Engineering of Viscoelasticity

Danton Gutierrez-Lemini
Engineering
Viscoelasticity
Engineering Viscoelasticity
Engineering
Viscoelasticity
123
Special Products Division
Oil States Industries, Inc.
Arlington, TX
USA
Additional material to this book can be downloaded from http://extras.springer.com/
ISBN 978-1-4614-8138-6
DOI 10.1007/978-1-4614-8139-3
ISBN 978-1-4614-8139-3
(eBook)
Springer New York Heidelberg Dordrecht London

Library of Congress Control Number: 2013942016
Springer Science+Business Media New York 2014
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or
information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed. Exempted from this legal reservation are brief
excerpts in connection with reviews or scholarly analysis or material supplied specifically for the
purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the
work. Duplication of this publication or parts thereof is permitted only under the provisions of
the Copyright Law of the Publishers location, in its current version, and permission for use must
always be obtained from Springer. Permissions for use may be obtained through RightsLink at the
Copyright Clearance Center. Violations are liable to prosecution under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of
publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for
any errors or omissions that may be made. The publisher makes no warranty, express or implied, with
respect to the material contained herein.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)
To Lolita, the love of my life; whose support,

faith and perseverance have made this book
possible
To my daughters, Gabriela and Paula,
who made parenthood such a blessing!
Preface
Elastic solids and viscous fluids are two types of engineering materials whose
response to loads, almost everyone, either seems to understand or takes for
granted. Then there are materials whose response to loads combines the features of
both elastic solids and viscous fluids. Not surprisingly, these materials are called
viscoelastic and are a little trickier to understand than elastic solids or viscous
fluids. The engineering discipline that developed to provide a rigorous mathematical framework to describe the behavior of such materials is called viscoelasticity. This book presents a comprehensive treatment of the theory and
applications of viscoelasticity.
Polymers are viscoelastic materials. The term polymer has been around since
Berzelius used the word polymeric in 1832; at a time when chemists were still
unsure of the structure of even the simplest of molecules. Today, it is hard to
imagine our world without polymers. Polymers and polymeric-based products are
commonplace in virtually every industry. This is unquestionably true of the
aerospace, rubber, oil, automotive, electronics, construction, piping, and appliances industries; and many more. Yet, despite the fact that viscoelasticity has been
taught in universities for several decades, providing the necessary tools to design
with polymers, today many polymeric-based products are still designed as if the
materials involved were elastic. One reason for this practice is that viscoelasticity
has been taught exclusively at the graduate level, yet most practicing professionals
lack an advanced degree in engineering; and those with advanced degrees, never
studied viscoelasticity, because the subject is usually an elective one.
If one thinks about it, the basic design courses, such as machine design, structural steel design, reinforce concrete design, and so on, are taught at the undergraduate level. The foundation of all these courses is mechanics of materialsthe
strength of materials of oldwhose mastery requires a background in statics and
some differential and integral calculus. If truth be told, the derivations of the design
equations for viscoelastic materials are the same as for elastic solids; and the
resulting expressions, virtually identical. The difference lies in that for viscoelastic
materials the relationship between stress and strain is not an algebraic product of a
constant modulus and the strain, as it is for elastic solids, but is given by a special
vii
viii
Preface
type of productcalled convolutionbetween a function which represents the

modulus, and the time derivative of the strain. The point being made here is that it is
just as demandingperhaps only a tad moreto learn the art of designing with
viscoelastic materials, as it is to learn the mechanics of elastic solids.
This book is intended to help Academia close the gap that exists between
current practice and the proper way to designing with polymers, as well as to serve
as a text on the theory of viscoelasticity. The book accomplishes these goals by
presenting a self-contained, rigorous, and comprehensive treatment of all the
topics that are relevant to the mechanical behavior of viscoelastic materials, and by
providing all the background in mathematics and mechanics that are central to
understanding the subject being presented.
As will be seen, Chaps. 17 could be used to teach a complete course on
viscoelasticity at the undergraduate level. These chapters cover the theory in the
one-dimensional form needed for design. The same chapters, complemented with
Appendix Awhich provides the mathematical background used in the derivations of the theorycontain all that a practicing professional would need to master
the art of designing with polymers.
All equations in these chapters are developed from first principles, without
presuming previous knowledge of the subject matter being presented. This
approach is followed for two reasons: first, because it is necessary for readers who
have no formal training in mechanics of materials; and second, because it provides
a method to follow when the use of popular engineering shortcuts, like the use
integral transform techniques, might not be clear.
The contents of Appendix Bwhich provides an introduction to tensors and an
overview of solid mechanics together with Chaps. 811, are written with the
graduate student in mind. A graduate-level course in viscoelasticity could be
taught with this material, and perhaps selected sections of earlier chapters, as a
continuation to an undergraduate course.
Arlington, TX, USA
Acknowledgments
I would like to express my deep appreciation to Michael Luby, Engineering Editor

at Springer, for believing in the manuscript; and because his easy and efficient
business manner gave me the encouragement to pursue the writing of this book. I
would also like to thank Merry Stuber, Michaels Assistant Editor, for her kindness, patience, and professional advice during the editorial process. Thanks are
also due to Scott Sykes, for his invaluable help in formatting the original manuscript. Special thanks go to the reviewers of the original manuscript, for having
endured its reading before I had a chance to review it myself, after letting it
simmer for a while.
ix
Contents
Fundamental Aspects of Viscoelastic Response . . . . . . . . .

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 The Nature of Amorphous Polymers . . . . . . . . . . . . . .
1.3 Mechanical Response of Viscoelastic Materials . . . . . .
1.3.1
Material Response to Step-Strain Loading . . . .
1.3.2
Material Response to Step-Stress Loading . . . .
1.3.3
Material Response to Cyclic Strain Loading . .
1.3.4
Material Response to Constant Strain
Rate Loading . . . . . . . . . . . . . . . . . . . . . . . .
1.4 Energy Storage and Dissipation . . . . . . . . . . . . . . . . .
1.5 Glass Transition and Regions of Viscoelastic Behavior .
1.6 Aging of Viscoelastic Materials . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
1
3
4
5
7
10
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
11
12
16
20
21
Constitutive Equations in Hereditary Integral Form

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Boltzmanns Superposition Principle . . . . . . . . .
2.3 Principle of Fading Memory . . . . . . . . . . . . . .
2.4 Closed-Cycle Condition. . . . . . . . . . . . . . . . . .
2.5 Relationship Between Modulus and Compliance
2.5.1
Elastic Relationships . . . . . . . . . . . . . .
2.5.2
Convolution Integral Relationships . . . .
2.5.3
Laplace-Transformed Relationships . . .
2.6 Alternate Integral Forms . . . . . . . . . . . . . . . . .
2.7 Work and Energy . . . . . . . . . . . . . . . . . . . . . .
2.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
23
24
25
30
34
37
38
39
41
43
46
48
52
Constitutive Equations in Differential Operator Form

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Fundamental Rheological Models . . . . . . . . . . . .
3.2.1
Linear Elastic Spring . . . . . . . . . . . . . .
3.2.2
Linear Viscous Dashpot . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
53
54
54
55
57
xi
xii
Contents
3.3
Rheological Operators . . . . . . . . . . . . . . . . . . . . . . . .
3.3.1
Fundamental Rheological Operators . . . . . . . .
3.3.2
General Rheological Operators. . . . . . . . . . . .
3.3.3
Rheological Equations in Laplace-Transformed
Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.4
Initial Conditions for Rheological Models . . . .
3.4 Construction of Rheological Models . . . . . . . . . . . . . .
3.5 Simple Rheological Models . . . . . . . . . . . . . . . . . . . .
3.5.1
KelvinVoigt Solid . . . . . . . . . . . . . . . . . . . .
3.5.2
MaxwellWiechert Fluid . . . . . . . . . . . . . . . .
3.6 Generalized Models . . . . . . . . . . . . . . . . . . . . . . . . .
3.6.1
Generalized Kelvin Model . . . . . . . . . . . . . . .
3.6.2
Generalized Maxwell Model . . . . . . . . . . . . .
3.7 Composite Models . . . . . . . . . . . . . . . . . . . . . . . . . .
3.7.1
Standard Linear Solid . . . . . . . . . . . . . . . . . .
3.7.2
Three-Parameter Fluid. . . . . . . . . . . . . . . . . .
3.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4
.....
.....
.....
59
60
61
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
63
63
67
69
69
74
79
79
81
83
84
88
89
91
......
......
93
93
......
94
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Constitutive Equations for Steady-State Oscillations . . . .

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Steady-State Constitutive Equations from Integral
Constitutive Equations. . . . . . . . . . . . . . . . . . . . . . .
4.3 Steady-State Constitutive Equations from Differential
Constitutive Equations. . . . . . . . . . . . . . . . . . . . . . .
4.4 Limiting Behavior of Complex Property Functions . . .
4.5 Energy Dissipation . . . . . . . . . . . . . . . . . . . . . . . . .
4.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
102
105
109
110
112
Structural Mechanics . . . . . . . . . . . . . . . . . . . . .
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Bending. . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.1
Hereditary Integral Models . . . . . . .
5.2.2
Differential Operator Models . . . . . .
5.2.3
Models for Steady-State Oscillations.
5.3 Torsion . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.1
5.3.2
5.3.3
Models for Steady-State Oscillations.
5.4 Column Buckling . . . . . . . . . . . . . . . . . . . .
5.4.1
5.4.2
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
113
113
115
117
118
119
120
121
122
123
124
125
127
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Contents
5.5
Viscoelastic Springs . . . . . . . . . . . .
5.5.1
Axial Spring . . . . . . . . . . .
5.5.2
Shear Spring . . . . . . . . . . .
5.5.3
Bending Spring . . . . . . . . .
5.5.4
Torsion Spring . . . . . . . . . .
5.6 ElasticViscoelastic Correspondence.
5.7 Mechanical Vibrations. . . . . . . . . . .
5.7.1
Forced Vibrations . . . . . . . .
5.7.2
Free Vibrations . . . . . . . . .
5.8 Problems . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . .
xiii
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
128
129
131
132
134
136
138
139
144
145
148
Temperature Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2 Time Temperature Superposition . . . . . . . . . . . . . . . .
6.3 Phenomenology of the Glass Transition . . . . . . . . . . .
6.4 Effect of Pressure on the Glass Transition Temperature
6.5 Hygrothermal Strains . . . . . . . . . . . . . . . . . . . . . . . .
6.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
149
149
150
157
159
160
161
163
Material Property Functions and Their Characterization. . . . .

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.2 Experimental Characterization . . . . . . . . . . . . . . . . . . . . .
7.2.1
Constant Strain Test . . . . . . . . . . . . . . . . . . . . . .
7.2.2
Constant Stress Test . . . . . . . . . . . . . . . . . . . . . .
7.2.3
Constant Rate Test . . . . . . . . . . . . . . . . . . . . . . .
7.2.4
Dynamic Tests . . . . . . . . . . . . . . . . . . . . . . . . . .
7.3 Analytical Forms of Constitutive Functions . . . . . . . . . . . .
7.3.1
Material Property Functions in Prony Series Form .
7.3.2
Material Property Functions in Power-Law Form . .
7.4 Inversion of Material Property Functions. . . . . . . . . . . . . .
7.4.1
Approximate Inversion of Material Property
Functions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.4.2
Exact Inversion of Material Property Functions . . .
7.5 Numerical Characterization of Material Property Functions .
7.5.1
Numerical Characterization of the Shift Function . .
7.5.2
Numerical Characterization of Modulus
and Compliance . . . . . . . . . . . . . . . . . . . . . . . . .
7.6 Source Data Files and Computer Application. . . . . . . . . . .
7.6.1
Source Data Files . . . . . . . . . . . . . . . . . . . . . . . .
7.6.2
Computer Application . . . . . . . . . . . . . . . . . . . . .
7.7 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
165
166
166
166
169
169
170
170
170
172
172
.
.
.
.
.
.
.
.
173
175
179
180
.
.
.
.
.
.
.
.
.
.
.
.
182
187
187
187
188
191
xiv
Contents
Three-Dimensional Constitutive Equations . . . . . . . . . . . . . .

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2 Background and Notation . . . . . . . . . . . . . . . . . . . . . . .
8.3 Constitutive Equations for Anisotropic Materials . . . . . . .
8.4 Constitutive Equations for Orthotropic Materials . . . . . . .
8.5 Constitutive Equations in Integral Transform Space . . . . .
8.6 Integral Constitutive Equations for Isotropic Materials . . .
8.7 Constitutive Equations for Isotropic Incompressible
Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.8 Differential Constitutive Equations for Isotropic Materials.
8.9 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
193
193
194
196
199
202
204
.
.
.
.
.
.
.
.
.
.
.
.
210
211
213
217
Isothermal Boundary-Value Problems . . . . . . . . . . . . . .

9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.2 Differential Equations of Motion . . . . . . . . . . . . . .
9.2.1
Balance of Linear Momentum . . . . . . . . . .
9.2.2
Balance of Angular Momentum . . . . . . . . .
9.3 General Boundary-Value Problem . . . . . . . . . . . . .
9.4 Quasi-Static Approximation . . . . . . . . . . . . . . . . . .
9.5 Classification of Boundary-Value Problems . . . . . . .
9.5.1
Traction Boundary-Value Problem . . . . . . .
9.5.2
Displacement Boundary-Value Problem . . .
9.5.3
Mixed Boundary-Value Problem . . . . . . . .
9.6 Incompressible Materials . . . . . . . . . . . . . . . . . . . .
9.7 Materials with Synchronous Moduli . . . . . . . . . . . .
9.8 Separation of Variables in the Time Domain . . . . . .
9.9 Integral-Transform Correspondence Principles . . . . .
9.10 The Poissons Ratio of Isotropic Viscoelastic Solids .
9.11 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10 Wave Propagation . . . . . . . . . . . . . . . . . .
10.1 Introduction . . . . . . . . . . . . . . . . . . .
10.2 Harmonic Waves . . . . . . . . . . . . . . .
10.2.1 Materials of Integral Type . . .
10.2.2 Materials of Differential Type
10.3 Shock Waves . . . . . . . . . . . . . . . . . .
10.3.1 Materials of Integral Type . . .
10.4 Problems . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
219
219
220
220
221
221
223
224
224
224
225
226
227
228
231
234
235
238
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
239
239
240
241
242
246
249
254
257
259
Contents
11 Variational Principles and Energy Theorems. . .

11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . .
11.2 Variation of a Functional. . . . . . . . . . . . . .
11.3 Variational Principles of Instantaneous Type
11.3.1 First Castigliano-Type Principle . . .
11.3.2 Second Castigliano-Type Principle .
11.3.3 Unit Load Theorem . . . . . . . . . . .
11.4 Reciprocal Theorems . . . . . . . . . . . . . . . .
11.4.1 Static Conditions . . . . . . . . . . . . .
11.4.2 Dynamic Conditions . . . . . . . . . . .
11.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xv
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
261
261
262
263
264
268
273
279
279
280
282
282
12 Errata to: Engineering Viscoelasticity . . . . . . . . . . . . . . . . . . . . .
E1
Appendix A: Mathematical Background. . . . . . . . . . . . . . . . . . . . . . .
283
Appendix B: Elements of Solid Mechanics . . . . . . . . . . . . . . . . . . . . .
325
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
351
Fundamental Aspects of Viscoelastic

Response
Abstract
This chapter describes the molecular structure of amorphous polymers, whose

mechanical response to loads combines the features of elastic solids and viscous
fluids. Materials that respond in such manner are called viscoelastic, and their
mechanical properties have an intrinsic dependence on the time and temperature at which the response is measured. To put this into context, the chapter
compares the nature of the response of elastic, viscous, and viscoelastic
materials to several types of loading programs, examining the physical nature of
their mechanical properties, their behavior regarding energy conservation, and
the phenomenon of aging. The topics treated in this chapter provide the
neophyte and casual reader with a good understanding of what viscoelastic
materials are all about.
Keywords
Aging Amorphous Compliance Creep Cross-linked Crystalline Delta

function Dirac Elastic Energy Equilibrium Fluid Glass Glassy
Heaviside Hooke Long term Modulus Newton Polymer Relaxation
Strain Stress Temperature Transition Viscous Viscoelastic
1.1

Introduction
As our intuition tells us, different materials respond differently to external agents.
Our experience also indicates that there are materials which, depending on how the
stimuli are applied, can respond either as solids or fluids, or can display behavior
that combines the characteristics of both. Silly putty, for instance, will bounce just
like an elastic solid if thrown against a hard surface, but will extend slowly and
continually when held solely under the action of its own weight. Materials whose
mechanical response to external agents combines the characteristics of both elastic
D. Gutierrez-Lemini, Engineering Viscoelasticity, DOI: 10.1007/978-1-4614-8139-3_1,
1 Fundamental Aspects of Viscoelastic Response
solids and viscous fluids are called viscoelastic. Not surprisingly, viscoelastic
materials are a little trickier to understand than elastic solids or viscous fluids.
The engineering discipline that developed to provide a mathematical framework to describe the behavior of viscoelastic materials is called viscoelasticity.
This book presents a comprehensive treatment of the theory of viscoelasticity,
explaining, in the present chapter, the nature of the response of viscoelastic
materials to external loads, and in subsequent chapters, how to model that response
mathematically, how to design with viscoelastic materials, and how to establish
the material properties needed to describe their mathematical models.
This chapter begins by discussing an important class of materials, known as
amorphous polymers, whose mechanical response to external stimuli combines the
features of elastic solids and viscous fluids. The terms polymer and polymeric
material have been in use since Berzelius coined the word polymeric in 1832, at
a time when chemists were still unsure of the structure of even the simplest of
molecules. Today, it is hard to imagine our world without polymers. Polymers and
polymeric-based products are commonplace in virtually every industry. This is
unquestionably true of the rubber, oil, aerospace, automotive, electronics, construction, piping, and appliances industries and many more.
The mechanical properties of polymers, such as tensile modulus, or tensile
strength, have an intrinsic dependence on the time and temperature at which the
response is measured. As will be seen, most viscoelastic properties exhibit a steep
gradient in the neighborhood of a temperature termed the glass transition temperature. So much so that the graphs of property functions for this type of materials
are typically displayed in double-logarithmic scales to allow encompassing the two
or more orders of magnitude difference between their extreme values.
The nature of the response of elastic, viscous, and viscoelastic materials to
several types of loading programs is examined next. It is then learned that viscoelastic solids and fluids will respond markedly differently only at temperatures
above the glass transition temperature, or when their response is measured after a
sufficiently long-time-following application of the load. It is learned that viscoelastic solids have non-zero long-term, equilibrium modulus and compliance,
while viscoelastic fluids have to have zero long-term modulus, or, correspondingly, infinite compliance, to allow for viscous flow under sustained load. Two
tests are identifiedconstant step strain and constant step stresswhich can be
used in an elementary manner to establish, respectively, the relaxation modulus
and creep compliance of viscoelastic substances, at fixed temperature. The
response of viscoelastic materials to constant rate loading is also compared to that
of elastic solids and viscous fluids.
The behavior of elastic solids, viscous fluids, and viscoelastic materials
regarding energy conservation and dissipation is also examined in some detail. The
discussion will show that while elastic solids have the capacity to store, in the form
of fully recoverable energy, all the work put into them, and viscous fluids dissipate
all of the work of the external agents, viscoelastic materials store part of the energy
that is put into them, and dissipate the rest as thermal energy, by raising their
internal temperature.
1.1
Introduction
Finally, it will be observed that viscoelastic materials may undergo aging,

which is a continuous change in properties with elapsed time. Since materials are
usually not put to use right after manufacture, aging materials make it necessary to
keep two time scales: one to measure age and the other to measure the time at
which load application starts.
1.2
The Nature of Amorphous Polymers
There are many materials, especially the so-called organic amorphous polymers,1
whose behavior is of viscoelastic type. An amorphous polymer is made up of longchain molecules. A typical polymeric chain may be comprised of several thousand
molecules, strung together in a linear, chain-like fashion [1]. Amorphous polymers
may be subdivided into uncross-linked and cross-linked, depending on the way in
which their molecules are connected.
In uncross-linked amorphous polymers, such as unvulcanized natural rubber
and hard and soft plastics, the individual long-chain molecules are randomly
intertwined with each other but are not chemically bonded together, as indicated in
part a of Fig. 1.1. The worm-like structure of uncross-linked polymers is, therefore, not permanent. As temperature increases, some chain disentanglement takes
place and whole molecules, or segments of polymeric molecules tend to slide past
each other. This allows the polymer to experience large deformations and, possibly, viscous flow [2].
There are several mechanisms by which polymer chains can be connected to
one another to form a continuous network. Vulcanization utilizes sulfur as the
bonding agent, which randomly attaches a chain to a number of neighboring
(a)
(b)
(c)
Fig. 1.1 Schematic representations of amorphous polymers. a Uncross-linked polymer. b Endlinked polymer. c Cross-linked polymer
Although polymers may be classified as crystalline and amorphous, the mechanical response of
crystalline polymers cannot be described by the theory in this book; thus, only amorphous
polymers are treated here.
chains, possibly at several points along its length, by means of strong covalent
bonds. This process results in a relatively permanent, three-dimensional network
structure which restrains molecules from freely slipping past each other, thus
eliminating viscous flow. In all molecular networks, some loose ends of molecules
attach to the network only at single points. In general, however, chemical bonding
of polymeric molecules results in several chains, typically three or four, joining at
the same locations, as illustrated in parts b and c of Fig. 1.1.
1.3
Mechanical Response of Viscoelastic Materials
According to the previous discussion, a material may be classified as viscoelastic if

its response to external stimuli combines the characteristics of elastic and viscous
behavior. Hence, the manner in which a viscoelastic substance would respond to
an external agent can be guessed by examining the response of two identical
specimensone, made of an elastic solid and the other, of a viscous fluidto the
same stimulus, and by imagining that the behavior of the viscoelastic substance
would lie somewhere in between.
For simplicity and clarity, one-dimensional specimens are used to examine the
response of elastic solids and viscous fluids: a uniaxial bar is used for the elastic
solid, and a hydraulic piston, or dashpot,2 for the viscous fluid, as shown in Fig. 1.2.
Also, in our experiment, we would apply a force to either of these specimens, as
suggested in the figure, and use the change in their lengths as a measure of their
response. However, for convenience, we normalize these quantities, and let r stands
for normal stress (force reckoned per unit of original cross-sectional area of the
specimen) and use e to denote normal strain (change in length per unit original
length). In addition, we use E for the elastic modulus of the solid, and g for the
coefficient of viscosity of the fluid, and assume their values to be constant.3
Fig. 1.2 Mechanical models
of elastic solid (uniaxial bar),
and viscous fluid (uniaxial
dashpot), with externally
applied force (F)
F
F
Viscous
fluid
F
bar
F
dashpot
A dashpot is a mechanical device which resists motion by viscous friction. The resisting force
is directly proportional to the velocity; and, acting in the opposite direction, slows the motion and
absorbs energy.
3
The values of E and g of real materials typically depend on temperature, among other things.
1.3
On the previous definitions and assumptions, we introduce the one-dimensional

versions of the stressstrain laws for the elastic solid and the viscous fluid. For the
linearly elastic solid under one-dimensional conditions, we use Hookes stress
strain law:
rt E et
1:1
Alternatively, Newtons law of viscosity is used to define the constitutive

behavior of a linearly viscous fluid:
rt g
1.3.1
d
et
dt
1:2
Material Response to Step-Strain Loading
We consider the response to a constant strain history of a one-dimensional article

which is made either of an elastic, viscous, or viscoelastic substance. A strain of
magnitude eo is assumed suddenly applied at time t = 0 and held constant
thereafter. Mathematically, we express such a strain history by means of the
Heaviside or unit step function, H(), which is identically zero for all negative
values of its argument and equals one when its argument is positive (cf. Appendix
A) and write:
et eo Ht
According to (1.1), the response of a straight bar of a linearly elastic solid to a

step strain would be
rt E eo Ht
In words, a uniaxial bar of an elastic solid would respond instantaneously to the

suddenly applied strain, and the corresponding stress would remain in the material
for as long as the strain is held.
The response of a linearly viscous fluidin a dashpot, saycan be established
by using (1.2) and taking the time derivative of (a). In doing so, note that the
generalized derivative of the unit step function is the Dirac delta function (cf.
Appendix A), so that:
rt g eo dt
The delta function is zero everywhere and becomes infinite where its argument
vanishes. Thus, (c) indicates that to produce an instant strain of finite magnitude on
a viscous fluid would require an infinitely large stress, and that the stress would
disappear immediately after application of the strain.
Per our elementary definition of viscoelastic behavior, we would expect the

response of a viscoelastic substance to lie somewhere between those of the elastic
solid and the viscous fluid. On the grounds of its elastic component, it is logical to
expect that a viscoelastic bar will respond with an instantaneous stress proportional
to the applied strain; and, that to accommodate its viscous behavior, it also seems
logical that the initial stress in the bar should decay, rather than remain constant
as it would in a purely elastic systembut that such decay should perhaps not be
instantaneous, as in a linearly viscous fluid, but should depend on the elapsed time.
The experiment just described is presented in Fig. 1.3.
Fig. 1.3 Qualitative
material responses to stepstrain loading. a Step strain.
b Elastic solid. c Viscous
fluid. d Viscoelastic material
(a)
= o
t
(b)
= E o
t
(c)
= o(t)
t
(d)
= M(t) o
t
Since the applied stimulus in our experiment is constant, the response of the
viscoelastic bar to a step-strain history could be mathematically described in the
following form:
rt Mteo
1:3
According to the previous discussion, M(t) must be a decreasing function of

time, or at least non-increasing function of time, as suggested in Fig. 1.4. This
material property function is called the relaxation modulus.
As will be discussed later on, not only is the relaxation modulus, M, a function of
elapsed time, t, as indicated in (1.3), but it also depends very strongly on temperature, T. Thus, in general, M = M(t, T). However, just as with E and g, temperature
dependence is omitted from (1.3) because it has been assumed, for the time being
only, that a uniform constant temperature is maintained in all experiments.
1.3
Fig. 1.4 Qualitative time

dependence of relaxation
modulus
M(t)
The constant strain history prescribed in (a) is known as a stress relaxation test
and is typically employed to establish the relaxation modulus, M(t); since from
(1.3):
Mt
rt
eo
1:4
Specifically, to establish the relaxation modulus at a constant temperature:

Select a displacement that would lead to an adequate strain for the test
specimen.4
Apply the selected target displacement as rapidly as possible and hold it
constant.
Record the force history necessary to maintain the prescribed displacement.
Use (1.4) with the corresponding definitions of stress and strain for the specimen to compute the relaxation modulus as a function of elapsed test time.
According to this terminology and the response shown in Fig. 1.3, elastic
materials do not relax. Thus, their characteristic relaxation timethe time it takes
the material to complete its relaxation processis infinite. Viscous fluids, on the
other hand, relax completely and instantly: their relaxation times are zero. On this
basis, one can expect that viscoelastic materials will have finite, non-zero relaxation times. Also, by analogy with elastic solids and viscous fluids, viscoelastic
solids are expected to relax to non-zero stress, while viscoelastic fluids should
relax to zero stress, as indicated in the Fig. 1.5.
1.3.2
Material Response to Step-Stress Loading
Here, we consider the response of the elastic and viscous one-dimensional test
pieces to a constant stress of magnitude, ro, which is suddenly applied at time
t = 0 and held constant thereafter.
By definition, the relaxation modulus of linear viscoelastic substances is assumed independent

of strain level. Thus, the magnitude of the prescribed displacement to use in a relaxation modulus
test should not be important. However, the relaxation modulus of materials capable of large
strains typically depends on strain level, and the selection of the test displacement should be
based on the end use of the material.
Fig. 1.5 Distinction

between viscoelastic solids
and fluids. a Step-strain
history. b Viscoelastic solid.
c Viscoelastic fluid
(a)
= o
t
(b)
= Ms (t)o
t
(c)
= Mf (t)o
t
rt ro Ht
According to Hookes law, Eq. (1.1), an elastic solid would respond instantly,
with the step strain:
et
ro
Ht
E
Per Eq. (1.2), a viscous fluid would continue to strain for as long as the applied
stress is sustained:
et
ro
t
g
Now the response of a viscoelastic specimen is expected to lie between that of

the elastic and viscous test specimens. It should, therefore, exhibit an instantaneous strain proportional to the applied stress, like a solid would; but, similar to a
fluid, its strain would grow with the passage of time.5 The behavior just described
is shown in Fig. 1.6.
It seems appropriate then, to call a viscoelastic substance a solid if, under
constant load it creeps to a non-zero, finite strain, and to call it a fluid if its strain
response does not seem to approach a finite limit. The distinction between viscoelastic solids and fluids is shown schematically in Fig. 1.7.
At this point we avoid the temptation to describe viscoelastic response to sustained stress using
the reciprocal of the relaxation modulus, as, say: e(t) = [1/M(t)]ro. Suffice it to say that doing so
would simply define viscoelastic behavior as little more than elastic.
1.3

material responses to stepstress loading. a Step-stress
history. b Elastic solid.
c Viscous fluid.
d Viscoelastic material
(a)
= o
t
(b)
= o /E
t
(c)
= (o /)t
t
(d)
= C(t)o
Fig. 1.7 Distinction

between viscoelastic solids
and fluids. a Step-stress
history. b Viscoelastic solid.
c Viscoelastic fluid
(a)
= o
t
(b)
Equilibrium strain
= Cs (t)o
t
(c)
Unlimited strain
= Cf (t)o
t
Since the applied load in this experiment is constant, the time variation of the
response of a uniaxial bar of viscoelastic material can be mathematically described
in the form:
et Ct ro
1:5
According to the foregoing arguments the function C(t), called the creep
compliance, must be an increasing, or, at the very least a non-decreasing function
of time; as shown in Fig. 1.8.
10
Fig. 1.8 Conceptual time

dependence of creep
compliance
C(t)
The constant stress history prescribed in (d) is known as a creep test, and is
typically employed to establish the creep compliance, C(t); since from (1.5):
Ct
et
ro
1:6
Specifically, to establish the creep compliance at a constant temperature:

Select a force that would lead to an adequate stress level for the test specimen.6
Apply the selected target force as rapidly as possible and then hold it constant.
Record the displacement history produced by the prescribed load.
Use (1.6) with the corresponding definitions of stress and strain for the specimen to compute the creep compliance as a function of elapsed test time.
As will be discussed at length in a later chapter, the creep compliance and the
relaxation modulus are not reciprocals of each other, except under special conditions, but they are not completely independent of each other, either.
1.3.3
Material Response to Cyclic Strain Loading
We now examine the stress response of a linear elastic solid and a linear viscous
fluid to the sinusoidal, cyclic strain history:
et eo sinxt
As before, we use expressions (1.1) and (1.2) to establish, respectively, the

responses of the elastic solid and viscous fluid, as:
rt E eo sinxt
rt g eo x cosxt
By definition, the creep compliance of linear viscoelastic materials is assumed independent of

stress level. Thus, the magnitude of the stress to use in a creep test should not be important.
However, the compliance of nonlinear materials depends on stress level, so that the selection of
the test stress level should be made carefully.
1.3

material responses to cyclic
strain. a Cyclic strain history.
b Elastic solid. c Viscous
fluid. d Viscoelastic material
11
(a)
= o sint
t
(b)
= E o sint
t
(c)
= o cost
t
(d)
= A(t) o sin( t+)
These expressions indicate that, while the stress in the elastic solid is in phase
with the strain, the stress in the viscous material is exactly 90 out of phase with it.
It is then reasonable to expect that the response of a viscoelastic material to cyclic
harmonic loading will lie anywhere between 0 and 90 out of phase with its
loading. This observation is indicated in Fig. 1.9.
1.3.4
Material Response to Constant Strain Rate Loading
We next examine the stress response of the linear elastic solid and linear viscous
fluid to the constant strain rate history:
et R t
Using this relation and (1.1), the response of the solid is seen to be:
r t E R t
The response of the linear viscous fluid follows from (j) and (1.2), as:
rt gR
These expressions indicate that while the stress in the elastic solid is directly
proportional to the strain, the stress in the viscous material is equal to the rate of
straining. One may then argue that the stress response of a linear viscoelastic
12

material responses to constant
strain rate loading:
comparison of responses at
one loading rate. a Strain
c Viscous fluid.
(a)
= Rt
t
(b)
= ERt
t
(c)
= R
t
(d)
= f (t)
t
material, having to lie between a constant and a linear function of time, would
have to be described by a function that increases less than linearly with time, as
indicated in Fig. 1.10.
Clearly, without knowing more about the material at hand, a graphic response
in a form such as that shown in Fig. 1.10d would not suffice to definitely identify a
material as viscoelastic. That type of response is also characteristic of nonlinear
elastic behavior. The point to emphasize here, regarding elastic and viscoelastic
response to constant strain rate loading is that, an elastic material will react with
the same stress to a given strain, irrespective of how long it takes to reach that
strain. A viscoelastic material, on the other hand, will respond with a stress that
depends on how long it takes to apply the strain. This aspect of material behavior is
depicted in Fig. 1.11.
1.4
Energy Storage and Dissipation
As with other aspects of viscoelastic behavior, we should require that the response
of any viscoelastic material regarding energy conservation should lie between those
of an elastic solid and a viscous fluid. As will be shown shortly, the total amount of
work performed on an elastic solid by external agents is stored in it in the form of
internal energy, which is fully recoverable upon removal of the external agents.7 By
7
This means that if the external agents that put work into an elastic solid were completely
removed, the internal energy stored in the solid could be used to perform an amount of workon
another system, sayequal to the work that was put into the solid in the first place.
1.4

material responses to constant
strain rate loading:
comparison of responses at
two loading rates. a Strain
c Viscous fluid.
13
1 = R1t
(a)
2 = R2t
t
(b)
1 = ER1t
2 = ER2t
t
(c)
1 = R 1
2 = R 2
t
1 (t)
(d)
2 (t)
t
contrast to an elastic solid, a viscous fluid has no capacity to store energy, and all
work performed on viscous fluids is lost or, more properly, dissipated.
By our definition of viscoelastic behavior, a viscoelastic material, be it a solid
or a fluid, should be expected to be able to store, and have available for recovery,
at least part of the energy put into it, while it should dissipate the rest. It therefore
seems logical to postulate that, irrespective of the constitution of the material at
hand, the total work performed on a body by external agents, Wext, should be
equal to the work of the internal forces, Wint, minus the work, Wdiss, dissipated in
the process. This statement embodies the first law of thermodynamics, on the
conservation of energy, that: Energy can neither be created nor destroyed, but
only transformed.
According to the previous discussion, linearly elastic solids must exhibit no
dissipation, while linearly viscous fluids must dissipate all energy put into them.
That is:
Wdiss
0;
for
Wext ; for
linearly
linearly
elastic
viscous
solids
fluids
1:7
The energy balance equation may be cast in the form:

Wext Wint Wdiss
1:8
14
Although we present energy conservation and dissipation in a broader sense in

Appendix B, in what follows we examine the balance of energy using onedimensional models: the bar as a linearly elastic solid, and the dashpot as a linearly
viscous fluid.
We consider an experiment in which the test piece is fixed at one end, while its
other end is first pulled a certain amount and then moved back to its original
position.8 Under these circumstances, the work of the external forces is exactly
zero, irrespective of the material constitution; since, by definition:
Wext
P2
Fext duext
P1
P1
Fext duext 0
P1
where, F and u denote force and displacement, respectively. In this case (1.8) takes
the simpler form:
Wdiss Wint
To calculate the amount of energy dissipated in the process, we note that,

similar to Wext, the work of the internal forces is given by:
Z
Wint
P2
Fint duint
e2
Ar Lde V
e1
P1
e2
r de
e1
This expression uses that, in a one-dimensional system: r = F/A, de = du/L,

and V = AL, for the stress, strain, and systems volume, respectively.
To apply (c) to the bar of a linearly elastic solid, we use the stressstrain Eq.
(1.1), which leads to:
Wint V
e2
r de V E
e1
e1
e1
e1
1
e de V E e2 e1 0
2
Taking this into (b) yields the first of (1.7): Wdiss = 0; that a linearly elastic
solid absorbs all external work as internal energy and does not dissipate any
energy.
Noting that the constitutive Eq. (1.2) for a linearly viscous fluid involves strain
rate, and not strain, we first transform (c) to accommodate this fact. In doing so, it
is assumed, without loss of generality, that the total duration of the experiment is
t*. Thus:
Wint V
e2
r de V g
e1
tt
t0
de de
dt V g
dt dt
tt
t0
2
de
dt
dt
In general, the loading and unloading paths do not have to be the same, as long as the test piece
is returned to its original position.
1.4
15
Fig. 1.12 Example 1.1
u
R
-R
t
ta
tb
The quantity under the integral sign in (e) is never negative, vanishing only for
the trivial case when the strain rate is identically zero in the interval of integration.
Therefore:
Wint V g
tt
t0
2
de
dt 0
dt
This result, together with (b) leads to:

Wdiss Wint V g
tt
t0
2
de
dt 0
dt
Thus, as asserted in the second part of (1.7): a linearly viscous fluid dissipates
all energy put into it.
Example 1.1 Determine the total energy that would be dissipated by applying the
displacement history shown in Fig. 1.12 to a dashpot with a linearly viscous fluid
of viscosityg.
Solution:
To calculate the total energy dissipated in the process we evaluate the evolution of
the strain rate, de/dt = (1/L)du/dt during the process:
de
dt
R;
0 t\ta
R; ta t tb
and take the result into (g), to get:

