%)
more rapidly because melt migrates and accumulates.
Moreover, magma mobilization in static models is pre- 20
dicted to occur in the hottest part of the DCHZ, irrespect-
ive of emplacement style (Fig. 9b). Consequently, static 15
models fail to capture the impact of emplacement rate on
the temperature at which magma mobilization occurs, 10
and hence magma composition. Magma temperatures pre-
dicted by static models are generally higher than those
5
observed in dynamic models, so the magmas comprise less Melt Migration
silicic melt that is broadly andesitic in composition No Melt Migration
(Fig. 9c).
Static models typically predict melt productivity that is Intra-Accretion (c)
similar to, or lower than, that predicted by the dynamic 25
models, depending upon emplacement rate and style
Productivity (%)
(Fig. 11). Productivity here is defined as the volume of 20

evolved (rhyolitic) melt that is produced for a given
volume of intruded basalt; in this study, we define evolved
15
melt to be crustal or residual melt (or a mixture of the
two) present in regions of the DCHZ where the tempera-
ture is less than 1200 K (Fig. 3), and calculate the product- 10
ivity as the ratio of the porosity-weighted total thickness
of melt present at temperature below 1200 K to the total 5
Melt Migration
thickness of intruded basalt, calculated at the accumula-
No Melt Migration
tion time. Productivity generally increases with emplace-
ment rate, from c. 10% at low emplacement rates to 425% 5 10 25 50
at high emplacement rates, and is generally higher for Emplacement Rate (mm/a)
over-accretion than for under- or intra-accretion. In static Fig. 11. Productivity of evolved melt, for (a) over-accretion, (b),
models, productivity remains fixed at c. 10^12%, regardless under-accretion and (c) intra-accretion, as a function of basalt em-
of emplacement rate. Bergantz (1989) estimated productiv- placement rate.
ities of 40^60%, but these were for tonalite^metapelite
protoliths, and his definition of productivity included all
melt, regardless of composition. magma. However, as discussed below, lateral migration of
Melt productivity values of the order of 10% suggest evolved magma through the high-porosity melt layers can
that basalt sills totalling 10 km in thickness must be collect the evolved magma generated from a DCHZ of
emplaced to produce a c. 1km thick layer of evolved larger areal extent into a localized ascent zone comprising
2019

Fig. 12. Schematic block diagram of a DCHZ after the formation of a high-porosity layer of evolved melt. The 3-D geometry of this melt layer is
poorly understood, but will be primarily dictated by the upward migration rate of the solidus isotherm, rather than the dynamics of flow
through the deformable mush. Despite the relative motion of melt and matrix, heat transfer is dominated by conduction, so the migration rate
of the solidus isotherm is unlikely to vary laterally by a significant amount (Jackson & Cheadle, 1998). Consequently, the high-porosity layers
are likely to be laterally continuous. Evolved melt may flow laterally through the high-porosity layers and be drained via a localized ascent
zone of dykes, faults or fractures.
dykes, faults or fractures (Fig. 12). For example, the granitic (equilibrium melting or crystallization) or that melt or solid
(sensu lato) magma required to fill a 100 km3 circular is instantaneously removed as it forms (fractional melting or
pluton of 300 m thickness and 20 km diameter, sourced crystallization). Neither of these end-member models cap-
from a circular DCHZ of diameter 40 km, can be formed tures the coupled chemical and physical processes within a
from only 800 m thickness of intruded basalt. Moreover, mushy DCHZ, where melt migrates along grain edges
the results presented here demonstrate that the magma through a steep thermal gradient, with chemical exchange
can be generated in only a few hundred thousand years. between melt and matrix.
A further complication is that the composition of the
Compositional consequences of magma that leaves the DCHZ may reflect the mixing of
melt migration melt sourced from the crust and intruded sills, depending
Magma composition is complicated by melt segregation, upon the emplacement style and rate. Magmas that form
whichcan affect meltchemistry inwaysthat are not predicted within the overlying crust contain both crustal and re-
by simple batch or fractional melting models and experi- sidual melt, and so will have chemical signatures that indi-
ments, because the melt migrates through, and chemically cate a mixed crustal and mantle origin. The model results
equilibrates with, matrix at progressivelylower temperatures suggest that contamination of crustal melt by residual
(Rushmer & Jackson, 2006; Getsinger et al., 2009). melt from basalt crystallization is an inevitable conse-
Equilibrium and fractional melting or crystallization quence of melt segregation in the DCHZ; indeed,
models, and variants of these, are always associated with a magmas that leave the DCHZ will always contain some
physical model (or assumption) for the separation of melt residual melt, but may contain little crustal melt. This is
from solid; the simplest end-members are that melt remains likely to be particularly manifest in the isotopic compos-
in contact with solid until the point of physical removal ition of the melts, which will show a hybrid mantle and
2020
crustal character. Moreover, when they are initially present, because the partial melt is less dense than its crys-
formed, the magmas contain a high proportion (40^60%) talline host; however, several studies have suggested that
of residual crystals. Depending upon the relative time- melt segregation in the crust is primarily driven by deform-
scales of magma extraction and crystal settling, some of ation induced by tectonic stresses. In these largely qualita-
these residual crystals may be entrained and leave the tive models, motivated by field observations of
DCHZ, further modifying the bulk composition of the migmatites, melt is assumed to flow along grain edges, col-
magma (Chappell et al., 1987). lecting in dilatant sites such as shear bands, tension gashes
Reactive transport is well known to produce chemical and boudins (e.g. Sawyer, 1991, 1994; Brown, 1994; Brown
complexity, especially for trace elements (Reiners, 1998). et al., 1995; Rabinowicz & Vigneresse, 2004). We agree that
Our results suggest that the chemical evolution of melts in melt segregation in a layered protolith may be driven by
the deep crust is likely to deviate significantly from the tectonic stresses; however, tectonic stresses are not required
concepts of fractional or equilibrium crystallization of