Wdiss Vg
"Z
ta
0
2
Z tb 2 #
de
de
dt
dt VgR2 tb
dt
dt
ta
16
Fig. 1.13 Typical relaxation

modulus as a function of
temperature; showing the
regions of viscoelastic
behavior for typical
viscoelastic solids and fluids
log M (to)
Glassy region
Mg
Transition
region
Rubbery
plateau
Viscous flow
____
-----
solid
fluid
Me
Direction of increase in volume and molecular mobility
1.5
Glass Transition and Regions of Viscoelastic Behavior
Experiments with viscoelastic materials indicate that:

The reported values of the relaxation modulus and creep compliance depend on
the time scale of the observations (10 s modulus, 10 min modulus, etc.).
Most properties of viscoelastic materials depend strongly on temperature.
Some properties change very drastically when the temperature is near a critical
value, called the glass transition temperature, usually denoted as Tg.
The influence of temperature and time of observation on viscoelastic propertiesthe time and temperature dependence of functions such as the modulus, M, or
the creep compliance, C, of previous discussionsis emphasized with the explicit
notations M (t, T) and C (t, T). The relaxation modulus, M, of a typical amorphous
polymer is used in Fig. 1.13 to exemplify the temperature dependence of viscoelastic properties in general. A logarithmic scale is employed to accommodate the
strong dependence of modulus with temperature. In such graphical representations,
it is always necessary to indicate exactly to what test time the reported property
corresponds (t = to, in this case).
The transition region falls in a narrow temperature range; and the temperature
in the middle of this region is called the glass transition temperature, Tg. At
temperatures well below the glass transition, in the so-called glassy region, an
amorphous polymer is an organic glass: a hard and brittle plastic with a high
modulus, Mg, called the glassy modulus. At such low temperatures, the polymer
chains are essentially frozen in fixed positions. At temperatures around Tg, in
the glass-transition region, whole molecules, or segments of polymeric molecules
are somewhat free to jump from one site to another, and the polymer responds
with a modulus that changes very sharply with temperature. At temperatures above
the glass transition but below the melting point, in what is called the rubbery
plateau, molecular mobility increases and segments of polymeric chains reorient
relative to each other. In this region, cross-linked polymers would relax to a moreor-less constant equilibrium modulus, Me, behaving like rubbery elastic solids
1.5
Fig. 1.14 Fixed-time creep

compliance as a function of
temperature, showing the
regions of viscoelastic
behavior for typical solid and
fluid polymers
log C(to)
17
Transition
region
fluid
Ce
solid
Glassy region
Rubbery
plateau
Viscous flow
Cg
of low modulus. In fact, Me may be several orders of magnitude smaller than the
glassy value, Mg. In this region, uncross-linked polymers eventually disentangle,
with entire molecular segments sliding past one another, giving rise to viscous
flow [3].
Similar observations apply to the creep compliance function, C (t, T). The
temperature dependence of the creep compliance function of a polymer looks very
much like the mirror image of the relaxation modulus about a line parallel to the
temperature axis, as indicated in Fig. 1.14.
The creep compliance function of an uncross-linked polymer typically lacks the
rubbery plateau, exhibiting continued creep, followed by failure. This is indicated
by a dotted line in the figure.
Also, to the elastic glassy and equilibrium moduli, Mg and Me, correspond
glassy and equilibrium compliances, Cg, and Ce, respectively. Since modulus and
compliance of any linearly elastic material are reciprocals of each other, it follows
that both: Cg = 1/Mg, and Ce = 1/Me. In the transition region, however:
C (t, T) = 1/M(t, T); although the two functions are inverses of each other, in a
sense to be explained in Chap. 2. In fact, it will be proven there, that in general,
C(t, T) M(t, T) B 1 [4].
As pointed out in Sect. 1.3, the relaxation modulus is a decreasingor, at least,
non-increasingfunction of time. In a double-logarithmic plot, log(M) versus
log(t), the shape of its graph resembles that of log(M) versus T. Entirely similar
observations hold for the creep compliance, as indicated in Fig. 1.15.
The fact that the graphs of the material functions M and C versus temperature
have the same general shape as those versus elapsed time suggests a relationship
between time and temperature for viscoelastic substances. The detailed nature of
this relationship is contained in the timetemperature superposition principle that:
in amorphous polymers, time and temperature are interchangeable. This principle
is explained in Chap 6. In loose terms, as far as material property functions are
concerned, the principle implies that short test times correspond to low
18
Fig. 1.15 Fixed-temperature
relaxation modulus and creep
compliance as functions of
test time, showing regions of
viscoelastic behavior for a
typical cross-linked polymer
log M( T )
Glassy region
Transition
region
Rubbery
plateau
or
log C(T)
log( t )
temperatures and long test times to high temperatures, and vice versa. This means
that reducing the test temperature is equivalent to shortening the test time; and
conversely, increasing test temperature is equivalent to extending the test time.
Two additional qualitative observations regarding polymers were reported by
R. Gough, in 1805, and confirmed experimentally by J. P. Joule, in 1857, at Lord
Kelvins insistence, [5] that:
Rubber heats up on stretching.
A loaded rubber band contracts on heating.
The first observation shows that rubber dissipates energy in the form of heat, as
our discussion of energy dissipation suggested viscoelastic materials would. The
second observation, known as the Gough-Joule effect, indicates that, unlike metals,
the modulus of rubber experiences a relative increasestiffening upwith
temperature.
Example 1.2 Draw a sketch of the response of a linear viscoelastic solid to the
step-strain program shown in Fig. 1.16. Consider: (a) response in the glassy
region, and (b) response in the transition region.
Solution:
(a) Response in the glassy region
In the glassy region, the behavior of any viscoelastic substance is elastic, with
glassy modulus, Mg, and glassy compliance, Cg = 1/Mg. For this reason, its
response must be directly proportional to the applied step strain. This is
indicated in Fig. 1.17.
(b) Response in the transition region
Here, we use that: the response of any viscoelastic substance to a suddenly
applied stimulusinstantaneous loading or instantaneous unloadingis
always elastic with glassy modulus Mg, and glassy compliance, Cg = 1/Mg.
Additionally, its response to a constant strain in the transition region follows
the shape of the relaxation modulus M(t), which is a decreasing function of the
time elapsed since the strain was applied. On unloading, the same is true, only
1.5
Fig. 1.16 Example 1.2:

step-strain loadunload event
19
o
t2
t1

response of a linear
viscoelastic solid to a stepstrain loadunload event in
the glassy region
(t) = Mg o
t1
viscoelastic solid to a stepstrain loadunload event in
the transition region
t2
Mg o
Mg o
t1
t2
the sign of the response is reversed. Thus, the stress first drops by Mgeo
becoming negativeto then undergo a reverse relaxation, recovering monotonically toward zero. This behavior is indicated in Fig. 1.18.
Example 1.3 Draw a sketch of the response of a linear viscoelastic solid to the
step-stress program shown in Fig. 1.19. Consider: (a) response in the equilibrium
zone, and (b) response in the transition region.
Solution:
(a) Response in the rubbery equilibrium region
In the rubbery region, a viscoelastic material will respond like an elastic solid
with modulus, Me, and compliance, Ce = 1/Me. Therefore, its response must
be directly proportional to the applied loading. This is indicated in Fig. 1.20.
20

step-stress loadunload event
o
t1
viscoelastic solid to a stepstress loadunload event in
the rubbery region
t2
(t) = Ceo
t1
t2
(b) Response in the transition region

Here, we use that: the response of any viscoelastic substance to a suddenly
applied stimulusinstantaneous loading or instantaneous unloadingis
always elastic with glassy modulus Mg, and glassy compliance, Cg = 1/Mg.
Additionally, its response to a constant stress in the transition region follows
the shape of the creep compliance, C(t), which is an increasing function of the
time elapsed since the stress was applied. On unloading, the same is true, only
the sign of the response is reversed: the strain first drops by first drops by Cgro,
and then recovers monotonically toward zero. This behavior is indicated in
Fig. 1.21.
1.6
Aging of Viscoelastic Materials
Aging is a phenomenon observed in many viscoelastic substances. It may be

defined as any change in constitutive or failure properties with time. There are two
types of aging: physical, reversible aging, which is due to thermodynamic

viscoelastic solid to a stepstress loadunload event in
the transition region
Cgo
Cgo
t1
t2
1.6
Aging of Viscoelastic Materials
21
processes; and chemical aging, which is caused by irreversible chemical reactions

in the material. Although their origins differ, the macroscopic manifestations of
physical and chemical aging are very similar. From a mechanical point of view,
material aging frequently manifests itself as an increase in modulus and a
reduction of creep strain [6].
As indicated earlier in this chapter, the response of a non-aging viscoelastic
substance to an external stimulus changes with the passage of time because its
material properties, such as relaxation modulus and creep compliancedepend on
the time at which the observations are made during an experiment. However, such
response is independent of the time at which the experiment is started, and is the
same every single time the same experiment is repeated, provided the experiment
is the same each time.9 Put another way, the material properties of a non-aging
material are fully described by functions which depend on a single time scale,
which measures only the time the material is under load, irrespective of when it
was manufactured.
By contrast, the properties of a material susceptible to aging change with the
passage of time even in the absence of external agents. That is, the response of an
aging material, measured at a specified time during an experiment, depends on the
time elapsed since the material was manufactured and the time it spends under
load; and hence, on the time at which the experiment is started. For this reason,
two time scales are required to unambiguously describe the material property
functionsand responseof an aging material. One time scale is needed to keep
track of the time since the manufacturing of the material, and the other time scale
is needed to keep track of the time the material is under load.
References
1. A.N. Gent (ed.), Engineering with Rubber, How to Design Rubber Components (Jos. C. Huber
KG, Diessen/Ammersee, 2002), pp. 3840
2. L.E. Malvern, Introduction to the Mechanics of a Continuous Medium (Prentice-Hall Inc.,
1969), pp. 306327
3. J.J. Aklonis, Introduction to Polymer Viscoelasticity (Wiley, 1983), pp. 4452
4. A.C. Pipkin, Lectures on Viscoelasticity Theory (Springer, 1986), pp. 1415
5. H. Morawetz, Polymers: The Origins and Growth of a Science (Dover, 1985), pp. 3536
6. A.D. Drosdov, Finite elasticity and Viscoelasticity, A course in the Nonlinear Mechanics of
Solids (World Scientific, Singapore, 1996), pp. 229234
Assuming it were possible to produce exactly the same non-ageing material every time
without lot-to-lot variabilitythis result would hold true even if the specimens came from
material lots of different age.
Constitutive Equations in Hereditary

Integral Form
Abstract
Materials respond to external load by deforming and straining, and by

developing stresses. The internal stresses corresponding to a given set of strains
depend on the constitution of the material itself. For this reason, the rules that
permit calculation of internal stresses from known strains, or vice versa, are
called constitutive laws or constitutive equations. There are two equivalent
ways to describe the mathematical relationships between stresses and strains for
viscoelastic materials. One form uses integrals to define the constitutive
relations, while the other relates stresses and strains by means of differential
equations. Starting from Boltzmanns superposition principle, this chapter
develops the integral form of the one-dimensional constitutive equations for
linearly viscoelastic materials. This is followed by a discussion of the principle
of fading memory, which helps to define the acceptable analytical forms of the
material property functions. It is then shown that the closed-cycle condition
(i.e., that the steady-state response of a non-aging viscoelastic material to a
periodic excitation be periodic) requires that the material property functions
depend only on the difference of their arguments. The chapter also examines the
relationships between the relaxation modulus and creep compliance functions
in the physical time domain as well as in Laplace-transformed space. Various
alternative forms of the integral constitutive equations often encountered in
practice are discussed as well.
Keywords
Boltzmann Constitutive Convolution Creep Cycle Equilibrium Fading

Glassy Hereditary Isothermal Laplace Long-term Matrix Memory
Operator Principle Relaxation Symbolic



23
24
2.1
2 Constitutive Equations in Hereditary Integral Form
Introduction
Materials respond to external stimuli by deforming and straining, that is by

changing their shape or size, and by developing stresses. The internal stresses
corresponding to a given set of strains depend on the constitution of the material
itself. For this reason, the rules that permit calculation of internal stresses from
known strains, or vice versa, are called constitutive laws, or, constitutive equationswhen such relationships are known in analytical form. The terms stress
strain or strainstress relations or equations, are widely used to emphasize that the
first variable is expressed in terms of the second.
There are two equivalent ways to describe the mathematical relationships
between stress and strain for linear viscoelastic materials. One way uses integrals
to define these relations, while the other relates stresses and strains through linear
ordinary differential equations. In this chapter, we develop the integral form of
constitutive equations, leaving for Chap. 3 the discussion of their differential
counterparts. All the developments are presented in great mathematical detail but
to motivate the proofs, some physical insight is also provided. The level of
mathematical detail used to present the subject matter and the exercises in this
chapter is intended to give the reader the confidence necessary to engage in
independent research, irrespective of the field of interest.
For clarity of presentation, only non-aging materials under isothermal conditions are treated in this and subsequent chapters, until Chap. 6, where the
dependence of material properties on temperature is examined. All material
functions referred to here are thus presumed independent of age and available at
the constant temperature implied in the discussions. The dependence of material
property functions on temperature will be omitted but assumed understood.1
This chapter starts from Boltzmanns superposition principle and develops the
integral form of the one-dimensional constitutive equations for a linearly viscoelastic substance. This is followed by a discussion of the principle of fading
memory, which helps to define the acceptable forms of relaxation and compliance
functions. It is then shown that the closed-cycle condition (that the steady-state
response of a non-aging viscoelastic material to a periodic excitation be periodic)
requires that the material property functions depend only on the difference of their
arguments, and all transients die out. The chapter also examines various relationships between the relaxation modulus and creep compliance functions, both in the
time domain and in Laplace-transformed space. Alternative forms of constitutive
equations often encountered in practice are also discussed. We conclude the chapter
with a discussion of how to evaluate the work done by external agents acting on a
linear viscoelastic material. This topic of great practical use, since, as shown in
Chap. 1, viscoelastic materials dissipate as heat, some of the energy that is put into
them, and hence polymeric materials are often used in industry to dissipate energy.
On this assumption, for instance, M(t) and C(t) will be used for M(t,T) and C(t,T), respectively.
2.1
Introduction
25
Fig. 2.1 Stress response to a

step strain applied at the time
the test clock is started
2.2
Boltzmanns Superposition Principle
By definition [c.f. Chap. 1], the tensile relaxation modulus, M(t,T), at any time t,
and fixed temperature T describes how the stress varies with time under a stepstrain load. To fix ideas, imagine a one-dimensional bar of a linearly viscoelastic
material after it is subjected to a strain of magnitude eo ; suddenly applied at the
start of an experiment and held constant thereafter. As seen in (Fig. 2.1), in
accordance with Eq. (1.3), the stress response, rt; of the bar to the applied step
strain would be given by:
0;
for t\0
a
M teo ; for t 0
By the definition of the Heaviside step function H, that: H(t) = 0, for negative
values of its argument, while H(t) = 1, whenever its argument is zero or positive,
one can rewrite (a) in the form [c.f. Appendix A]:
rt
rt M t H teo
Now assume that exactly the same experiment as that described by (a) or (b)
were to be carried out using the same material but applying the loading t1 units of
time after starting the clock. Also assume that all loading2 and environmental
conditions would be the same in both cases. If the material did not age, all its
relevant property functions would be exactly the same in both experiments.
2
The terms load and loading are used in their broader sense to include tractions, or stresses,
as well as displacements, or strains. The exact meaning should be clear from the context in which
the term is used.
26
Fig. 2.2 Stress response to a

step strain applied t1 units of
time after the test clock is
started
t1
t1
Consequently, exactly the same response would be observed in the second

experiment as in the first, but with a time delay t1, as indicated in Fig. 2.2.
Similarly to (a) and (b), the stress response could now be expressed, respectively, as follows:
rt
0; for t\t1
M t t1 eo ; for t t1
rt M t t1 Ht t1 eo
c
d
It is an easy matter to extend these results to arbitrary load cases. As suggested

in Fig. 2.3, any piecewise continuous function of time may be approximated by a
series of step functions; with each subsequent step adding an incremental amount
to the previous step. Using (c), then, the response to the kth incremental step strain,
Dek, which is taken to occur at time tk+1, would be:
Drk t M t tk Dek ;
Fig. 2.3 Approximation of a

continuous function as a finite
series of incremental step
functions
t tk
tk-1
tk
2.2
27
According to Boltzmanns principle, the response to each incremental load is

independent of those due to the other incremental loads, and the response to the
complete load history, as idealized through the series of incremental step-loads,
equals the sum of the individual responses:
rt
XN
Drk t
k1
XN
k1
M t tk Dek ;
t tk
Dividing and multiplying the right-hand side of (f) by the time interval,
Dtk = tk - tk-1, between successive steps, and using the properties of the Heaviside step function, yields:
rt
N
X
Drk t
k1
N
X
Mt tk
k1
Dek
Dtk ; t tk
Dtk
Passing to the limit as N increases without bound and the size of successive
intervals is made vanishingly small:
rt
lim
lim N!1
N
X
Drk t lim
N
X
N!1
tk !s k1
k1
Mt tk
Dek
Dtk ;
Dtk
t tk
Since this process turns the discrete set tk into a continuous spectrum, we use
the letter s to denote it and arrive at3 (see, for instance, [1]):
rt
Zt
drt
Zt
Mt s
d
esds
os
To allow the strain to have a step discontinuity at time t = 0+, we add (a) and
(i) and write:
rt Mte0
Zt
Mt s
d
esds
os
2:1a
The term M(t) e(0+) may be taken inside the integral, using that e(0-) : 0,
because:
Z0
0
d
Mt s esds Mt
os
Z0
d
esds Mte0
ds
0
The notation x+ is used to signify a value of x that is just larger than x. Similarly, x- means a
value of x just less than x.
28
Hence, (2.1a) may be alternatively expressed as:
rt
Zt
Mt s
0
d
esds
os
2:1b
Had we chosen the applied action to be a stress instead of strain history, entirely
similar arguments would have led to the strainstress forms:
et Ctr0
Zt
Ct s
d
rsds
ds
2:2a
et
Zt
Ct s
d
rsds
ds
2:2b
0
Equations in (2.1a, b) and (2.2a, b) show that the response of a viscoelastic

substance at any point in time depends not only on the value of the action at that
instant, but also on the integrated effect, or complete history of all past actions. In
other words, the response at the present instant inherits the effects of all past
actions. For this reason, viscoelastic materials are also frequently called hereditary
materials; and viscoelasticity, hereditary elasticity.
Example 2.1 The (one-dimensional) viscoelastic response to a constant strain-rate
loading, e (t) = Rt, may be expressed in the elastic form: r(t) = Eeff(t)e (t).
Derive an expression for Eeff(t), the constant-rate effective modulus, for a viscoelastic substance.
Solution:
Assume the relaxation modulus of the viscoelastic material to be M(t) and
compute its stress response with (2.1a), using that de (s)/ds = d(Rs)/ds = R, and
introducing the change of variables t - s = u, to arrive at:
rt Mte0 R
Zt
0
Mt sds R
Zt
Mudu
0
Multiplying and dividing this expression by t, recalling that e(t) = R t, and reordering:
2.2
1
rt Rt
t
Zt
29
2 t
3
Z
1
Mudu 4
Mudu5et Eeff tet
t
With the obvious definition of the constant-rate effective modulus, Eeff:
Eeff t
1
t
Zt
Mudu
2:3
This expression can be used to evaluate the stress response of a viscoelastic

material to constant strain-rate loading, by means of the elastic-like expression:
r(t) = Eeff(t) e (t).
Had the roles of strain and stress been reversed, we would have employed (2.2a)
to derive the following definition of the constant-rate effective compliance:
1
Deff t
t
Zt
Cudu
2:4
As before, this can be used to determine the strain at any specified time, of a
viscoelastic material subjected to constant-rate stress, using the elastic-like form:
e (t) = Deff(t) r (t).
Example 2.2 Obtain the instantaneous response of a viscoelastic material with
relaxation modulus, M(t), to a general strain history e(t).
Solution:
We evaluate the stress response using expression (2.1a) at t = 0, to get:
rt M 0e0 Mg e0
2:5
In similar fashion, (2.1b) would yield the instantaneous strain response to an

arbitrary stress history r(t), as:
et C0r0 Cg r0
2:6
This example indicates that the instantaneous, impact, or glassy response of a

non-aging viscoelastic material is elastic, with operating properties equal to its
glassy modulus, or its glassy compliance, depending on whether strain or stress,
respectively, is the controlled variable.
Example 2.3 Obtain the equilibrium response of a viscoelastic substance with
relaxation modulus, M(t), to a general strain history e(t).
30
Solution:
We evaluate the stress response using expression (2.1a) as t ? ?:
r1 lim
Zt
t!1
2 0
3
Zt
Z
de
de
de
Mt s ds lim 4 Mt s ds
Mt s ds5
t!1
ds
ds
ds
0
0
Noting that e(t) : 0, t \ 0:
r1 M1e0 M1 lim
Zt
t!1
de
ds
ds
Or, after canceling like terms, since the integral evaluates to: e(?) - e(0), and
M(?) is the equilibrium modulus Me:
r1 Me e1
2:7
By the same procedure, starting with (2.2a), it is found that the long-term strain
response to an arbitrary stress history, r(t), is given as:
e1 Ce r1
2:8
This example indicates that the long-term response of a non-aging viscoelastic

material is elastic, with operating properties equal to either its long-term or
equilibrium modulus, or its long-term or equilibrium compliance, depending on
whether strain or stress is the controlled variable.
2.3
Principle of Fading Memory
Loosely speaking, we say that a material has fading memory if the influence of an
action on its response becomes less important as time goes by. Accordingly, the
mathematical implications of the fading memory hypothesisoften called principlecan be established by loading and unloading a viscoelastic system, and
monitoring its response after the load is removed. Before establishing the consequences of the principle of fading memory on a rigorous basis, we develop them by
examining the response of a viscoelastic material to the relaxation and creep
experiments; with which we are already familiar. The results of these experiments
are the relaxation modulus and the creep compliance. As discussed in Chap. 1, the
general shapes of these functions are as shown in Fig. 2.4.
2.3
Fig. 2.4 Functional forms of

the stress relaxation modulus
and creep compliance of a
viscoelastic material used to
explain the fading memory
hypothesis
31
M(t)
C(t)
Creep Compliance
Relaxation modulus
The functional forms shown in the figure indicate that the fading memory
hypothesis should require that the relaxation modulus be a monotonically
decreasing function of time, with monotonically decreasing slope. In similar
fashion, the creep compliance should be a monotonically increasing function of
time, with monotonically decreasing slope. We now proceed with the rigorous
proofs of these statements. To do that, we will take the applied action to be a step
strain of magnitude eo, applied to a one-dimensional viscoelastic system starting at
time t = 0 and ending at time t = t*: e(t) = eo[H(t) - H(t - t*)].
Expression (2.1a) will be used to establish the corresponding response. Before
we proceed, we put (2.1a) in a form more suitable to our purposes, integrating it by
parts and writing the resulting derivative of the modulus in terms of the time
difference, t - s; thus:
rt M0et
Zt
oMt s
esds
ot s
2:9
Inserting the step-strain load into this expression leads to the response after the
load is removed (t [ t*):
rt M0eo Ht Ht t
Zt
oMt s
Hs Hs t eo ds;
ot s
t [ t
j
By the definition of the Heaviside unit step function, the term inside the first set
of brackets is zero. The other term in the expression may be evaluated using the
mean-value theorem of integral calculus4 [c.f. Appendix A]:
rt t f
oMt kt
gHkt Hkt t eo ;
ot kt
t [ t ;
The mean value theorem of integral calculus states that

b a; 0\k\1:
Rb
a
0\k\1
f xdx b af a k
32
Since kt* \ t*, the second Heaviside step function inside the brackets vanishes, so
that:
rt t f
oMt kt
geo ;
ot kt
t [ t ;
0\k\1
For the influence of an action removed at t = t* to eventually disappear, so that

r ? 0, it is necessary that:
lim f
t!1
oMt kt
g 0;
ot kt
8t \1;
0\k\1
Or, equivalently:
o
lim f Mtg 0
ot
2:10
t!1
Otherwise, the material would retain permanent memory of the effect of the
applied load, and the process would induce irreversible changes.
As may be seen from (2.9), the derivative, qM(s)/qs, of the relaxation function
with respect to its argument acts as a weighting factor on the applied action, e. For
the effect of the action to be less and less pronounced with the passage of time, it is
necessary that the weighting factor be a monotonically decreasing function of its
argument. That is,

o

Mt o Mt ;

ot
ot
tt2
tt1
t2 [ t1
2:11
Also, as experimental evidence shows [c.f. Chap. 1]:

jMtjt2 jMtjt1 ;
t2 [ t1
2:12
In similar fashion, repeating the previous arguments with a step stress applied at
t = 0 and removed at t = t*, leads to the following requirements for the creep
compliance function:
o
lim f Ctg 0
t!1 ot

o

Ct o Ct ;

ot
ot
tt2
tt1
jCtjtt2 jCtjtt1 ;
2:13
t2 [ t1
t2 [ t1
2:14
2:15
2.3
33
Geometrically, then, the fading memory hypothesis simply requires that the
relaxation modulus and creep compliance be monotonically decreasing and
increasing functions of their arguments, respectively, and also that the absolute
values of their slopes decrease monotonically. In addition, as indicated in Chap. 1,
experimental observations indicate that:
The relaxation modulus decreases with observation time and is bounded by the
glassy modulus for fast processes and by the equilibrium modulus for very slow
processes.
The creep compliance increases with observation time and is bounded by the
glassy and equilibrium compliances for very fast and slow processes,
respectively.
The fading memory principle embodied in (2.10)(2.15), together with the
experimental observations, requires that the general forms of the relaxation and
creep compliance functions be as shown in Fig. 2.4.
Example 2.4 As an application of the fading memory principle, we evaluate the
stress responses of a viscoelastic material to two arbitrary loading programs,
e1(t) and e2(t), which reach the same constant value, e*, at time t* and remain at that
level from that point on, as indicated in Fig. 2.5.
Solution:
Use (2.1a) to evaluate the response as t ? ?, splitting the integration interval
from 0+ to t*, and t* to ?; and note that the derivatives of the strain histories
e1(t) and e1(t) vanish after t = t* to write:
r1 1 M1e1 0 lim
Zt
t!1
Mt s
de1
ds M1e
ds
Mt s
de2
ds M1e
ds
r2 1 M1e2 0 lim
Zt
t!1
0
Fig. 2.5 Example 2.4: Two

arbitrary loading histories
which become identical and
constant after a finite time
1( t )
2( t )
t*
34
Since the integrals evaluate to M(?)[e1(t*) - e1(0+)] and M(?)[e2(t*) e2(0+)], and also, e1(t*) = e2(t*) = e*, it follows that: r1 1 r2 1 M1
e : Or, using the alternate notations for the long-term or equilibrium modulus
M(?) : M? : Me:
r1 M1 e Me e
2:16
Proceeding in an entirely similar fashion, but using (2.2a), one would find that
the long-term, equilibrium strain response to an arbitrary stress history, r(t) would
be given by:
e1 C1 r Ce r
2:17
These expressions clearly show that a viscoelastic material would remember

only that the loading got to e*or r*, for that matterbut not how it got there.
That is, after sufficiently long, a viscoelastic material will have effectively forgotten the details of the loading history; in agreement with the principle of fading
memory.
2.4
Closed-Cycle Condition
This section examines the mathematical consequences of the physical expectation

stated in Chap. 1, that the response of a linear viscoelastic material to harmonic
loading ought to be harmonic, of the same frequency as the excitation, but out of
phase with it. This so-called closed-cycle condition, that: the steady-state
response to harmonic loading also be harmonic, is satisfied by materials that do
not age. That is, by materials whose property functions depend only on one
timescale: the time measured from when the load was first applied, irrespective of
the time elapsed since their manufacturing.
As will be shown in what follows, the closed-cycle condition requires that the
kernels of the constitutive integrals, M and C, depend only on the difference of
their arguments, and also, that all transients die out. In other words, the closedcycle condition requires that M t; s Mt s; and Ct; s Ct s; as has
been assumed without proof in our derivations, so far. A physical proof of this
implication of the closed-cycle condition can be constructed rewriting (2.1a), say,
using M(t, s), in place of M(t - s), in order to remove the assumption made so far
in our derivations that the kernel of the constitutive equation depends only on the
difference of its arguments:
rt M0et
Zt
0
o
fMt; sgesds
os
2.4
35
The first term in this expression is simply the instantaneous value of the stress
response. The second term, the hereditary component, is calculated as follows. In
the time interval between s and s ? ds of the past, the strain was e(s). Since the
material is assumed to be linear, its memory of this past action should be
proportional to the product e(s) and the duration of the action; that is: e(s) ds;
producing the stress: oso Mt; s esds: If the material does not age, its properties
must be independent of the time when the experiment starts. For this to be the case,
the kernel Mt; s can only be a function of the difference t - s. Clearly, the same
is true of the creep compliance. In particular, and for this reason, such kernels are
called difference kernels.
Proceeding now with the mathematical proof, we evaluate the stress response to
a periodic strain of period p : et p et, using (a):
rt p M0et p
Ztp
o
Mt p; sesds
os
Next, introduce the change of variable s = s0 ? p and use the stated periodicity of
the applied strain, p : et p et, to write:
rt p M 0et
Zt
o
M t p; s0 pes0 ds0
os0
p
Splitting the interval of integration from -p to 0, and from 0 to t; and afterward

replacing the new variable of integration, s0 with the original symbol s, for simplicity, get:
Z0
rt p M0et
o
Mt p; s pesds
os
p
Zt
o
Mt p; s pesds
os
c
Now, use that: rt M0et
Rt
0
rt p rt
Zt
o
os Mt; sesds;
to cast (c) in the form:

o
o
Mt; s Mt p; s p esds
os
os
Z0
p
o
Mt p; s pesds
os
36
It then follows that, for the response to be periodic, that is, for r(t ? p) = r(t):
Zt

o
o
Mt; s Mt p; s p esds 0;
os
os
8et
Together with:
Z0
o
Mt p; s pesds 0
os
p
Condition (e) implies that:

Mt; s Mt p; s p 0
Differentiating this expression with respect to p, and setting p = 0, afterward,

leads to5:

o
o
Mt p; s p
Mt p; s p 0
ot p
p0 os p
p0
The general solution of this equation is an arbitrary function of t - s, as is easily

verified by direct substitution. Consequently:
Mt; s Mt s
2:18
According to the fading memory principle, condition (e) is met for arbitrary
excitations only in the limit as t ? ?, if the kernel |qM/qt | of the integral is
bounded, as indicated by relation (2.10). This means, additionally, that the lower
limit in the integral in (b) must be taken as -?; and that the approximation:
Zt
o
Mt p; s pesds
os
Zt
o
Mt p; s pesds
os
2:19
1
holds only for sufficiently long times. Otherwise, the response to a periodic
excitation, even of a non-aging material, will be non-periodic.
Here, use is made of the total derivative:
d
dp f x; y
dy
o
dx
o
ox
f x; y dp
oy
f x; y dp
:
2.4
37
Fig. 2.6 Side-by-side

comparison of relaxation
modulus and creep
compliance
M ( t ) or C( t )
C(t)
M(t)
t
Summarizing: for the response of a viscoelastic material to a cyclic excitation

to also be periodic, its material property functions must depend on the difference
between current time and loading time.6 That is, the closed-cycle condition (that
the response to periodic excitation be periodic) can only be satisfied by non-aging
materials.
2.5
Relationship Between Modulus and Compliance
Expressions (2.1a, b) and (2.2a, b) relate stresses to strains, through the corresponding relaxation modulus and creep compliance. This suggests that the two
expressions may be combined in some form to obtain the relationship between the
two property functions. Before we go into the mathematical details of this, we use
what we have learned already about these two material functions and compare
their forms side-by-side in Fig. 2.6.
As suggested by the figure, it is reasonable to expect that the values of C and
M at t = 0, as well as at sufficiently long times, might be reciprocals of each other.
Although one could argue that C (t) M(t) & 1 elsewhere, the figure shows that, in
general, the creep compliance and the relaxation modulus are not reciprocals of
each other. As will be shown subsequently, the values of the relaxation modulus
and its creep compliance, for the extreme cases of glassy and equilibrium response
are indeed reciprocals of each other, as they would be for elastic solids. However,
unlike for elastic materials, the relaxation modulus is not the reciprocal of the
creep compliance.
A relationship between relaxation modulus and creep compliance may be
derived using (2.1b) to evaluate the stress response to a step-strain history e(t) = eo
H(t), together with the fact that dH(t)/dt = d(t) [c.f. Appendix A]; thus:
rt
Zt
Mt seo dsds Mteo
0
Material property functions which depend on the difference between current and loading time
are known as difference kernels.
38
Putting this result into (2.2b), taking eo outside the integral, and using that e(t) = eo
H(t):
2
et 4
Zt
3
d
Ct s Msds5eo eo Ht
ds
0
Canceling out eo produces the first form of the relationship between a relaxation
modulus and its corresponding creep compliance:
Zt
Ct s
d
Msds Ht
ds
2:20
0
Proceeding in the reverse order, applying a step stress r(t) = ro H(t), and then
calculating the corresponding strain response, the result would be:
Zt
Mt s
d
Csds Ht
ds
2:21
0
As stated earlier, (2.20) and (2.21) show that, in general, the relaxation modulus
and creep compliance are not reciprocals of each other. Additional, practical
information can be gained by examining the behavior of these expressions as time
approaches 0 and ?; as well as by invoking the consequences of the fading
memory principle.
Before proceeding, we note that the integrals in (2.20) and (2.21) correspond to
a special class of integrals known as Stieltjes convolutions. Convolution integrals
are presented in the next section in the context of viscoelasticity and are fully
discussed in Appendix A. As shown in the Appendix, by the commutative property
of convolution integrals, Eqs. (2.20) and (2.21) are mathematically equivalent and
either one could have been derived from the other.
2.5.1
Elastic Relationships
The relationships between a relaxation modulus and its creep compliance, corresponding to short and long term are obtained by taking the limit of either (2.20) or
(2.21), as t ? 0+ and t ? ?, respectively. To do that, (2.20) is rewritten by
splitting its integration interval into two intervals going from 0- to 0+, and 0+ to t:
Z0
0
d
Ct s Msds
ds
Zt
0
Ct s
d
Msds Ht
ds
2.5
39
The relationship between the short-term property functions is developed by

letting t ? 0; noting that the first integral evaluates to M(0+)C(0+) and that the
second integral vanishes. Proceeding thus, and using the notation M(0+) = Mg, and
C(0+) = Cg, to denote glassy quantities, leads to:

Mg 1 Cg
2:22
In similar fashion, taking the limit of either (2.20) or (2.21) as t ? ?, and using
the notation M(?) = Me, and C(?) = Ce, to denote equilibrium properties:
Me 1=Ce
2:23
It is left as an exercise for the reader to derive (2.23).

The last two expressions show that, as pointed out at the beginning of the
section, in the extreme cases of short-term (or glassy) and long-term (or equilibrium) response, a relaxation function and its compliance counterpart are indeed
reciprocals of each other, just as for elastic materials.
The monotonic nature of the modulus and compliance functions, stated in
(2.12) and (2.15), can be used to establish a relationship between them which also
shows that modulus and compliance are not, in general, simple inverses of each
other [2].
Indeed, using (2.21), say, with the facts that M(t) is a monotonically decreasing
function of its argument, so that M(t - s) C M(t), for all s C 0; and
M(t) = C(t) : 0, for t \ 0, there results:
Ht
Zt
0
o
Mt s Csds Mt
os
Zt
0
o
Cs MtCt
os
That is:
MtCt 1
2:24
Which, as stated before, shows that in general: M(t) = 1/C(t).
2.5.2
Convolution Integral Relationships
The mathematical relationships listed in (2.20) and (2.21) are known as Stieltjes
convolution integrals [c.f. Appendix A]. Formally, the Stieltjes integral of two
functions, / and w, is defined as [3]:
40
Zt
ut s dws
ut s
d
wsds u dw
ds
1
In which u(t) is assumed continuous in [0,?); w(t), vanishes at -?; and the form
on the far right is used when the argument, t, is understood.
In line with the mathematical structure of relaxation and compliance functions,
the further assumption is made that u and w vanish for all negative arguments,
which allows splitting the interval of integration from -? to 0-, and from 0- to t,
to write, more simply:
ut s dws
Zt
ut s
d
wsds
ds
0
Alternatively, integrating by parts:
ut s dws utw0
Zt
ut s
d
wsds
ds
As shown in Appendix A, under the stated restrictions on the functions

involved, the convolution integral is commutative, associative and distributive.
Thus, for any three well-behaved functions, f, g, and h:
f ggf
f g h f g h f h h
f g h f g f h
Based on their definition, the convolution integral allows writing viscoelastic
constitutive equations in elastic-like fashion. Corresponding to (2.1a, b) or (2.2a, b),
for instance, we write:
rt M t s des M de
2:25
et Ct s drs C dr
2:26
Additionally, corresponding to (2.20) and (2.21), above:

Ct s dM s C dM Ht
2:27
M t s dC s M dC Ht
2:28
2.5
41
These expressions clearly show that the relaxation modulus and creep compliance
are, in general, not mere inverses, but convolution inverses of each other. In
addition, the viscoelastic relations in (2.25) and (2.26) look exactly like elastic
constitutive equations, if the operation of multiplication is replaced by that of
convolution. Using this fact, it is straightforward to write down the viscoelastic
constitutive counterparts of any given elastic constitutive equations. This is done
by simply replacing the elastic property of interest (modulus or compliance) with
the corresponding viscoelastic property, and ordinary multiplication with the
convolution operation between the material property function and the applied
action (strain or stress).
Example 2.5 Write the viscoelastic version of the three-dimensional constitutive
equations of a linear isotropic elastic solid which has its stressstrain equations split
into a spherical and a deviatoric part as follows7: rS 3KeS ; rDij 2GeDij ;
i; j 1; 3
Solution:
Although three-dimensional constitutive equations will be discussed at length in
Chap. 8, this exercise is meant to get the reader comfortable with writing the viscoelastic counterparts of elastic constitutive equations. So, whatever the meaning of
the symbols involved, replace the elastic products with convolutions to write
the results directly: rS t 3Kt s deS s; rDij t 2Gt s deDij s;
i; j 1; 3:
2.5.3
Laplace-Transformed Relationships
Since linear viscoelastic constitutive equations correspond to convolution integrals, one may apply the Laplace transform to convert them into algebraic equations. As explained in Appendix A, any piecewise continuous function, f(t), of
exponential orderthat is, bounded by a finite exponential functionhas a
Laplace transform, f s; defined as:
Lff tg f s
Z1
est f tdt
The 3 9 3 stress and strain matricesindeed any square matrix of any ordermay be split into
a spherical and a deviatoric part. The spherical part is a diagonal matrix with each of its three
non-zero entries equal to the average of the diagonal elements of the original matrix. Therefore,
any one of its non-zero entries may be used to represent it. The deviatoric part of the matrix is, by
definition, the matrix that is left over from such decomposition. This decomposition is discussed
fully in Appendix B.
42
Some properties of the Laplace transform are presented in Appendix A. We list

the following two and use them to transform convolution integrals in the time
domain, t, into algebraic expressions in the transform variable, s.
Transform of first derivative:
Transform of the convolution:
d
Lf f g sf s f 0
dt
Lff gg f sgs
Indeed, applying these expressions to the convolution forms (2.25) and (2.26),
respectively, results in the following algebraic form of the constitutive equations:
rs sMses
2:29
es sCsrs
2:30
The same results would have been obtained if the Laplace transform had been
applied to the original stressstrain and strainstress equations, (2.1a, b) and
(2.2a, b). For example, if the Laplace transform is applied to both sides of (2.1a),
the relationship in (2.29) would be obtained, after collecting terms as follows:
rs Mse0 Ms ses e0 sMs es
The advantage of taking the Laplace transform of viscoelastic constitutive equations is that the transformed expressions involve only products of the transform of
the material property function of interest (modulus or compliance) and the Laplace
transform of the input functionstrain or stress, just like elastic constitutive
equations do. In other words, the Laplace transform converts a viscoelastic constitutive equation into an elastic-like expression between transformed variables.
Conversely, if each material property in an elastic constitutive relation is replaced
by its Carson8 transform and each input variable in it is replaced by its Laplace
transform, the resulting expression must stand for the Laplace transform of the
corresponding viscoelastic constitutive equation. Thus, as in the case of the convolution notation, this equivalence between elastic constitutive relations and the
Laplace transform of viscoelastic equations allows one to write down the transformed viscoelastic constitutive equations directly from the elastic ones. This
equivalence forms the basis of a so-called elasticviscoelastic correspondence
principle, which is presented in Chap. 9.
Example 2.6 Use the elasticviscoelastic correspondence to write down the viscoelastic version of the three-dimensional constitutive equations of the linear
isotropic elastic solid of Example 2.5.
8
The s-multiplied Laplace transform of a function is simply called the Carson transform of the
function.
2.5
43
Solution:
Using the elasticviscoelastic correspondence, write the Laplace transform of
eDij ; i; j 1; 3: The vis eS ; r
S 3sK
Dij 2sG
the elastic expressions as: r
coelastic constitutive equations are obtained taking the inverse Laplace transform
of the forms given. Thus, rS 3K deS ; rDij 2G deDij ; i; j 1; 3:
The relationship between the relaxation modulus, M, and the creep compliance,
C, in the transformed plane, can be obtained either by applying the Laplace
transform to (2.27) or (2.28), or by combining the algebraic expressions (2.29) and
(2.30). In either case, there results:
MsCs
1
s2
2:31
Example 2.7 The relaxation modulus of a viscoelastic solid is given by M t

Me M1 eat : Use expression (2.31) and Laplace transform inversion to obtain its
creep compliance, assuming the latter is a function of the form:
C t Ce C1 ebt :
Solution:
According to (2.31), the creep compliance function would be given by the

Hence, we first evaluate this
inverse Laplace transform of the function 1=s2 Ms:
s Ce =s
function, then invert it, and equate it to the Laplace transform C
C1 =s b; of the desired creep compliance. Proceeding thus, using the table of
s
transforms included in Appendix A, and simplifying, there results: s2 M
sa
: Expanding this rational function into its partial fractions, as
sMe a Me M1 s
explained in Appendix A; using the notation Me ? M1 = Mg, simplifying and
C1
equating the result to the Laplace transform of C(t), there results: Cse
s b
1=Me Me Mg =Me Mg
: Equating coefficients of the corresponding powers of

s
s aMe =Mg
s yields: Ce = 1/Me, C1 = (Mg - Me)/(MeMg), b = Me a/Mg.
More general methods of approximate and exact inversion of material property
functions given as sums of exponential functions are presented in Chap. 7.
2.6
Alternate Integral Forms
Depending on preference, and the application at hand, the integral constitutive

equations for viscoelastic substances may be written in several different ways. The
mathematical operations that are used to transform one constitutive form into
44
anothermost typically, integration by partsrequire that the material property

functions involved, and their time derivatives, be bounded. On occasion, the
transformations also assume that the material property functions vanish identically
for all negative time.
For ease of reference, we list Boltzmanns equation (2.1b), where it was noted
that e (t) : 0, for t \ 0, allowed us to write [4]:
rt
Zt
Mt s
de
ds
ds
2:32a
0
On the physical expectation that M(t) be bounded for all values of time, and
requiring that e (t) ? 0, as t ? -?, which is satisfied, since both e (t) ? 0, and
de/dt : 0, for t \ 0, one may extend the lower limit of integration to -?, in
(2.32a), without altering its value. Thus,
rt
Zt
Mt s
de
ds
ds
2:33a
1
Another useful form is obtained integrating (2.32a) by parts and simplifying:
rt M0et
Zt
o
Mt sesds
os
2:34a
Using the notation M (0) : Mg, and introducing the normalized function
m(t) : M(t)/Mg:
8
<
rt Mg et
:
Zt
0
9
=
o
mt sesds
;
os
2:35a
Using the notation rg (t) : Mg e (t), and taking Mg inside the integral, produces
the form:
rt rg t
Zt
o
mt srg sds
os
2:36
An important application of this is in the derivation of constitutive equations for

materials that are termed hyper-viscoelastic. Equations for hyper-viscoelastic
materials are derived from those of hyper-elastic materials. A material is termed
hyper-elastic, if there exists a potential function of the strains, say, W, such that
each individual stress component in such a material may be computed as the
2.6
Alternate Integral Forms
45
derivative of W with respect to the corresponding strain [5]. Since both the glassy
or short-term and the equilibrium or long-term responses of a viscoelastic solid are
elastic, either can be used to define the potential function. We proceed by using the
glassy response; thus
r g t
o
Wg et
oet
2:37
The stressstrain law in (2.36) would then take the equivalent form:
rt
o
Wg et
oet
t
0
o
o
mt s
Wg esds
os
oes
2:38
Another form, which allows a generalization to non-linear viscoelasticity, is

derived by introducing the strain relative to the configuration at time t :
erel t; s et es: Using that erel t; 0 et in (2.35a) yields:
8
<
rt Mg erel t
:
Zt
0
9
=
o
mt serel sds
;
os
2:39
Constitutive Eqs. (2.34a, b), (2.35a, b), (2.36) and (2.39) are also frequently
written in terms of integral operators, using convolution integral notation, but the
exact form of the kernel (i.e., the derivative of the relaxation function) is not
disclosed. With the obvious definitions, those equations would read:
rt Mg fet Ct s esg
2:40a
rt Mg f1 Ct sget
2:40b
rt rg t Ct s rg s
2:41a
rt f1 Ct sgrg t
2:41b
rt Mg erel t; 0 Mg Ct s erel t; s
2:42
Example 2.8 Use (2.37) and (2.38) to develop the stressstrain law for a hyperviscoelastic material having normalized relaxation function, m = a ? (1 - a)e-t/g,
if it is known that its glassy response can be established from the potential function
of the strains Wg = Ee2(t).
Solution:
The hyper-viscoelastic form is derived by putting the given functions into
(2.37) and (2.38) directly. Evaluating (2.37) first: rg t oeo Wg et E et:
46
Taking this result and m into (2.38) and re-arranging: rt E et
Rt o
0
os fE a
1 aets=g g esds:
Comparing this with (2.34a) shows that hyper-viscoelastic material in question
is linearly viscoelastic with relaxation modulus: M t Ea 1 aet=g :
We end this section by presenting some of the constitutive equations in strain
stress form which are the exact counterparts of the foregoing expressions. These
strainstress forms are derived by reversing the roles of stress and strain in the
arguments that led to the previous forms. For instance, the strainstress equations
analogous to (2.32a)(2.35a) are:
et
Zt
dr
ds
ds
2:32b
dr
ds
ds
2:33b
o
Ct srsds
os
2:34b
Ct s
0
et
Zt
Ct s
1
et C0et
Zt
0
8
<
et Cg rt
:
2.7
Zt
0
9
=
o
ct sssds
;
os
2:35b
Work and Energy
Under the action of external agents, be these loads or displacements, a deformable

body will change its configuration and, if not properly restrained, undergo largescale motion. At any rate, as the points of application of the external agents move,
workdefined as the product of force and displacementis performed on the
body. To develop an expression for the work performed on a body by the external
agents, we use a uniaxial specimen of constant cross-sectional area, A, initial
length, l, and volume, V, that is loaded at its ends by either a displacement u or a
force F. With this, the rate of work of the external forcesthat is, force times
displacement ratecan then be expressed as:
dW
du
F
dt
dt
2:43
2.7
Work and Energy
47
Multiplying and dividing by the specimens volume V = Al, and using that the
strain is given e = u/l, we cast the previous expression in the form:
dW
du F
du=l
de
F Al
V r
dt
dt A
dt
dt
2:44
The total work performed during a time interval (0, t) is given by the integral:
Wjt0
Zt
dW
ds
ds
Zt
Fs
du
ds
ds
2:45
Combining (2.44) and (2.45) and dividing the result by the specimens volume, V,
produces the work per unit volume, WV, that is input into the system:
WV jt0
1=VWjt0
Zt
rs
des
ds
ds
2:46
In practical applications, we insert an appropriate form or another of the constitutive equation, such as (2.1b) and write (2.46) as:
WV jt0
Z t Zs
Ms s
des des
ds
ds
ds
ds
2:47
s0 s0
Suitable functions (bounded and piecewise continuous) allow interchanging the

order of integration. Before doing this, we note that the relaxation modulus, M(t),
is defined only for positive values of time. One can continue it to negative values
of its argument in an arbitrary manner. In particular, it is sometimes convenient to
assume M(t) either as an even or an odd function of time, that is,
M t M t
Fig. 2.7 Region of

integration used for change of
variables in work expression
2:48
s
2
s = t
=s
= t
48
M t M t
2:49
Here, we take M to be an even function of time. With this, the integral in (2.47) in
the s-s plane is taken over the area of a right triangle with base of length t, whose
hypotenuse is the ray s = s, as indicated in Fig. 2.7.
Using (2.48) and noting that the square 1234 in the figure is made up of two
triangles of equal area, and that on account of (2.48), the value of M is the same on
points that are symmetrically located about the diagonal of the square, there
results:
1
WV jt0
2
Zt
s0
2
4
Zt
3
des 5 des
ds
ds
Ms s
ds
ds
2:50
s0
The expressions derived here apply equally to any other one-dimensional pair
or work-conjugate quantities, such as shearing force and deflection, torque and
twist angle, or bending moment and rotation.
2.8
Problems
P.2.1 Determine the constant-rate effective modulus, Eeff t; of a one-dimensional

solid made of a viscoelastic material whose relaxation modulus is:
Mt Ee E1 et=s1 :
Answer : Eeff t Ee
i
E1 h
1 et=s1
1=s1
Hint: Use M(t) with the defining expression derived in Example 2.1 and carry out
the indicated integration.
P.2.2 Use convolution notation to derive the relationship between relaxation
modulus and creep compliance.
Answer : Mt s dCs Ht
Hint: Combine (2.25) and (2.26) to get rt Mt s des Mt s
dfCs s drsg; then, use that: rt ro H t and thus drt ro dt to
evaluate the convolution integral inside the braces, and obtain: ro M t s
dCsro ; from which the desired result follows.
P.2.3 As presented in Chap. 7, a popular analytical form used to represent relaxation
functions consists of a finite sum of decaying exponentials, which in the literature is
usually referred to as a DirichletProny series or, more simply, Prony series:
2.8
Problems
49
Mt Me
N
X
Mi et=si ;
Me 0;
si [ 0; 8i
and : Mi ;
i1
In this expression, Me represents the equilibrium modulus, which is zero for a

viscoelastic liquid [c.f. Chap. 1]. Also, although the sis represent relaxation times
of the material and are thus material properties, in practice, they, as well as Me and
the coefficients Mi, are all established by fitting the Prony series to experimental
data. Prove that such forms satisfy the requirements of fading memory.
Hint:
(a) Evaluate the derivative of the series as t ? ? to show it satisfies (2.10).
o
Mt
lim
t!1 ot
lim
t!1
N
X
Mi
1
si
)
e
t=si
(b) Compare the values of the function at t2 [ t1, to prove that the Prony series is
a monotonically decreasing function, in accordance with (2.12).
Mt2
N
X
Mi et2 =si
N
X
Mi et1 =si Mt1 ;
8 t2 [ t1
(c) Evaluate the derivative of the series at t2 [ t1 and prove that the absolute value
of its derivative is also monotonically decreasing, satisfying (2.11).

X
o
X
N
N

o
M
M

i
i
t2 =si
t1 =si
Mt

e
e

Mt ;

ot

ot
s
s
i
i
tt2
tt1
1
1
8 t2 [ t1
P.2.4 As discussed in Chap. 7, the power-law form: M t Me Mt 1 at p is

also used to represent the relaxation function of viscoelastic solids. Show that this
form satisfies the requirements of fading memory. In this expression, Me, Mt, a and
p, are all positive.
Hint:
(a) Proceed as in P.2.3 and evaluate the derivative of the given power-law form as
t ? ? to show it satisfies (2.10).
o
lim
M t
t!1 ot
pMt
1
lim
t!1
a 1 t=a1p
)
0
50
(b) Compare the values of the given function at t2 [ t1, to prove that this powerlaw form is a monotonically decreasing function, in accordance with (2.12).
t2
t1
M t2 Me Mt 1 p Me Mt 1 p ;
a
a
t2 [ t1
(c) Evaluate the derivative of the function at t2 [ t1 and prove that the absolute
value of its derivative is also monotonically decreasing, satisfying (2.11).

o

M t pMt 1 t2 1p pMt 1 t1 1p o M t ;
a

ot

ot
a
a
a
t2
t1
t2 [ t1
P.2.5 Compute the steady-state response of the one-dimensional solid of P.2.1 if it

is subjected to the cyclic strain history e(t) = eocos(xt).
"
Answer : rt Ee
E1 xs2
1 xs2
eo cosxt
E1 xs
1 xs2
eo cosxt
Hint: Take the strain history into (2.1b); use integration-by-parts twice; simplify,
xs
t=s
and discard the transient term: E1 eo 1xs
to obtain the desired result.
2 e
P.2.6 Repeat problem P.2.5 if the cyclic strain history is e(t) = eosin(xt).
"
Answer : rt Ee
E1 xs2
1 xs2
#
eo sinxt
E1 xs
1 xs2
eo cosxt
Fig. 2.8 Problem 2.7
t*
2t *
2.8
Problems
51
Hint: Take the strain history into (2.1b); use integration-by-parts twice; simplify,
xs
t=s
to obtain the desired result.
and discard the transient term E1 eo 1xs
2 e
P.2.7 A uniaxial bar of a viscoelastic solid with relaxation modulus M(t) is subjected to a constant-rate loadunload strain history, as shown in Fig. 2.8. Prove
that a non-zero stress will exist in the bar at the time when the strain reaches zero
at the end of the load-unload cycle.
Hint:
Rt;
t t
; evaluate (2.1b) at t = 2t*. Split

Rt Rt t ; t t
the integration interval into two parts: from 0 to t* and t* to 2t* and introduce a
R t
R 2t
change of variables to arrive at the result: r2t R 0 Esds R t Esds:
Proceed as in Example 2.1 and multiply
and divide this expressioni by t* to cast the
h
R
R 2t
t
result into the form: r2t Rt t1 0 Esds t1 t Esds : The quantities
Using that et
inside the brackets are the average values of the relaxation function in the
respective intervals of integration. Because the relaxation modulus is a monotonically decreasing function of time, the first integral inside the brackets is
numerically larger than the second. This proves that, while e2t is zero, r2t is
negative.
P.2.8 In Chap. 1, it was pointed out that in an elastic solid the stress corresponding
to a given strain will always be the same, irrespective of the time it takes to apply
the strain, and that contrary to this, the stress in a viscoelastic material will depend
on the rate of straining, and hence, on the time it takes the strain to reach a
specified value. Considering two constant strain-rate histories, e1(t) = R1t and
e2(t) = R2t, derive an expression for the duration, t2, at which a viscoelastic
material subjected to a strain history e2(t) = R2t would develop the same stress
response as it would after t1 units of time under the strain history e1(t) = R1t1.
Answer : t2
Eeff t1 R1
t1
Eeff t2 R2
Hint: Proceeding as in Example 2.1, evaluate (2.1b) for each constant strain-rate
load and arrive at rt1 Eeff t1 R1 t1 and rt2 Eeff t2 R2 t2 ; where the average
or effective modulus, Eeff t; is given by (2.3). The result follows from these
relations.
P.2.9 The work per unit volume, WV (t), performed by external agents acting for
t units of time on a uniaxial bar of a viscoelastic material is given by: WV t
Rt
0 rsdes: Evaluate the work per unit volume, done in a complete cycle, on a
bar of a viscoelastic materials with relaxation modulus M(t), if the applied excitation is e(t) = eosin(xt).
52
Answer : WV pro eo sind

Hint: Using that the response to a periodic excitation will be periodic and of the
same frequency as the excitation, but out of phase with it, let d be the phase angle,
and take the response to be r(t) = rosin(xt ? d). Insert the stress and strain into
R tp
the expression for the work per unit volume, and write: WV t r0 sin
xs deo x cosxsds; where p = 2p/x is the period. Now use trigonometric
identities to expand the circular function sinxs d sinxs cosd
cosxs sind; perform the integration, using the periodicity of the circular
functions, and simplify to arrive at the result. As will be explained in Chap. 4, the
phase angle, d, is a characteristic of the materials relaxation modulus.
P.2.10 Repeat Problem P2.9 using the periodic strain history e(t) = eocos(xt).
Answer: The result is the same as for a sine function history: WV pro eo sind
Hint: Proceed as in Problem 2.9.
References
1. A.D. Drosdov, Finite elasticity and viscoelasticity; a course in the nonlinear mechanics of
solids, World Scientific, pp. 267271, 279283 (1996)
2. A.C. Pipkin, Lectures on viscoelasticity theory (Springer, New York, 1986), pp. 1416
3. H.-P. Hsu, Fourier analysis (Simon and Schuster, New York, 1970) pp. 8892
4. R.M. Christensen, Theory of viscoelasticity, 2nd edn. (Dover, New York, 2003), pp. 39
5. L.E. Malvern, Introduction to the mechanics of a continuous medium (Prentice-Hall,
Englewood, 1963) pp. 278290, 282285
Constitutive Equations in Differential

Operator Form
Abstract
The mechanical response of a viscoelastic material to external loads combines

the characteristics of elastic and viscous behavior. On the other hand, as we
know from experience, springs and dashpots are mechanical devices which
exhibit purely elastic and purely viscous response, respectively. It is then
natural to imagine that the equations that relate stresses to strains in a
viscoelastic material could be represented with an appropriate combination of
equations which relate stresses to strains in springs and dashpots. To develop
this idea, Sect. 3.2 examines the response of the linear elastic spring and linear
viscous dashpot to externally applied loads. The response equations for these
simple mechanical elements are formalized in Sect. 3.3, with the introduction of
so-called rheological operators. As it turns out, because combinations of springs
and dashpots require the addition and multiplication of constant and first
derivative operators, it turns out that the constitutive equation of general
arrangements of springs and dashpots, such as are needed to reproduce observed
viscoelastic behavior, must be represented by linear ordinary differential
equations whose order depends on the number, type, and specific arrangement
of the springs and dashpots. The physical significance of the coefficients in the
resulting differential equations is examined also, and the proper form of the
initial conditions established. As will be seen, the mere presence or absence of
some of the coefficients of a differential equation reveals whether the particular
arrangement of springs and dashpots it represents will model fluid or solid
behavior, and whether it will exhibit instantaneous, elastic response. A general
approach to establishing rheological models is presented in Sect. 3.4, and
applied in Sects. 3.5 through 3.7 to develop the differential equations, and
examine the behavior of simple and general rheological models.
Keywords
Compliance Creep Damper Dashpot Deviatoric Fluid Kelvin Laplace

53
54
3 Constitutive Equations in Differential Operator Form
Maxwell Model Modulus Pressure Recovery Relaxation Spherical

Spring Strain Stress Relaxation Retardation Solid Spectrum
3.1
Introduction
As indicated in Chap. 1, the mechanical response of a viscoelastic material to

external loads combines the characteristics of elastic and viscous behavior. On the
other hand, as we know from experience, springs and dashpots are mechanical
devices exhibiting purely elastic and purely viscous response, respectively. It is
then natural to imagine that the equations that relate stresses to strains in a viscoelastic material could be represented with an appropriate combination of
equations which relate stresses to strains in springs and dashpots. To develop this
idea, Sect. 3.2 examines the response of both the linear elastic spring and linear
viscous dashpot to externally applied loads. The response equations for these
simple mechanical elements are formalized in Sect. 3.3, with the introduction of
so-called rheological operators. Because the mathematical combinations of springs
and dashpots require the addition and multiplication of constant and first-derivative operators, it turns out that the constitutive equation of general arrangements of
springs and dashpots, which are needed to capture real viscoelastic behavior, must
be represented by linear ordinary differential equations whose order depends on
the number, type, and specific arrangement of the springs and dashpots. The
physical significance of the coefficients in the resulting differential equations is
also examined, and the proper form of the initial conditions is established. As will
be seen, the mere presence or absence of some of the coefficients of a differential
equation can reveal whether the particular arrangement of springs and dashpots it
represents will model fluid or solid behavior and whether it will exhibit instantaneous, elastic response. A general approach to establishing rheological models is
presented in Sect. 4 and then applied to develop the differential equations and
examine the behavior of simple, generalized, and degenerate rheological models,
in Sects. 3.5 through 3.7, successively.
3.2
Fundamental Rheological Models
There are two basic rheological models in linear viscoelasticity: the linear elastic
spring and the linear viscous damper. The constitutive equations of elastic springs
and viscous dashpots are expressed in terms of force and displacement, and force
and displacement rate, respectively.1 Here, an appropriate scale is assumed and
forces are replaced with stresses and displacements with strains. Also, in
1
We use axial springs and dashpots in the derivations, but emphasize that the corresponding
relations for shear stress and shear strain have the same mathematical form.
3.2
55
Fig. 3.1 General loading

with step discontinuity
f(t)H(t-to)
fo
to
examining the response of the basic rheological elements to general loading (either
stress or strain), it is assumed that the load may have a finite jump discontinuity at
the time of its application, but is continuous thereafter, as depicted in Fig. 3.1.
Note that, by definition:
f tH t to
3.2.1
0;
t to
f t; t to
Linear Elastic Spring
The physical (mechanical) model and free body diagram of the linear elastic spring
are shown in Fig. 3.2. Its constitutive equation is presented next, followed by its
response to general strain and stress loading.
3.2.1.1 Constitutive Equation

The mechanical behavior of a linear elastic spring is governed by Hookes law:
stress is directly proportional to strain. Using r to represent the stress, e the
strain, and E the modulus as the constant of proportionality, we write the
constitutive equation of the spring in stressstrain and strainstress forms, as
rEe
3:1a
e 1=E r
3:1b
Fig. 3.2 Mechanical model

and free body diagram of
elastic spring
56
3.2.1.2 Response to Strain Loading

In this case, the controlled variable is strain, the loading takes the form e(t)=
f(t)H(t - to), and the response of the elastic spring to such load is from (3.1a):
rt E ftHt to
This shows that the response of an elastic spring is instantaneous and remains nonzero for as long as the applied strain is non-zero: a known characteristic of the
behavior of solid materials.
The relaxation modulus of an elastic spring, defined as the stress response to a
step strain, e(t) = eo H(t - to), reckoned per unit applied strain, is obtained from
(a), using f t eo
M t to rt=eo E H t to
3:2a
M t E H t:
3:2b
or, simply
3.2.1.3 Response to Stress Loading

The controlled variable for this case is stress. The loading takes the form r(t)=
f(t)H(t - to), and the response of the spring is obtained using (3.1b), as
et 1=E rt H t to
In other words, the stain response of an elastic spring is instantaneous,

remaining non-zero for as long as the applied stress remains non-zero. This is a
characteristic of solid behavior.
The creep compliance of an elastic spring, defined as the strain response to a
step stress, r(t) = ro H(t - to), measured per unit of applied stress, is obtained
from (b), using f t ro
C t to et=ro 1=EH t to
3:3a
Ct 1=EH t
3:3b
More simply,
In particular, the response of the linear elastic spring to stress and strain recovery
tests is easily evaluated using (a) and (b), setting f t fo H t to H t t1 ,
with fo replaced by eo or ro, respectively. The two experiments and corresponding
responses are depicted in Fig. 3.3.
3.2
Fig. 3.3 Recovery response

of elastic spring. a Stress
recovery. b Strain recovery
57
(a)
(b)
= Eo
= (1/E)o
3.2.2
Linear Viscous Dashpot
The physical model and free body diagram of the linear viscous damperor
dashpotare presented in Fig. 3.4. Its constitutive equation is discussed next and
used afterward to examine the mechanical response of the dashpot to general stress
and strain loading.

The mechanical behavior of a linear dashpot follows Newtons law: stress is
directly proportional to strain rate. As before, r and e will denote stress and strain,
respectively. Therefore, using g to represent the dashpots constant viscosity, the
stressstrain form of the constitutive equation for the linear dashpot becomes
r t g
d
et
dt
3:4a
Dispensing with integration, for the time being, we rewrite the constitutive
equation for the linear dashpot in strainstress form, by reversing the order of
terms in the equation:
d
1
et rt
dt
g
3:4b
Fig. 3.4 Mechanical model

and free body diagram of
linear dashpot
, d /dt
58

The response of the linear dashpot to the general strain history
et f tH t to , shown in Fig. 3.1, is obtained by taking it into (3.4a) and
performing the indicated differentiation using the properties of the unit step and
delta functions2 (see Appendix A):
d
d
et g f tH t to
dt

dt
d
g
f t H t to f tdt to
dt

d
g H t to f t f to dt to
dt
r g
Note that, because the response contains an impulse function at t = to, it would
take an infinite stress to impose an instantaneous strain on a linear viscous dashpot.
Once the strain is imposed, however, the stress would decay instantly to the value
of the derivative of the strain history.
The relaxation modulus of the linear viscous dashpot, being defined as the stress
response to a step-strain history e(t) = eoH(t - to), and measured per unit applied
strain, is obtained from (a), noting that f = f (to) = eo and df/dt = 0; hence,
Mt to g dt to
3:5a
Mt g dt
3:5b
or
In other words, the relaxation modulus of a linear viscous dashpot is an impulse

function.
3.2.2.3 Response to Stress Loading:

The response of the linear viscous dashpot to a general stress history is obtained by
taking the controlled variable, r(t) = f(t)H(t - to), into (3.4b), and integrating the
resulting expression between 0 and t; using the shifting property of the Heaviside
function3 (see Appendix A):
1
e t e 0
g
Z
0
1
f sH s to ds H t to
g
Using the initial condition, e(0) = 0:
2
3
In particular, that dH(t)/dt = d(t); f(t)d(t - to) = f(to)d(t - to).

Rt
Rt
Specifically, that 0 f sH s to ds H t to to f sds; for to [ 0.
to
f sds
3.2
59
(a)
(b)
t
to
t1
to
t1
= o (t-to)
= o(t1-to)/
t
= - o (t-t1)
Fig. 3.5 Response of viscous damper. a Stress recovery. b Strain recovery
1
et Ht to
g
f sds
to
The creep compliance follows from this by setting f (t) : f t eo and performing the required integration:
1
C t to
g
eo ds
to
t to
Ht to
g
3:6a
or
t
Ct ; for t [ 0
g
3:6b
The response of the linear viscous dashpot to the standard stress and strain
recovery tests is evaluated using (a) and (b), setting f t fo H t to
H t t1 , with fo replaced by eo or ro, respectively. The experiments and corresponding responses are depicted in Fig. 3.5.
3.3
Rheological Operators
Rheological operators arise naturally from the form of the constitutive equations of
the spring and dashpot models. Indeed, these equations listed in stressstrain form
as (3.1a, b) can be expressed symbolically as resulting from a mathematical
operation on the strain history. For the linear elastic spring, this interpretation
60
actually defines the modulus as the constant operator in question, [E], and multiplication of real numbers (i.e., modulus times strain) as the operation. Similarly,
the stress response of the linear viscous damper may be thought of as resulting
from applying the first-order derivative operator, [gd/dt] to the strain history, e(t).
This idea is developed further in what follows.
3.3.1
Fundamental Rheological Operators
Following the previous reasoning, we rewrite the stressstrain constitutive equations of the spring and dashpot using the two fundamental rheological operators,
[E] and [gd/dt], as
rt Eet
3:7a

d
r t g e t
dt
3:8a
The usefulness of working with rheological operators stems from the fact that
rheological operators may be considered as algebraic entities which follow the
rules of addition and multiplication of real numbers and enjoy the properties of
derivatives of functions of real variables. In more formal language, the basic
rheological operators are as follows:
Linear, because they are homogeneous and additive:
E[ae1(t) ? be2(t)] = aEe1(t) ? bEe2(t);
gqt[ae1(t) ? be2(t)] = a gqte1(t) ? bgqte2(t); etc.
Commutative, because they can be added or multiplied together in any order:
Ee(t) + gqte(t) = [E + gqt]e(t) = [gqt + E]e(t) = gqte (t) +Ee(t);
[E][gqt]e(t) = [E][gqte(t)] = Egqte(t) = gEqt e(t) = gqt[Ee(t)]; etc.
Associative, because they can be grouped together in any order:
E1e(t) ? E2e(t) +a gqtEe(t) = [E1 +aEgqt]e(t) ? E2e(t)
= [E1 ? E2]e(t) +aEg qte(t); etc.
Distributive, because their products distribute their sums and vice versa:
[E1 ? g1qt] [E2 ? g2qt]e(t) = [E1E2 ? E1g2qt ? g1 E2qt ? g1g2qtqt]e(t); etc.
Put another way, rheological operators can be added or multiplied together in
any order; subtracted from one another; and, being careful about it, even symbolically divided by one another. The trick to performing these algebraic manipulations with rheological operators is to bear in mind that each operator or operator
expression acts on a function and its meaning is deciphered proceeding from right
to left. For instance, in the product of the operator g qt and the constant operator
E, the result must equal g E ot E g ot , irrespective of which operator
appears initially on the left. This is so, because the product should be interpreted as
g ot E f E g ot f , and not as the differential operator qt applied first to
3.3
61
the constant operator E and then to the function, f, which would, incorrectly, yield
the zero operator g ot E 0, applied on f.
It is important to note that when the functions involved (stress or strain) depend
both on position and on time, as is typical in two and three dimensions, it is more
proper to use the partial derivative operator, qt, in place of the total derivative
operator dtd . For ease of notation, we use the partial derivate symbol almost
exclusively, even in the one-dimensional case. In keeping with this notation,
k
dk
dtk and ot will be used to denote the kth-order derivative operator with respect to
the independent variable, t.
With the above in mind, we divide (3.7a) and (3.8a) by the corresponding
operator and arrive at the constitutive equations of the linear spring and linear
viscous damper in strainstress form:
1
rt
E
3:7b
1
rt
got
3:8b
et
et
3.3.2
General Rheological Operators
Since the constitutive relations for the spring and dashpot involve derivatives of
both the stress and the strain, it is logical to expect that the constitutive equations
corresponding to rheological models made up of more elaborate combinations of
linear springs and linear dashpots should include higher-order derivatives of the
stress and the strain. That this should be so is due to the fact that, whatever the
complexity of a given spring-and-dashpot arrangement, its rheological equation
has to be constructed through the addition, subtraction, multiplication, and division
of fundamental rheological operators, Ei and gjqt, whose parameters, Ei and gj,
depend on the specific properties of the basic elements they represent. The result is
that the constitutive equation of a general rheological model is always an
expression of the form:
po r p 1
d1
dm
d1
dn
r

p
r
q
e
q
e

q
e
m
o
1
n
dt1
dtm
dt1
dtn
3:9a
and, by the very process that leads to this expression, it may be argued that
m B n because the orders of the differential operators, P and Q are established
by products of the fundamental rheological operators, one of whichthe
springsis balanced regarding differentiation; the other onethe dashpots
contains a derivative of strain which is of order one higher than that of stress.
62
If qo = 0, the constitutive equation will contain only derivatives of strain, and

at least the lowest of them must be non-zero, for a stress to develop in the
system. Since the stress would then depend on strain rate, the rheological
equation would represent fluid behavior.
If qo = 0, both the stress and the strain would approach finite values r? and e?,
as time increases without limit (t ? ?), but all derivatives in (3.9a) would
vanish. In the limit, a non-zero stress, r? = (qo/po)e?, would remain, for as long
as the strain does, and vice versa. Such rheological models represent elastic, solid
behavior with long term, or equilibrium modulus, M(?) = po/qo : M? : Me,
and also equilibrium compliance C(?) = po/qo : C? : Ce, which is the
reciprocal of the equilibrium modulus.
Expression (3.9a) may be rewritten in a number of forms as follows:
Using the summation symbol and introducing the zero-derivative operator,
d0
dt0 1:
X

Xm dk
n
dj
r
et
p
q
k k
k0
j0 j dt j
dt
3:9b
Or, as we shall usually do, in symbolic form, using linear differential operator
notation, introducing the stress and strain operators, P and Q, respectively, as
Pr Qe; or: Pr Qe
3:9c
With the following obvious definitions of the operators P and Q,

P po
d0
d1
d2
dm

p
1
2
m
dt0
dt1
dt2
dtm
3:10a
Q qo
d0
d1
d2
dn
q1 1 q2 2 q n n
0
dt
dt
dt
dt
3:10b
Finally, by the properties of linear differential operators, stated earlier, we write

the symbolic stressstrain and strainstress equations of a general rheological
model, as
r Q=Pe
3:11a
e P=Qr
3:11b
3.3
3.3.3
63
Rheological Equations in Laplace Transformed Space
Under certain conditions of continuity and boundedness of the functions involved,

it is possible and often advantageous to transform an ordinary linear differential
equation into an algebraic one. As discussed in Appendix A, this can be done by
means of an integral transformation such as the Laplace or the Fourier transform.4
The Laplace transform is used in what follows to convert the constitutive equation
(3.9a) into an algebraic equation between transformed stress and transformed
strain. As usual, the letter s will be used to denote the transformed variable, and an
over-bar will denote a transformed quantity.
Accordingly, applying the Laplace transform to the general constitutive equation (3.9a), using at-rest conditions for both the stress and the strain, yields (see
Appendix A):
pm sm r qo e q1 s e qn sn e
p1 s r
po r
3:12a
or after factoring out the transformed stress and strain and including the transform
variable as argument of the functions involved, for emphasis
ses
s
P
r s Q
3:12b
po p 1 s p 2 s 2 p m s m
Ps
3:13a

Qs
q o q 1 s q2 s 2 q n s n
3:13b
where
are algebraic quantities, the symbolic expressions (3.11a, b) take

and Qs
Since Ps
a true algebraic meaning in Laplace-transformed space, becoming, respectively:
3.3.4
s
Q
s es
r
P s
3:14a
s
P
es
s
r

Q s
3:14b
Initial Conditions for Rheological Models
As expression (3.9a) indicates, the constitutive equation of a general rheological

model is an ordinary linear differential equation of order m, if stress is the dependent
4
The Fourier transform is used in later chapters as an efficient means of solving steady-state
oscillation problems.
64
variable, and strain, the controlled variable, and of order n, if the roles of stress and
strain are reversed. Its general solution requires a set of independently prescribed
initial conditions, which must be equal in number to the order (m or n) of the equation
[1]. Before we present the general procedure to establishing the correct initial
conditions for general rheological differential equations, we examine a simple case.
Example 3.1 Establish the initial conditions for the rheological model po rt
p1 ot rt q1 ot et if it is subjected to a step strain history e(t) = eoH(t).
Solution:
In particular, note that we need to establish the value r0 of the stress at
t = 0+. To do this, we integrate the differential equation of the model in the
R 0
interval 0 ; 0 and write po 0 rsds p1 r0 r0 q1 e0
e0 .
Using that the first integral in this expression is zero, for any bounded stress,
and that so are r0 and e0 , leads to the initial condition: r0
q1 =p1 e0 q1 =p1 eo .
k
We note that, each term in the general rheological equation is of the form dtd k f ,
where f stands for either stress or strain. Also, although m B n, we assume the
differential equation is of order n. In other words, we assume pn = 0, but set
pn = 0, in case m \ n.
We also allow the dependent variable (the one which is the object of the
differential equation) to have jump discontinuities whenever the controlled
variable has them. This is convenient for analytical purposes, as it permits the
derivation of succinct mathematical expressions for creep and stress relaxation
experiments.
With the previous notes in mind, we proceed as follows:
1. Assume that the controlled variable (either r or e) and all its time derivatives,
up to and including its nth derivative, are given at t to and are zero at t to .

f to ; o1t f to ; o2t f to ; . . .; ont f to ; 6 0;
f r or e
2. Assume that the dependent variable (either r or e) and all its time derivatives,
up to and including the nth derivative, may have a non-zero value immediately
after the controlled variable is applied. That is,
3. Integrate the differential equation n times, between to and to . In doing this,
note that after n-iterated integrations, all derivatives of order k B n become
continuous. This means that, after n integrations, the respective values of each
integrated function are the same at to and to , and so their integrals between to
and to vanish. Then, after n integrations, obtain that
pn rto qn eto
3.3
65
There are two cases to consider:

Case 1 m = n, and therefore, pn = 0.
(a) If stress is the dependent variable, the initial condition on stress is
rto
qn
et
pn o
This implies that to a strep strain of magnitude eto , the material will respond

with a step stress of magnitude qpnn e to . In other words, the material exhibits
instantaneous elasticity with glassy (impact) modulus M 0 qpnn Mo Mg :
(b) If strain is the dependent variable, the initial condition on it is
eto
pn
rt
qn o
This implies that to a step stress of magnitude rto , the material will respond

with a step strain of magnitude pqnn r to . The material exhibits instantaneous
elasticity, with compliance: C 0 pqnn Co Cg .
Clearly, the instantaneous modulus and compliance are reciprocals of each
other.
Case 2 m \ n, that is: pn = 0.
(a) if stress is the dependent variable, the result after m integrations would be
pm rto qm eto qm1
d
et
dt o
For a discontinuous strain, this implies that the stress would be infinite at the
point of application of the strain, because the derivative of a step function is the
Dirac delta function (see Appendix A):
rto ! 1
In other words, if the higher-order derivative of the stress is less than that of the
strain, and the stress is the dependent variable, the material will not be able to
respond to an instantaneously applied strain.
(b) If strain is the dependent variable, we would integrate the differential equation
m times to get the initial condition on strain. The resulting functions of stress
would be continuous and would integrate to zero in the interval between to
and to . The proper initial condition on strain would then have to be that
eto 0
66
In this case, the material has zero instantaneous elastic response and does not
respond instantaneously to an instantaneously applied stress.