for segregation. Indeed, we show that melt segregation
magma bodies that have prevailed since the time of occurs rapidly even when buoyancy is the only driving
Bowen (1922). Compositional evolution of melts is a func- force. For example, Rabinowicz & Vigneresse (2004) esti-
tion of both chemistry (phase equilibria) and physics mated timescales of 30 kyr^0·3 Myr for deformation-
(melt migration). Chemical processes do not occur alone, driven, small-scale melt migration into veins a few metres
separate from physical processes during differentiation; apart; here we predict that buoyancy-driven compaction
rather, chemical and physical processes are intimately can lead to large-scale magma mobilization over similar
related and this relationship should be captured in math- timescales at basalt emplacement rates 42 mm a1. It
ematical models of differentiation. Interaction of melts should be noted that the model of Rabinowicz &
derived by crystallization of mantle-derived basalt and Vigneresse (2004) does not directly couple phase change
partial melting of the crust is a common phenomenon in to the local thermodynamic conditions; rather, a melting
DCHZ, which can account for the observation that
rate is imposed that does not account for the migration of
almost all granitic (sensu lato) rocks display some evidence
melt through a thermal gradient within the source region.
of a mixed origin (Pitcher, 1993). Crustal contamination is
There is also no attempt to relate the composition of the
typically ascribed to an assimilation^fractional crystalliza-
melt to the local thermodynamic conditions.
tion (AFC) process (De Paolo, 1981; Spera & Bohrson,
Studies of migmatites are usually confined to the mid- to
2001). The ‘magma intrusion þ crystallization þ partial
upper crust, where the country rock is broadly metasedi-
melting þ melt migration þ melt mixing’ process described
mentary in composition. In contrast, we consider melting
in our model is a form of AFC, and the model provides a
in lower crustal sections, comprising primitive amphibo-
physical mechanism to support the chemical processes
lite. At these depths and temperatures, melt segregation
associated with AFC. Assimilation occurs during partial
processes are more likely to resemble those in the upper
melting of the over- and underlying crust, and mixing of
mantle than in the mid- to upper crust. Moreover, migma-
melt from the intrusion and the crust occurs as melt mi-
grates. Furthermore, chemical evolution of the melt occurs tites have been described as ‘failed’ granites, because the
as it migrates into, and equilibrates with, cooler regions of evolved melt has remained in the source region as leuco-
the hot zone. Numerous published models show that bulk some, rather than migrating to higher crustal levels
(or wholesale) assimilation of crustal rocks can rarely be (e.g. White & Chappell, 1990; Clemens & Mawer, 1992).
justified on thermodynamic grounds (e.g. Hodge, 1974; One striking aspect of our model is that evolved melt is
Huppert & Sparks, 1988; Bergantz, 1989; Bergantz & very efficiently collected and mobilized. If melt extraction
Dawes, 1994; Petford & Gallagher, 2001; Annen & Sparks, is not efficient, then very large volumes of source material
2002; Dufek & Bergantz, 2005; Annen et al., 2006; are required to produce the observed volumes of evolved
Glazner, 2007); however, partial assimilation is consistent magma in the mid- to upper crust, whilst also accounting
with the predictions of these models and also with geo- for the volume of leucosome left behind in the source
chemical data, exemplified in the original treatment of De region. In our model, solid material left behind after the
Paolo (1981) and numerous subsequent papers (e.g. Grove DCHZ finally cools and solidifies is predicted to comprise
et al., 1988, 1997; Hildreth & Moorbath, 1988; Musselwhite a refractory, anhydrous blend of plagioclase þ pyrox-
et al., 1989; De Paolo et al., 1992). Further refinement of our ene quartz, olivine, garnet, magnetite and ilmenite,
models, to embrace chemical evolution, particularly for with little evolved melt; this mineralogy matches that of
trace elements and their isotopes, is clearly called for. many exposed granulite terrains (e.g. Clifford et al., 1981;
Clemens, 1990; Pin, 1990; Vielzeuf et al., 1990). These ter-
Buoyancy-driven versus deformation-driven rains can retain mesoscopic layering structures, which in-
melt segregation dicate that the matrix was not disrupted significantly
In this study we have assumed that buoyancy is the only during melt segregation (Clemens, 1990), consistent with
force driving melt migration. Buoyancy forces are always the model predictions.
2021
Differentiation of the continental crust CONC LUSIONS

Compaction-driven melt migration leads to bi-directional The model results we present here suggest that chemical
flow of relatively low-temperature, evolved melts upwards differentiation in deep crustal hot zones (DCHZ), asso-
and their refractory residua downwards, yielding progres- ciated with the repeated intrusion of mantle-derived
sive enrichment of the overlying country rocks in basalt into amphibolite lower crust, is driven by reactive
low-temperature, evolved components over distances of a flow of melt along grain boundaries through a steep ther-
kilometre or so and timescales that can be significantly mal gradient. The DCHZ comprises a crystalline mush
less than 1 Myr. The evolved melt accumulates until the for most of its life, and melt segregation processes are
melt fraction reaches the SLT and a mobile magma forms. required to collect and accumulate melt from along grain
Subsequent magma ascent is probably via dykes, fractures boundaries until a magma forms that can ascend through
and/or faults, which may rapidly drain the evolved melt the crust via fractures or dykes. The model results suggest