4. Having established the initial value of the dependent function, r to or e to ,
we now integrate the differential equation n - 1 times, between the same two
limits to and to , and arrive at the following equation, in which the only
unknown is the first derivative of the dependent variable (stress or strain):

d
d
to pn rto qn1 e to qn e to
dt
dt
n1 r
Use this expression together with the given values of the controlled variable and
its first derivative, and the value already established for the initial condition on the
dependent variable, to solve for the initial condition sought.
(a) If stress is the dependent variable, expression (h) yields that
d qn1 qn d pn1
r t
e t
e to
r to
pn
pn dt o
pn
dt o
Every term on the right-hand side of this expression is known. This means that
we could, in principle, use this expression directly to establish the correct initial
condition on the first derivative of stress. For convenience, we replace the initial
condition on stress by its value in terms of applied strain, collect terms (multiplying and dividing the first term on the right-hand side by the pn), and arrive at
d qn1 pn qn pn1 qn d
r t
e t
e to
p2n
pn dt o
dt o
(b) If strain is the dependent variable, the roles of stress and strain are reversed.
Then, using (j) and inverting the roles of the ps and qs yield
d pn1 qn pn qn1 pn d
e t
r t
r to
q2n
qn dt o
dt o
To establish the proper initial conditions for higher-order derivatives of the

dependent variable, one proceeds in the same fashion. For instance, to establish the
initial conditions for the second derivative, we integrate n - 2 times between to
and to and, after some algebra, arrive at the following expressions.
When stress is the dependent variable,
d2 qn2 qn1 d qn d
e t
e t
r t
e to
dt2 o
pn
pn dt o
pn dt o
pn2 pn1 d1
r to
r t

pn
pn dt1 o
3.3
67
Similarly, when strain is the dependent variable,

d 2 pn2 pn1 d pn d
e t
r to
r t
r t
dt2 o
qn
qn dt o
qn dt o
qn2 qn1 d1
e to
e t

qn
qn dt1 o
Example 3.2 Use the above rules to decipher the significance of the coefficients of
the rheological model of Example 3.1: po rt p1 ot rt q1 ot et and to
establish the correct initial conditions, if the model is subjected to a step-stress
history r(t) = roH(t).
Solution:
On the basis of the coefficients present in P and Q, this model represents a fluid, as
qo = 0, and, because m = 1 = n, it will exhibit instantaneous elastic response
with glassy compliance Cg pq11 and its proper initial condition should be
e0 pq11 ro .
Example 3.3 Given the following sets of non-zero coefficients of the stress and
strain operators P and Q of a set of rheological models, identify which model
represents a fluid, which a solid, and which will exhibit instantaneous elastic
response.
(a) po and qo
(b) po and q1
(c) po, p1, p2 and qo, q1, q2
(d) po, p1, p2, p3 and q1, q2, q3
Solution
(a) This model represents a solid because qo = 0 and has instantaneous elastic
response because m = n.
(b) This model represents a fluid, because qo = 0 and does not have instantaneous
elastic response because m \ n.
(c) This model represents a solid because qo = 0 and has instantaneous elastic
response because m = n.
(d) This model represents a fluid because qo = 0 and also exhibits instantaneous
elastic response because m = n.
3.4
Construction of Rheological Models
As mentioned before, the usefulness of mechanical constitutive equations in differential operator form lies in that they make it easy to construct the corresponding
equations for arbitrary combinations of basic elements. The procedure to
accomplish this may be summarized in three steps as follows:
68
1. Identify which elements in the system are acting in parallel with each other, and
which ones are in series, and write the constitutive equations for the components of each group of rheological elements accordingly.
(a) For elements in parallel, write their constitutive equations in stressstrain
form, ri = (Qi/Pi)ei and note that all elements in parallel with each other
experience the same strain and that the total stress in the collection is equal
to the sum of the stresses in the individual elements. This is a consequence
of force balanceas shown in the free body diagram in Fig. 3.6.
Thus, for subcollections in parallel, write
r
ri
X Qi
Pi
ei
X Qi
Pi
3:15a
For sets of elements in series write the constitutive equations in strainstress

form, ei = (Pi/Qi) ri, and note that elements in series with each other will carry the
same stress, but the total strain in the collection is equal to the sum of the strains in
the individual elements. This is a consequence of force balance, as shown in the
free body diagram in Fig. 3.7.
Thus, for a sub-collection in series, write:
e
X Pi
Qi
ri
X Pi
Qi
3:15b
2. Identify whether the overall assembly corresponds to a system in parallel or to

one in series and convert the equations of the subcollections accordingly.
(a) To stressstrain form (ri = [Qi/Pi]e), if the overall assembly is in parallel.
(b) To strainstress form (ei = [Pi/Qi]r), if the overall assembly is in series.
3. Add the responses (stress or strain) of all the subcollections of elements to
define the constitutive equation of the assembly.
1 = 2= i
Fig. 3.6 Force balance for a set of rheological units in parallel
3.4
Construction of Rheological Models
69
Fig. 3.7 Force balance for a

set of rheological units in
series
Naturally, the constitutive equations for isolated springs and dashpots would be
written directly in either stressstrain or strainstress form, depending on which
form is adequate for the problem at hand.
3.5
Simple Rheological Models
The simplest, non-trivial combination of basic rheological elements consists of one

spring and one dashpot, either in series or in parallel with each other. The former
arrangement is known as a Maxwell model and the latter as a KelvinVoigt model.
Both models are widely used in theoretical studies because of their simplicity and
not necessarily because they provide good representations of real viscoelastic
behavior. As will be seen shortly, the Kelvin arrangement exhibits the behavior of
a solid and the Maxwell unit behaves like a fluid. These models are alternatively
referred to as the Kelvin, or KelvinVoigt solid, and the Maxwell or Maxwell
Wiechert fluid, respectively.
3.5.1
KelvinVoigt Solid
This model consists of one spring and one dashpot connected in parallel, as
depicted in Fig. 3.8. In what follows, we derive the constitutive equation of the
model and then apply it to obtain the response of the model to strain and stress
loading; in doing so, the corresponding relaxation and creep compliance functions
are developed.
70
Fig. 3.8 KelvinVoigt

model and force balance
S= D=

We apply the operator procedure following the three suggested steps:
1. We first identify two subcollections of elements (in this case, the spring and
the dashpot) and write their individual constitutive equations as follows:
For the spring, rS E eS
For the dashpot, rD g ot eD
2. The overall Kelvin model is an arrangement in parallel, so we leave the
equations of each subset in stressstrain form, using that the strain is the same
in each unit, that is: es = ed = e.
For the spring, rS E e
For the dashpot, rD got e
3. We now add the responses (stresses) of the separate (subsets of) elements and
collect terms to write
r rS rD E e got e E got e
In unencumbered form, to identify the stress and strain operators P and Q

r E got e
3:16a
As seen from this expression, m = 0 \ 1 = n; po = 1, p1 = 0; qo = E; and

q1 = g. We then conclude that the Kelvin unit has no instantaneous response
(m \ n); behaves like a solid (qo = 0); and admits the initial condition
rto eto dt to , if in strain controlthis behavior is consistent with what
would be expected of the spring and dashpot combination
the
p1 in parallel,
where

0
dashpot would lock at high strain rates, and, e to q1 r to g r to 0, if
stress were the controlled variable.
We next evaluate the response of the Kelvin model to general stress and strain
loading histories of the type shown in Fig. 3.1, featuring a step discontinuity at
t = to, and apply the solutions to the special cases of the creep and stress recovery
3.5
71
experiments, examined earlier in connection with the individual spring and

dashpot.

In this case, the controlled variable is strain, so that e(t) = f(t)H(t - to); the
rheological Eq. (3.16a) is not a differential equation in stress. To assess the response
of the model, we simply carry out the indicated operations, using the properties of
the Heaviside and delta functions, presented in Appendix A,5 to get

d
a
r Ef tHt to g Ht to f t f to dt to
dt
This expression can be used to establish the relaxation modulus of the Kelvin
model. In this case, f(t) : eo, df/dt = 0, and, by the definition of the relaxation
modulus as the stress response to a constant strain of magnitude eo, measured per
unit of applied strain: M(t) = r(t)/eo, (a) yields
Mt to EHt to gdt to
3:17a
Mt EHt gdt
3:17b
Or, more simply:
With this, the response of the model to the stress recovery test, e(t) : eoH(t to) - eoH(t - t1), shown conceptually in Fig. 3.9a, may be obtained directly, as
rt fEHt to gdt to EHt t1 gdt t1 geo

In this case, the controlled variable is stress, and the rheological Eq. (3.16a) is a
general differential equation of first order, for which an integrating factor can be
found by the method presented in Appendix A. To this end, we rearrange the
differential Eq. (3.16a) and cast it in the standard form of the linear differential
equation of first order:
de E
1
e r;
dt g
g
t[0
3:16b
Following the procedure discussed in Appendix

R A, and using the notation
1=sc ds
t
sc : g/E, obtain an integrating factor u e
et=sc and apply it to the
differential equation and, changing the dummy variable of integration on the righthand side, rewrite it as
Specifically, that f tHt f t Ht f tdt and that f tdt to f to dt to :
72
(b)
(a)
o
to
t1
to
t1
o /E
E o
t
Fig. 3.9 Response of the Kelvin model to stress and strain recovery experiments. a Stress
recovery response. b Creep recovery response
h
i
f s
H s to ds
d et=sc et es=sc
g
Integrating this expression from 0- to t, using that e(0+) = 0, rearranging, and

inserting the exponential et=sc inside the integral6 produce the solution sought, as
et Ht to
ets=sc
to
f s
ds
g
In particular, according to this expression, the response of the model to a step

stress, or creep loading, f(t)H(t - to) : r(t) = ro H(t - to), noting that H(0 to) = 0, would be
et Ht to
ets=sc
to
i
tto
ro
1h
ds
1 e sc Ht to ro
g
E
Consequently, the creep compliance of the Kelvin model, defined as the strain
response measured per unit stress, turns out to be
C t to
i
tto
1h
1 e sc ;
E
t[0
This can be done because t is not the variable of integration.
3:18a
3.5
73
Or, simply
C t
i
1h
1 et=sc ;
E
3:18b
t[0
The parameter sc : g/E, introduced here, has the dimensional units of time and
is referred to as creep or retardation timethe creep time of the Kelvin
modelbecause it is associated with a creep function.
As will be evident from the treatment that follows, when the springs and
dashpots are linear, that is, when their own constitutive equations are linear and
the higher-order derivatives of the stresses and strains entering their differential
equations are equal, the rheological models have equivalent hereditary integral
forms (see, for instance, [2, 3].
An important alternate form of the general response presented in (d) may be
obtained in terms of the creep compliance. This is done integrating expression
(d) by parts, setting to = 0, for simplicity, so as not to carry the factor H(t - to);
using u : r and dv ets=sc ds:
i
1h
rt r0 et=sc
et
E
ets=sc
d
rs ds
ds
Adding and subtracting r(0+)/E to the right-hand side of this expression and
taking the sum [r(t) - r(0+)]/E inside the integralchanging the sign of the
integral to account for the introduction of this termresult in
et
i
1h
1 et=sc r0
E
i d
1h
1 ets=s rs ds
ds
0 E
t
Using the creep compliance (3.18a), this expression may be cast in integral
form, as
et Ctr0
Ct s
d
rs ds
ds
3:19
Integrating by parts again, and collecting terms, (3.19) can be expressed in the
hereditary integral form involving the derivative of the creep compliance of the
model:
et C0rt
d
Ct s rs ds
ds
0
3:20
The response of the Kelvin model to the creep or strain recovery experiment
described by r(t) : roH(t - to) - roH(t - t1) is shown in Fig. 3.9b and may be
obtained using (3.19) or (3.20). Using (3.20), for instance, the response, e1(t), to
the first part of the load, roH(t - to), is
74
e1 t C0ro H t to
d
Ct s ro H s to ds
0 ds
The integral is evaluated using the following property of the unit step function
(c.f. Appendix A):
Z
H s to f s ds H t to
f s ds
to
and the fact that the integral of the derivative of a function d=dsCt s is the
function Ct s itself. Proceeding thus, and canceling terms, the expression for
e1(t) becomes
e1 t Ct to ro H t to
The response of the Kelvin model to a general creep recovery experiment is, by
linearity, the sum of its individual responses to the loading and unloading parts;
hence,
et C t to ro H t to C t t1 ro H t t1
Or, explicitly, in terms of the creep compliance (3.18a),

et
i
i
1h
1h
1 etto =sc ro H t to 1 ett1 =sc ro H t t1
E
E
Example 3.4 Find the steady-state response of a Kelvin unit having spring and
viscosity parameters G and g, respectively, to a sinusoidal strain history
e(t) = eosin(xt)
Solution:
Use of the loading history and the corresponding constitutive equation listed in
(3.16a) leads to rt Geo sin xt gxeo cos xt, which, not having any transient
terms in it, is the steady-state response of the Kelvin unit to the cyclic strain
history.
3.5.2
MaxwellWiechert Fluid
The Maxwell model consists of a single spring and a single dashpot connected in
series, as depicted in Fig. 3.10. We derive the constitutive equation of the model
first and then apply it to evaluate the response of the model to general strain and
stress loading.
3.5
75
Fig. 3.10 Maxwell

Wiechert model and force
balance
S
E

The constitutive equation of the Maxwell model is established following the threestep procedure. Accordingly,
1. We first identify two subcollections of elements (in this case, the single spring
and the dashpot) and write down the corresponding constitutive equations.
For the spring: rS E eS
For the dashpot: rD g ot eD
2. he overall Maxwell model is an arrangement in series; for this reason, we
rewrite the equations of each unit in its strainstress form and use that the stress
is the same in each individual unit: rs = rD = r
For the spring, eS E1 r
For the dashpot, eD go1 t r
3. We now add the responses (strains) of the separate subcollections of elements
and use operator algebra to obtain
e eS eD
1
1
E got
r
r
r
E
got
Egot
Clearing off fractionsmultiplying throughout by Egot to remove the operator

from the denominator on the right-hand sideproduces the constitutive equation
in the standard form (P r = Q e):
E got r Egot e
3:21a
As seen from this expression: m = 1 = n, po = E, p1 = g, and qo = 0 and

q1 = E g. This indicates that the Maxwell model has instantaneous response
of the form
(m = n); represents a fluid (qo = 0); and admits
initial conditions

rto Eeto under applied strain; and e to E1 r to , if the controlled
variable is stress.

In this case, the controlled variable is strain, so that e(t) = f (t)H(t - to); we use
the rheological Eq. (3.21a), which is an ordinary linear differential equation of first
order. To solve it by the procedure in Appendix A, we introduce, for convenience,
the time parameter sr : g/E and rewrite the constitutive equation, as
76
dr 1
d
r E f t
dt sr
dt
3:21b
In accordance with the solution procedure in Appendix A, we find

R dt=sr
et=sr , as an integrating factor for the equation and, changing the dummy
ue
variable on the right-hand side, write it in total differential form, as
h
t=sr
d e
rt e
s=sr
df
E H s to ds f sds to ds
ds

b
Upon integration between 0- and t, using the properties of the impulse and unit
functions:
et=sr rt r0 H t to
Ees=sr
to
df
ds
dt0
Eesr f sds to ds

Using that r(0-) = 0, and f to e to , multiplying throughout by et=sr ,
taking this factor inside the integral, and regrouping:

rt etto =sr Ee to
Eets=sr
to
de
ds
ds
Using this to evaluate the response of the model to a step-strain load,

e(t) = eoH(t), noting that, in this case, de/dt = 0, in the interval of integration,
yields
h
i
rt Eetto =sr eo ;
t [ to
From this, the relaxation modulus, defined as the stress response measured per
unit of applied strain, is found to be
M t to Eetto =sr ;
t [ to
3:22a
Or, more simply:

M t Eet=sr ;
3:22b
t[0
Setting to = 0 in (d) and combining it with (3.22b) produces the following

hereditary integral form of the stressstrain law of the Maxwell model:
rt Mte0
Mt s
0
d
esds
ds
3:23a
3.5
77
A second hereditary integral form of the constitutive equation can be derived

from this expression. Indeed, integrating it by parts, once, using u = M(t - s) and
dv = de/ds and collecting terms:
rt M 0et
d
Mt sesds
dt
3:23b
In this case, the time parameter, sr, is called a relaxation time because it is
associated with a relaxation function. The relaxation time of a Maxwell unit
represents the time it takes the stress to decay by about 63 %. This is so, because
the relaxation modulus, evaluated at t = sr (or t - to = sr), is e-1 E & 0.37E.
The response of the Maxwell model to the general stress recovery experiment
described by e(t) : eoH(t - to) - eoH(t - t1) is shown in Fig. 3.11a and may be
obtained using one of the expressions in (d), (e), or (3.23a, b). Using (d) and the
properties of the unit step and impulse functions, and collecting terms:
h
i
rt E etto =sr H t to ett1 =sr H t t1 eo

In this case, the controlled variable is stress, so that r(t) = f(t)H(t - to), and the
differential equation is recast as
d
1 dr 1
et
r
dt
E dt g
Before substituting the applied stress, we integrate this expression between

0- and t, using that e(0-) = 0 and r(0+) = 0, and collect terms to write
(a)
(b)
o
to
t1
to
t1
o /E
o /E
E
t
Fig. 3.11 Response of the Maxwell model to stress and strain recovery experiments. a Stress
recovery response. b Creep recovery response
78
1
1
e t r t
E
g
rsds
This expression may be cast in hereditary integral form through integration by

parts and some algebraic manipulations. Integrating it by parts once, choosing
u = r, and dv = ds; performing the required operations, bearing in mind that the
loading, r(t), has a discontinuity at to, and collecting terms, leads to

Z t
1 t
s dr
r t
ds
et
E g
g
ds

to

Adding and subtracting the term 1=Er to and taking the term in brackets
inside the integral lead to
et
Z t
to

1 t s dr
ds
E
g ds
Splitting the integral from to to to and from to to t and using that
r(t) = f (t)H(t - to)

et

Z t
1 t to
1 t s df
f to
ds;
g
E
g ds
to E
t [ to
The response of the model to a constant stress load f (t) : ro, that is, r
(t) = roH(t - to) can be easily established from this expression, noting that the
derivative of the loading history vanishes inside the interval of integration, as

et

1 t to
Ht to ro
g
E
The result, divided by the applied stress is, by definition, the creep compliance
of the Maxwell model:

C t to

1 t to
Ht to
g
E
3:24a
Or, more simply:

1 t
C t
Ht
E g
3:24b
The hereditary integral form of the strainstress constitutive equation of the

Maxwell model, which we set out to find, may be obtained taking (3.24a) into (k).
For clarity, however, we set to = 0, to write
3.5
79
et C tr0
Ct s
dr
ds
ds
3:25
A second form of hereditary integral constitutive equation for the Maxwell

model, which features the derivative of the creep compliance under the integral
sign, is readily obtained from this last relation through integration by parts.
Selecting C(t - s) = u, and (de/ds)ds = dv, and proceeding thus, produces that
et C0rt
d
Ct srsds
ds
3:26
The response of the Maxwell model to the general creep recovery experiment
described by rt ro H t to ro H t t1 is shown in Fig. 3.11b and may be
obtained using (3.25), and the creep compliance, (3.24b), noting that dr/ds : 0 in
the integration interval; thus,

et
3.6

1 t to
1 t t 1
H t t o
H t t1 ro
E
g
E
g
Generalized Models
The concepts introduced earlier are used here to develop the constitutive equations
for two models involving multiple Kelvin units in series and multiple Maxwell
units in parallel. For obvious reasons, the first type of arrangement is called a
generalized Kelvin model, while the second type of arrangement is known as a
generalized Maxwell model.
3.6.1
Generalized Kelvin Model
A generalized Kelvin model or KelvinVoigt model is a collection of Kelvin units

in series, plus an isolated springor an isolated dashpot, as shown in Fig. 3.12.
In accordance with the operator method, the constitutive equation of this model
may be established as the sum of the strainstress, series form of the equations
of the individual Kelvin elements. In detail:
1. The subcollections of the overall model are the individual Kelvin elements. The
ith subset in the arrangement has a constitutive equation given by (3.16a), as
ri Ei gi ot ei
a:1
80
Fig. 3.12 Generalized

KelvinVoigt model
E0
E1
E2
En
for a spring and dashpot in parallel; and

ro Eo eo
a:2
for the isolated spring

2. which in strainstress form become
ei
1
ri
Ei g i o t
b:1
for the Kelvin unit, and

eo 1=Eo ro
b:2
for the isolated spring.

3. The constitutive equation for the whole assembly is established by adding the
contributions of all the KelvinVoigt subsets, recalling that ri = r:

1 Xn
1
r
e
i1 i
i1 E g o
Eo
i
i t
Xn
3:27a
Applying the minimum common multiple of the denominators of the operators;

setting go = 0, to include the isolated spring as a degenerate Kelvin unit, switching
the order of the members of the equation:
Yn
Xn Yn
g
o
r
Ei gi ot e
E
i
t
j
0
i
io
i0
j 6 i
3:27b
The linear operator form, P r = Q e, is obtained by performing the indicated operations and collecting like terms. The result is

d1
dm
d1
dn
po p1 1 pm m r qo q1 1 qn n e
dt
dt
dt
dt
3:28
3.6
Generalized ModelsModels
81
Examination of the general constitutive equation (3.27a) or (3.28)7 reveals the

following:
(a) If none of the spring and dashpot coefficients are zero, the order, m, of the
higher-order derivative of the stress would be one less than that of the strain
(m = n - 1 \ n). In the context of the analysis of the order of the derivatives
of P and Q, this means that a generalized Kelvin model with no degenerate
units would have no initial elastic response (m \ n).
(b) If the arrangement hasat leastone isolated spring (that is, one Kelvin unit
with zero viscosity), the operator Q will contain a non-zero constant operator,
qo = 0; the rheological model will represent a solid.
(c) If the arrangement has one Kelvin unit with no spring, the operator Q will contain
a zero constant operator, qo = 0; the rheological model will represent a fluid.
Because the generalized Kelvin model is an arrangement in series, its creep
compliance may be obtained directly as the sum of the creep compliance functions of
each individual Kelvin unit. Thus, using (3.18b), (3.27a) and that Ci (t) = ei(t)/ro:
C t
C i t
i
1 Xn 1 h
t=sci
1

e
;
i1 E
Eo
i
s ci
gi
Ei
3:29
In this expression, 1/Eo is the instantaneous or glassy compliance and is

denoted by Cg. Also, in molecular dynamic studies, the finite set of pairs (si, 1/
Ei), including the pair (0, Cg), is called a discrete retardation spectrum; the
notation Li : 1/Ei is typically used. With this in mind, in the limit of infinitely
many Kelvin units, the summation becomes an integral and the result is a
continuous creep or retardation spectrum [4].
C t Cg
h
i
Ls 1 et=s ds
3:30
3.6.2
Generalized Maxwell Model
A collection of Maxwell units in parallel, as depicted in Fig. 3.13, is referred to as

a generalized Maxwell model, or generalized MaxwellWiechert model. Per the
operator approach, the corresponding mechanical constitutive equation may be
constructed as the sum of the constitutive equations in stressstrain form of all its
Maxwell elements.
In detail:
1. The overall assembly may be considered as a collection of Maxwell elements, in parallel, rather than start at the individual spring and dashpot sets in
7
Because of the explicit summation and multiplication symbols present in (3.1b), it is easier to
use that expression to discern the order of the constitutive equation and the nature of its
coefficients.
82
Fig. 3.13 Generalized

MaxwellWiechert model
E1
1
E2
En
E
n
series. We use (3.21a) as the basic constitutive equation and write it for a
generic Maxwell element:
Ei gi ot ei Ei gi ot ri
c:1
ro E1 eo
c:2
with

2. Since the overall arrangement is in parallel, we express it in stressstrain form:
ri
Ei g i o t
ei
Ei gi ot
d:1
and:
ro E1 eo
d:2

3. Adding the contributions of all the elements in the model and using that ei = e:
r

Xn
Ei g i o t
e
ri E1
i1 E g o
i
i t
3:31a
Multiplying throughout by the minimum common multiple of the denominators

of the operators on the right side, and rearranging:
3.6
Generalized ModelsModels
83
hYn
hYn
i
i
E
g
o
E
g
o
r
E
e
j
t
1
j
t
j
j
j1
j1
3
2
Y
X

n
7
6 n
4 i1 Ei gi ot
j 1 Ej gj ot 5e
j 6 i
3:31b
After performing the indicated operations, and simplifying, the explicit operator
equation, P r = Q e, is seen to be of the same form as for the generalized
Kelvin model. In this case, however, examination of the general constitutive
equation (3.31a) reveals that
(a) The order of the highest order derivative of the stress would be the same as
that of the strain; irrespective of whether the system has an isolated spring or
not. This means that a generalized Maxwell model will always exhibits
instantaneous response.
(b) If the arrangement has at least one isolated spring, the operator Q will contain
a non-zero constant, qo = 0; and the rheological model will represent a solid.
(c) If the arrangement has one Maxwell unit without a spring, the operator Q will
contain a zero constant, qo = 0; and the rheological model will represent a
fluid.
Because the generalized Maxwell model is an arrangement in parallel, its
relaxation function is the sum of the relaxation functions of each Maxwell unit.
Thus, using (3.22b), (3.31a) and that M(t) = r(t)/eo:
M t E1
M i t E1
Ei et=sri ;
sr i
gi
Ei
3:32
This finite sum is called a discrete relaxation spectrum. As the number of

Maxwell units increases without limit, the summation becomes an integral, giving
rise to the continuous relaxation spectrum:
M t E1
Eset=s ds
3:33
3.7
Composite Models
Composite models are those formed by judicious combinations of springs and

dashpots, which not only are more complex than the Maxwell and KelvinVoigt
units, but are capable of reproducing more realistic viscoelastic behavior. Clearly,
the choices are very many, and what follows restricts attention to models with
three parameters, representing, respectively, solid and fluid behavior.
84
Fig. 3.14 Standard linear

solid. a From generalized
Kelvin model. b From
generalized Maxwell model
(a)
(b)
Eo
EM
EK
3.7.1
Standard Linear Solid
The two equivalent versions of this model are shown in Fig. 3.14. One of them
consists of a Kelvin unit in series with a spring. It is a degenerate form of a
generalized KelvinVoigt model. The other one is composed of a Maxwell element in parallel with a spring, which is a degenerate form of the generalized
Maxwell model. As will be shown, the three-parameter model in Fig. 3.14 exhibits
solid behavior and is thus also referred to as the standard linear (viscoelastic) solid.
The constitutive equation of the standard linear viscoelastic solid is derived
considering it as a degenerate case of the generalized Maxwell model, depicted in
part (a) of Fig. 3.14. The mathematical equivalence between the two models is
demonstrated later, as part of the examples.
For a Kelvin unit in series with a spring, rK EK gK ot eK and rS Eo eS ,
and the strains are additive: e = eK ? eS; therefore, by the operator method, the
differential equation of the standard linear solid shown in Fig. 3.14a is given by

1
1
r
e
E K gK o t Eo
3:34
As may be readily verified, this equation could have been written directly as a
particular case of the generalized Kelvin model given in (3.27a). After clearing off
fractions, using operator algebra and rearranging:
Eo EK r gK
dr
de
Eo EK e Eo gK
dt
dt
3:35a
This equation may be cast in general form:

po r p 1
dr
de
qo e q1
dt
dt
3:35b
If the following notation is used,

p o Eo E K ; p 1 g K ; q o Eo E K ; q 1 Eo g K
3:35c
3.7
Composite Models
85
These relationships show that not all model parameters, pi and qi, in the generic
equation of the model are independent.
Examination of (3.35a) indicates that, because m = 1 = n, the model exhibits
glassy response, with modulus Mo = q1/p1 : Mg, and will admit initial conditions
of the form: por(to) = qoe(to). In addition, as asserted before, the model is a solid,
because qo = 0, and has long-term modulus M? = qo/po = Eo EK/(Eo ? EK).
Also, the derivative terms in its constitutive equation indicate the model captures
creep response. These characteristics of the standard linear solid will be demonstrated by evaluating its response to the stress recovery and creep recovery tests. In
so doing, it should be noted that the constitutive equation is symmetric or balanced
in the stress and the strain. This means that the form of its solution to a strain
history is of the same form as its response to a stress history of the same type and
vice versa.

In this case, the loading is of the form e(t) = f(t)H(t - to); the response of the
model is obtained as the general solution to the differential Eq. (3.35b), which, for
clarity, is rewritten as
dr 1
qo
q1 de
r e
;
p1
p1 dt
dt s
s
p1
po
Or, after applying the integrating factor et/s and casting the equation in differential form:
t q t0
q1
q1 0
0 df
o
d es r e s f t0 H t0 to dt0 H t0 to ets
ds ets f tdt0 to dt0
p1
p1
p1
dt0
Integrating between 0- and t; using that r(0-) = 0; and recalling the properties
of the unit and impulse functions:
Z t

Z
qo
q1 t t0 =s df 0
q1
t0 =s
0
0
e f t dt
e
dt eto =s f to
e rt H t to
p1 t o
p 1 to
p1
dt0
t=s
Multiplying throughout by e-t/s; taking the exponential inside the integrals; and
collecting terms:
rt H t to
e
to
tt0 =s

qo 0
q1 df
q1
dt0 etto =s f to
f t
p1
p1 dt0
p1
This expression may be used to obtain the response of the model to the stress
recovery experiment e(t) = eoH(t - to) - eoH(t - t1). Because of linearity, the
load is split into two parts e1(t) = f1(t)H(t - to) : eoH(t - to) and
86
e2(t) = f2(t) H(t - t1) : -eoH(t - t1); the response to each part, calculated
separately; and the results added together to obtain the total response. Noting that
f1(t) : eo, and df1/dt = 0; using (h) to evaluate the response to the first part of the
load leads to the following responseafter performing the indicated integrations,
recalling that s : p1/po, and rearranging
r 1 t

qo
q1 qo tto =s
e
Ht to eo

po
p1 po
From this, the relaxation modulus of the standard linear solid follows, as
M t to

qo
q1 qo tto =s
e

;
po
p1 po
s

qo
q1 qo t=s
e ;

po
p1 po
p1
po
p1
po
3:36a
Or:
M t
s
3:36b
The response, r2(t), to the second part of the loading would be identical to r1(t),
but of opposite sign. Therefore,

i q
qo h
1 tto =s
tto =s
H t to ro eo
1e
rt r1 t r2 t
e
po
p1
h

i q
qo
1 tt1 =s
tt1 =s
H t t1 eo

1e
e
po
p1
Or, in terms of its physical parameters,

n
h
i
o
rt M1 1 etto =s Mg etto =s H t to eo
n
h
i
o
tt1
tt1
M1 1 e s Mg e s H t t1 eo ; s
gK
E o EK
This solution is shown schematically in Fig. 3.15. As the figure indicates, the
response of the standard linear solid features all aspects of the stress response of a
Fig. 3.15 Stress recovery

response of three-parameter,
standard linear solid
Mgo
o
to
t1
to
t1
Mg o
3.7
Composite Models
87
linear viscoelastic solid; it has instantaneous elastic response upon loading and
unloading; relaxation under constant strain; and delayed recovery after unloading.

In this case, the loading is of the form r(t) = f(t)H(t - to), and strain becomes the
controlled variable. Because of the symmetry of the constitutive equation (e) in
stress and strain, all the arguments used for the case of strain loading may be
repeated interchanging the roles of stress and strain, as well as interchanging the
parameters po and p1 with qo and q1. This leads to the following relations:
de 1
po
p1 dr
e r
;
q1
q1 dt
dt k
C t to
k
q1
qo

po
p1 po tto =k
e

;
qo
q1 qo

po
p1 po t=k
e

;
C t
qo
q1 qo
k
k
l
q1
qo
3:37a
q1
qo
3:37b
And, lastly, to
h

i p
po
1 tto =k
tto =k
et
1e
e
H t to ro
qo
q1
h

i p
po
1 tt1 =k
tt1 =k
H t t1 ro

1e
e
qo
q1
And, in terms of the models physical parameters,
et

i
1 h
1 tto =k
H t to ro
1 etto =k
e
M1
Mg

i
1 h
1 tt1 =k
H t t1 ro ;

1 ett1 =k
e
M1
Mg
k
gK
EK
As depicted in Fig. 3.16, the creep response of the standard linear solid exhibits
all behavioral characteristics of the creep response of viscoelastic solids.
Example 3.5 Show that the model of Fig. 3.14b, also called the Zener model, is
mathematically equivalent to that in Fig. 3.14a.
Solution:
Since this model is a particular case of the generalized Maxwell
model, its
h
i
gM ot
e
constitutive equation is given by (3.31a), with n = 1; thus, r E1 EEMMg
M ot
which upon rearrangement may be recast as
88
Fig. 3.16 Creep response of

standard linear solid.
a Elastic, impact response.
b Creep. c Elastic recovery.
d Delayed recovery
c
d
a
t1
EM r gM
t2
t1
dr
de
EM E1 e gM EM E1
dt
dt
t2
3:35d
This expression is mathematically equivalent to (3.35a), which proves the point.

As may also be seen, equating the corresponding coefficients of the two equations
reveals that the parameters of the two versions of the standard linear solid are
related as follows:
gK g M ;
3.7.2
Eo EM E 1 ;
1
1
1
:
E K EM E 1
Three-Parameter Fluid
There are two equivalent versions of this model; they are shown in Fig. 3.17. As
seen in the figure, one version of the model consists of a Kelvin unit in series with
a dashpot. It is a degenerate form of a generalized KelvinVoigt model. The other
one is composed of a Maxwell element in parallel with a dashpot, which is a
special case of the generalized Maxwell model. As will be shown shortly, the
three-parameter model exhibits fluid behavior and is thus referred to as the threeparameter fluid.
As in the case of the three-parameter solid, the constitutive equation of each
version of the three-parameter fluid shown Fig. 3.17 may be derived either from
the equation of the generalized Kelvin model or from the equation of the Maxwell
model. For instance, the equation of the model in part (a) of the figure results from
equation (3.27a), using n = 2, E1 = 0, g1 = go, and removing the term 1/Eo,
which
to account for an isolated spring. This leads to
h had been introduced
i
e
1
g0 ot
1
EK g
K ot
r. After clearing off fractions, using operator algebra, and
rearranging
EK r g0 gK
dr
de
d2 e
E K go go gK 2
dt
dt
dt
3:38
Accordingly, m = 1 \ 2 = n, so the model does not exhibit impact response.

Also, qo = 0, which indicates the model represents a fluid, as claimed earlier.
3.7
Composite Models
89
Fig. 3.17 Three-parameter

fluid. a From generalized
Kelvin model. b From
generalized Maxwell model
(a)
(b)
EM
EK
3.8
Problems
P.3.1 Determine the initial conditions for the model of Example 2, if the controlled variable is stress and the loading is defined by r(t) = roH(t).
Answer: e(0+) = (p1/q1)ro
Hint: Use expression (d) of Sect. 3.3.4, with to 0 and r(0+) = ro, to get the
desired result. Note that since m = 1 = n, this model has instantaneous elasticity with compliance p1/q1, from which the result follows.
P.3.2 Obtain the response of a Kelvin unit to the constant strain rate history e(t) =
Rt.
Answer: rt ERt gR
Hint: The easiest way to solve this problem is by direct application of (3.16a) to
the given strain history. Doing this produces the result.
P.3.3 Find the steady-state response of a Maxwell unit with spring and viscosity
parameters G and g, respectively, to a sinusoidal strain history
e(t) = eosin(xt)h
i
Answer: rt G
xs2
1xs2
sinxs 1xs
cosxs eo
xs2
Hint: Take the strain history f(t) = eosin(xt) and E = G, s = g/G into Eq. (3.21b)
and use integration by parts, twice, to resolve the integral. The full solution
shouldh be
i
r t G
xs2
1xs2
sinxs 1xs
cosxs eo 1xs
et=s eo . From which the
xs2
xs2
steady-state solution follows dropping the exponential, transient term. This

problem is solved more efficiently by the methods of Chap. 4.
P3.4 Show that the model in Fig. 3.14b is mathematically equivalent to that in
Fig. 3.14a.
Hint: Use either the operator method or Eq. (3.31a) for a generalized
Maxwell
h
i
gM ot
model, with n = 1, E1 = EM, and g1 = gM, to write r E1 EEMMg
ot e and
M
resolve the expression to arrive at EM gM ot r E1 EM EM gM ot e;

which, being of the same form as (3.35a), proves the point.
90
E2
E1
P.3.5 Derive the constitutive equation for the four-parameter model shown in
Fig. 3.18, treating it as a degenerate generalized Kelvin model. Use physical
arguments regarding how the springs and dashpots of the arrangement
would behave under a constant stress to show that the system would respond
like a fluid and that it provides for initial elastic response. This arrangement
is known as the four-parameter or Burgers fluid.
2
Answer: Pr = Qe, with: P E1 E2 E1 g2 E2 g3 E1 g3 dtd g2 g3 dtd 2 , and:
2
Q E1 E2 g3 dtd E1 g2 g3 dtd 2 . Since m = 2 = n, this model exhibits instantaneous elastic response with impact, or glassy modulus Mg = q2/p2 = E1. Also,
since qo = 0, the model represents a fluid. On physical grounds alone, under
instantaneous strain, the dashpots will lock, but the isolated spring, will
respond elastically. In the limit of extremely long time under load, the isolated
dashpot will extend like a fluid would.
Hint: use constitutive equation (3.27a) for the generalized Kelvin model with
E3 = 0 and g1 = 0, to arrive at the result.
P.3.6 Develop the differential equation for the rheological model depicted in
Fig. 3.19 adjacent sketch, show that it behaves as a solid, and determine its
instantaneous
and long-term
moduli.

Answer: E1 E2 g1 dtd r E2 E3 E3 E1 E1 E2 E2 E3 g1 dtd e.
In this case, m = 1 = n; the model exhibits instantaneous elastic response with
glassy modulus Mg = q1/p1 = E2 + E3. Also, since qo = 0, the model repre3 E1 E1 E2
sents a solid with long-term modulus, M1 qpoo E2 E3 E
E1 E2
Hint: Note that this model is composed of a standard linear solid in parallel with an
isolated springh and add the
i corresponding constitutive equations in stressstrain
form, rstd
E1 E2 E2 g1 ot
E1 E2 g1 ot
e, and rspring = E3e, respectively, and arrive at the
desired result.
P.3.7 The simple exponential function M t Eeat was fitted to the results of a
shear relaxation test of a polymer. Use the Laplace transform to obtain the
corresponding creep compliance and explain why the relaxation function
given models fluid response.
E1
E2
1
E3
3.8
Problems
91
Answer: Ct E1 Ea t
1
1

Hint: Use the relationship Cs
s2 Ms
and the Laplace transform Ms
E sa
of

1 1
a 1
the relaxation modulus to arrive at the expression Cs

2 . The
E s
E s
inverse Laplace transform of this yields the desired result. The relaxation
function given will model fluid response because it has zero long-term modulus,
which is a characteristic of fluid behavior. In addition, the corresponding creep
compliance indicates that the model used for this polymer will exhibit unlimited
flow under a sustained constant stress (see the behavior of a Maxwell model).
P.3.8 The three parameters of a standard linear solid model were fitted to the
results of a shear test of a rubber. Using the Laplace transform, determine
the corresponding hshear relaxation
modulus.
i
Answer: M t qpoo
q1
p1
qpoo et=s ;
s pp1o
Q and the Laplace transform of the differential

Hint: Use the relationship M
sP
qo q1 s . Factor p1
operators P = po ? p1qt and Q = qo ? q1qt to write M
spo p1 s
out of the denominator; introduce the relaxation time s = p1/po; and use partial
qo =po q1 =p1 q0 =po , from
fraction expansion (see Appendix A), to arrive at M
s
s1=s
which the desired result follows by inversion to physical space.
P.3.9 A viscoelastic material of Kelvin type obeys the law rt Eet g dtd et.
Find its steady-state response to the cyclic strain history e = eosin(xt).
Answer: rt Eeo sinxt gxeo cosxt.
Hint: Insert the given strain history into the analytical model and perform the
needed operations to reach the result.
P.3.10 Repeat Problem P.3.9 for a viscoelastic material of Maxwell type whose
stressstrain law is described by dtd rt 1s rt E dtd et.
n
o
xt2
xt
Answer: rt E eo 1xt
2 sinxt
2 cosxt
1xt
Hint: Insert the cyclic strain history into the differential equation of the model, using
the integrating factor u n et=s and perform integration byoparts, twice, to arrive a t
the result rt E eo
xt2
1xt2
Eeo
t=s
sinxt 1xt
cosxt 1xt
. After a
2 e
xt2
sufficiently long time, the last term becomes negligible, and the steady-state
solution listed is reached.
References
1. D.L. Kreider, R.G. Kuller, D.R. Ostberg, Elementary Differential Equations (Addison-Wesley,
Reading, 1968), pp. 6669
2. R.M. Christensen, Theory of Viscoelasticity, 2nd edn. (Dover, NY, 2003), pp. 1420
3. A.D. Drosdov, Finite Elasticity and Viscoelasticity (World Scientific, Singapore, 1996),
pp. 250255
4. J.D. Ferry, Viscoelastic Properties of Polymers, 3rd edn. (Wiley, NY, 1980), pp. 6067
Constitutive Equations
for Steady-State Oscillations
Abstract
Although the viscoelastic constitutive equations in either integral or differential

form apply in general, irrespective of the type of loading, or the point in time at
which the response is sought, it is possible to derive from them constitutive
equations of a form especially well suited to steady-state oscillations. This
chapter uses complex algebra to transform the integral and differential
constitutive equations of viscoelasticity, defined in the time domain, into
algebraic expressions in the complex plane. The chapter also examines the
relationships between the material property functions defined in the time
domain, and their complex-variable counterparts, and examines the problem of
energy dissipation during steady-state oscillations, important in the design of
mounts for vibratory equipment, among others.
Keywords

Steady state Complex Real Imaginary Harmonic Excitation Frequency

Storage Loss Constitutive Euler Fourier Argand Dirichlet Laplace
Newton
Transform
Modulus
Compliance
Amplitude
Integral
Differential Energy

4.1

Introduction
It is often necessary to evaluate the steady-state response of a viscoelastic material

to harmonic loading. Although the constitutive equations in either integral or
differential form (c.f. Chaps. 2, 3) apply irrespective of the type of loading and the
point in time at which the response is sought, it is possible to derive from them
constitutive equations of a form especially well suited to steady-state oscillatory

93
94
Constitutive Equations for Steady-State Oscillations
conditions. Such equations, expressed in terms of complex quantities, enter the

problem of harmonic oscillations in a natural way.
This chapter uses Eulers formula: ejh = cos(h) ? jsin(h), and complex algebra1 to represent the steady-state harmonic excitations imposed on a viscoelastic
material, and their steady-state response to such loads. As will be seen, this
transforms the integral and differential constitutive equations, defined in the time
domain, into simple algebraic expressions in the complex plane.2 As it turns out,
the resulting equations in the complex plane are of exactly the same forms as the
corresponding Laplace-transformed equations and hence are of the same forms as
their elastic counterparts. This similarity is due to the fact that the constitutive
equations in complex-variable form may be obtained through an integral transformation, called the Fourier transform, of the general constitutive equations,
which is equivalent to the Laplace transform for functions typically occurring in
viscoelasticity (see Appendix A).
The chapter also examines the relationships between the time-domain material
property functions and their complex-variable counterparts. This treatment is
followed by an evaluation of the energy dissipated per cycle. There, a simple
analysis shows that the energy dissipated by a viscoelastic material in a cycle is
given by an expression that resembles that for energy storage in linear elastic
materials; in that it is proportional to the product of some adequate measure of
modulus and squared strain; or compliance and squared stress.
Finally, although there are several excellent books on viscoelasticity dealing
with the present subject, we list only three as references: Ferrys seminal book on
properties of polymers [1], Flgges classical text on viscoelasticity [2], and the
more recent treatment by Wineman and Rajagopal [3].
4.2
Steady-State Constitutive Equations from Integral

The aim here is to develop constitutive equations which apply specifically to

steady-state oscillatory loads, where the controlled variable, say strain, is of the
form e(t) = eosin (xt), or e(t) = eocos(xt); and where both, eo and x, representing,
respectively, the amplitude of the excitation and its frequency, are real numbers.
Noting that Eulers exponential form of a complex number can be used to represent a sine or a cosine function of its argument, we introduce the complex strain
e* (jxt) (c.f. Appendix A):
e jxt eo ejxt eo cosxt j sinxt
4:1
p
In the main body of the text, the symbol j is used to denote the imaginary unit: j 1.
2
Because the frequency of the excitations becomes the independent variable in this transformation, the resulting equations are said to be defined in the frequency domain.
1
4.2
Steady-State Constitutive Equations
95
where the real part of the complex excitation, Re[e*(jxt)], would represent excitations of cosine type; and the imaginary part, Im[e*(jxt)], would stand for excitations of sine type.
The compex stress, r*, is obtained by inserting (4.1) into (2.32a), setting the
lower limit to -?, in accordance with the closed-cycle condition for steady-state
response to oscillatory loads. Carrying out the differentiation under the integral
sign, introducing the change of variable u = t s, and regrouping, using that
e jxt eo ejxt , leads to:
r jxt eo jx
Zt
Mt sejxs ds 4jx
1
Z1
3
Muejxu du5e jxt
Or:
2
r jxt 4jx
Z1
3
Muejxu du5e jxt M jxe jxt
4:2
This expression shows that, just as in an ideally linear elastic material, where the
stress response is directly proportional to the applied strainand vice versathe
steady-state complex stress in a viscoelastic substance subjected to harmonic
excitation is directly proportional to the complex strain. As discussed in Appendix
A, the factor of proportionalityin fact, a function of frequencyis termed the
complex modulus, M*, and is defined by (4.2), as:
M jx jx
Z1
Mtejxt dt
4:3
The integral in (4.3) is the Fourier transform of the relaxation modulus, M(t). In
general, the Fourier transform of an arbitrary function f(t) will exist if the function
is piecewise continuous, has a finite number of finite discontinuities,and is absolutely integrablei.e., the integral of its absolute value is bounded.3 As discussed
in Appendix A, the Fourier transform and the inverse Fourier transform, referred to
as the Fourier transform pair, are, respectively, defined as:
F ff tg f jx
1
Z
ejxt f tdt
4:4a
f t
11
Z
ejxtf jxdx
p0
These are the so-called Dirichlet conditions.
4:4b
96
In this manner, the complex modulus, M*, is the jx-multiplied Fourier trans jx. The relaxation
form of the relaxation function: M jxF fM tg jxM
function M(t) may be recovered from (4.3) dividing through by jx and applying
the inverse Fourier transform [4]:
1
Mt
p
Z1
M jx jxt
e dx
jx
4:5
Example 4.1 Obtain the analytical form of the complex modulus of an isotropic
viscoelastic solid with shear relaxation modulus given by the following finite sum
of exponentials:
Gt Ge
n
X
Gk et=sk
k1
Solution:
The complex modulus is obtained by putting this expression into (4.3) and
carrying out the indicated integrations. Thus:
G jx jx
Z1 "
Ge
jx4Ge
Z1
0
n
X
k1
Gk
#
Gk e
t=sk
ejxt dt
k1
20
G jx Ge
n
X
ejxt dt
n
X
Z1
Gk
k1
1jxs
s k t
k
dt5
n
n
X
X
jxsk
xsk 2
xsk
Ge
Gk
j
Gk
2
1 jxsk
1 xsk
1 xsk 2
k1
k1
It is common practice to express the complex modulus in Cartesian form; that is

in terms of its real and imaginary components, which are denoted, respectively, as
M 0 and M 00 :
M jx M 0 x jM 00 x
4:6
As with any other complex number, the graphical representation of M* in the

Argand plane, which is shown in Fig. 4.1, leads to the polar form of M* in terms of
its amplitude, kM k, and phase angle, dM (see Appendix A):
M jx kM kejdM
4:7a
4.2
97
Fig. 4.1 Graphical

representation of the complex
modulus in the Argand Plane
q
kM k M 0 2 M 00 2
4:7b
tan dM M 00 =M 0
4:7c
M 0 kM k cos dM ;
M 00 kM k sin dM
4:7d
For reasons that will become clear later, M 0 and M 00 are also frequently referred to
as the storage and loss modulus, respectively; and tan(dM), as the loss tangent.
The following important alternate form of (4.2), results from combining it with
(4.1) and (4.7 a):
r jxt kM keo ejxtdM ro ejxtdM ; ro kM keo
4:8
This expression defines, quite naturally, the amplitude of the modulus, M , as the
ratio ro/eo, of the stress amplitude to the strain amplitudeas for linear elastic
materials. In addition, this expression and (4.1) state that to a complex strain
e jxt eo ejxt there corresponds a complex stress r jxt ro ejxtdM ; which
is of the same frequency as the controlled strain, but out of phase with it by dM
radians.
The algebraic relationships between the complex modulus M* and its real and
imaginary components, M 0 and M 00 , result from taking (4.7a) into (4.3); using
Eulers formula to split the integral into its real and imaginary parts; separating the
equilibrium modulus, so that Mt t Mt Me and grouping:
0
M jx Me x
Z1
Mt t sinxtdt
4:9a
00
M jx x
Z1
Mt t cosxtdt
4:9b
These are the Fourier sine and cosine transforms, respectively, of the relaxation
modulus, M. The corresponding inverse transforms are given by:
98
Mt t
2
p
Z1
M 0 jx Me
sinxtdx
x
4:10a
2
Mt t
p
Z1
M 00 jx
sinxtdx
x
4:10b
These expressions indicate that the Cartesian components, M 0 and M 00 , of the

complex modulusM* are interdependent, since the relaxation modulus M, may be
obtained from either component.
Example 4.2 Determine the storage and loss moduli, and the loss tangent of a
material with relaxation modulus E(t) = Ee-t/s.
Solution:
The complex modulus of the material may be obtained directly from (4.3).
After carrying out the integration, multiplying and dividing the result by (1 - jx),
2
jxs
jxs
and collecting real and imaginary parts, the result is M E 1jxs
E xs1
,
2
2
xs
xs
The
with the following Cartesian components: M E 1xs
2 ; M E
1xs2
0
00
00
1
loss tangent is obtained from these components and (4.7a), as: tan d M
M0 xs
The strainstress constitutive equations applicable to steady-state conditions
can be established by reversing the roles of stress and strain in the previous
derivations. In this case, similar to (4.1), the complex form of the given excitation,
r(t) = ro cos(xt), or r(t) = ro sin(xt), would be:
r jxt ro ejxt ro cosxt j sinxt
4:11
Inserting this into (2.33a) and proceeding as before, leads to the strainstress
equation:
e jxt C jxr jxt
4:12
With the complex compliance C*:

C jx jx
Z1
Ctejxt dt
4:13
In addition, combining (4.2) and (4.12) produce that:

M jxC jx 1
4:14
4.2
99
This shows that the complex modulus and complex compliance are inverses of
each other: M* = 1/C*.
Moreover, in view of (4.7a) and (4.13):
C jx
1
1
ejdM kC jxkejdM kC jxkejdC

M jx kM jxk
4:15
From this expression, it then follows that:

kC jxk
4:16
kM jxk
dC dM
4:17
That is, the amplitudes of the complex modulus and complex compliance are
reciprocals of each otheras for linear elastic materials; and the phase angle, dC,
of the complex compliance is the negative of the phase angle, dM, of the complex
modulus. Because of this last result, a negative sign is typically used in front of the
imaginary part, C 00 , of the Cartesian form of the complex compliance, and the
same phase angle d : dM used for the modulus and compliance. Then, corresponding to (4.6), the complex compliance is usually written as:
C jx C 0 jx jC 00 jx
4:18
The analogous of (4.7a) results from this expression, the definition of magnitude of
a complex number, and the relationship between its components; thus:
jjC jjjx jjC jjejd
jjC jj
4:19a
q
C 0 2 c00 2
4:19b
tan d C 00 =C 0
0
4:19c
00
C jjC jj cos d; C jjC jj sin d
4:19d
Likewise, combining (4.12) and (4.15) with the complex stress r*(t) = ro e
given in (4.11), yields the following result, in analogy with (4.8):
e jxt kC kro ejxtd eo ejxtd ;
eo kC kro
jxt
4:20
Just as in the case of the complex modulus, this expression defines the amplitude
of the complex compliance,kC k, as the ratio eo/ro, of the amplitudes of the strain
and stress, as for linear elastic materials.
100
Other relationships involving the creep compliance, such as their Fourier sine
and cosine transforms are entirely similar to those for the relaxation modulus.
Specifically, using Ct t Ct Ce :
0
C jx Ce x
Z1
Ct t sinxtdt
4:21a
00
C jx x
Z1
Ct t cosxtdt
4:21b
Ct t
2
p
Z1
C0 jx Ce
sinxtdx
x
4:22a
2
Ct t
p
Z1
C 00 jx
sinxtdx
x
4:22b
In practice, all functions involved in the response of viscoelastic materials

such as excitations, stress or strain response, and material propertiesare of
Dirichlet type; and hence, their Fourier transforms exist. It is then permissible to
apply the Fourier transform to viscoelastic relationships, such as constitutive
equation. In particular then, the Fourier transform of Eq. (2.33a) and (2.33b), for
instance, would lead to (c.f. Appendix A):
(
d
F frtg F
Mt s esds
ds
1
(
Zt
d
F fetg F
Ct s rsds
ds
1
Zt
)
jxejx
jx jxM
)r
4:23
jx
) ejx jxC
rjx
4:24
Combining these expressions one may arrive at the following additional result:
jx
jxC
M
1
jx2
4:25
Expressions (4.23) to (4.25) are of exactly the same form as (2.292.31),

derived in Chap. 2, taking the Laplace transform of constitutive equations (2.32).
As explained in Appendix A, there is a close relationship between the Laplace and
Fourier transforms; and it is to emphasize this fact, that the term1/(jx)2, in (4.25),
has been left as such, instead of as -1/x2.
4.2
101
As emphasized in Chap. 1, except for very long observation times, the response of
a viscoelastic solid is indistinguishable from that of a viscoelastic fluid. Consequently, the behavior of viscoelastic fluids under cyclic loading can be evaluated
using either one of the complex property functions, M* or C*, derived earlier. Quite
often, however, in reference to Newtons constitutive equation for a viscous fluid:
r = gde/dt, the complex viscosity, g g0 jg00 , is used for this purpose. The
relationship between the complex modulus and complex viscosity can be easily
established introducing complex quantities, as before. Hence, if e*(jxt) = eoejxt, and
r*(jxt) : g* de*/dt = (jx)g*e*. Then, since r*(jxt) : M*e*(jxt), it follows that:
g g0 jg00
M
jx
4:26
The Cartesian components of the complex viscosity follow from this and (4.6), as:
g0
M 00
;
x
g00
M0
x
4:27
This expression shows that the phase relationships for the complex viscosity and
complex modulus are exactly the opposite of one another. In addition, (4.26) and
(4.3) define the complex viscosity as the Fourier transform of the relaxation
modulus:
g
M 1
Z
ejxt M tdt
jx
0
4:28
In the limit of very slow frequency excitations (x ? 0), this expression leads to
what is termedfor this reasonthe zero shear-rate viscosity, go, as:
g 0 go
1
Z
M tdt
4:29
The zero shear-rate viscosity enters the compliance function of viscoelastic fluids,
in the term t/go.
Example 4.3 Determine the steady-state response of the viscoelastic substance of
Example 4.2 to the harmonic excitation: e(t) = eocos(xt).
Solution:
The steady-state response may be computed with the constitutive equation
r M e , and the excitation in complex form e jxt eo ejxt eo cosxt
j sinxt. Since the applied strain corresponds to the real part of the complex strain,
the response will be the real part of the resulting complex stress; that is rt
Refr jxtg: Putting Gt Get=s in (4.3) produces the complex modulus, after
102
integrating and reordering, as:

2
R1
jxs
jx
G jx jx ej-t Get=s dt 1jxs
G G xs
.
1xs2
0
Taking the complex modulus and the complex applied strain into (4.2) yields
the complex stress as:
2
Geo
r jxt G jxe jxt 1xs
2 xs jxscosxt j sinxt:
Or, in full:
n
Geo
r jxt 1xs
xs2 cosxt xs sinxt jxs cosxt
2
xs2 sinxtg:
The steady-state response to the actual applied loading, et Refe g, then
2
Geo
becomes: rt Refr jxtg 1xs
2 xs cosxt xs sinxt
It is worth pointing out that because of the linearity of the Fourier transform, the
steady-state response of a viscoelastic system to a sum of excitations is equal to the
sum of the steady-state responses to the individual excitations. The simplest way to
proceed in this case is to establish the steady-state response of the system to each
cosine or sine component of the excitation and add the corresponding real or
imaginary part of the complex responses.
Example 4.4 Determine the steady-state response of a viscoelastic material having
shear relaxation modulus G(t), to the harmonic excitation: e(t) = e1osin(x1t) ?
e2ocos(x2t).
Solution:
The solution may be readily obtained as the sum of the steady-state response to
the two strain histories, e1 jx1 t e1o ejx1 t and e2 jx2 t e2o ejx2 t , representing the
actual excitations e1(t) = e1osin(x1t) and e2(t) = e2ocos(x2t). Using complex
algebra, let M M 0 jM 00 , and calculate the steady-state responses r1
M jx1 e1 jx1 t and r2 M jx2 e2 jx2 t. Then, take the imaginary part of r1
because of the sine functionand the real part of r2 because of the cosine
functionto arrive at the result:
rt G0 x1 sinx1 t G00 x1 cosx1 te1o
G0 x2 cosx2 t G00 x2 sinx2 te2o
4.3
Steady-State Constitutive Equations from Differential

As indicated in the introduction, the constitutive equations for steady-state conditions can also be derived from constitutive equations in differential form. This is
done in the same manner as before, by:
Introducing a complex excitation, e* = eoe jxt, say, to represent the controlled
variable
4.3
103
Invoking the closed-cycle condition to take the response as harmonic too, with
the same frequency as the excitation, but out of phase with it: r ejxtdM .
Taking the excitation and response into differential constitutive Eq. (3.9),
repeated here, for reference:
po r p1
d1
dm
d1
dn
r

p
r
q
e
q
e

q
e
m
o
1
n
dt1
dtm
dt1
dtn
Proceeding as indicated, performing the required operations and collecting terms,

yields:
h
i
h
i
po p1 jx1 pm jxm r qo q1 jx1 qn jxn e 4:30a
More succinctly:
P r Q e
4:30b
P po p1 jx1 pm jxm
4:31a
Q qo q1 jx1 qn jxn
4:31b
With the definitions:
Since (4.30b) is an algebraic relation involving complex numbers, it may be

rewritten in either of the following forms:
r
Q
e
P
4:32a
e
P
r
Q
4:32b
Proceeding as in the previous section yields the complex material property functions, M* and C*, and the relationship between them:
r M e ) M
Q
P
4:33a
e C r ) C
P
Q
4:33b
M C 1
4:34
The practical implication of expressions (4.33) is that to determine the complex

modulus or the complex compliance of any given rheological model, it suffices to
104
evaluate P* and Q*, in accordance with (4.31), and use complex algebra to remove
the complex quantities from the resulting denominator. The following example
illustrates this point.
Example 4.5 Derive the material property functions of a Maxwell fluid for steadystate conditions.
Solution:
1
d
Use (3.21b): dr
dt sr r E dt f t, and identify: po = (1/sr), p1 = 1, qo = 0,
q1 = E. With them, obtain P* = (1/sr)(1 ? jxsr), and Q* = jxE. Then, apply Eq.
2
Ejxsr
r jxsr
(4.33a) to establish M 1jxs
Exs
. The complex compliance, C*
r
1xs 2
r
1=Eo
1
1
1
may be determined directly from (4.33b), as: C 1jxs
Eo jxs Eo jxs Eo Eo j xs .
It is worth pointing out that in accordance with (4.3) and (4.13) the complex
material property functions of rheological operators may be obtained directly from
their relaxation or creep compliance functions.
Example 4.6 Use the relaxation modulus of the standard linear solid to derive its
complex modulus M*.
Solution:
The relaxation modulushof the standard
or three-parameter solid was obtained in
i
(3.36b), as: M t qpoo
q1
p1
qpoo et=sr ; sr pp1o ; where the index r has been
appended to the relaxation time, forh clarity.

i Taking this into (4.3) and performing
q1 qo
2
p1 po xsr jxsr
.
the integrations yields: M qpoo
1xs 2
r
Alternatively, using that qo/po : M? : Me, q1/p1 : Mo : Mg (Sects. 3.3.2,

3.3.4):
h
i
Mo M1 xsr 2 jxsr
M M1
2
1
h
i xsr
M1 Mo xsr 2
Mo M1 xsr
j
2
1 xsr 2
1 xsr
In similar fashion, using (3.37b):C t pqoo
(4.13), leads, after some algebra to: C
p1
q1
i
pqoo et=sc ; sc qq1o and applying
C1 Co xsc 2
1xsc 2
o C1 xsc
j C1xs
2
c
4.4
4.4
Limiting Behavior of Complex Property Functions
105
Since the complex modulus and compliance are related to their time-domain
counterparts through expressions such as (4.3), it is reasonable to expect that their
asymptotic values M*(?) and M(?), or M*(0) and M(0), etc., are related in some
manner. That such values are related may be seen from a physical perspective. For
instance, since the glassy response (t = 0) is elastic and M(0) is a constant, and
since x is proportional to 1/t, one should expect that M*(?) = M(0).
To fix ideas, we obtain the relationships between the asymptotic values of the
moduli and use (4.3) for that effect. Before applying the limiting process, we
integrate the equation by parts:
Z1
M jx M0
ejxt
o
Mtdt
ot
Taking the limit of this expression as x ? ? confirms the physical argument

made before:
lim M jx M0
x!1
4:35
Or, using the Cartesian components and equating real and imaginary parts,
respectively:
M 0 1 M0;
M 00 1 0
4:36
To obtain the limit as x ? 0, we introduce the change of scale g = xt, and write:
Z1
o
g
M dg
og x
o
g
M dg M0 M1 M0
og x
M jx M0
ejg
From which, follows:

lim M jx M0
Z1
g!0
Finally, since x ? 0 as g ? 0:
lim M jx M1
x!0
Or, equivalently, in terms of the real and imaginary components:
4:37
106
Ce
C(j)
M(j)
M
M(j)
C(j)
C
Fig. 4.2 Real and imaginary components of the complex property functions of a viscoelastic solid.
M x ! 0 M t ! 1 M1 Me ; and M x ! 1 M t ! 0 Mo Mg . Therefore: M 0 0 Me ; M 00 0 0; and M 0 1 Mg ; M 00 1 0 . Also: C x ! 0 1=Me
C 0 0 Ce ; thus: C 00 0 0; and C x ! 1 1=Mg C 0 1 Cg ; and so; C 00 1 0
M 0 0 M1;
M 00 0 0
4:38
Starting with Eq. (4.13) and using arguments entirely similar as with the complex
modulus, the following relationships are obtained describing the limiting behavior
of the complex compliance. The analogs of (4.35) to (4.38) are, thus:
lim C jx C0
x!1
C0 1 C0;
C00 1 0
lim C jx C1
x!0
C0 0 C1;
C00 0 0
4:39
4:40
4:41
4:42
Expressions (4.35) to (4.42) indicate that the glassy and equilibrium complex
moduli and compliances are elasticas they should bein agreement with the
properties of the relaxation and creep functions, discussed in Chap. 2. Typical
graphs for the real and imaginary components of the complex property functions,
including the loss tangent are as shown in Fig. 4.2 for a typical viscoelastic solid.
4.4
107
C(j)
M
M(j)
M(j)
C(j)
C
Fig. 4.3 Real and imaginary components of the complex property functions of a typical viscoelastic
fluid M x ! 0 M t ! 1 0; and M x ! 1 M t ! 0 M0 Mg : Therefore M 0 0
0; M 00 0 0 and M 0 1 Mg and M 00 1 0: C x ! 0 C t ! 1 ! 1; thus
C 0 0 1; C 00 0 1 and C x ! 1 C t ! 0 1=Mg Cg C 0 1; C 00 1 0
Regarding viscoelastic fluids, we know, from Chaps. 1 and 2, that the nature of
their response to loads is in general indistinguishable from that of viscoelastic
solids, except in the long run, because viscoelastic fluids have zero equilibrium
modulus. For this reason, in particular, tan d0 M 00 =M 0 ! 1. Consequently,
the graphical forms of the complex material property functions of a typical viscoelastic fluid are as shown in Fig. 4.3.
Quite naturally, the graphical representations of the complex material property
functions of a rheological model depend on the specific arrangements of its elements. Because the mechanical response of the standard linear solid to loads
exhibits all the characteristics of a viscoelastic solid, it follows that its complex
property functions should resemble those shown in Fig. 4.2, whereas those of a
fluid should be as in Fig. 4.3.
The complex modulus and compliance of the Maxwell fluid are depicted in
Fig. 4.4. The representations in the figure are constructed taking the indicated
limits of the results of Example 4.5.
The complex modulus and compliance of a Kelvin solid are displayed in
Fig. 4.5, which was constructed using the results of Problem P.4.3.
108
C(j)
M(j)
C(j)
M(j)
Fig. 4.4 Real and imaginary components of the complex properties of a Maxwell fluid
M x ! 0 0; M x ! 1 Mg ; C x ! 0 j1; C x ! 1 1=Mo 1=Mg
Therefore :M 0 0 0; M 00 0 0; M 0 1 Mg ; M 00 1 0; C 0 0 0; C 00 0 1;
C 0 1 Cg ; C 00 1 0; with : Cg 1=Mg
C
C(j)
M(j)
C(j)
M
M(j)
Fig. 4.5 Real and imaginary components of the complex properties of a Kelvin Solid
C x ! 1
M x ! 0 M1 Me ; M x ! 1 j1; C x ! 0 1=M1 Ce ;
0!M 0 0 Me ; M 00 0 0; M 0 1 M 00 1 1C 0 0 Ce ; C 00 0 0; C 0 1
C 00 1 0
4.5
Energy Dissipation
4.5
109
Energy Dissipation
In Chap. 1, we stated the physical expectation that viscoelastic substances should

have some capacity to store energy and some to dissipate it. This section proves
that statement by means of a simple experiment.
The ability of a material to store or dissipate energy may be ascertained subjecting a one-dimensional specimen of the material to a full load-unload cycle.
Without loss of generality, we consider a uniaxial specimen subjected to the
excitationDl(t) = Dlosin(xt), which divided by the original length of the specimen
produces the strain e(t) = eosin(xt). The test is then carried out for a complete
cycle, so that td = 2p/x. Using Eq. (2.46):
tZd
WV jt0d 1=VWjt0 r
0
de
ds
ds
4:43
and noting that the stress corresponding to the applied strain is r(t) = ro
sin(xt +d), leads to:
WV jt0d
2p=x
Z
r o eo
d
sinxt d sinxtdt ro eo x sin d
dt
2p=x
Z
1 cos 2xtdt a
which integrates to:

WV jt0d pro eo sin d
4:44
This result indicates that maximum dissipation occurs when d = p/2; which
according to the discussions in Chaps. 1, 2 and 3, corresponds to fluid behavior. By
contrast, zero or no dissipation, corresponding to solid, elastic response, occurs at
d = 0.
Using (4.8): ro = ||M*|| eo, and noting that k M k sin d M 00 , puts (4.44) in the
form:
WV jp0 p e2o M 00 jx
4:45a
In words: energy dissipation depends on M 00 ; which, for this reason, is termed

the loss modulus. By contrast, zero dissipation, and hence fully recoverable elastic
response occurs at d = 0, when: M M 0 . For this reason, M 0 is called the storage
modulus.
Alternatively, substituting o = ||C*||ro, in accordance with (4.20), into (4.45a),
and noting that, per the definition in (4.19): kC k sin d C 00
WV jp0 p r2o C 00 jx
4:45b
110
The same remarks can be made about C 0 and C 00 , relative to energy storage or
dissipation, respectively, as were made in relation to M 0 and M 00 . Thus, C 0 is the
storage compliance, and C 00 , the loss compliance. Furthermore, since the total
energy dissipated in a cycle has to be positive, in accordance with the closed-cycle
condition, expressions (4.45a) show that M 00 jx [ 0 and C 00 jx [ 0. This last
inequality is one reason why the complex compliance is written with a negative
sign in front of C 00 .
Example 4.7 Derive an expression for the maximum energy per unit volume
dissipated per cycle by a material with constitutive equation of the standard linear
solid type.
Solution:
C1 Co xsc 2
o C1 xsc
Use the results of Example 4.6: C 1xs 2 j C1xs
, together with
2
c
(4.45b) to write:
W
V
o C1 xsc
pr2o C1xs
; where, C? : po/qo, Co : p1/q1, and
2
c
sc : q1/qo.
4.6
Problems
P.4.1. Generalize the result of Example 4.2 to the case when the relaxation
function is given by a finite sum of exponentials, such as that in Example
4.1.
Answer:
G0 jx Ge
n
X
i1
Gi
xi si 2
1 xi si
;
2
G00 jx
n
X
i1
Gi
xi si
1 xi si 2
Hint: Use the results of Example 1 and follow the approach in Example 2.
P.4.2. Verify the result of Example 4.3 by direct integration of the hereditary
integral.
Hint: Start with the integral constitutive equation (2.33a), and apply integration
by parts twice to derive the steady-state response.
P.4.3. Derive the material property functions of a Kelvin solid appropriate to
steady-state conditions.
Answer:
M M1 1 jxsK ;
C

1
1 jxsK
q1
qo
; sK ; M 1
2
M1 1 xsK
p1
po
Hint: Write (3.16a) in operator form: por = (qo ? q1qt)e, and obtain P* = po,
and Q* = qo ? jxq1 = qo(1 ? jxsK); where sK : q1/qo is the relaxation time
of the model. Use this and (4.33a) to get M qpoo 1 jxsK M1 1 jxsK .
4.6
Problems
111
Establish the complex compliance by inverting this relation: C M11
1jxs
1xs2
P.4.4. Derive the expression for the maximum energy dissipated per cycle by a
material with relaxation function of Prony-Dirichlet type, like that in P.4.1,
under oscillations of amplitude eo.
Answer:
n
X
W
x k sk
pe2o
Gk
V
1 xk sk 2
k1
Hint: Use the results of P.4.1 together with (4.45a) to arrive at the desired result.
P.4.5. Calculate the maximum energy dissipated per cycle by a bar of a standard
linear solid under a uniaxial cyclic strain of amplitude eo?
o M1 xsr
; where: M? : qo/po, Mo : q1/p1, and sr : p1/po.
Answer: pe2o M1
xs 2
r
Hint: Use the results of Example 4.6, that: M
M1 Mo xsr 2
1xsr 2
o M1 xsr
j M1xs
,
2
r
together with (4.45a) to arrive at the quoted result.

P.4.6. Obtain the complex viscosity of the standard linear solid of Problem P.4.5.
Answer:
h
i
2
M
M
xs
1
o
r
Mo M1 sr
h
i
g jx
j
2
2
1 xsr
x 1 xsr
Hint: Insert the expression for M* derived in Example 4.6 into Eq. (4.26) and

00
0
(4.27)for the complex viscosity: g jx Mjx Mx j Mx , to arrive at the result.
P.4.7. Calculate the complex compliance for a viscoelastic solid whose creep
n
P
compliance is C t Ce Ci et=si .
i1
Answer:
C jx Ce
n
X
i1
Ci
xi si 2
1 xi si
j
2
n
X
i1
Ci
xi si
1 xi si 2
Hint: Insert C(t) into (4.13), performing the indicated integrations and simplifying to obtain the result.
P.4.8. Calculate the work done per cycle of oscillation under a stress
r(t) = rosin(xt), for a bar made of the material of Problem P.4.7.
Answer:
n
X
W
x i si
pr2o
Ci
V
1 xi si 2
i1
112
W
V
Hint: Use (4.45b):
p r2o C 00 jx and the result: C 00 jx
n
P
i1
x i si
Ci 1x
,
s 2
i i
from Problem P.4.7, to arrive at the desired result.