that has accumulated within the melt layer and allow it to that evolved (granitic, sensu lato) magmas form over time-
migrate to higher crustal levels (Fig. 12; Petford et al., 1993, scales of c. 10 kyr^1 Myr for typical time-averaged basalt
1994). At present, our models terminate when the SLT is emplacement rates in arcs (2^20 mm a1). Purely thermal
reached somewhere within the DCHZ; evolution of the models, which omit melt segregation processes, predict
DCHZ following magma mobilization and ascent of the that magma formation occurs over timescales that are typ-
magma to shallower levels in the crust are not addressed. ically an order of magnitude greater, and that the magma
Yet magma extraction is not terminal to the development composition is less evolved (broadly andesitic).
of a DCHZ. Further emplacement of basalt sills into the That evolved (granitic sensu lato) magmas may originate
DCHZ simply involves interaction with a more refractory in DCHZ is not a notable conclusion; it has long been
residuum. Magma mobilization occurs repeatedly as melt recognized that ‘partial re-melting to a sufficient degree to
accumulates in high-porosity layers, which are then produce batholithic quantities of magma must normally
drained into fractures as the magma ascends through the occur in very deep crustal or mantle environments which
crust. Thus there is a cyclical release of magma from the are rarely revealed by erosion, so that the processes of seg-
DCHZ, the frequency of which is dictated by the non-linear regation and collection of granitoid magmas thus become
interaction of heat and mass transfer processes leading to matters for speculation’ (Pitcher, 1979, p. 643). What is not-
the formation of melt layers, rather than the frequency of able here is that the coupled physical and chemical pro-
basaltic sill emplacement. Indeed, many cycles of basalt em- cesses of crystallization within the intruded basalt,
placement may lead to only a single cycle of evolved melting of the surrounding crust, and reactive flow of the
magma release. Once basaltic magma emplacement ceases, melts through the crystalline mush have a profound
the residuum cools and the density increase may be suffi- impact on the timescale of evolved magma formation and
cient to cause gravitation instability and recycling back into the composition of the magma; moreover, these processes
the mantle (e.g. Jull & Kelemen, 2001; Dufek & Bergantz, yield two contrasting styles of evolved magma formation.
2005). Thus evolved magma generation in DCHZ, followed In one, melt accumulation occurs in the overlying crust,
by ascent of the magma via dykes, fractures or faults to shal- and the resulting magma is composed of both crustal melt
lower crustal levels, provides a mechanism for differentiation and residual melt that has migrated upwards out of the
of the continental crust, while recycling of the dense re- intruded basalt; in the other, melt accumulation occurs in
siduum maintains total crustal thickness. the basalt layer, and the resulting magma is composed pri-
A final issue is that of basalt sill emplacement sequence. marily of residual melt. Static models, which omit melt mi-
It is clear that the evolution of a DCHZ during under- gration, fail to capture the migration of residual melt
and over-accretion differs appreciably, particularly at from the intruded basalt into the overlying crust, and
mid- to low emplacement rates. We consider these as from the underlying crust into the intruded basalt, which
end-member scenarios. What is now required is a better leads to mixing of melts.
understanding of the physics of sill emplacement to enable Magma composition is complicated by melt migration,
us to constrain the likely emplacement sequence and its because the melt migrates through, and chemically equili-
evolution with time. Recent models (e.g. Kavanagh et al., brates with, the matrix at progressively lower tempera-
2006) suggest that rheological contrasts play an important tures; moreover, mixing of melt sourced from the crust
role in controlling sill emplacement. As a DCHZ develops, and intruded sills yields chemical signatures that indicate
the location of rheological contrasts will change with time a mixed crustal and mantle origin. Cross-contamination
owing to melt migration. Consequently, it is likely that the of crustal and residual melts appears to be an inevitable
depth of sill emplacement changes with time, perhaps consequence of melt segregation in DCHZ, and is consist-
with the interfaces between regions of high and low melt ent with assimilation^fractional crystallization (AFC)
fraction providing rheological contrasts that control and meltingâssimilation^storage^hybridization (MASH)
emplacement. models. Previous thermal models have largely neglected
2022
the contribution of residual melt, focusing primarily on Bergantz, G. W. & Dawes, R. (1994). Aspects of magma generation
crustal melts formed by partial melting. Moreover, the and ascent in continental lithosphere. In: Ryan, M. P. (ed.)
Magmatic Systems. San Diego, CA: Academic Press, pp. 291^317.
complex dynamics of a DCHZ are unlikely to be captured
Bottinga, Y. & Weill, D (1972). The viscosity of magmatic silicate
by conventional models of fractional or equilibrium melt- liquids: a model for calculation. American Journal of Science 272,
ing and crystallization. As noted by (Pitcher, 1979, p. 90), 438^475.
‘in searching for the origin of granites, it is tempting to Boudreau, A. & Philpotts, A. R. (2002). Quantitative modeling of
view them as purely chemical systems’. Our results suggest compaction in the Holyoke flood basalt, Hartford Basin,
that the physics of melt segregation is likely to be just as Connecticut. Contributions to Mineralogy and Petrology 144, 176^184.
important in the chemical evolution of DCHZ as the Bowen, N. L. (1922). The behavior of inclusions in igneous magmas.
Journal of Geology 30, 513^570.
phase equilibria of the intruding basalt and surrounding
Brasse, H. & Soyer, W. (2001). A magnetotelluric study in the Southern
country rocks. Chilean Andes. Geophysical Research Letters 28, 3757^3760.