P.4.9. Derive a general expression for the steady-state response of a linear visN
P
coelastic material to a cyclic stress rt
rok sinxk t.
k1
Answer:
et
N
X
rok C0 jxk sinxk t C 00 jxk cosxk t
k1
Hint: The solution may be readily obtained as the sum of the steady-state
response to each of the stress histories, rk jxk t rok ejxk t . Proceeding as in
Example 4.4, let Ck C jxk C0 jxk C00 jxk jCk0 jCk00 , calculate
the steady-state strain responses ek Ck ek and add them together. Since all
components of the excitation are sine functions, the solution sought corresponds to the imaginary part of the total complex strain thus obtained.
P.4.10. Assuming that p 2p=x1 and xk kx1 ; fork 1; 2; . . .; N, calculate the
work done in the period, p, by the stress history of Problem P.4.9 acting on
a viscoelastic bar having creep compliance Ct.
Answer:
WV jtp
t
N
X
pr2ok C00 jxk ;
p 2p=x1 ;
xk kx1 ;
j
p
1
k1
P
Hint: Take the stress history rt Nk1 rok sinxk t given in Problem P.4.9,
P
and the derivative dtd et Nk1 rok xk C0 jxk cosxk t C 00 jxk sinxk t
of the strain response calculated there, into (4.43). Perform the indicated
algebra, and use the orthogonality of the sinusoidal functions over an interval
spanning the fundamental period (see Appendix A) to arrive at the result.
References
1. J.D. Ferry, Viscoelastic Properties of Polymers (Wiley, New York, 1980), pp. 1114
2. W. Flgge, Viscoelasticity, 2nd edn (Springer, New York, 1975), pp. 95120
3. A.S. Wineman, K.R. Rajagopal, Mechanical Response of Polymers, an Introduction
(Cambridge University Press, Cambridge, 2000), pp. 115147
4. R. M. Christensen, Theory of viscoelasticity,2nd edn (Dover, New York, 1982), pp. 2126
Structural Mechanics
Abstract
This chapter is devoted to structural mechanics, developing the theories of

bending, torsion, and buckling of straight bars, and presenting a detailed
account of vibration of single-degree-of-freedom viscoelastic systems, including vibration isolation. A balanced treatment is given to stressstrain equations
of integral and differential types, and to stressstrain relations in complexvariable form, which are applicable to steady-state response to oscillatory
loading. All equations in this chapter are developed from first principles,
without presuming previous knowledge of the subject matter being presented.
This approach is followed for two reasons: first, because it is necessary for
readers without a formal training in mechanics of materials; and secondly,
because it provides the readereven one with formal training in classical
engineeringwith a method to follow when the use of popular shortcuts, like
the integral transform techniques, might be questionable or unclear.
Keywords

Static Bending Torsion Buckling Navier Hereditary Integral

Differential Steady-state Euler Critical Creep Spring Correspondence
Mass Free Forced Vibration Amplification Transmissibility
5.1

Introduction
Previous chapters have examined the various forms available of the stressstrain
laws for uniaxial conditions, using a straight bar of viscoelastic material as the
physical specimen to which these laws applied. This chapter expands the one-

113
114
dimensional applications to examine the quasi-static1 bending, torsion, and

buckling of structural members made of homogeneous, linear, isotropic viscoelastic materials. Free and forced vibrations of one-dimensional viscoelastic
spring-mass systems are also examined in detail, including the important topic of
vibration isolation. As will be explained, the case of free vibrations is considerably
more complicated than that of forced vibration, to which the methods of Chap. 4
apply directly. For free vibrations, then, only the solution outline is given. For
completeness of presentation as well as for the benefit of readers without formal
training in mechanics of materials, all topics are treated in great detail.
As will be seen, the relations derived for viscoelastic materials have the same
forms as their elastic counterparts. The kinematic relations and the equilibrium
equations are established for materials whose stressstrain law is in integral as well
as in differential operator form. The approach used to derive the equilibrium
equations is as follows (c.f. Chaps. 2 and 3):
1. Establish the equilibrium equation relating the pertinent mechanical element,2
F, to the corresponding stress, r:
Fx; t Axrx; t
2. Introduce the stressstrain law:

a. For materials of integral type, rx; t Et s des (c.f. Chap. 2).3
Thus,
Fx; t AxEt s des
b. For materials of differential type apply operator P[] to both sides of the
equilibrium equation, and use that P[r] = Q[e], to write (c.f. Chap. 3)
PFx; t AxQe
3. Use the pertinent kinematic relation to replace the generalized strain, e, with the
corresponding generalized displacement to arrive at the desired form.
That is, such that the external stimuli vary sufficiently slowly that no significant inertial effects
develop.
2
The mechanical elements that may act on a bars cross section are two bending moments, one
torsional moment, one axial force, and two shear forces.
3
In this chapter, to avoid confusion, the letter M is used to denote the moment of the external
forces, while E will denote relaxation modulus. As in previous chapters, C will still denote
compliance.
5.1
Introduction
115
(a)
(c)
q
V
(+)
x
V
(b)
(d)
u
y
dA
M
x
dA
x
x
Fig. 5.1 Bending of a straight beam. Kinematics of deformation. a Initial geometry. b Deformed
geometry. c Sign convention. d Force resultant at x
5.2
Bending
This section examines the mechanical response of initially straight viscoelastic

beams subjected to transverse loads only.4 The derivations assume that the lateral
deflection and in-plane rotation of the beam are small; and that, in accordance to
Naviers assumptions, plane sections initially normal to the axis of the beam
remain plane and normal to the beams axis during deformation [1].
Regarding geometry, the axis of the beam is taken along the x-axis; the y- and zaxes are assumed principal centroidal axes of inertia of the cross section; the y-axis
directed upward, and the z-axis defined by the right-hand-screw rule, as shown in
Fig. 5.1. To keep the presentation simple, all transverse loads will be assumed to
act in the xy plane, along the y-direction.
As can be seen from Fig. 5.1b, the axial displacement, u(x,y,z,t), of the bar
depends only on the axial, x, and transverse, y, coordinates. For this reason, in the
sequel, we omit the out-of-plane coordinate, z. Hence,
ux; y; z; t ux; y; t ytan/; 8z 2 Ax
5:1
Generalization to beam-columns, which include axial loads, is straightforward and is left as an

exercise.
116
For small deflections, v(x,t), and rotations, u, of the bars axis: tan
u u ov=ox; thus,
ux; y; t y/ y
o
vx; t; 8z 2 Ax
ox
5:2
Introducing the axial strain in the bari.e., the change in length per unit
original lengthleads to:
ex; y; t
o
o2
ux; y; t y 2 vx; t yv00 x; t; 8z 2 Ax
ox
ox
5:3
Figure 5.1d shows the internal stresses r and s which develop on a generic
cross section to balance the externally applied loads. Taking moments with respect
to point O, located on the generic section at station x, we write
dx
M x; t qx; tdx
2
yrx; y; tdA
A

sx; y; tdA dx 0
In the limit as dx ? 0, this expression produces the resultant moment acting on

the cross section at a generic axial station, x, as
Mx; t
5:4
rx; y; tydA
A
The following relations result from force and moment equilibrium taken
between any two stations separated an infinitesimal distance Dx along a beams
axis, as depicted in Fig. 5.2, and do not contain material properties.
Indeed,
a. Moment equilibrium about the bottom left-hand corner of the elemental length
of beam leads to
V DV Dx M DM M qDx
Dx
0
2
Fig. 5.2 Force and moment

equilibrium of a differential
beam element
M+ M
V
V+ V
x
5.2
Bending
117
Neglecting terms of order higher than the first,

Vx; t
o
Mx; t
ox
Just like for elastic solids, this expression is independent of material properties.
b. Equilibrium of forces requires that V qDx V DV 0, which, neglecting
higher-order terms yields:
qx; t
o
Vx; t
ox
As for elastic materials, this expression is also independent of material

properties.
5.2.1
Hereditary Integral Models
Introducing the strainstress form e(t) = C(t - s) * dr(s) of the constitutive

equation into (5.3), and convolving the result with the relaxation modulus, leads to
(c.f. Chap. 2)5
rx; y; t yEt s ov00 x; s
5:5
Combining (5.4) and (5.5), and separating the spatial and time operations
produces
2
3
Z
Mx; t 4 y2 dA5Et s ov00 x; s
Or, more succinctly,

Mx; t IxEt s ov00 x; s
5:6
This is the relaxation integral analogue of the elastic expression

M = Iz E v00 , in which Iz(x) represents the second moment of areaalso called
area moment of inertiaof the beams cross section at station x.
To solve for the deflection function, v(x,t), of the beam, (5.6) is convolved with
the tensile compliance function, C(t), of the beams material (c.f. Chap. 2). Proceeding thus, and rearranging,
Note that the same result may be reached by taking the kinematic relation (5.3) into the stress
strain constitutive equation r(t) = E(t - s)*de(s).
118
o2 v 1
Ct s oMs
ox2 I
5:7a
This expression is also often written in terms of the rotation angle, u = qv/qx,
as
ou 1
C t s dMs
ox I
5:7b
The relation between the normal stress and the external moment is obtained
combining (5.5) and (5.6). This is accomplished by first rewriting (5.5), as
Et s ov00 x; s
rx; y; t
y
Inserting this result on the right-hand side of (5.6), and rearranging, leads to
rx; y; t
Mx; t
y;
Ix
8z 2 Ax
5:8
This expression is the same as that for an elastic beam in bending; as it should
be, since the bending stress is independent of material constitution.
5.2.2
Differential Operator Models
Proceeding as suggested in the introduction, we apply the operator Q to the strain

displacement relation (5.3) and use that P[r] = Q[e] to write
Prx; y; t Qex; y; s yQv00 x; t; 8z 2 Ax
5:9
Applying the operator P to expression (5.4), combining the result with (5.9),
and separating the spatial and temporal parts, leads to the counterparts of (5.6) and
(5.7a, b), respectively,
PMx; t IxQv00 x; t
Qv00 x; t
PMx; t
Ix

o
PM x; t
ux; t
Q
ox
Ix
5:10
5:11a
5:11b
5.2
Bending
119
(a)
(b)
P(t) = Po
y
M(0,t)
F(t)
x
V(0,t)
To derive the counterpart of (5.8), first rewrite (5.9) as Qv00 x; t Prx;y;t

.
y
Then, combine the result with (5.10) to get Prx; y; t PMx;t
Ix y. Applying the
operator P1 to both sides of this expression leads to the elastic form in (5.8).
Example 5.1 A cantilevered beam of length L and uniform cross section with
moment of inertia I is made of a viscoelastic material with tensile relaxation
modulus E(t) and is loaded as shown in Fig. 5.2. Determine the deflection at its tip.
Solution:
As indicated in Fig. 5.3, force and moment equilibriumas in the statics of
elastic solidsproduce the reactions V(0,t) = F(t) and M(0,t) = -LF(t), on the
beams support. Using these results and the sign convention, shown in Fig. 5.1,
produces the moment M(x,t) = -LF(t) ? F(t)x, at station x. Taking this into
(5.7a), integrating the resulting differential equation twice using that
v(0,t) = v(0,t)
0, and i inserting F(t), yields, after some algebra,
h =

L3 3 x 2 1 x 3
vx; t 3I 2 L 2 L C t s dFs. The value of this expression at
3
x = L, gives the result vL; t L3I Ct s dFs.
5.2.3
Models for Steady-State Oscillations
The interest here is on the steady-state lateral deflections of straight viscoelastic

beams of uniform cross section, which are made of materials of either integral or
differential type and are subjected to cyclically varying loads of frequency, x. To
develop the appropriate relations, we take the complex forms M*(x, jxt) =
Mo(x)ejxt and v*(x, jxt) = vo(x)ej(xt+d), of the cyclic moment and the bending
deflection, respectively, and insert them into (5.7a) for materials of hereditary
integral type, and (5.11a) for materials of differential type, proceeding as in
Chap. 4.
For materials of integral type, convolve (5.7a) with the tensile relaxation
modulus E, say, to cast it in the form: M t IEt s ov00 x; s. Then, replace
M and v with their complex forms and proceed as in Chap. 4 to get6
jxt
Here E jx jx R 1
Etdt jxF Et; where F Et is the Fourier transform of
0 e
E(t).
120
Fig. 5.4 Torsion of a

straight bar of circular cross
section. Kinematics of
deformation
T
r
x
M jxt IE jx
d2
v x; jxt
dx2
5:12a
In similar fashion, inserting M* and v* into (5.11a), and rearranging the terms,
leads to7
M jxt I
Qjx o2
o2

v
x;
jxt
IE
jx
v x; jxt
Pjx ox2
ox2
5:12b
The expression on the far right uses (4.33) with E QP ; as the corresponding
modulus.
5.3
Torsion
The torsional response of an initially straight bar of circular cross section is

considered in what follows. Because of mechanical and geometric symmetry, a
straight circular bar subjected to pure torsion deforms in such a way that initially
plane cross sections, which are normal to the bars axis before deformation, rotate
without distortion and remain plane and normal to the bars axis after deformation
[2]. Therefore, any straight line initially parallel to the axis of the bar and located
at any radius from the bars axis will remain straight during deformation. As a
consequence, the shear strain, c, at any radial distance, r, from the center of the
cross section of the bar must be proportional to r, as indicated in Fig. 5.4.
The kinematics of deformation in this case leads to the following relation
between the angle of twist, w, and the shear strain, c:
cr; u; t x r wx; t
As shown in Chap. 4, Qjx=Pjx jx F Et E jx.
5:13
5.3
Torsion
121
Differentiating this expression with respect to x, and introducing the angle of

twist per unit length, h, as
hx; t
dw
dx
5:14
We arrive at the relation:

cr; u; x; t r
dw
rhx; t
dx
5:15
Moment equilibrium at axial station, z requires that

Tx; t
Z
rsr; x; tdA
5:16
Next, the kinematic and equilibrium relations in (5.15) and (5.16) will be
combined with the constitutive equations in hereditary and differential operator
forms.
5.3.1
Use (5.13) and the convolution form of the constitutive equations for a linear
homogeneous viscoelastic material with shear relaxation modulus G(t):
s G dc, to write (c.f. Chap. 2):
sr; xt Gt s ocr; x; s rGt s ohx; s
5:17
Combining this with the equilibrium Eq. (5.16) yields

Tx; t
r 2 dAGt s ohx; s JxGt s ohx; s
5:18
This is the viscoelastic equivalent of the elastic relation T = J Gh; and, just as
in the elastic case, J(x) represents the polar moment of inertia of the cross section
at station, z.
If the twist per unit length is the prescribed loading, (5.18) may be inverted by
convolving it with the shear creep compliance function CG to read:
hx; t
o
1
wx; t
CG t s dTx; s
oz
J x
5:19a
122
This expression is also often written in terms of the angle of twist, w, using
(5.14), as
ow
1
CG t s oTx; s
ox Jx
5:19b
The relationship between shear stress and external torque is obtained by

replacing the expression G dh s=r, obtained from (5.18), into (5.19a). The
result is identical to that for the torsion of an elastic bar:
sr; x; t
Tr; tr
Jx
5:20
As with the bending of straight beams, the stress is independent of material

constitution.
5.3.2
The stressstrain relations governing the torsion of bars of material of differential

type are derived using that P[s] = Q[c] with expression (5.15). Thus,
Psr; x; t Qcr; x; s Qrhx; t rQhx; t
5:21
The relationship between torque and angle of twist per unit length is obtained
by applying operator P[] to both sides of (5.16), and using (5.21):
PTz; t
rPsr; x; tdA
r 2 Qhx; tdA
Performing the integration, leads to the final form in terms of the second polar
moment of areapolar moment of inertiaJ of the cross section at axial station x.
PTx; t JxQhx; t
5:22a
And, in terms of the total twist, w:

PT x; t JxQ
o
wx; t
ox
5:22b
Finally, expression (5.20), which shows that the shear stress is independent of
material properties, may be derived taking Q[h] = P[s]/r, from (5.21), into (5.22a)
and rearranging to get Ps PT
Jx r, which, after integrationi.e., after application
-1
of the inverse operator P to both sides, yields (5.20).
5.3
Torsion
123
Example 5.2 A straight bar with polar moment of inertia J, and length L, is
subjected to torsional moments M(t) at its ends. Derive an expression for the
total angle of twist if the bars material is a standard linear solid. Determine the
rotation after to units of time if the applied torque is the step function
M(t) = MoH(t).
Solution:
The solution sought may be obtained solving (5.22a) for h w=L: To do this,
first set Qw LPM=J; and then insert the operators P = po ? p1qt and
Q = qo ? q1qt for the material at handh (c.f. Chap. i3), as well as the controlled
dw
dt
M pq11 dM
dt , with s = q1/qo. This first

order differential equation has an integrating factor u = et/s, and thus, d wet=s
h
i
0
t 0
p1 dM
Let =s po
0
t=s
M
LJ R et =s
0
J
q1
q1 dt dt [c.f. Appendix A]. Assuming w0 0; wte
0
h
i
po
p1 dM
Mo L
0
0
Mt
.
Using
M(t)
=
M
H(t),
at
t
=
t
,
leads
to
w
dt
o
o
o
J
q1
q1 dt0
h

i
po
p1
po
to =s
.
q0 q1 q0 e
variable, M. This leads to
5.3.3
1s w LJ
po
q1
Models for Steady-State Oscillations
The interest here is on the steady-state torsional deflections of straight viscoelastic

beams of uniform circular cross section, made of materials of either integral or
differential type, subjected to cyclically varying loads of frequency, x. Following
the same procedure as was used to develop the steady-state equations for beam
bending, represent the complex torque and rotation as T*(x, jxt) = To(x)ejxt and
w*(x, jxt) = wo(x)ej(xt+d), respectively, and insert them into (5.19a) for hereditary
integral type materials, and (5.22a, b) for materials of differential type, and proceed as in Chap. 4.
For materials of integral type, convolve (5.19b) with the shear relaxation
modulus G, say, to cast it in the form: T t JGt s d dw
dx x; s. Then, replace
T and w with their complex forms and proceed as in Chap. 4 to get
T jxt JG jx
o
w x; jxt
ox
5:23a
In similar fashion, inserting T* and w* into (5.22b), and rearranging the terms,
leads to
T jxt J
Qjx o
o
w x; jxt JG jx w x; jxt
Pjx ox
ox
5:23b
The expression on the far right uses (4.33a), with G QP , as the corresponding
modulus.
124
5.4
Column Buckling
This section discusses the stability of an initially straight axially compressed

viscoelastic bar of constant cross section. The axial load is assumed independent of
position along the bars axis. As will be shown, under constant sustained loading, a
viscoelastic column may remain stable indefinitely, become instantaneously
unstable, or turn unstable only after enough time has elapsed since the load was
applieda phenomenon referred to as creep buckling. A bar made of an elastic
material, by contrast, can only either remain stable, or become unstable instantly.
The critical load of a compressed bar is that at which its initial configuration
ceases to be stable. An unstable configuration is an equilibrium configuration that
may differ only very slightly from the initial, static equilibrium configuration. For
this reason, to determine the critical load of a column, one must find the load under
which equilibrium is possible in a configuration that is lightly different from the
undistorted one [3].
The pin-ended column of Fig. 5.5 will be used to develop the subject matter. In
so doing, it will be assumed that the axial load is applied in some time-dependent
manner until it reaches its target value, No, which will be held constant thereafter.
Using the free-body diagrams in the figure to establish equilibrium in the slightly
bent configuration yields
N t vx; t Mx; t
v0; t 0
5:24
vl; t 0
Fig. 5.5 Equilibrium of pinended column in slightly bent

configuration. a Initial
configuration. b Equilibrium
in a slightly deformed state
(a)
(b)
N(t) = P f(t)
No
x
v
M
No
N
5.4
Column Buckling
Fig. 5.6 Temporal

dependence of axial load and
lateral deflection of an
initially straight viscoelastic
column
125
N
No
f ( t )
t
v (x , t)
Axial
Lateral
deflection
deflection
v( x , t ) = V (x) C (t) No
t
t
The response of the bar prior to the load reaching its target value is immaterial
for the present discussion. Hence, the origin of the time scale is shifted to the time
when the target load No is reached. Under these conditions, the lateral deflection is
controlled by the materials creep function, so that v(x,t) = V(x)C(t)No, as indicated in Fig. 5.6 (c.f. Chaps. 2 and 3), in which V(x) represents the magnitude of
the lateral deflection v(x,t).
5.4.1
In this case, using that v(x,t) = V(x)C(t)No with (5.6) and (5.24) leads to
No Ct No Vx IxEt s dCsNo V 00 x 0
v0; t 0
vl; t 0
5:25
After canceling out the common term No, using that E*dC = H(t) (c.f. Chap. 2),
and rearranging, this take the form:
No Ct Ix
V 00 x
0
V x
5:26
The only way in which the algebraic sum of two functions of different variables
can be equal to zero is either if both functions are zeroa trivial caseor if both
126
functions equal the same constant. Calling this constant b2 leads to the following
characteristic and evolution equations, respectively:
V 00 x 1=Ib2 Vx 0
V0 0
Vl 0
5:27
No Ct b2
5:28
Introducing the notation k2 b2 =I, the non-trivial solution of (5.27) is:

V x A1 sinkx A2 sinkx, which, in view of the boundary conditions, yields
the following Eigen-values of the equation:
b
np
; n1
k pffiffi
l
I
b
5:29a
pffiffi np
I ; n1
l
5:29b
The critical loads are obtained combining (5.26), (5.28), and (5.29), replacing
No with Ncr, for consistency with common usage, as
Ncr t
n2 p2 I
;
l2 Ct
n1
5:30
From this, using the relations E(0) : Eg = 1/C(0), and E(?) : Ee =

1/C(?), the following asymptotic expressions may be obtained (c.f. Chap. 2):
lim Ncr t Ncrg
n2 p 2
n2 p2
IE0 2 IEg ;
2
l
l
lim Ncr t Ncre
n2 p2
n2 p2
IE1 2 IEe ;
2
l
l
t!0
t!1
n1
n1
5:31
5:32
These are the short-term or glassy, Ncrg, and long-term or equilibrium, Ncre,
Euler loads. Quite often, the critical Euler load of interest corresponds to n = 1.
The physical meaning of the above results is that as follows:
a. A viscoelastic column will be asymptotically stable if the applied load does not
exceed its long-term critical Euler load, Ncre
b. A viscoelastic column will become asymptotically unstable if the applied load
is larger than its long-term critical Euler load, Ncre
c. A viscoelastic column will be instantaneously unstable if the applied load is
larger than its short-term critical Euler load, Ncrg.
In addition, if the viscoelastic column does not buckle instantly after reaching
the target load, the presence of the compliance function, C(t), in (5.30) allows one
5.4
Column Buckling
127
to estimate the time, measured from that point on, that it would take the column to
buckle.
Example 5.3 Find the instantaneous and long-term buckling loads of a straight
column 100 cm long, and rectangular cross section 15 cm by 10 cm, if it is made
of a viscoelastic material with modulus E(t) = 3.5 ? 8.5e-t/0.25 MPa.
Solution:
The critical load depends directly on the second moment of area and is a minimum for the smallest value of the latter. The minimum centroidal moment of inertia
1
for the column is I 12
bh3 =15 1,000/12 = 1,250 cm4 = 1.250 10-5 m4. The
long-term modulus of the material is 3.5 MPa; and its instantaneous modulus,
3.5 ? 8.5 = 12 MPa. Inserting these values in (5.31) and (5.32), and setting n = 1,
produces the critical loads: Ncrg & 1.48 kN and Ncre & 0.43 kN.
5.4.2
When the constitutive equations are given in differential operator form, applying
operator P to both sides of (5.24), and invoking (5.10) produces
PNtvx; t PMx; t IQv00 x; t
Proceeding as before, using N(t) = No, v(x,t) = V(x)C(t), and rearranging,
produces the same characteristic Eq. (5.27), and Eigen-values (5.29), as for
materials of integral type. In this case, the evolution equation, equivalent to (5.28),
takes the form:
No
b2
QCt=PCt
5:33
The critical load is obtained combining (5.29) and (5.33) to give

Ncr
n2 p2
IQC t=PCt
L2
5:34
As seen from this last expression, the limiting values of the critical load depend
on the specific form of the differential operators.
As discussed in Chap. 3, for rheological models whose operators P and Q are of
the same order, Q[C(t)]|t=0 = qnC(0) and P[C(t)]|t=0 = pnC(0). Likewise, for
models with non-zero long-term modulus, Q[C(t)]|t=? = qoC(?), and
P[C(t)]|t=? = poC(?). Taking these values into (5.33), using the notation
Eg : qn/pn, for the impact or glassy modulus, and Ee : qo/po for the long-term,
or equilibrium modulus, produces the same expressions as (5.31) and (5.32).
128
Eo
EK
Example 5.4 Find the instantaneous and long-term buckling loads for a straight
column of length L, and minimum moment of inertia, I, made of a viscoelastic
material of standard linear solid type, with parameters as shown in Fig. 5.7.
Solution:
The solution is given by (5.31) and (5.32), which require the glassy and
equilibrium moduli. These are obtained from the differential operators
P = (Eo ? EK) ? gKqt, and Q = (EoEK) ? EogKqt [see Eq. (3.35a)], and the
relationships Eg : qn/pn = Eo; and Ee : qo/po = (EoEK)/(Eo ? EK). Hence,
I
EK
Ncrg p2 Eo I=L2 , and Ncre p2 EEooE
2 .
K L
5.5
Viscoelastic Springs
Analytical models of viscoelastic springs find many important practical applications, both under quasi-static and dynamic conditions. Composite elastomeric
bearings, for instance, are made up of alternating layers of elastomers and elastic
materials, such as is shown conceptually, in Fig. 5.8. The elastomeric layers are
called pads, and the elastic layers, shims. In general, each layereither pad or
shimmay have a different thickness.
Since the elastic materials used for the shims are orders of magnitude stiffer
than the elastomers, all deformation in these bearings is essentially due to that of
the pads. Idealizing each pad as a massless spring of the appropriate type (axial,
shear, bending, or torsion), one may construct a multidimensional structural model
of the bearing from a combination of viscoelastic springs connected by rigid
interleaves. The same is true of analytical models for more elaborate parts.
Fig. 5.8 Flat-pad composite

elastomeric bearing and
possible types of external
loads
N
T
M
endplate
V
shim
pads
endplate
5.5
129
In this section, the equations of static equilibrium are invoked to develop the
forcedeflection relationships for viscoelastic springs of axial, shear, bending, and
Fig. 5.9 Geometry, material
properties, and loading of
viscoelastic spring models
E, A, L
Axial spring
N(t), u(t)
G, A, h
V(t), v(t)
Shear spring
E, I, L
Bending spring
M(t), (t)
G, J, L
Torsion spring
T(t), (t)
torsion type. For simplicity, straight bars of uniform cross section are used to
represent axial, bending, and torsional springs, and a plate to idealize a shear
spring, as shown in Fig. 5.9.
Fig. 5.10 Axial spring
E, A, L
N(t), u(t)
A (t)
5.5.1
N(t)
Axial Spring
The bar-like idealized model of an axial spring and the corresponding free body
diagram are shown in Fig. 5.10.
Use of the figure to establish force equilibrium leads to
N t Art
Assuming the deflection at the end of the spring to be u(L,t) = uo f(t) the axial
deflection at station x along the bars axis takes the form: u(x,t) = (uo/L) x f(t),
producing the axial strain:
ex; t oux; t=ox uo =L f t uL; t=L
These expressions will now be used to develop the forcedeflection relationshipand stiffnessfor an axial spring in terms of its integral, differential, and
steady-state constitutive equations (c.f. Chaps. 2, 3 and 4).
a. Hereditary integral springs. In this case, substitution of r = E*de into (a), using
(b), leads to
130
N t
A
Et s dus KN t s dus
L
5:35a
Hence, the operational form of the spring stiffness becomes:

KN t s d
A
E t s d
L
5:35b
b. Differential operator springs. Here, applying the operator P to both sides of (a),
using (b) and P[r] = Q[e] on the right-hand side produce:
PN t
A
Qut
L
5:36a
The spring stiffness is only indirectly defined by this expression. Its Laplace
transform yields a form more directly comparable to its elastic counterpart:
K N s
AQs=Ps
L
5:36b
c. Springs for steady-state oscillations. Introducing the complex strain e* = u*/L,

and complex force, N*, and integrating the constitutive equation (c.f. Chap. 4),
results in the complex modulus E* and produces the relationship:
N jx
A
E jxu jx KN jxu jx
L
5:37a
This defines the complex stiffness of axial springs made of materials of integral
type, as
KN jx
AE jx
L
5:37b
For rheological models, use of N* and u* with (5.36a) and (4.33a) leads to
KN jx

A Qjx
AE jx
L Pjx
L
5:37c
5.5
5.5.2
131
Shear Spring
The idealized model of a shear force spring is shown in Fig. 5.11. Using it to
establish force balance leads to
Fig. 5.11 Shear spring
G, A, h
V(t), v(t)
V t Ast
c
Assuming the shear deflection at the end of the spring is given by
v(L,t) = vo f(t), and using the thickness, h, of the spring, yields the shearing
strain:
cx; t
vx; t
h
Expressions (c) and (d) will be used to develop the forcedeflection relationshipand stiffnessfor a shear spring in terms of its integral, differential, and
steady-state constitutive equations (c.f. Chaps. 2, 3 and 4).
a. Hereditary integral springs. In this case, taking (d) and s = G*dc into (c) leads
to
V t
A
Gt s dvs KV t s dvs
h
5:38a
With this, the operational form of the spring stiffness becomes:

K V t s d
A
G t s d
h
5:38b
b. Differential operator springs. Here, applying the operator P to both sides of (c)
and using (d) with P[s] = Q[c] on the right-hand side produces
PV t
A
Qvt
h
5:39a
The spring stiffness is only indirectly defined by this expression; but, its
Laplace transform yields a form that is directly comparable to its elastic
counterpart, as
132
K V s
AQs=Ps
h
5:39b
c. Springs for steady-state oscillations. Here, we introduce the complex strain

c* = v*/h, and complex force V*, integrate the constitutive equation (as in
Chap. 4), to obtain the complex modulus G*, and arrive at the relationship:
V jx
A
G jxv jx KV jxv jx
h
5:40a
This defines the complex stiffness of shear springs made of materials of integral
type, as
KV jx
AG jx
h
5:40b
For rheological models, use of V* and v* with (5.39a) and (4.33a) leads to
KN jx
5.5.3

A Qjx
AG jx

L Pjx
h
5:40c
Bending Spring
The idealized model of a bending spring is shown in Fig. 5.12. The pertinent
relations for this case were derived earlier as Eq. (5.7), for constitutive equation of
integral type, and (5.11a, b), for materials of differential type.
a. Hereditary integral springs. The loaddeflection relationship in this case is
obtained by integrating (5.7b) with respect to position, x, along the axis, noting
that M is a function of time only. Setting u(0) = 0, this leads to
L
ut Ct s dM s
I
Taking the convolution of this expression with the tensile relaxation modulus
and rearranging, the following form is obtained, which resembles the elastic
relation:
Fig. 5.12 Bending spring
E, I, L
M(t)
M(t),
M(t)
5.5
133
I
M t Et s dus KM t s dus
L
5:41a
This defines the spring stiffness in operational form as

I
KM t s d Et s d
L
5:41b
b. Differential operator springs. The loaddeflection equation for a viscoelastic

beam of differential type in bending was obtained as (5.11b). Integrating it with
respect to x, noting that M and u are independent of position along the axis,
setting u(0) = 0, and reordering terms leads to
PM t
IQut
L
5:42a
The spring stiffness is only indirectly defined by this expression; but its Laplace
transform yields a form directly comparable to its elastic counterpart, as
K M s
IQs=Ps
L
5:42b
c. Springs for steady-state oscillations. In this case, introduce the complex rotation
u* (jxt) : uoejxt, integrate (5.7a) with respect to x,and effect the convolution
integral with respect to time of as in Chap. 2. Recognizing that now the applied
moment is a complex quantity, M*, and replacing C* with 1/E*, in accordance
with (4.14), leads to the form:
I

jxu jx
M jxt E jxu jxt KM
L
5:43a
This defines the complex stiffness of bending springs of materials of integral

type, as
I

KM
jx E jx
L
5:43b
For rheological models, use of M* and u* with (5.42a) and (4.33a) leads to
KN jx

I Qjx
IE jx

L Pjx
L
5:43c
134
5.5.4
Torsion Spring
The idealized bar-like model of a torsion spring is shown in Fig. 5.13. Equations
(5.19b) and (5.22b) will be used to derive the forcedeflection relationships for
viscoelastic springs of integral and differential types, respectively, as well as those
applicable to steady-state conditions.
a. Hereditary integral springs. The loaddeflection relationship for this case is
obtained by integrating (5.19b) with respect to position along the axis. Noting
that T is independent of the position coordinate, this leads to
o
L
wx; tdx wL; t w0; t CG t s dTs
ox
J
0
ZL
Setting w0 0, for reference, convolving this result with the shear relaxation
modulus, G, and rearranging, produces a form that is reminiscent of the elastic
equation T GJ
L w:
J
T t Gt s dws KT t s dws
L
5:44a
This defines the hereditary operator form of the stiffness of a viscoelastic

torsion spring, as
J
KT t s d Gt s d
L
5:44b
b. Differential operator springs. The loaddeflection equation for the viscoelastic

torsion spring of differential type is obtained integrating (5.22b) with respect to
axial position. This is done noting that T is independent of position, and
operators P and Q are functions of time only, which allows interchanging the
order of integration and differentiation. Proceeding thus, leads to

o
wx; t dx J fQwL; t Qw0; tg
PT tdx PT tL JQ
ox
o
o
ZL
ZL
Fig. 5.13 Torsion spring
G, J, L
T(t)
T(t),
T(t)
(t)
5.5
135
Setting w0; t 0, for reference, leads to the following differential operator

form of the torquetwist relationship for the viscoelastic torsion spring:
PT t
JQwt
L
5:45a
Again, as for other springs, the torsional spring stiffness is only indirectly
defined by this expression. However, taking its Laplace transform and rearranging yields that
Ts
JQs=Ps
ws K T sws
L
This defines the Laplace transform of the viscoelastic torsional spring stiffness:
K T s
JQs=Ps
L
5:45b
c. Springs for steady-state oscillations. Here, as with all other types of spring, the
approach is to introduce the complex twist w* (jxt) : woejxt, and complex
torque T* = Toej(xt+d), into (5.44a) for materials of integral type, and (5.45a),
for those of differential types.
Starting with (5.44a), and proceeding as in Chap. 4, leads to8:
J
T jxt E jxw jxt KT jxw jx
L
5:46a
From this follows the complex stiffness of torsional springs made of materials
of integral type:
J
KT jx G jx
L
5:46b
For rheological models, use of T* and w* with (5.45a) and (4.33) leads to
KT jx

J Qjx
J

G jx
L Pjx
L
5:46c
Example 5.5 The straight viscoelastic beam shown in Fig. 5.14 having length
L and uniform cross section with second moment of area I is subjected to a
sinusoidal concentrated load of frequency x at its midpoint. Determine the
8
Note that, just as E jxt jx F Et, for the uniaxial modulus in tension, the complex
shear modulus is defined by G jxt jx F Gt.
136
F(t)
x
F(t)/2
F(t)/2
complex stiffness of the beam corresponding to lateral deflections of its midpoint,

if the beams material is of differential type.
Solution:
Take the complex form of the lateral deflection as v* (x, jxt) = vo(x)e j(xt+h) into
(5.12b) and integrate it twice with respect x, using the boundary condition v* (0,
x3
F jxt
jxt) = 0, to arrive at the expression: IQjx=Pjxv x; jxt 12

d
v x; jxt xL=2 0, so
Ajxtx. By mechanical as well as geometric symmetry: dx
that Ajxt F jxtL2 =16. Inserting this result in the general solution, evaluating it at x = L/2, and rearranging it, yields the following forcedeflections relationship: F jxt 48IQjxL3=Pjx v L=2; jxt. Hence, the complex stiffness for

mid-span deflections is KL=2
jxt 48IQjxL3=Pjx.
5.6
ElasticViscoelastic Correspondence
All expressions derived previously could have been established directly from their
elastic counterparts, invoking the correspondence between elastic and viscoelastic
relationships introduced in an elementary fashion in Chaps. 2 and 3. According to
this version of the so-called elasticviscoelastic correspondence, replacing each
quantity by its transform and each material constant entering an elastic formula by
its Carson transform9 produces the transform of the corresponding viscoelastic
relationship. The inverse transformation of the latter yields the viscoelastic relation
being sought. By this approach, elastic expressions that do not involve elastic
constants are valid, in exactly the same form, for viscoelastic materials.10 In
applying this method to materials whose constitutive equations are of differential
type, it is important to remember that sM QP (c.f. Chap. 3).
Example 5.6 Use the correspondence principle and the elastic relationship
EIy00 = M, for the deflection y(x) of a beam of Young modulus E and second
moment of area I, subjected to a bending moment M(t), to establish the pertinent
The Carson transform of a function is the transform-variable multiplied transform of the

function.
10
This is true for all materials, irrespective of their constitution.
5.6
ElasticViscoelastic Correspondence
137
relationship for a viscoelastic beam. Assume that the constitutive equation of the
beam may be given in either integral or differential form.
Solution:
Start by writing the Laplace transform of the given expression as sEIy00 M.
Then, reorder it to read: IEsy00 M, which makes it easier to associate with the
transform of the product of E and the time derivative of y00 : Then, apply the inverse
transform to arrive at the desired results:
a. For a material of integral type, the inverse transform gives
IEt s dy00 s Mt, which, as expected, is Eq. (5.6).
b. For a material of differential type, replace sE QP to get I QP y00 M; then,
rewrite it as IQy00 PM and take the inverse Laplace transform to arrive at
expression (10): PM IQy00 :
Example 5.7 Use the elasticviscoelastic correspondence and the elastic relationship: Pcr p2 EI=L2 ; for the critical Euler load of a simply supported elastic
column, of Young modulus E, length L, and second moment of area, I, to establish
the instantaneous and long-term critical loads for a viscoelastic solid of differential
type.
Solution:
Write the elastic expression in transform space as Pcr p2 sEI=L2 and use that
sE QP : Then, invoke the Initial- and Final-Value theorems for the Laplace
transform: lim sf s lim f t; and lim sf s lim f t; to arrive at the results
s!1
t!0
s!0
t!1
(c.f. Appendix A): Pcrg = p2Q(0)/P(0)/L2 : p2Eg/L2 and Pcre = p2Q(?)/P(?)/