Brown, M. (1994).The generation, segregation, ascent and emplacement
of granitic magma: The migmatite-to-crustally derived granite con-
AC K N O W L E D G E M E N T S nection in thickened orogens. Earth-Science Reviews 36,83^130.
We gratefully acknowledge the significant contribution of Brown, M. & Rushmer, T. (1997). Consequences of deformation-
the editor and the reviewers to the successful publication assisted melt segregation: New views from the field and the labora-
of the paper. tory. In: Holness, M. (ed.) Deformation-enhanced Melt Segregation and
Metamorphic Fluid Transport. Mineralogical Society Series. London:
Chapman & Hall, pp. 111^139.
FUNDING Brown, M., Averkin, Y. A. & McLellan, E. L. (1995). Melt
segregation in migmatites. Journal of Geophysical Research 100,
Funding was obtained from the Leverhulme Trust under
15655^15679.
grant F/00182/AY. C.A., R.S.J.S. and J.D.B. received Castrucchio, A., Rust, A. & Sparks, R. S. J. (2010). Rheology and flow
European Research Council Advanced Grants. of crystal-rich bearing lavas: insights from analogue gravity cur-
rents. Earth and Planetary Science Letters 297, 471^480.
Chappell, B. W. (1984). Source rocks of I- and S-type granites in the
R EF ER ENC ES Lachlan Fold Belt, southeastern Australia. Philosophical Transactions
Alonso-Perez, R., Mu«ntener, O. & Ulmer, P. (2009). Igneous garnet of the Royal Society of London, Series A 310, 693^707.
and amphibole fractionation in the roots of island arcs: experimen- Chappell, B. W. & White, A. J. R. (2001). Two contrasting granite
tal constraints on andesitic liquids. Contributions to Mineralogy and types: 25 years later. AustralianJournal Of Earth Sciences 48, 489^499.
Petrology 157, 541^558. Chappell, B. W., White, A. J. R. & Wyborn, D. (1987). The importance
Annen, C. (2009). From plutons to magma chambers: thermal con- of residual source material (restite) in granite petrogenesis.
straints on the accumulation of eruptible silicic magma chambers Journal of Petrology 28, 1111^1138.
in the upper crust. Earth and Planetary Science Letters 284, 409^416. Chmielowski, J., Zandt, G. & Haberland, C. (1999). The central
Annen, C. (2011). Implications of incremental emplacement of magma Andean Altiplano^Puna magma body. Geophysical Research Letters
bodies for magma differentiation, thermal aureole dimensions and 26, 783^786.
plutonism-volcanism relationships. Tectonophysics 500, 3^10. Clarke, D. B. (1992). Granitoid Rocks, 1st edn. London: Chapman &
Annen, C. & Sparks, R. S. J. (2002). Effects of repetitive emplacement Hall, 283 p.
of basaltic intrusions on thermal evolution and melt generation in Clemens, J. D. (1990). The granulite^granite connexion. In:
the crust. Earth and Planetary Science Letters 203, 937^955. Vielzeuf, D. & Vidal, Ph. (eds) Granulites and Crustal Evolution.
Annen, C., Blundy, J. D. & Sparks, R. S. J. (2006). The genesis of inter- NATO ASI Series C 311, 25^36.
mediate and silicic magmas in deep crustal hot zones. Journal of Clemens, J. D. & Mawer, C. K. (1992). Granitic magma transport by
Petrology 47, 505^539. fracture propagation. Tectonophysics 204, 339^360.
Ariskin, A. A., Barmina, G. S., Ozerov, A. Y. & Nielsen, R. L. (1995). Clemens, J. D. & Vielzeuf, D. (1987). Constraints on melting and
Genesis of high-alumina basalts from Klyuchevskoi Volcano. magma production in the crust. Earth and Planetary Science Letters
Petrology 3, 449^472. 86, 287^306.
Atherton, M. P. (1993). Granite magmatism. Journal of the Geological Clifford, T. N., Stumfl, E. F., Burger, A. J., McCarthy, T. S. & Rex, D.
Society, London 150, 1009^1023. C. (1981). Mineral-chemical and isotopic studies of Namaqualand
Atherton, M. P. & Petford, N. (1993). Generation of sodium-rich granulites, South Africa: A Grenville analogue. Contributions to
magmas from newly underplated basaltic crust. Nature 362, Mineralogy and Petrology 77, 225^250.
144^146. Connolly, J. A. D. & Podladchikov, Y. Y. (1998). Compaction driven
Bachmann, O. & Bergantz, G. W. (2004). On the origin of crystal-poor fluid flow in viscoelastic rock. Geodinamica Acta 11, 55^84.
rhyolites: extracted from batholithic crystal mushes. Journal of Cooper, R. F. & Kohlstedt, D. L. (1984). Solution^precipitation
Petrology 45, 1565^1582. enhanced diffusional creep of partially molten olivine-basalt aggre-
Barnes, H. (1999). The yield stressça review or ‘panta rei’çevery- gates during hot pressing. Tectonophysics 107, 207^233.
thing flows? Journal of Non-Newtonian Fluid Mechanics 81, 133^178. Costa, A. (2005). Viscosity of high crystal content melts: Dependence
Beard, J. S. & Lofgren, G. E. (1991). Dehydration melting and on solid fraction. Geophysical Research Letters 32, L22308, doi:10.1029/
water-saturated melting of basaltic and andesitic greenstones and 2005GL024303.
amphibolites at 1, 3, and 6·9 kb. Journal of Petrology 32, 365^401. Costa, A., Caricchi, L. & Bagdassarov, N. (2009). A model for the rhe-
Bergantz, G. W. (1989). Underplating and partial melting: implications ology of particle-bearing suspensions and partially molten rocks.
for melt generation and extraction. Science 245, 1093^1095. Geochemistry, Geophysics, Geosystems 10, Q03010.
2023
Crisp, J. A. (1984). Rates of magma emplacement and volcanic output. elastic models. Journal of Volcanology and Geothermal Research 166,
Journal of Volcanology and Geothermal Research 20, 177^211. 47^75.
Cull, J. P., O’Reilly, S. Y. & Griffin, W. L. (1991). Xenolith geotherms Grove, T. L. & Kinzler, R. J. (1986). Petrogenesis of Andesites. Annual
and crustal models in Eastern Australia. Tectonophysics 192, 359^366. Review Of Earth And Planetary Sciences 14, 417^454.
Dell’Angelo, L. N., Tullis, J. & Yund, R. A. (1987). Transition from dis- Grove, T. L., Donnelly-Nolan, J. M. & Housh, T. (1997). Magmatic
location creep to melt-enhanced diffusion creep in fine-grained processes that generated the rhyolite of Glass Mountain, Medicine
granitic aggregates. Tectonophysics 139, 325^332. Lake volcano, N California. Contributions to Mineralogy and Petrology
De Paolo, D. J. (1981). Trace element and isotopic effects of combined 127, 205^223.
wallrock assimilation and fractional crystallization. Earth and Grove, T. L., Kinzler, R. J., Baker, M. B., Donnelly-Nolan, J. M. &
Planetary Science Letters 53, 189^202. Lesher, C. E. (1988). Assimilation Of Granite by Basaltic Magma
De Paolo, D. J., Perry, F. V. & Baldridge, W. S. (1992). Crustal versus at Burnt Lava Flow, Medicine Lake Volcano, Northern
Mantle Sources of Granitic Magmas ^ A 2-Parameter Model California ^ Decoupling of Heat And Mass Transfer. Contributions
Based on Nd Isotopic Studies. Transactions of the Royal Society of to Mineralogy and Petrology 99, 320^343.

Edinburgh-Earth Sciences 83, 439^446. Grove, T. L., Baker, M. B., Price, R. C., Parman, S. W., Elkins-
Deves, M., Tait, S., King, G. C., Grandin, R. & Tapponnier, P. (2010). Tanton, L. T., Chatterjee, N. & Mu«ntener, O. (2005). Magnesian
Continental magmatism by shear heating at geometric complex- andesite and dacite lavas from Mt. Shasta, northern California:
ities on fault systems, American Geophysical Union, Fall Meeting products of fractional crystallization of H2O-rich mantle melts.
2010, abstract T23A-2244. Contributions to Mineralogy and Petrology 148, 542^565.
Drummond, B. J. & Collins, C. D. N. (1986). Seismic evidence for Hacker, B. R. & Abers, G. A. (2004). Subduction Factory 3: An Excel
underplating of the lower continental crust in Australia. Earth and worksheet and macro for calculating the densities, seismic wave
Planetary Science Letters 79, 361^372. speeds, and H2O contents of minerals and rocks at pressure and
Ducea, M. N. & Saleeby, J. B. (1998a). The age and origin of a thick temperature. Geochemistry, Geophysics, Geosystems 5, Q01005,
maficûltramafic keel from beneath the Sierra Nevada batholith. doi:10.1029/2003GC000614.
Contributions to Mineralogy and Petrology 133(1^2), 169^185. Heath, E., Turner, S. P., Macdonald, R., Hawkesworth, C. J. & van
Ducea, M. N. & Saleeby, J. B. (1998b). A case for delamination of the Calsteren, P. (1998). Long magma residence times at an island are
deep batholithic crust beneath the Sierra Nevada, California. volcano (Soufrie're, St. Vincent) in the Lesser Antilles: evidence
International Geology Review 40(1), 78^93. from U-238^Th-230 isochron dating. Earth and Planetary Science
Dufek, J. & Bergantz, G. W. (2005). Lower crustal magma genesis and Letters 160, 49^63.
preservation: a stochastic framework for the evaluation of basalt^ Hersum, T. G., Marsh, B. D. & Simon, A. C. (2007). Contact partial
crust interaction. Journal of Petrology 46, 2167^2195. melting of granitic country rock, melt segregation, and re-injection
Dungan, M. A. & Davidson, J. (2004). Partial assimilative recycling of as dikes into Ferrar Dolerite Sills, McMurdo Dry Valleys,
the mafic plutonic roots of arc volcanoes: an example from the Antarctica. Journal of Petrology 48, 2125^2148, doi:10.1093/petrology/
Chilean Andes. Geology 32, 773^776. egm054.
England, P. C. & Thompson, A. B. (1984). Pressure-Temperature-Time Hildreth, W. & Moorbath, S. (1988). Crustal contributions to arc mag-
paths of regional metamorphism I. Heat transfer during the evolu- matism in the Andes of Central Chile. Contributions to Mineralogy
tion of regions of thickened continental crust. Journal of Petrology and Petrology 98, 455^489.
25, 894^928. Hirose, K. & Kawamura, K. (1994). A new experimental approach for
Feeley, T. C. & Davidson, J. P. (1994). Petrology of calc-alkaline lavas incremental batch melting of peridotite at 1·5 GPa. Geophysical
at Volca¤n Ollague and the origin of compositional diversity at Research Letters 21, 2139^2142.
Central Andean stratovolcanoes. Journal of Petrology 35, 1295^1340. Hodge, D. S. (1974). Thermal model for origin of granitic batholiths.
Feeley, T. C. & Hacker, M. D. (1995). Intracrustal derivation of Nature 251, 297^299.
Na-rich andesitic and dacitic magmasçan example from Volca¤n Holbrook, W. S., Lizarralde, D., McGeary, S., Bangs, N. & Diebold, J.
Ollague, Andean Central Volcanic Zone. Journal of Geology 103, (1999). Structure and composition of the Aleutian island arc and
213^225. implications for continental crustal growth. Geology 27, 31^34.
Garrison, J. M. & Davidson, J. P. (2003). Dubious case for slab melting Hollister, L. S. & Crawford, M. L. (1986). Melt-enhanced deform-
in the Northern volcanic zone of the Andes. Geology 31, 565^568. ation; A major tectonic process. Geology 14, 558^561.
Getsinger, A., Rushmer, T., Jackson, M. D. & Baker, D. (2009). Huppert, H. E. & Sparks, R. S. J. (1988). The generation of granitic
Generating high Mg-numbers and chemical diversity in tonalite^ magmas by intrusion of basalt into continental crust. Journal of
trondhjemite^granodiorite (TTG) magmas during melting and Petrology 29, 599^624.
melt segregation in the continental crust. Journal of Petrology 50, Irvine, T. N. (1974). Petrology of the Duke Island Ultramafic
1935^1954. Complex, Southeastern Alaska. Geological Society of America,
Gill, J. (1981). Orogenic Andesites and Plate Tectonics. Berlin: Springer, Memoirs 138.
390 p. Jackson, M. D. & Cheadle, M. J. (1998). A continuum model for the
Giordano, D. & Dingwell, D. B. (2003). Non-Arrhenian multicompo- transport of heat, mass and momentum in a deformable, multicom-
nent melt viscosity: a model. Earth and Planetary Science Letters 208, ponent mush, undergoing solid^liquid phase change. International
337^349. Journal of Heat and MassTransfer 41, 1035^1048.
Giordano, D., Russell, J. K. & Dingwell, D. B. (2008). Viscosity of Jackson, M. D., Cheadle, M. J. & Atherton, M. P. (2003).
magmatic liquids: A model. Earth and Planetary Science Letters 271, Quantitative modeling of granitic melt generation and segregation
123^134. in the continental crust. Journal of Geophysical Research 108,
Glazner, A. F. (2007). Thermal limitations on incorporation of wall 2332^2353.
rock into magma. Geology 35(4), 319^322. Jackson, M. D., Gallagher, K., Petford, N. & Cheadle, M. J. (2005).
Grosfils, E. B. (2007). Magma reservoir failure on the terrestrial pla- Towards a coupled physical and chemical model for tonalite^
nets: assessing the importance of gravitational loading in simple trondhjemite^granodiorite magma formation. Lithos 79, 43^60.
2024
Jull, M. & Kelemen, P. B. (2001). On the conditions for lower Morton, K. M. & Mayers, D. F. (2005). Numerical Solution of Partial
crustal convective instability. Journal of Geophysical Research 106, Differential Equations. Cambridge: Cambridge University Press.
6423^6446. Muntener, O., Keleman, P. B. & Grove, T. L. (2001). The role of H2O
Katz, R. F. (2008). Magma dynamics with the enthalpy method: during crystallization of primitive arc magmas under uppermost
Benchmark solutions and magmatic focusing at mid-ocean ridges. mantle conditions and genesis of pyroxenites: an experimental
Journal of Petrology 49, 2099^2121. study. Contributions to Mineralogy and Petrology 141, 643^658.
Kavanagh, J. L., Menand, T. & Sparks, R. S. J. (2006). An experimen- Murase, T. & McBirney, A. R. (1973). Properties of some common ig-
tal investigation of sill formation and propagation in layered elastic neous rocks and their melts at high temperatures. Geological Society
media. Earth and Planetary Science Letters 245, 799^813. of America Bulletin 84, 3563^3592.
Kay, R. W. & Mahlburg-Kay, S. (1991). Creation and destruction of Musselwhite, D. S., De Paolo, D. J. & McCurry, M. (1989). The evolu-
lower continental crust. Geologische Rundschau 80, 259^278. tion of a silicic magma systemçisotopic and chemical evidence
Kimbell, G. S. & Richards, P. C. (2008). The three-dimensional litho- from the Woods Mountains Volcanic Center, Eastern California.
spheric structure of the Falkland Plateau region based on gravity Contributions to Mineralogy and Petrology 101, 19^29.