L2 : p2Ee/L2.
Example 5.8 Use the elasticviscoelastic correspondence and the elastic relationship, v PL3 =3EI; for the deflection under the point of load application of a
cantilever elastic beam that is subjected to a concentrated load at its free end, to
establish the deflection of a viscoelastic beam of hereditary integral type. In the
elastic expression, E, L, and I represent the Young modulus of the elastic material,
and the length and second moment of area of the beam.
Solution:
Invoke the elasticviscoelastic correspondence to write the transform of the
elastic relationship as v PL3 =3sEI:. Rewrite it to cast it in the more convenient
form: Esv PL3 =3I and take the inverse transform to get Et s dvs
L3
3I
Pt: Convolving this expression with the tensile compliance function CE pro3
duces the desired result: vt L3I CE t s dPs:
138
Fig. 5.15 Viscoelastic

spring-mass system
F* ( j t)
K* ( j )
u* ( j t)
F*
K * u*
m
5.7
Mechanical Vibrations
This section examines the mechanical response of single-degree-of-freedom

springs-mass systems subjected to excitations which are harmonic functions of
time, and in which the springs are made of homogeneous, linear isotropic viscoelastic materials. Attention will be limited to single-degree-of-freedom (SDOF)
systems, consisting of a mass element attached to a single massless spring element
of one of the types described previously. For simplicity and ease of representation,
all derivations are based on the axial spring-mass system depicted in Fig. 5.15 in
terms of complex quantities.
Because an axial spring-mass system is used in the derivations, the equation of
motion
balance of linear momentum,
P is established from* the complex
Fi m
u , assuming that u = uoej(xt-d). As may be seen from Fig. 5.15, this
leads to11
F jxt K jxu jxt x2 mu jxt
5:47
The same approach is applicable to other types of spring-mass systems and

leads to entirely similar expressions. In particular, it should be clear that the
expressions that would result for the shear spring-mass system would be identical
to those of the axial case, except that spring height, h, would replace spring length,
L, and shear, instead of axial relaxation modulus, would enter the equations.
The expressions for bending and torsional spring-mass systems are likewise
similar to those of the axial spring-mass system, except that as follows:
a. Axial deflection is replaced by bending, u, or torsional rotation, w, respectively
or rotational
b. Axial acceleration, u ; is replaced by angular acceleration, w

respectively.
acceleration, w
c. The mass element, m, is replaced by bending, IM, or torsional, JM, mass
moments of inertia, respectively
d. The equation of motion is established from the balance of angular momentum:
11
In general, because the steady-state viscoelastic input and response are out of phase, whenever
the complex-controlled variable is, say, c* = coejxt, the corresponding response variable is of the
form r* = roej(xt+d) where the sign of the phase angle is positive if the response is of stress type,
and negative, if it is of strain type.
5.7
139
G *, J, L
T * ( j t) ,
T * (t)
KT*
; or
Mi IM u

Ti JM u
( j t)
(t)
Example 5.9 Establish the equation governing the torsional vibrations of the
SDOF viscoelastic spring-mass system shown in Fig. 5.16, assuming that the mass
is concentrated in a disk with centroidal polar mass moment of inertia of magnitude JM, and the spring material is of integral type, with shear relaxation modulus G.
Solution:
Take w*(jxt) = woe jxt into Newtons second law for angular motion to obtain
P
. Noting that (j)2 = -1, leads to T K w x2 JM w ; or,
Ti jxt JM w
T
rearranging: x2 JM w KT w T
5.7.1
Forced Vibrations
Proceeding as in Chap. 4, we replace the forcing function F(t) and displacement

u(t) with the complex quantities F*(jxt) : Foejxt, and u* : uoej(xt-h). Inserting
them in (5.47), rearranging, and canceling the common factor, produces

K jx mx2 uo Fo ejh
5:48
In this case, Fo, m, and x, are all real quantities, as are the component K 0 s and
K of K . Using this, we factor K 0 out of the left-hand side of (5.48), use that,
according to (4.7c), and the geometric properties of the axial viscoelastic spring:
00
tand
E00 AE00 =L K 00
0

AE0 =L
E0
K
5:49
Fo =K 0

ejh
1 b2 jtand
5:50
and rearrange the result to read:

uo
where b is newly defined, as
140
b2
mx2
K0
5:51
Since uo is real, the imaginary part of (5.50) must equal zero. This condition
yields the phase angle, h, in terms of b and tand E00 =E0 K 00 =K 0 , as
tan h
tan d
1 b2
Using the definition of the tangent function, this expression can be represented
as in Fig. 5.17:
Fig. 5.17 Geometric
representation of phase angle
(1
) + tan 2
2 2
tan
2
(1 - )
As seen in the figure,

tan d
sin h qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ;
2
1 b2 tan2 d

1 b2
cos h qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ;
2
1 b2 tan2 d
Equating the real parts of (5.50) and using the previous relations for the sine
and cosine functions produces the following expression for the displacement
amplitude, uo, as
Fo =K 0
uo rffihffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiiffiffi

2
1 b2 tan2 d
5:52
Because Fo =K 0 may be interpreted as a static displacement under the load

amplitude Fo, the quantity:
1
AF rffihffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiiffiffi

2
1 b2 tan2 d
5:53
is called the amplification factor, and is one of the three main formulas used in
the analyses and design of viscoelastic spring-mass systems. Since tand 6 0,
expression (5.53) shows that the amplification factor for viscoelastic materials is
never infinite, not even when the frequency ratio, b, equals one.
As mentioned earlier, all the previous expressions are of the same form as those
for a viscously damped elastic spring-mass system. In the present case, the
5.7
141
frequency ratio, b, is defined by (5.51), and the damping factor, 2f b, of the elastic
case [4] is replaced by the loss tangent; that is
2fb tan d
5:54
In this expression, f is the critical damping ratioi.e., the ratio of the systems
actual damping to the critical damping. Critically damped systems do not oscillate.
Also, it is common practice to refer to the real part, K 0 , of the complex stiffness,
K , as dynamic stiffness [5]; and, in analogy with elastic systems, to use the
quantity:
xN
pffiffiffiffiffiffiffiffiffiffiffi
K 0 =m
5:55
as a measure of the springs natural frequency. On this note, we put expression

(5.51) in the form:
b x=xN
5:56
Expression (5.55) is the second formula required in vibration analysis and

design of viscoelastic systems.
The third formula used in vibration analysis concerns the maximum force that is
transmitted to the support of a spring-mass system. This force may be computed
from the expression F* = K*u*, using (5.50) and the fact that u* : uoej(xt-h).
Indeed, the magnitude, F is jF j jK u j jK jju j, which, noting that
jejxth j 1, yields that jF j jK jjuo j. Introducing K 00 K 0 tand, one may write
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
jK j K 0 1 tan2 d, and using this with (5.50) results in:
Fomax
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 tan2 d
rhffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiiffiffi Fo

2
1 b2 tan2 d
5:57
The magnitude of the ratio of the maximum force, Fomax , transmitted to the
support, to the force Fo , that would be transmitted to the support under static
conditions, is the transmission ratio, TR, of the system. The quantity 1 - TR is
called transmissibility. The transmission ratio is given by:

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2

1 tan d
TR rhffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiiffiffi
2

1 b2 tan2 d

5:58
The expressions for the amplification factor and the transmission ratio for
elastic and viscoelastic materials are similar in appearance. However, the damping
and forcing frequency for elastic materials are decoupled; whereas for viscoelastic
142
materials, they are not, because tand, which is responsible for damping in viscoelastic materials, depends on forcing frequency, intrinsically.
The graphs of the amplification factor and the transmission ratio for singledegree-of-freedom rheological models of viscoelastic systems are identical as hose
for viscously damped elastic systems. In them, the systems damping is separated
from the forcing frequency, through the term 2f(x/xN). These graphs, like those
shown in Fig. 5.18, are strictly valid only for rheological models with one dashpot,
like the Maxwell, Kelvin, and standard linear solid and fluid.
6.0
= 0 .0 0
Amplification Factor ( A F)
Fig. 5.18 Amplification

factor for rheological model
with one dashpot
5.0
= 0 .1 0
4.0
3.0
= 0 .2 0
2.0
= 0 .3
30
= 0 .55 0
1.0
0.0
0
Frequency Ratio (
N)
For viscoelastic systems in general, the dependence of tand on frequency and

temperature is as shown in Fig. 5.19 for a lightly damped natural rubber compound. The corresponding amplification factor is presented in Fig. 5.20.
Example 5.10 A rubber compound used in a shock mount application has a loss
tangent tan(d) = 0.05 at a frequency of 6 Hz. What will the maximum expected
Fig. 5.19 Frequency

dependence of the loss
tangent of a lightly damped
natural rubber
0.12
0.10
tan
0.08
0.06
0.04
140 F
77 F
0.02
0.00
0.0
1.0
2.0
3.0
Frequency ratio (
4.0
n)
5.0
Fig. 5.20 Amplification

factor for a lightly damped
natural rubber compound
143
18.0
16.0
Amplification Factor ( A F )
5.7
14.0
12.0
140 F
77 F
10.0
8.0
6.0
) 1/2
4.0
2.0
0.0
0.0
1.0
2.0
3.0
Frequency ratio ( /
4.0
5.0
n)
A
G(t)
h
cmin
displacement of this shock mount be if it is designed to support a package of

weight Wo?
Solution:
According to (5.52), the maximum deflection occurs at resonance, when b = 1.
0
o =K
Since Fo = Wo, this expression yields umax
Ftand
20 Wo =K 0 .
o
Example 5.11 To avoid damage during operation, an electronic instrument is
supported inside its container by two shear mounts, as shown in Fig. 5.21.
The shear relaxation modulus of the rubber used for the mounts may be
represented by G(t) = 0.4 ? 1.40e-t/0.06 MPa, with t measured in seconds. The
mass of the instrument is 100 kg, and each pad has plan area A = 25 cm2 and
thickness h = 1 cm. Determine the minimum vertical clearance needed to avoid
damage to the instrument at an operating frequency of 6 Hz.
Solution:
The vertical clearance needed to avoid the instrument from hitting the bottom
wall of its container must be larger than the sum of the long-term shear deflection, ve,
of the mounts plus the maximum dynamic deflection, vmax
o , during operation. The
long-term shear deflection is given by ve = W/Ke where W = 100 9.81
& 980 N, Ge = 0.4 MPa = 0.4 N/mm2 and Ke = AGe/h : 2(25 102 mm2)
(0.4 N/mm2)/(10 mm) = 200 N/mm. With this, ve = 980/200 mm & 4.9 mm.
Note that this produces a long-term shearing strain of 0.5/1 = 50 %, which is well
below the capability of typical elastomers. Now the maximum dynamic deflection is
144
given by (5.52). This requires K 0 , b = x/xn, and tand K 00 =K 0 . From P.4.1, using
2
xs
x = 2p f = 2p 6 & 37.7 rad/s, one may obtain G0 Ge Gt 1xs
2
xs
1:571 MPa; and G00 Gt 1xs
2 0:518Mpa. This gives tand & 0.330. There0
2
K
fore,K 0 AG
h 785:5N=mm. Using (5.55) with this value gives xn m

3
2
785:510
7855 rad
and b2 = (x/xn)2 & 0.181. Evaluating (5.52) gives the
s
100
ffiffiffiffiffiffiffiffiffiffiffi2ffiffiffiffiffiffiffiffiffiffiffi 1:4 mm. This and the
maximum dynamic deflection: vmax
pffiffiffiffi980=785:5
o
10:181 0:332
long-term deflection yield a minimum vertical clearance of cmin C 0.63 cm. In

practice, a safety factor of 2 is not unusual; leading to cmin & 1.25 cm.
5.7.2
Free Vibrations
This is the type of motion with which a bodyviscoelastic or otherwisewould

respond to a momentary perturbation from its initial position of static equilibrium.
In such case, the response of the system would be due entirely to inertia effects,
and the motion would have to be such that the points where the system is supported (a single point in our spring-mass system) would not move, in addition, all
externally applied forces would be identically zero, except for their temporary
application to start the motion. In formal terms, free vibrations of a viscoelastic
body require zero boundary conditions and zero initial conditions:
u0; t 0; F l; t 0; t [ 0
5:59
The method of Chap. 4 developed for steady-state response to oscillatory

excitations does not apply here. For the present case, however, the equation of
motion F m
u may be written using that e(t) = u(t)/L, and F(t) = Ar(t), in
equation (2.40a). Thus,
mut
AMg
fut Ct s usg 0
L
5:60
Dividing by m, identifying AMg =Lwith the glassy stiffness of the spring-mass

system:
Kg AMg =L
5:61
and introducing the following logical notation for the glassy frequency, xg :
Kg mx2g
5:62
ut x2g ut Ct s us 0
5:63
transforms (5.60) into
5.7
145
Unlike the case of forced vibrations, posed by (5.48), Eq. (5.63) is an integral
equation in the unknown displacement function u(t). The solution to equations of
this type may be sought, for instance, by means of the Laplace transform (see
Appendix A).
5.8
Problems
P. 5. 1. Derive the equation for the defection of a straight viscoelastic beam of

uniform cross section, subjected to the loads shown in Fig. 5.22.
Fig. 5.22 Problem 5.1:
geometry and loading
y
q(x,t)
N(t)
N(t)
x
E(t), L, I
As a matter of terminology, when the axial force is tension, the beam is called a
tie, and when it is compression, beam-column.
Answer: IEt s dvIV x; s N tv00 x; t qx; t
Hint: Establish the equations of force and moment equilibrium of a differential
element of the beam located at an arbitrary axial station, x, as indicated in
Fig. 5.23.
differential beam element
q
M+ M
N
v
N
V+ V
M V
x
d
d
d
This should yield dx
V x; t qx; t and dx
M x; t N dx
vx; t V x; t.
Combine these two expressions and insert (5.6) into the result to obtain the
solution.
P. 5. 2. Determine the instantaneous and long-term deflections of the loaded end
of the beam in Example 5.1.
3
0L
1L
Answer: vL; 0 F3E
; vL; 1 F3E
gI
eI
Hint: The impact and equilibrium responses correspond, respectively, to the
3
limits as t ? 0, and t ? ? of the solution: vL; t L3I C t s dFs, listed
in the example. Take the stated limits and use that C(0) = Cg = 1/Eg and
C(?) = Ce = 1/Ee, for the impact or glassy properties and the long-term or
146
equilibrium properties (c.f. Chap. 2). Substitute these values in the loaddeflection
formulas to arrive at the results.
P. 5. 3. Determine the deflection of the loaded end of the problem given in
Example 5.1, for a material of differential type.
3
3
oL
Answer: QvL; t L3I PFt F3I
PFt
Hint: Insert the expression M(x, t) = -LF(t) ? F(t)x, for the moment at station
x, into (5.10) and integrate the resulting differential equation twice with respect to
x, using the boundary conditions v(0,t) = v(0,t) = 0. Insert the external load,
F(t) = Fo f (t), and evaluate the expression at x = L, to reach the result.
P. 5. 4. Determine the instantaneous and long-term deflections of the loaded end
of the beam in Problem P.5.3, if the beams material is a standard linear
solid.
3
0L
1L
Answer: vL; 0 F3M
; vL; 1 F3M
;
gI
eI
Hint: The impact and equilibrium responses correspond, respectively, to the
3
limits as t ? 0, and t ? ? of the solution QvL; t L3I PFt, obtained in
P.5.2. Since the material parameters have not been specified, assume the forms:
P = po ? p1qt and Q = qo ? q1qt. Take the stated limits and use that the impact
and long-term moduli for the standard linear solid are, respectively, Mg = q1/p1
and Me = qo/po (c.f. Chap. 3). Substitute these values in the loaddeflection formulas to get the results.
P. 5. 5. The pipe with uniform circular cross section shown in Fig. 5.24 is made of
a viscoelastic material having tensile relaxation modulus E(t) and shear
relaxation modulus G(t) = E(t)/3. Find the complex stiffness corresponding to vertical deflections of the free end of the pipe, assuming its
polar moment of area ish J.
i
Answer: K jx 3E jxJ
1
2L3ab 2L3bc 9Lab L2bc
Hint: The vertical deflection of point c is the sum of three components. (a) The
deflection at point b of cantilevered beam ab: vcb F jxtL3ab =3E I. (b) The
deflection of segment bc due to the twist of segment ab under the torque
Lbc F*: vcw Lbc w Lbc MT Lab =G J. (c) The deflection of cantilevered beam
b-c due to load F*(jxt): vcc F L3bc =3E I. Add these three quantities together,
and using that I = J/2, and G = E/3, express the result in the form:
F jxt K jx vc jxt, to obtain K*.
P ( t)
c
5.8
Problems
147
P. 5. 6. The beam of Example 5.1, of length L, and second moment of area I, is

made of a viscoelastic material of Kelvin type having spring stiffness
E and viscosity parameter g. If the concentrated load at its free end is
F(t) = Rt, determine the rotation of the point of load application.

L2
Answer: uL; t 2EI
Rt Rs 1 et=s ; s Eg
Hint: Use (3.16a): r E got e to get P : 1 and Q E got , and insert
o

x;t
ux; t PMIx
, to write EIu0 x; t Ig oto u0 x; t
these into (5.11b): Q ox
Mx; t. Then, evaluate M x; t Ftx FtL, as in Example 5.1, and inte2
grate once with respect to x, to obtain EIux; t Ig oto ux; t Ft x2 F tLx.
Evaluate this expression at x = L to get EIuL; t Ig oto uL; t RL2 t=2. Divide
through by Ig and define s : g/E to arrive at the differential equation:
ou
1
2
t=s
ot s u RL t=2. Use the integrating factor u e ; integrate by parts once and
simplify to arrive at the result.
P. 5. 7. A beam of length L, second moment of area I, built in at one end, and
made of a viscoelastic solid of hereditary integral type, is subjected to a
load F(t) = Fosin(xt) at its free end. Find the steady-state deflection of
the beams tip.
3
oL
C 0 x sinxt C 00 x cosxt
Answer: vL; t F3I
Hint: First, introduce the complex load: F jxt Fo ejxt F , and using
2
(5.12a): M jxt IE jx ddx v x; jxt, with M jxt F x F L, integrate
twice with respect to x, and making use of the boundary conditions:
3
v0 0; t v0; t 0; 8t, to arrive at v L; jxt L3I FE . Use that F
Fo ejxt Fo cosxt jsinxt together with (4.18): 1=E C C0 jC 00 ,
perform the indicated operations and collect terms, to arrive at the result.
P. 5. 8. A circular bar of uniform cross section with polar moment inertia, J, and
length L, which is made of a viscoelastic solid of integral type, is fixed at
one end while its other end is subjected to a cyclic torque T(t) =
Tosin(xt). Find the steady-state rotation of the tip of the bar.
Answer: wt TJo L C0 sinxt C00 cosxt
Hint: Rewrite (5.46a) as w jxt LJ C jxT jxt; then, introduce the complex
torque T jxt To ejxt To cosxt jsinxt, and using (4.14) and (4.18):
1=E C C0 jC 00 , perform the indicated operations and simplify. Since the
torque actually applied corresponds to the imaginary part of the complex torque, the
solution sought is the imaginary part of the result thus obtained.
P. 5. 9. Calculate the energy dissipated per cycle for the bar of Problem P.5.8.
Answer: WV p LJ To2 C 00 x
Hint: Apply (2.37) in the form Wjtp
t
tp
R
t
Tt dw
dt dt, with p = 2p/x using the
rotation wt TJo L C 0 sinxt C00 cosxt calculated in Problem P.5.8, and the
applied torque T(t) = Tosin(xt). Carry out the integration invoking the orthogonality of the sinusoidal functions [see Appendix A] and arrive at the result.
148
P. 5. 10. A simply supported viscoelastic beam of length L has constant cross

section with second moment of area I and supports a concentrated mass
M, at its midpoint. If the concentrated mass is acted on by the cyclic load
F(t) = Focos(xt), and the self-weight of the beam is ignored in comparison, calculate the steady-state oscillation of the attached mass, in
terms of the creep compliance of the beams material.
3
oL
Answer: vt F48I
C 0 x cosxt C 00 x sinxt
Hint: Introduce the complex load: F Fo ejxt Fo cosxt jsinxt, toge
jxt 48IQjxL3=Pjx, and the fact that the
ther with the result of Example 5.5: KL=2

48IE =L3 . With
ratio Qjx=Pjx E jxt to get the complex stiffness KL=2

these, obtain the complex deflection of the concentrated mass as v F =KL=2
.

Substitute the complex forms of F and KL=2 , using (4.14) and (4.18):
1=E C C0 jC 00 ; perform the indicated operation, collect terms, and, noting
that the applied load is the real part of the complex load, select the real part of the
result to arrive at the solution sought.
References
1. E. Volterra, J.H. Gaines, Advanced Strength of Materials (Prentice-Hall, Inc., Englewood
Cliffs, 1971), pp 257268
2. J.T. Oden, Mechanics of Elastic Structures (McGraw-Hill Book Co., New York, 1967),
pp 3033
3. A. Chajes, Principles of Structural Stability Theory (Prentice-Hall, Englewood Cliffs, 1974),
pp 13
4. R. M. Christensen, Theory of viscoelasticity, 2nd edn. (Dover, New York, 1982), pp. 2126
5. A.N. Gent, Engineering with Rubber (Hanser, Munich 2001), pp. 7387
Temperature Effects
Abstract
This chapter examines the simultaneous dependence of material property

functions on time and on temperature. The timetemperature superposition
principle and the concept of timetemperature shifting are introduced first. The
dependence of the glass transition temperature both on the time of measurement
and on surrounding pressure is examined in detail. The integral and differential
constitutive equations are then generalized to include thermal strains and strains
due to changes in humidity. Two ways used in practice to represent the material
property function that accounts for thermal strains are considered: independent
of time and time dependent.
Keywords

Time-temperature superposition Shift Time Frequency Scale Master

Thermorheologically simple WLF Transition Thermal Volume Pressure
Temperature Moisture Hygrothermal
6.1

Introduction
As indicated in Chap. 1, the specific values of some properties, such as the

relaxation modulus, creep compliance, or strain capability of viscoelastic materials, depend on the time and the temperature at which the measurements are taken.
The discussions up to this point have assumed that all required material property
functions were available at the temperature of interest. Temperature was simply
thought of as a parameter rather than as an independent variable on which material
properties depend and was thus omitted from explicit consideration.

149
150
Temperature Effects
This chapter examines the dependence of material property functions on time

and temperature. The timetemperature superposition principle and the concept of
timetemperature shifting, stating that increasing temperature is equivalent to
shortening the timescale of observation, and vice versa, are introduced first. This is
followed by a physical interpretation of the glass transition, where viscoelastic
behavior is most pronounced. The dependence of the glass transition temperature
on time of measurement is pointed out, and its dependence on surrounding pressure
examined in some detail. The hereditary constitutive equations are then generalized
to include thermal strains. Two ways used in practice to represent the material
property function that accounts for thermal strains are considered: independent of
time, as for elastic solids, and time dependent, as other viscoelastic properties.
6.2
Time Temperature Superposition
Several constitutive and failure properties of viscoelastic substancessuch as

relaxation modulus, creep compliance, tensile strength, fracture resistance, and so
onare time and temperature dependent. In other words, the values of some
properties depend on the time and temperature at which the measurements are
taken. This is indicated conceptually in Fig. 6.1 for a generic property, P, such as
the relaxation modulus or the tensile strength of a viscoelastic material [c.f.
Chap. 1]. The same can be said of the dependence on frequency and temperature of
some viscoelastic properties, as depicted in Fig. 6.2, for the real part of the
complex modulus [1].
In principle, timetemperature characterization of a viscoelastic property would
require that the pertinent testsstress relaxation modulus, sustained load to failure, and so onbe carried out at several temperatures in the expected range of
application, and the values be recorded for periods long enough to span that range.
In similar fashion, the frequencytemperature characterization of a viscoelastic
property function would require tests at several constant temperatures in which the
forcing frequency spans the selected range of application.
Such direct approach to characterizing viscoelastic properties is not only timeconsuming and expensive, but also impractical. Fortunately, however, viscoelastic
materials of the type dealt within this textamorphous polymerscan trade either
Fig. 6.1 Typical time and

temperature dependence of a
generic property, P, such as
the relaxation modulus of an
amorphous viscoelastic
polymer (t = time;
T = temperature;
T1 \ T2 \ T3)
T3 T2
t1
T1
log t
t2
t3
6.2
Fig. 6.2 Typical frequency

and temperature dependence
of the storage modulus, M0 , of
an amorphous viscoelastic
polymer (x = frequency;
T = temperature;
T1 \ T2 \ T3)
151
T1
T2
T3
log
time or frequency for temperature and vice versa, which permits reducing or
extending the actual timescale of observation through a corresponding change in
test temperature.
Indeed, some functions describing properties of viscoelastic materials can be
constructed by testing over smaller time intervals and at different temperatures than
used in the direct approach and by shifting the curves describing those functions to
one of the curves obtained at an arbitrarily selected temperature of interest.
Experimental observation confirms that, within reason, a functions so constructed
overlaps the function that would be obtained, were the tests carried out at the
selected temperature and for the duration covered by the composite curve [2].
The process just described is shown schematically in Figs. 6.3 and 6.4. In the
first of these figures are shown four curves of the same property function. Three of
the curves are shorter than the fourth one because the time of observation used to
measure them was shorter. The shorter curves correspond to tests carried out at
temperatures T1, T2, and Tr, while the longer-duration curve was tested at temperature Tr.
In Fig. 6.4, the short curves at temperatures T1 and T2 are shifted to the
right and left, respectively, to overlap the short curve that was tested at the same
temperature as the long curve. As the figure indicates, by judiciously shifting test
data collected at different temperatures, and for adequately short durations, one
can construct the material property function at one of the test temperatures; which
would thus be valid for a longer duration than that of the individual curves. For
obvious reasons, the process described is referred to as timetemperature shifting
or timetemperature superposition.
Fig. 6.3 Typical test data:
viscoelastic property
measured with short-duration
(laboratory timescale) tests at
different temperatures and
with a long-duration test at
one of the temperatures
T2
Tr
T1
log t
152
Fig. 6.4 Time-temperature

shifting procedure.
a Reference temperature, Tr,
is preselected. b All curves
are then shifted toward it
Temperature Effects
T2
Tr
T1
log t
Clearly, the reference temperature, Tr, to which the partial material property
functions are shifted, is arbitrary. Hence, by keeping track of the shiftthe horizontal distance between a curve and that at the reference temperatureone can,
equally well, construct the material property function corresponding to any other
temperature among those used in the tests. Such curves succinctly represent the
material property function for all temperatures and all times in the range of
observation times covered by their extended composite forms; and are thus called
master curves for the properties in question.
To get a feel for the usefulness of timetemperature shifting, consider the data
in Fig. 6.5, which show the results of relaxation modulus testing of a natural
rubber, carried out at three temperatures. As may be seen in the figure, each test
lasted close to 1,000 s (about 20 min); yet, when the data were shiftedin this
case to 22.8 Cthe experimental time of observation at that temperature was
extended to about 1 million seconds (about 11 days), as indicated in Fig. 6.6. In
practical terms, this indicates that, where it is not for the timetemperature
superposition principle, one would have had to test the material at 22.8 C, for
slightly over 11 days to get the same information as obtained by means of just
three 20s duration tests, performed at the three temperatures listed in the figure.
2.0
1.8
Modulus (MPa)
Fig. 6.5 Relaxation

modulus data for a natural
rubber compound tested at
three temperatures
1.6
82.2 C
1.4
22.8 C
1.2
-1.11 C
1.0
0.8
0.6
0.4
0.2
0.0
-2.0
-1.0
0.0
1.0
2.0
logtime (sec)
3.0
4.0
Fig. 6.6 Relaxation

rubber compound tested at
three temperatures and
shifted to 22.8 C
153
2.0
1.8
1.6
Modulus (MPa)
6.2
-1.11C
1.4
22.8C
1.2
82.2C
1.0
0.8
0.6
0.4
0.2
0.0
-8.0
-6.0
-4.0
-2.0
0.0
2.0
4.0
6.0
8.0
log time (sec)

Fig. 6.7 Definition of time
temperature shift function, aT,
as a shift from time at
temperature T, to time at
temperature Tr
P|Tr
T
Tr
log aT
log t|Tr log t|T
log t
In the present case, the principle was actually corroborated by performing two
1,000-h (over 40 days) tests at the reference temperature of 22.8 C.
It is not infrequent for practitioners in the field of viscoelasticity to be requested
by their customers to demonstrate that the timetemperature superposition principle applies to the polymeric materials at hand. In some cases, such as for strategic and tactical missiles and even the Space Shuttle programs, corroboration
tests, lasting several years, on polymeric materials have been carried out.
The timetemperature shift process can be formalized easily. This is done with
reference to Fig. 6.7, which shows two partial curves of a material property
function collected at temperatures, T and Tr, respectively. Without any loss of
generality, Tr may be taken as the reference temperature and the other curve is
shifted toward it.
In Fig. 6.7, the timetemperature shift allows the property, P, measured at time
t and temperature, T, (denoted t|T to emphasize the temperature T) to be established
from test data collected at temperature Tr. In other words, the value sought can be
read off the curve at temperature Tr and time t|Tr. Based on the figure, the passage
from t|T to t|Tr is provided by the shift aT (T, Tr):
logaT T; Tr log tjT log tjTr
6:1
154
Temperature Effects
so that

log tjTr log tjT log tjT log tjTr log tjT logaT T; Tr
6:2a
Or, after rearranging
log tjTr
tjT
log
aT T; Tr

6:2b
By its definition, aT (T = Tr, Tr) = 1. In addition, it is required that its slope, that
d
aT [ 0 [1]. The above
is, its derivative with respect to temperature, be positive: dT
expressions encompass the timetemperature superposition principle, that:
t units of time at temperature T are equivalent to t/aT(T,Tr) units at temperature Tr.
The timetemperature shift function, aT, for a viscoelastic polymer is depicted

on a semi-logarithmic timescale in Fig. 6.8.
Fig. 6.8 Typical form of the
timetemperature shift
function, aT, for amorphous
polymers
Tr
There are several analytical forms that capture the behavior shown in the figure.
Perhaps the most widely used expression to represent the shift function, aT, is the
WLF equation, named after its developers [3]. Save for a different interpretation of
the parameters C1 and C2, entering the equation, which were once thought by the
proponents to be universal constants but are now taken more as curve-fit values,
the WLF equation is expressed as1
logaT T; Tr
C1 T Tr
C2 T Tr
6:3
From the preceding discussion, a master curve is expressed as a function of time

at an arbitrarily selected reference temperature. To obtain the value of the property
at an arbitrary time t and temperature Tin the range of the teststhe shift factor,
aT (T, Tr), is calculated first, and the value being sought of the property function is
read off the master curve at t/aT (Tr), that is,
Note that, for the purposes of converting between different temperature scales, C1 is nondimensional, but C2 has the dimensions of temperature.
6.2
155

Pt; T P
t
; Tr
aT T; Tr

6:4
In particular, application of the timetemperature superposition principle to the

relaxation modulus, M, and creep compliance, C, produces that
t
; Tr
M t; T M
aT T; Tr

C t; T C
t
; Tr
aT T; Tr

6:5

6:6
Also, to simplify notation, or when it is clear from the context, the reference
temperature, Tr, is usually omitted from the shift function so that aT (T) or even aT
is used instead of aT (T,Tr).
The timetemperature superposition principle applies to constitutive functions
expressed in the frequency domain, as well. The analytical form of the principle in
that case can be established using (6.4) in expression (4.3) for the complex
modulus, M* and (4.13), for the complex compliance, C*. For the complex
modulus, this would yield
M jx; T jx
1
Z
Mt=aT ; Tr ejxt dt:
t0
Introducing the change of variables u = t/aT, and regrouping, produces:

M jx; T jxaT
1
Z
Mu; Tr ejxaT u du
6:7a
t0
In other words,
M jx; T M jxaT ; Tr
6:7b
Entirely similar considerations, starting with (4.13), result in

C jx; T C jxaT ; Tr
6:8
Materials for which the timetemperature superposition principle (6.4) holds are
generically called thermorheologically simple [1]. The constitutive properties of
these materials possess master functions, valid for any constant temperature state in
the range of the test data used to establish the shift function. For this reason, if the
relaxation modulus and creep compliance of a thermorheologically simple material
are, respectively, M(t) and C(t), the timetemperature superposition principle, (6.4),
156
Temperature Effects
can be used to write constitutive equations such as (2.1a) and (2.2a), using master
material property functions which are valid for all constant temperature states, as

t
Zt
ts
d
esds
e 0 M
aT T; Tr
a
ds
T;
T
T
r
0
6:9

t
Zt
ts
d
rsds
r0 C
aT T; Tr
a
ds
T;
T
T
r
0
6:10
rt M

et C
Although the timetemperature superposition principle was derived for constant

temperature, it can be formally extended to non-constant temperature states by
requiring the shift function to depend on time and position. This is done by first
ts
splitting the time difference argument aT T;T
of the property function of interest in
r
ts
t
s
the equivalent form: aT T;Tr aT T;Tr aT T;Tr and then introducing the dependence on time and position, as needed. On this basis, shifted time, nt, is generalized
as follows:
nt
Zt
ds
aT Tx; s; Tr
6:11
This notation allows writing constitutive equations for non-constant temperature states in succinct fashion. The constitutive equations for a thermorheologically simple material with master modulus M and master creep compliance C,
equivalent to (6.9) and (6.10), which are applicable to varying temperatures, take
the following forms:
rt M nt e0
Zt
et C nt r0
Zt
d
esds
ds
6:12
d
rsds
ds
6:13
Mnt ns
Cnt ns
Summarizing, the viscoelastic property functions of any thermorheologically

simple material at a prescribed temperature and selected time may be obtained
from the corresponding properties at any other temperature, used as a reference,
dividing the time of interest by the value of the factor for shifting from the selected
temperature to the reference temperature.
The reason why a viscoelastic materialsuch as a solid propellant or a rubber
compoundcan trade time for temperature, and vice versa, lies in its molecular
structure. As pointed out in Chap. 1, a polymeric material is a network of
molecular chains tied to one another by chemical bonds at discrete locations along
their lengthsthe cross-links. The chains are in constant, Brownian motion, which
6.2
157
speeds up when the temperature goes up and slows down when the temperature is
lowered. This type of behavior confers upon the polymer an internal clock with
which it measures external events. At high temperaturesand fast molecular
motionsthe materials internal clock beats fast and interprets actual elapsed time
as being longer than it really is. The reverse is true at low temperatures. The shift
factor introduced above represents the number of clicks of the observers clock per
click of the materials internal clock at the selected temperature [4].
6.3
Phenomenology of the Glass Transition
The glass transition temperature is a very important concept in a variety of

applications of viscoelastic materials. In Chap. 1, it was identified as a temperature, Tg, somewhere in the transition region, where properties change drastically
[see Figs. 1.13 and 1.14]. There are several practical methods to establish Tg. Each
method is based on the fact that the property used in the selected test (heat
capacity, free volume, specific heat, viscosity, loss tangent) changes rather drastically near the glass transition. The method discussed here (calorimetry) is based
on the volumetric thermal expansion of a polymer sample of known initial volume
and mass, subjected to a uniform temperature change. The sample is quenched
quickly to the target temperature and its volume and temperature recorded. The
resulting change in volume per unit of original volume versus temperature is
curve-fitted with two straight lines of different slopes, as shown in Fig. 6.9. The
temperature at which the break in slope occurs is defined as the glass transition
temperature Tg [5].
The total volume, v, of a sample of an amorphous polymer consists of the solid
volume, vo, occupied by the moleculesthis is the occupied volumeand the
interstitial volume, vf, of the spaces between moleculesthe free volume. That is,
v = vo ? vf. The fractional occupied volume, fo , and fractional free volume, ff ,
are, respectively, defined as fo : vo/v and ff : vf,/v.
Figure 6.9 shows the division of the total volume of a sample into occupied
volume and free volume. The volume occupied by the solid molecules is assumed
to increase uniformly with temperature, so that using avo as the coefficient of
thermal expansion for the occupied volume: vo = voi ? avo DT. By contrast, the
free volume remains constant up to Tg, but increases linearly thereafter, due to an
increase in the coefficient of free-volume thermal expansion, avf. That is,
vf
vfg ;

vfg avf T Tg ;
T\Tg
T Tg
6:14
This discontinuity in the coefficient of volumetric thermal expansion is called a

glass transition and the temperature associated with it, glass transition temperature.
The glass transition marks the onset of molecular processes that control viscoelastic behavior. Also, in particular, as indicated by the previous expression, the
158
Fig. 6.9 Glass transition
temperature defined by
thermal volume-change data
Temperature Effects
Free
Volume
vf
vo
Occupied
Volume
voi
Ti
Fig. 6.10 Effect of time of

measurement after quenching
on the value of the glass
transition. (Tg|t1 \ Tg|t2
whenever t1 t2)
Tg
v
t2 << t1
time t 2
time t 1
Tg | t1
Tg | t2
glass transition is a state of constant free volumea so-called iso-free-volume

state.
The slope discontinuity is not as abrupt as suggested by Fig. 6.9, but gradual; as
shown in Fig. 6.10, and extrapolation is usually needed to establish Tg. Also, as the
figure suggests, the glass transition temperature depends on the specific time at
which the measurements are taken, after quick quenching the sample to the target
temperature. In other words, the glass transition temperature is rate sensitive: the
shorter the time to take the measurement, the larger the value obtained for Tg [2].
The physical interpretation of the free-volume behavior of a viscoelastic
polymer is that at low temperature, the sample shrinks and thus restricts the motion
of the polymer chains. This manifests physically as an increase in modulus (the
glassy behavior). By contrast, at high temperature, the sample expands, increasing
free volume and molecular mobility. At the macroscopic level, this is seen as a
reduction in modulus (the rubbery behavior).
6.4
6.4
Effect of Pressure on the Glass Transition Temperature
159
Effect of Pressure on the Glass Transition Temperature
An important experimental observation about the glass transition temperature is

that its value increases with surrounding pressure. This may be explained rather
easily on purely physical grounds. Indeed, any external confining pressure on a
polymer sample will tend to reduce its free volume through squeezing; but because
the glass transition is an iso-free-volume state, the only way the volume can be
reduced is by increasing the glass transition temperatureshifting Tg to the right
in Fig. 6.9. The reduction in free volume due to a pressure change DP : PPr,
where Pr is an arbitrary reference pressure is given by the product Cvf DP; where
Cvf denotes the isothermal compressibility of free volume of the polymer. Adding
the reduction in free volume to (6.14) yields

vf vfg avf T Tg Cvf P Pr
6:15
The magnitude of the effect of a pressure change DP on Tg can now be

established evaluating (6.15) at two pressures P1 and P2 at the corresponding glass
transitions, Tg1 and Tg2, where vf1 = vf2. At that instant,

vf 1 vfg avf T1 Tg1 Cvf P1 Pr
6:15a

vf 2 vfg avf T2 Tg2 Cvf P2 Pr
6:15b
Similarly,
Subtracting (6.15a) from (6.15b), collecting terms, and rearranging produce the
expression:
Tg2 Tg1
Cvf
P2 P1
avf
6:16a
Identifying avf with the difference between the coefficients of thermal expansion
above and below Tg : avf Da; and Cvf with the difference between the compressibility above and below Tg : Cvf DC allows casting (6.16a) in the incremental form:
DTg
DC
DP
Da
6:16b
Although very little data are available in the literature, amorphous polymers
seem to obey this relation [2]. Also, because the WLF equation can be cast using
Tg as reference,2 it turns out that the shift function aT depends on pressure as well.