modelling. Journal of the Geological Society, London 165, 795^806. Petford, N. (1995). Segregation of tonalitic^trondhjemitic melts in the
Kinzler, R. J. (1997). Melting of mantle peridotite at pressures ap- continental crust: The mantle connection. Journal of Geophysical
proaching the spinel to garnet transition: Application to mid-ocean Research 100, 15735^15743.
ridge basalt petrogenesis. Journal of Geophysical Research 102(B1), Petford, N. (2003). Rheology of granitic magmas during ascent and
853^874. emplacement. Annual Review of Earth and Planetary Sciences 31,
Kinzler, R. J. & Grove, T. L. (1992). Primary magmas of mid- 399^427.
oceanic basalts 2: Applications. Journal of Geophysical Research 97, Petford, N. & Atherton, M. P. (1996). Na-rich partial melts from newly
6907^6926. underplated basaltic crust: The Cordillera Blanca Batholith, Peru.
Klein, E. M. & Langmuir, C. H. (1987). Global correlations of ocean Journal of Petrology 37, 1491^1521.
ridge basalt chemistry with axial depth and crustal thickness. Petford, N. & Gallagher, K. (2001). Partial melting of mafic (amphi-
Journal of Geophysical Research 92(B8), 8089^8115. bolitic) lower crust by periodic influx of basaltic magma. Earth
Klugel, A., Hansteen, T. H. & Galipp, K. (2005). Magma storage and and Planetary Science Letters 193, 483^499.
underplating beneath Cumbre Vieja volcano, La Palma (Canary Petford, N. & Koenders, M. A. (1998). Self-organisation and fracture
Islands). Earth and Planetary Science Letters 236(1^2), 211^226. connectivity in rapidly heated continental crust. Journal of
Kohlstedt, D. L. & Chopra, P. N. (1994). Influence of basaltic melt on Structural Geology 20, 1425^1434.
the creep of polycrystalline olivine under hydrous conditions. In: Petford, N., Kerr, R. C. & Lister, J. R. (1993). Dike transport of gran-
Ryan, M. P. (ed.) Magmatic Systems. New York: Academic Press, itoid magmas. Geology 21, 845^848.
pp. 34^56. Petford, N., Lister, J. R. & Kerr, R. C. (1994). The ascent of felsic
Krawczyk, C. M., Stiller, M. & Group, D.-B. R. (1999). Reflection magmas in dykes. Lithos 32, 161^168.
seismic constraints on Paleozoic crustal structure and Moho be- Petford, N., Cruden, A. R., McCaffrey, K. J. W. & Vigneresse, J.-L.
neath the NE German Basin. Tectonophysics 314, 241^253. (2000). Granite magma formation, transport and emplacement in
Kushiro, I. (2001). Partial melting experiments on peridotite and the Earth’s crust. Nature 408, 669^673.
origin of mid-ocean ridge basalt. Annual Review of Earth and Pharr, G. M. & Ashby, M. F. (1983). On creep enhanced by a liquid
Planetary Sciences 29, 71^107. phase. Acta Metallurgica 31, 129^138.
Leake, B. E. (1990). Granite magmas: their sources, initiation and con- Pin, C. (1990). Evolution of the lower crust in the Ivrea Zone: A model
sequences of emplacement. Journal of the Geological Society, London based on isotopic and geochemical data. In: Vielzeuf, D. &
147, 579^589. Vidal, Ph. (eds) Granulites and Crustal Evolution. NATO ASI Series C
Lezaeta, P. & Brasse, H. (2001). Electrical conductivity beneath the 311, 59^85.
volcanoes of the NW Argentinian Puna. Geophysical Research Letters Pitcher, W. S. (1979). The nature, ascent and emplacement of granitic
28, 4651^4654. magmas. Journal of the Geological Society, London 136, 627^662.
Lupulescu, A. & Watson, E. B. (1999). Low-melt fraction connectivity Pitcher, W. S. (1993). The Nature and Origin of Granites, 1st edn. Glasgow:
of granitic and tonalitic melts in a mafic crustal rock at 8008C and Blackie, 321 p.
1 GPa. Contributions to Mineralogy and Petrology 134, 202^216. Rabinowicz, M. & Vigneresse, J.-L. (2004). Melt segregation under
Macdonald, R., Hawkesworth, C. J. & Heath, E. (2000). The Lesser compaction and shear channeling: Application to granitic magma
Antilles volcanic chain: a study in are magmatism. Earth-Science segregation in a continental crust. Journal of Geophysical Research
Reviews 49, 1^76. 109, B04407, doi:10.1029/2002JB002372.
Majdanski, M., Kozlovskaya, E. & Grad, M. (2007). 3D structure of Ranalli, G. (1987). Rheology of the Earth: Deformation and Flow Processes
the Earth’s crust beneath the northern part of the Bohemian in Geophysics and Geodynamics, 2nd edn. London: Allen & Unwin,
Massif. Tectonophysics 437(1^4), 17^36. 366 p.
Marsh, B. D. (1981). On the crystallinity, probability of occurrence, Rapp, R. P. & Watson, E. B. (1995). Dehydration melting of metaba-
and rheology of lava and magma. Contributions to Mineralogy and salt at 8^32 kbar: Implications for continental growth and crust^
Petrology 78, 85^98. mantle recycling. Journal of Petrology 36, 891^931.
McKenzie, D. P. (1984). The generation and compaction of partially Reiners, P. W. (1998). Reactive melt transport in the mantle and geo-
molten rock. Journal of Petrology 25, 713^765. chemical signatures of mantle-derived magmas. Journal of Petrology
McKenzie, D. & Bickle, M. J. (1988). The volume and composition of 39, 1039^1061.
melt generated by extension of the lithosphere. Journal of Petrology Reubi, O. & Blundy, J. D. (2009). A dearth of intermediate melts at
29, 625^679. subduction zone volcanoes and the petrogenesis of arc andesites.
Miller, C. F., Watson, E. B. & Harrison, T. M. (1988). Perspectives Nature 461, 1269^1273.
on the source, segregation and transport of granitiod magmas. Reubi, O. & Nicholls, I. A. (2004). Magmatic evolution at Batur vol-
Transactions of the Royal Society of Edinburgh, Earth Sciences 79, 135^156. canic field, Bali, Indonesia: petrological evidence for polybaric
2025
fractional crystallization and implications for caldera-forming Spiegelman, M. & Kenyon, P. (1992). The requirements for chemical
eruptions. Journal of Volcanology and Geothermal Research 138, 345^369. disequilibrium during magma migration. Earth and Planetary
Rodriguez, C., Selles, D., Dungan, M., Langmuir, C. & Leeman, W. Science Letters 109, 611^620.
(2007). Adakitic dacites formed by intracrustal crystal fractionation Stanley, W. D., Finn, C. & Plesha, J. L. (1987). Tectonics and conductiv-
of water-rich parent magmas at Nevado de Longav|¤ volcano ity structures in the southern Washington Cascades. Journal of
(36·28S; Andean Southern Volcanic Zone, central Chile). Journal of Geophysical Research 92, 10179^10193.
Petrology 48, 2033^2061. Stratford, W. R. & Stern, T. A. (2006). Crust and upper mantle struc-
Rogers, G. & Hawkesworth, C. J. (1989). A geochemical traverse ture of a continental backarc: central North Island, New Zealand.
across the North Chilean Andesçevidence for crust generation Geophysical Journal International 166(1), 469^484.
from the mantle wedge. Earth and Planetary Science Letters 91, Tegner, C., Thy, P., Holness, M. B., Jakobsen, J. K. & Lesher, C. E.
271^285. (2009). Differentiation and compaction in the Skaergaard intru-
Rosenberg, C. L. & Handy, M. R. (2005). Experimental deformation sion. Journal of Petrology 50, 813^840.
of partially melted granite revisited: implications for the continen- Tepper, J. H., Nelson, B. K., Bergantz, G. W. & Irving, A. J. (1993).

tal crust. Journal of Metamorphic Geology 23, 19^28. Petrology of the Chilliwack Batholith, North Cascades,
Rudnick, R. L. & Fountain, D. M. (1995). Nature and composition of Washington: generation of calc-alkaline granitoids by melting of
the continental crust: a lower crustal perspective. Reviews of mafic lower crust with variable water fugacity. Contributions to
Geophysics 33, 267^309. Mineralogy and Petrology 113, 333^351.
Rushmer, T. (1991). Partial melting of two amphibolites: contrasting Ulmer, P. (2007). Differentiation of mantle-derived calc-alkaline
experimental results under fluid-absent conditions. Contributions to magmas at mid to lower crustal levels: experimental and petrologic
Mineralogy and Petrology 107, 41^59. constraints. Periodico di Mineralogia 76, 309^325.
Rushmer, T. (2001). Volume change during partial melting reactions: van der Molen, I. & Paterson, M. S. (1979). Experimental deformation
Implications for melt extraction, melt geochemistry and crustal of partially melted granite. Contributions to Mineralogy and Petrology
rheology. Tectonophysics 34(2/3^4), 389^405. 70, 299^318.
Rushmer, T. & Jackson, M. D. (2006). Impact of melt segregation on Vielzeuf, D., Clemens, J. D., Pin, C. & Moinet, E. (1990). Granites,
tonalite^trondhjemite^granodiorite (TTG) petrogenesis. granulites, and crustal differentiation. In: Vielzeuf, D. &
Transactions of the Royal Society of Edinburgh 97, 325^336. Vidal, Ph. (eds) Granulites and Crustal Evolution. NATO ASI Series C
Rushmer, T. & Miller, S. (2006). Melt migration in the continental 311, 59^85.
crust and generation of lower crustal permeability: inferences von Bargen, N. & Waff, H. S. (1986). Permeabilities, interfacial areas
from experimental studies and modeling. In: Brown, M. & and curvatures of partially molten systems: results of numerical
Rushmer, T. (eds) Evolution and Differentiation of the Continental Crust. computations of equilibrium microstructures. Journal of Geophysical
Cambridge: Cambridge University Press, pp. 430^454. Research 91, 9261^9276.
Sandrin, A. & Thybo, H. (2008). Seismic constraints on a large mafic Voss, M. & Jokat, W. (2007). Continentôcean transition and volumin-
intrusion with implications for the subsidence history of the ous magmatic underplating derived from P-wave velocity model-
Danish Basin. Journal of Geophysical Research 113, B09402, ling of the East Greenland continental margin. Geophysical Journal
doi:10.1029/2007JB005067. International 170(2), 580^604.
Sawyer, E. W. (1991). Disequilibrium melting and the rate of melt^re- Wager, L. R. (1961). A note on the origin of ophitic textures in the
siduum separation during migmatisation of mafic rocks from the chilled olivine gabbro of the Skaergard Intrusion. Geological
Grenville Front, Quebec. Journal of Petrology 32, 701^738. Magazine 98, 353^366.
Sawyer, E. W. (1994). Melt segregation in the continental crust. Geology White, A. J. R. & Chappell, B. W. (1990). Per magma ad migma
22, 1019^1022. downunder. Geological Journal 25, 221^225.
Seber, S., Barazangi, M., Ibenbrahim, A. & Demnati, A. (1996). Wickham, S. M. (1987). The segregation and emplacement of granitic
Geophysical evidence for lithospheric delamination beneath the magmas. Journal of the Geological Society, London 144, 281^297.
Alboran Sea and Rif^Betic mountains. Nature 379, 785^790. Wilson, J. R., Esbensen, K. H. & Thy, P. (1981). Igneous petrology of
Sen, C. & Dunn, T. (1994). Dehydration melting of a basaltic compos- the synorogenic Fongen^Hyllingen layered basic complex,
ition amphibolite at 1·5 and 2·0 GPaçimplications for the origin Southern Norwegian Caledonides. Journal of Petrology 22, 584^627.
of adakites. Contributions to Mineralogy and Petrology 117, 394^409. Wolf, M. B. & Wyllie, P. J. (1994). Dehydration-melting of amphibolite
Shaw, H. R. (1963). Obsidian^H2O viscosities at 1000 and 2000 bars in at 10 kbar: the effects of temperature and time. Contributions to
the temperature range 7008 to 9008C. Journal of Geophysical Mineralogy and Petrology 115, 369^379.
Research 68, 6337^6343. Xia, S.-h., Qiu, X.-l., Zhao, M.-h., Xu, H.-l. & Shi, X.-b. (2010).
Shirley, D. N. (1986). Compaction of igneous cumulates. Journal of Analysis of crustal average velocity and Moho depth beneath the
Geology 94, 795^809. onshoreôffshore transitional zone in the northern South China
Sisson, T. W., Ratajeski, K., Hankins, W. B. & Glazner, A. F. (2005). Sea. Journal of Tropical Oceanography 29(4), 63^70.
Voluminous granitic magmas from common basaltic sources. Zandt, G., Leidig, M., Chmielowski, J., Baumont, D. & Yuan, X. H.
Contributions to Mineralogy and Petrology 148, 635^661. (2003). Seismic detection and characterization of the Altiplano^
Smith, D. R. & Leeman, W. P. (1987). Petrogenesis of Mount St. Puna magma body, central Andes. Pure and Applied Geophysics
Helens dacitic magmas. Journal of Geophysical Research 92, 160(3^4), 789^807.
10313^10334. Zandt, G., Gilbert, H., Owens, T. J., Ducea, M., Saleeby, J. &
Spera, F. J. & Bohrson, W. A. (2001). Energy-constrained open-system Jones, C. H. (2004). Active foundering of a continental arc root be-
magmatic processes I: General model and energy-constrained as- neath the southern Sierra Nevada in California. Nature 431(7004),
similation and fractional crystallization (EC-AFC) formulation. 41^46.
Journal of Petrology 42, 999^1018.
2026

Egs 041

Uploaded by

Copyright:

Available Formats

Egs 041

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Egs 041

Uploaded by

Copyright:

Available Formats

JOURNAL OF PETROLOGY VOLUME 53 NUMBER 10 PAGES 1999^2026 2012 doi:10.

Melt Segregation in Deep Crustal Hot Zones:

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

RECEIVED JANUARY 28, 2011; ACCEPTED MAY 18, 2012

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

high-pressure differentiation within DCHZ is often pre- M E LT G E N E R AT I O N A N D

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

conservation equation (Katz, 2008). This allows the model

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Table 1: Nomenclature In ‘static’ melting (or crystallization) models and experi-

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Table 4: Physical parameters describing melt migration (Jackson et al., 2003)

Symbol Description Min.–max. values Values used Units

a matrix grain radius 5  104–5  103 2·5  103 m

Evolution of melt fraction through time

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

1 mm/a 5 mm/a 10 mm/a 25 mm/a

1.72 Ma 0.42 Ma 0.08 Ma

-31 base of -31

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

1 mm/a 5 mm/a 10 mm/a 25 mm/a

1.45 Ma 0.35 Ma 0.07 Ma

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

1 mm/a 5 mm/a 10 mm/a 25 mm/a

12.02 Ma 0.51 Ma 0.09 Ma

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

-32 Basalt Layer

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

-26 Amphibolite Crust (b) (e)

uncertainty in the matrix and, in particular, the melt vis- DISCUSSION

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

(Fig. 11). Productivity here is defined as the volume of 20

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Differentiation of the continental crust CONC LUSIONS

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

Downloaded from https://academic.oup.com/petrology/article/53/10/1999/1475184 by guest on 22 May 2023

You might also like

a matrix grain radius 5 104–5 103 2·5 103 m