C TT
The WLF equation was originally proposed as: logaT T; Tg C21TTgg ; implying different
C1 and C2 from those corresponding to the case when Tr = Tg.
160
Temperature Effects
This fact becomes important in applications of polymers subjected to large

hydrostatic pressures, such as in the case of elastomeric bearings. In general, a
change in shift factor changes the timescale and hence the modulus and the
compliance in accordance with the timetemperature superposition principle. This,
in turn, leads to a change in the polymers effective stiffness and consequently a
change in the response of the component.
6.5
Hygrothermal Strains
Deformation of polymers is possible not only under the action of mechanical

loads, but also upon change in temperature and upon absorption of moisture. The
associated strains are referred to as mechanical, eM, thermal, eT, and moisture, eH,
strains, respectively. The latter two are often generically termed hygrothermal
strains [6]. When the strains are small, linear superposition is valid; the total strain,
e, is obtained as the sum of the mechanical and hygrothermal strains. Thus,
et eM t eT t eH t
6:17
Hygrothermal strains do not produce stresses in bodies which are unconstrained

and free to deform and accommodate changes in temperature or moisture.
Mechanical strains, on the other hand, are always accompanied by internal stresses
in the material and are thus called stress-producing strains. Using this and the
appropriate mechanical constitutive equation from those discussed in Chaps. 2, 3,
and 4, it is straightforward to generalize the stressstrain laws of viscoelasticity to
account for hygrothermal strains. This is done by first rewriting Eq. (6.17) to
express the mechanical strain in terms of the total and hygrothermal strains and by
introducing the mechanical constitutive equation at hand. The mechanical strain is
thus
eM t et eT t eH t
6:18
For a viscoelastic material of relaxationintegral type, the constitutive equation

is obtained inserting into the convolution integral form, eM = C * dr, of the strain
stress Eq. (2.26): eM t et eT t eH t Ct s drs. Convolving the
relaxation modulus M, with this expression, using that M*dC = H(t), in accordance with (2.20) or (2.21) yields

rt M t s d es eT s eH s
6:19
For materials of differential operator type, for which, as presented in (3.9c),

P[r] = Q[e], the stressstrain equation generalized for hygrothermal strains is
obtained by applying Q to expression (6.18), replacing Q[eM] with P[r]:
6.5
Hygrothermal Strains
161

Prt Q et eT t eH t
6:20
To complete the formulation, it is necessary to define the manner in which

changes in temperature and moisture are related to hygrothermal strains. The
needed relationships are, quite naturally, constitutive equations. The general
thermomechanical constitutive equations of viscoelasticity can be derived from
rigorous thermodynamic theory [1]. The approach followed here is based on the
observations that
Under a change in temperature from a reference value, a viscoelastic material
deforms in a time-dependent creep-like fashion [7]. Therefore, the thermal
strain may be defined in creep integral form, as
eT t
Zt
0
a t s
d
DT sds;
ds
DT t T t To
6:21a
Quite frequently, for simplicity, a non-hereditary, time-independent coefficient

of thermal expansion is assumed, leading to a thermal straintemperature
constitutive equation of elastic type:
eT t aDT t
DT t T t To
6:21b
Under a change in moisture concentration from a reference value, a viscoelastic

material deforms in a time-independent fashion [6]. As a consequence, the
moisture-induced strain may be defined by and expression of the following
simple form:
eH t bDct Dct ct co
6:22
In this relationship, b is the swelling coefficient, c(t) is the moisture concentration at time t, and co is the reference moisture concentration.
6.6
Problems
1. The three coefficients of the WLF version of the shift function of a natural
rubber are Tr = 22.8 C, C1 & 5.499, C2 = 74.96. What would the coefficients be for this function in degrees Fahrenheit?
Answer: Tr = 73.0 F, C1 & 5.499, C2 = 134.9 F
Hint: By dimensional homogeneity, since aT is dimensionless, the denominator is the sum of C2 and a temperature difference (T - Tr), and the
numerator is the product of C1 and the temperature difference (T - Tr), it
follows that C2 has to have dimensions of temperature (difference) and C1 be
dimensionless. Hence, C2(F) = 9/5 C2(C) & 134.9 F, and Tr(F) = 9/5
Tr(F) ? 32 & 73.0 F.
162
Temperature Effects
2. A flexible element is made of natural rubber whose master tensile stress

p
endurance function is given by a power law of the form: rt A atT
. To
certify the design, a sustained load test has to be performed at a uniform
temperature, T1, and stress level r1. The design will be deemed satisfactory if
the flexible element lasts at least t1 units of time. Since a successful design is
expected to last several years at the operational stress and temperature, the
manufacturer wants to shorten the timescale of the test. Derive an expression
that would allow changing the test duration in a meaningful way.
1=p
rdesign
a Ttest; Tr
Answer: ttest aTTTdesign;
tdesign
Tr
rtest
1=p
Hint: Solve for t from the master curve to get t aT T; Tr Ar
, evaluate this
for the design and desired test conditions, divide one expression by the other
and solve for the duration of the test. In practice, a suitable pair ttest ; Ttest
would be obtained by trial and error.
3. The coefficient of volumetric thermal expansion of a natural rubber compound
is a = 56010-5/ C, and its bulk modulus established at 35 MPa is
K = 1,520 MPa. Use these values as representative of the coefficient of freevolume thermal expansion af and the reciprocal of the isothermal compressibility of free volume, respectively, to estimate the change in glass transition
temperature that could be expected at a pressure of 80 MPa.
Answer: DTg 5:29
C
Hint: Use expression (6.16b) with the approximations Daf a 560
105 =
C and Dcvf 1=K 658 106 =MPa to arrive at the result. Actual
measured values of the ratio Dcvf =Daf for natural rubber are of the order of
0.240 C/MPa [2].
4. A slender bar made of a viscoelastic material of Kelvin type with elastic
modulus E and viscous constant g is held at one end and subjected to a strain
e(t) = eo sin(xt) at its other end. Determine the stress in the bar if the bar is
maintained at a constant thermal change, DT units above the stress-free temperature of the material, and its thermal response is elastic with coefficient of
expansion a.
Answer: rt Eeo sinxt gxeo cosxt EaDT
Hint: Combine (6.20) and (6.21b) with the constitutive equation of the Kelvin
solid given in (3.16a) to write rt E got eo sinxt aDT . Perform the
indicated operations and arrive at the stated result.
5. A slender bar of a viscoelastic material may be idealized as a standard linear
solid. If the bar is held fixed at both ends and subjected to a constant temperature change, DT units above its stress-free temperature determine the stress
that would be developed in the bar as a consequence of the restraint, assuming
elastic thermal expansion.

Answer: rt qo =p1 aDT 1 et=s ; s p1 =p0
Hint: Take the constitutive equation of the standard linear solid in the form
M
M
deM
is the mechanical
given in (3.35b): po r p1 dr
dt qo e q1 dt , where e
6.6
Problems
163
strain and combine it with (6.20) and (6.21b) to arrive at po r p1 dr

dt
d
qo e aDT q1 dt e aDT , in which e is now the total strain, which is
identically zero, because the bar is restrained. Hence, obtain the differential
equation: po r p1 dr
dt qo aDT: Cast this equation in the standard form:
po
qo
dr
t=s
dt p1 r p1 aDT; using the integrating factor u e , where s p1 =p0 ,
integrate the equation to arrive at the result.
References
1. R.M. Christensen, Theory of Viscoelasticity, 2nd edn. (Academic Press, Waltham, Massachusetts, 2003), pp. 9094, 9496, 7787
2. J.J. Aklonis, W.J. MacKnight, Introduction to Polymer Viscoelasticity, (Wiley, Hoboken,
1983), pp. 4447, 6265
3. M.L. Williams, R.F. Landel, J.D. Ferry, Viscoelastic properties of polymers. J. Am. Chem.
Soc. 77, 3701 (1955)
4. A.C. Pipkin, Lectures on Viscoelasticity Theory, (Springer, Berlin, 1972), pp. 98102
5. I.M. Ward, J. Sweeney, An Introduction to the Mechanical Properties of Solid Polymers, 2nd
edn. (Wiley, Hoboken, 2012), pp. 108113
6. S.W. Tsai, H.T. Hahn, Introduction to Composite Materials, (Technomic Pub. Co., Lancaster,
1980), pp. 329344
7. W.G. Knauss, I. Emri, Volume change and the nonlinearly thermo-viscoelastic constitution of
polymers, (Polymer Eng. Sci., 27, 1987), pp. 86100
Material Property Functions and Their

Characterization
Abstract
This chapter examines four topics of practical importance. It begins with an

introduction to material characterization testing, covering stress relaxation,
creep, constant rate, and dynamic tests. The chapter then introduces two types
of analytical forms, typically used to describe mechanical constitutive property
functions. One type, usually referred to as a Dirichlet-Prony series, is expressed
as a finite sum of exponentials; the other form is a power law in time. This
treatment is followed by a discussion of methods of inversion of material
property functions given in Prony series form; both exact and approximate
methods of inversion are presented. The chapter is completed with a discussion
of practical ways to establish the numerical coefficients entering the analytical
forms used to represent the WLF shift relation, the relaxation modulus, and the
creep compliance. The use of a computer application available with the book,
which was specifically developed to obtain the exact convolution inverse of
function in a Prony series form, is also presented and its use is illustrated by
means of some examples.
Keywords

Characterization Test Constitutive Kernel Numerical Inversion

Analytical Parameter Prony Dirichlet Power Law Hereditary
Logarithmic Transition Property function Relaxation Creep Compliance
Complex Laplace Transform Convolution Retardation Time Volterra
Polynomial Equation Roots Shift function Least squares



165
166
7.1
Material Property Functions and Their Characterization
Introduction
This chapter examines four topics of practical importance, beginning with an

introduction to material characterization testing, covering stress relaxation, creep,
constant rate, and dynamic tests. The chapter then introduces two types of analytical forms which are typically used to describe mechanical constitutive property
functions, such as relaxation modulus, creep compliance and coefficient of thermal
expansion. One form is expressed as a finite sum of exponentials, which is frequently referred to as a Dirichlet-Prony series. The other form is a power law
function of time. We then present both approximate and exact methods of finding
the convolution inverse of material property functions of Prony type. The chapter
ends with a discussion of practical ways to establish the numerical coefficients
entering the WLF function and the analytical forms used to represent the relaxation modulus and the creep compliance.
7.2
Experimental Characterization
The objective of experimental characterization is the construction of the kernel

functions that enter the constitutive equations. What follows restricts attention to
the tests used to establish the relaxation modulus and the creep compliance.
Alternatively, either one of these functions may be derived from the other, by
inversion, as discussed in Chap. 2 and in Sect. 7.4. Dynamic test methods used to
establish property functions for steady-state conditions are only briefly mentioned,
because the theory is treated extensively in Chap. 4.
7.2.1
Constant Strain Test
This test is generally known as stress relaxation. In it, a one-dimensional specimen

is subjected to a specified target strain, eo, which is applied as fast as possible, and
then held constant. This test is designed to yield the relaxation modulus,
M(t) = r(t)/eo, as the ratio of the stress response to the constant applied strain. The
theoretical loading is e = eoH(t); the ensuing stress is calculated using Eq. (2.32a):
Rt
rt 0 M t sdes=ds and that de/dt = eod(t), which imply that rt
Mteo : Consequently,
M t
rt
eo
7:1
Quite simply, then, what is required in a stress relaxation test is to keep a record of
the load as a function of time, and then divide that load by the cross-sectional area of
the test specimen and by the enforced target strain. Since total stress relaxation
typically occurs over a long time, and the difference between the short- and long-term
7.2
Fig. 7.1 Relaxation

modulus data collected at five
temperatures for a natural
rubber compound
167
16.0
14.0
82.2 C
48.9 C
22.8 C
-1.11 C
-17.8 C
Modulus (MPa)
12.0
10.0
8.0
6.0
4.0
2.0
0.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
log time (sec)
Fig. 7.2 Mismatch in

response between theory and
experiment in a stress
relaxation test. The
theoretical input is a step
strain but the actual test is a
ramp-and-hold strain
theory
o
loading
t
theory
test
response
t
moduli may be several orders of magnitude, it is convenient to plot the test data on a
semi-logarithmic or even, double-logarithmic scales. In practice, relaxation tests are
performed at five or six different temperatures, each some 15 C apart from its
neighbors, so as to allow a sufficient expansion of the laboratory timescale of
observation. To account for material variability, three to five replications are advisable. Table 7.1 presents relaxation modulus data for a natural rubber compound. The
data were collected at five temperatures and averaged over three replications.
These data are displayed in a semi-logarithmic scale in Fig. 7.1 and will be used
in later sections to demonstrate how to establish the coefficients of the WLF shift
function, as well as of two analytical forms of the relaxation modulus.
In closing, it is pointed out that since it is not possible to apply an instantaneous
strain, as the theory assumes, a discrepancy will always exist between theory and
experimental data. This situation is illustrated in Fig. 7.2.
168
Table 7.1 Relaxation modulus data for a natural rubber compound

Temperature
-17.8 C
-1.11 C
22.8 C
48.9 C
82.2 C
Test time, t (s)
Measured modulus, M(t) (MPa)
0.1
14.25
1.432
0.654
0.499
0.409
0.2
12.00
1.317
0.622
0.489
0.402
0.3
10.84
1.245
0.601
0.482
0.398
0.4
10.07
1.195
0.586
0.475
0.395
0.5
9.46
1.157
0.575
0.471
0.392
0.6
9.02
1.127
0.566
0.467
0.389
0.7
8.67
1.103
0.558
0.463
0.387
0.8
8.37
1.080
0.552
0.460
0.386
0.9
8.12
1.060
0.547
0.458
0.384
1.0
7.89
1.044
0.541
0.455
0.383
2.0
6.60
0.946
0.511
0.440
0.373
3.0
5.97
0.894
0.495
0.432
0.367
4.0
5.60
0.863
0.484
0.425
0.366
5.0
5.34
0.841
0.477
0.421
0.363
6.0
5.12
0.821
0.471
0.417
0.361
7.0
4.96
0.807
0.466
0.415
0.358
8.0
4.81
0.796
0.461
0.412
0.358
9.0
4.69
0.785
0.457
0.410
0.355
10.0
4.58
0.777
0.454
0.408
0.355
20.0
3.91
0.722
0.434
0.395
0.347
30.0
3.55
0.697
0.423
0.389
0.344
40.0
3.31
0.678
0.416
0.384
0.341
50.0
3.14
0.664
0.411
0.381
0.339
60.0
3.00
0.654
0.407
0.378
0.339
70.0
2.88
0.643
0.403
0.376
0.336
80.0
2.79
0.635
0.400
0.374
0.335
90.0
2.71
0.627
0.398
0.372
0.335
100.0
2.64
0.619
0.395
0.371
0.333
200.0
2.20
0.585
0.381
0.363
0.328
300.0
1.97
0.563
0.373
0.358
0.325
400.0
1.82
0.549
0.367
0.355
0.324
500.0
1.72
0.540
0.364
0.353
0.322
600.0
1.63
0.532
0.360
0.350
0.322
(continued)
7.2
169
Table 7.1 (continued)

Temperature
-17.8 C
-1.11 C
Test time, t (s)
Measured modulus, M(t) (MPa)
22.8 C
48.9 C
82.2 C
700.0
1.56
0.526
0.358
0.350
0.320
800.0
1.51
0.515
0.356
0.348
0.319
900.0
1.46
0.511
0.354
0.346
0.319
1,000.0
1.42
0.508
0.352
0.346
0.318
1,500.0
1.27
0.485
0.346
0.342
0.315
2,000.0
1.19
0.458
0.342
0.339
0.315
The source data shown in this table, and elsewhere in the text, may be accessed at
extras.springer.com
7.2.2
Constant Stress Test
This test is generally known as a creep compliance test. In a creep compliance test,
a one-dimensional specimen is subjected to a specified target stress, ro, that is
applied as fast as possible and then held constant. This test is designed to yield the
creep compliance, C(t) = e(t)/ro, as the ratio of the strain response to the constant
applied stress. The loading is r = roH(t); the strain is calculated per Eq. (2.2):
Rt
et 0 C t sdrs=ds, noting that: dr/dt = rod(t), and hence, that:
Rt
et 0 C t sdsds Ctro . Consequently,
C t
et
ro
7:2
Therefore, what is required in a creep compliance test is to keep a record of the

strain response as a function of time and divide it by the applied target stress.
Creep compliance test data are also plotted in double-logarithmic coordinates. A
typical graph of the creep compliance of a viscoelastic material is shown in Fig
1.14. Three analytical forms to represent one or more aspects of creep compliance
functions, which are the counter parts of those used to represent relaxation moduli,
are presented later on.
7.2.3
Constant Rate Test
Although the forcing function in this type of test can be either strain or stress,
strain is commonly used because it is easier to control under laboratory conditions.
In a constant strain rate test, a uniaxial specimen is subjected to the loading is
e(t) = Rt, in which the rate R is a selected constant. This test is designed to
produce the relaxation modulus at the selected temperature, by calculating the
Rt
response using Eq. (2.1) and that de/dt = R, one gets rt R 0 M t sds:
170
Introducing the change in variable u = t - s converts this expression into rt

Rt
R 0 M udu: Differentiating under the integral sign leads to dr
dt RM t: Now,
d
d
given that e(t) = Rt, and thus, dt R de ; results in
M t
7.2.4
dr
de eRt
7:3
Dynamic Tests
Material property functions appropriate for steady-state conditions

M ; M 0 ; M 00 ; C ; C 0 ; C00 ; tan d are obtained by means of dynamic tests.
Dynamic tests are carried out by forcing a sinusoidal excitation of varying frequency on a polymer sample that is held at a given temperature. There are several
methods of dynamic testing [1]. However, most of them target certain types of
deformation, such as torsion or bending, or a specific frequency range, and even
material stiffness.
In principle, any of the quantities associated with the either the complex
modulus jM j; M ; M 0 ; M 00 ; d; tand or complex compliance jC j; C ; C0 ; C 00 ;
d; tandcan be used in dynamic characterization testing. However, the absolute
values, |M*|and |C*|, and loss angle, d, are not typically used.
7.3
Analytical Forms of Constitutive Functions
Two types of analytical forms are examined in what follows which comply with
the fading memory hypothesis and closed cycle condition and, in addition, provide
elastic response under fast and slow processes, as discussed in Chap. 2. One of
these forms, referred to as a Dirichlet-Prony series, is expressed as a finite sum of
decaying exponentials; the other is a power law in time.
7.3.1
Material Property Functions in Prony Series Form
A Prony series is a finite sum of decaying exponentials in time. The individual

decaying exponentials allow modeling of relaxationby adding the exponentialsas well as creepby adding the complements to one, of each exponential
term. The following DirichletProny series is frequently used to represent the
relaxation modulus for viscoelastic solids:
M t M e
N
X
i1
Mi et=si
7:4
7.3
Analytical Forms of Constitutive Functions
171
The parameters si have dimension of time and are usually called time parameters or time constants. In a phenomenological sense, these time constants may be
thought of as characteristic relaxation times of the material, but in general have no
more meaning than that of curve-fit parameters. The coefficients Me and Mi are
real, positive constants.
Expression (7.4) implies the following relationship between the extreme values
of M(t):
M 0 M e
N
X
Mi Mg
7:5
i1
Here, as discussed in Chap. 1:

Me M1
6 0; for viscoelastic solids

0; for viscoelastic fluids
7:6
Another Prony series form frequently used to describe relaxation functions is in

terms of the glassy modulus, Mg, and may be derived from the previous one. Use
(7.5) to express Me in terms of Mg and insert the result into (7.4), to get:
M t M g
N
X
Mi 1 et=si
7:7
i1
By contrast, the creep compliance function increases from its glassy value, Cg,
as its argument increases or, conversely, decreases from its long-term value, Ce, as
t ? 0. With this, the creep compliance counterparts of (7.4) and (7.7) for solids
are, respectively,
C t C e
N
X
Ci et=ki
7:8
Ci 1 et=si
7:9
i1
C t Cg
N
X
i1
For viscoelastic fluids, it is usual to add the term t/g to represent Newtonian
viscous flow.
Just as for the relaxation function, the following expressions are derived from
(7.8) and (7.9); and here, too, the coefficients Ce and Ci, are real and positive:
Cg C0 Ce
N
X
i1
Ci
7:10
172
Ce C 1 Cg
N
X
Ci
7:11
i1
Before leaving this subject, it is important to point out that since the hereditary
form of thermal strains is expressed through a creep-like integral, the thermal
expansion creep function may be modeled using Prony series of the same type as
those in (7.8) or (7.9) [2].
7.3.2
Material Property Functions in Power-Law Form
A power law is an incomplete polynomial of fractional power. A popular power

law form used to represent relaxation functions is as follows:
Mg Me
p
M t M e
1 t=h
7:12
In this expression, the time constant h represents a value somewhere in the

middle of the transition region; the exponent, p, is a positive real value, and the
other symbols are as defined earlier. As can be seen, this form retrieves the glassy
and equilibrium values of the relaxation modulus. The analogous form for the
creep compliance is
Ce Cg
C t Cg
.
^ ^p
1h t
7:13
Here, the notations ^h and p^ for the corresponding time constant and fractional
power are used to emphasize that they are different from those for the relaxation
function.
For mathematical convenience, and in cases when only behavior in the transition region is of relevance, the following simplified forms of (7.12) and (7.13)
are used:
7.4
Mt Mt tp
7:14
Ct Ct t^p
7:15
Inversion of Material Property Functions
Two methods are presented which are especially useful to find the convolution
inverse of relaxation modulus or creep compliance functions when they are
expressed as finite sums exponentials. In both methods, the inverse function is
7.4
173
sought by enforcing the equivalent inversion requirements expressed in (2.20) and

(2.21) that
Z
d
M t s Csds M tCg
ds

0
0
C t s
d
M sds C tMg
ds
d
Csds Ht
ds
7:16a
d
M sds Ht
ds
7:16b
M t s
0
C t s
0
Here, H(t) represents the unit step function [c.f. Appendix A]. The difference
between the inversion methods lies in the manner in which the inversion
requirement is met. The first method is approximate and enforces the inversion
requirement in a least-squares sense in Laplace-transform space. The second
method enforces the requirement in exact form in the time domain.
7.4.1
Approximate Inversion of Material Property Functions
Several approximate methods of Laplace transform inversion have been proposed

in the literature. One of these, introduced by Schapery [3], is presented here,
because it is especially attractive for finding approximate inverse Laplace transform of functions in general and not necessarily related by a convolution integral.
In Schaperys approach, f s is taken as the known Laplace transform of the
function f(t) being sought. In addition, it is assumed that f(t) can be approximated
by a Prony series1:
f t fA t
N
X
Aj et=sj
7:17
j1
In this expression, the sjs are prescribed positive constants,2 and the coefficients Aj
are to be determined in such a manner as to minimize the total squared error
introduced by the approximation, fA(t), in the domain of f(t):
ef ; fA
f t fA t2 dt
This condition is always met by relaxation or creep compliance functions, because a series of
exponentials is complete in that it can represent any continuous function to any desired degree of
accuracy, if enough terms are used in the representation.
2
Typically, the time parameters are more or less arbitrarily taken at each of the several logarithmic cycles spanning the available data, such as at 10-5, 10-4,,103, 104 min, without worrying much about their relationship to any intrinsic response times (relaxation or creep) of the
material in question.
174
Minimization with respect to the coefficients, Ai, demands that oe=oAi 0; for all
i. Upon introducing (7.17) and simplifying, this requirement becomes
Z1
f t fA tet=si dt 0;
i 1; . . .; N
With the definition ofR the Laplace transform of an arbitrary function, f, of expo1
nential order: f s 0 est f tdt, (b) can be expressed as
fA s
f ss1=si
s1=si
Inserting the Laplace transform of the sum of exponentials in (7.17) leads to [c.f.
Appendix A]:
Ak
f ss1=si ;
s 1=sk s1=si
k1
n
X
i 1; . . .; n
7:18
Quite clearly, when all the sis are known, or assumed to be so, (7.18) represents a system of linear equations in the unknown coefficients Ak of the Dirichlet
Prony series. This system may be put in familiar matrix form as follows:
Fik fAk g fbi g;
Fik
1
si
1
;
s1k
bi f ss 1 ;
i k 1; . . .; n
7:19
si
To obtain the convolution inverse of any viscoelastic material property function

using this method, the target property function takes the place of the unknown
function, fA(t), while f(t) represents the known, source function. For the specific
case of relaxation and creep compliance functions, assume for definiteness that the
relaxation modulus, M(t), is the source function, f(t); the creep compliance, C(t), is
s 1=s2 ; yields:
sC
the target fA(t), being sought. Then, inverting (2.31): M
fA t C t L
1
1
2

s Ms

f t
7:20
1

Ms
7:21
In other words:
f s C
s; and f s
A
s2
As was evident in Chap. 3, the time parameters associated with relaxation

functions differ from retardation or creep times. Thus, when dealing with
7.4
175
relaxation moduli and creep compliance functions, the assumption made here, of
utilizing the same time parameters for both the known function and the inverse
function being sought, is not altogether correct. Another disadvantage of the
method is that there is no guarantee that all the Prony coefficients of the inverse
function will be positive, as they should be for real materials. The more general
case, which does not require that the time parameters of the inverse function be
known before hand, and which will yield positive coefficients for the Prony series
of the inverse function, is discussed subsequently.
7.4.2
Exact Inversion of Material Property Functions
In practice, mutually inverse property functions, such as the relaxation modulus

and creep compliance, are obtained from independent tests, from approximate
inverse relationships pertinent to the processes at hand or by enforcing the
inversion requirements (7.16a, b) in one form or another. The approach, here, is to
enforce that requirement exactly.
The two expressions (7.16a, b) may be cast in the standard form of Volterra
integral equations of the first kind in the unknown functions M(t) and C(t),
respectively [c.f. Appendix A]. Given that such equations always admit a solution,
it is possible, at least in principle, to obtain either function from the other.
Although Volterra integral equations of the first kind always admit a solution, it
may not be possible to obtain it exactly. However, when the functions involved are
represented by series of decaying exponentials, the target function is found by
satisfying the inversion requirement exactly.
For ease of computation, the reciprocal relaxation times, ai : 1/si, and the
reciprocal retardation times, bi : 1/ki, will be used. This and the physically
reasonable assumption that the Prony series of both the source and target functions
have the same number of terms, N, transform the Prony series listed in (7.4) and
(7.8) into the following forms:
Mt Me
N
X
Mi eai t
7:22
Cr ebi tr
7:23
i1
Ct Ce
NJ
X
r1
Since, save for a sign, both functions have the same analytical representation, it
does not matter which of the functions is known and which is not. To fix ideas, it is
assumed that the relaxation modulus, M(t), is the one available, and endeavor to
obtain from it the creep compliance, C(t).
176
For clarity, only a few intermediate steps of the lengthy but straightforward
derivation are presented next. Taking (7.22) and (7.23) into (7.16a) leads to
Me
NG
X
Mi e
ai t
Cg
i1
Zt
Me
N
X
Mi eai ts
i1
o
Csds 1
os
Integrating directly the first portion of the integral and by parts the second and
grouping like terms produce
Cg
N
X
Mi eai t Me Cg
i1
Cg
N
X
C r Me
r1
N
X
Mi e
ai t
i1
N
X
ai t
N
X
Cr ebr t Ct
r1
Zt
M i ai e
i1
fCg
N
X
Mi
i1
N
X
b
br s
Cr 1 e
ai s
ge ds 1
r1
Using (7.10) and (7.11), the last integral evaluates to

Zt
fCg
N
X
Cr 1 ebr s geai s ds
r1
N
X
Ce ai t
Cr
e 1
eai br t 1
ai
a

b
i
r
r1
c
Inserting this expression into the previous one, using relations (7.7) to (7.11),
together with (2.22): Mg = 1/Cg and (2.23): Me = 1/Ce, as appropriate, and
collecting like terms yield
N
X
i1
(
ai t
Mi e
N
X
Cr ai
Ce
a br
r1 i
)
N
X
r1
(
br t
Cr e
N
X
M i ai
Mg
a br
i1 i
)
d
Because the exponential functions eai t and ebr t are linearly independent,
expression (d) will be satisfied for all choices of the original function only if the
quantities in braces are independently equal to zero. This condition produces the
equations [4]:
Mg
N
X
Mi b
i1
Ce
br ai
N
X
Cr b
b ai
r1 r
0;
r 1; . . .; N
7:24
0;
i 1; . . .; N
7:25
7.4
177
The first of these relations contains only the time constants, br, as unknowns;
hence, it may be used to establish them. As a function of a single variable, b,
expression (7.24) involves N different terms, each one having (b - ai) in its
denominator. Hence, (7.24) is a polynomial equation, U(b) = 0, of degree N in b.
Ub Mg
N
X
Mi b
i1
b ai
7:26
The N roots of this equation are the characteristic times, br, of the target
function, C(t). Once the roots are known, Eq. (7.25) becomes a system of linear
algebraic equations in the N unknown coefficients, Cr, of the Prony series of the
target function. The solution of this linear system can be obtained with a linear
solver, after it is noted that the equilibrium value of the inverse function is known:
Ce = 1/Me.
The N roots, br, of the polynomial equation may be located by examining the
behavior of the polynomial U(b) as b approaches its extremes (0 and ?), as well
as when it approaches each characteristic time, ai, of the source function. These
considerations reveal that U tends to Mg and Me, as b approaches 0 and ?,
respectively. In addition, U tends to -? and +?, as b approaches ai from above
and from below, respectively. Since U is continuous in each subinterval (ai,ai+1),
the graph of U = 0 must be as depicted in Fig 7.3.
From this graph follows that the time constants ai and bi of the original and
inverse functions are nested. Explicitly,
ai \bi \ai1 ;
for i 1; N 1;
and : bN [ aN
7:27
This result is the mathematical description of the fact that the creep times, ki : 1/bi,
are shorter than the corresponding relaxation times, si : 1/ai. In practical terms,
Fig. 7.3 Behavior of the polynomial equation in the retardation times. M is the source function
and C, the target.
178
()
Ce
Cg
t
1
i+1
i +1
Fig. 7.4 Behavior of the polynomial equation in the relaxation times. C is the source function,
and M, the target
nesting of the time constants allows efficient use of bisection to obtain the characteristic times as the roots of the polynomial equation U(b) = 0.
As indicated before, the roles of functions M and C may be interchanged. If
C were the source function and M the target, (7.25) would be a polynomial
equation, W (a) = 0, in a, while (7.24) would represent the linear system in the
unknown Prony coefficients, Mi. In this case, one has that W (a = 0) = Cg and
W (a ? ?) = Ce. Also, W tends to +? and -?, as a approaches br from above
and from below, respectively, and is continuous in each subinterval (bi, bi+1).
Under these conditions, the graph of W = 0 mirrors that of U = 0, as indicated in
Fig. 7.4.
Example 7.1 The relaxation modulus of a viscoelastic material is given by M t
Me M1 ea1 t : Use the method presented here to find the exact creep compliance.
Solution:
Take the creep compliance in the form C t Ce C1 eb1 t : Although in this
simple case it is straightforward to obtain the coefficients without resorting to the
inversion method, we proceed as in the derivation of the method and insert the
analytical form of the given creep compliance into (7.16a, b) to write
Rt
M tCg 0 Me M1 ea1 ts b1 C1 eb1 s ds 1
We now replace M(t) in the first term by its analytical form, perform the
indicated integration, and collect terms to get

M1 Ce C1 M1 C1

b1
b1
Me C1 eb1 t 0
ea1 t M1 C1
a1 b 1
a 1 b1
Linear independence of the exponential functions implies the vanishing of each of

the expressions in brackets. Setting the expression inside the second bracket to
zero allows solving for the inverse retardation time, b1 Me a1 =Me M1
Me a1 =Mg : Now setting the expression inside the first bracket to zero, using the
7.5
Numerical Characterization of Material Property Functions
179
value of b1 just established and simplifying, produces that C1 M1 Cg =Me ; which

upon using that Cg 1=Mg ; may be expressed as C1 M1 =Me Mg :This is the
same value one would obtain from (7.11) and the condition Cg 1=Mg :
Accompanying this text is a computer application called Prony-Inverse,
which was specifically developed to invert DirichletProny series using the
method presented in this section. Prony-Inverse is available at extras.springer.com, and its use is explained in Sect. 7.6.
7.5
Numerical Characterization of Material Property

Functions
This section describes how to use a standard spreadsheet program to curve-fit

analytical models to material property function data. The models specifically
examined include the WLF equation to fit shift function test data, and the
DirichletProny series and power-law form to fit relaxation modulus and creep
compliance data. In each case, the data provided in Table 7.1 will be used to
exemplify the procedure. As it turns out, judicious treatment of the analytical
forms in each case allows expressing them as multi-linear functions of the
unknown coefficients, which can then be obtained by linear least-squares fitting.
Therefore, in general, we will consider a multi-linear function, z, of the NC variables xk:
z a1 x 1 a2 x 2 a3 x 3 aN C x N C
NC
X
ak x k
7:28
k1
When this expression is evaluated at a data point, (xi, zi), an error, ei, results. Using
(7.28), this error may be expressed as
ei zi
NC
X
ak xki
7:29
k1
The method of least squares is based on minimizing the sum of the squared errors
that result from the approximation. The mathematical condition for the summed
squared error to attain a minimum is the vanishing of the first partial derivatives of
the approximating function with respect to the unknown coefficients, aj:
N
p
o X
e2 0;
oaj i1 i
j 1; . . .; NC
7:30
Using (7.29) with this expression and collecting terms lead to the following
system of linear algebraic equations in the unknown coefficients:
180

Fjk fak g bj ;
Fjk
Np
X
xji xki ;
7:31a
j; k 1; . . .; Nc
j; k 1; . . .; Nc
7:31b
i1
bj
Np
X
xji zi ;
j; k 1; . . .; Nc
7:31c
i1
7.5.1
Numerical Characterization of the Shift Function
The procedure described in Sect. 6.2 to shift viscoelastic property function test
data may be readily implemented as follows:
1. Collect the available time-versus-property test data logti ; Pi at each of
several temperatures, Tk. Use the data provided in Table 7.1 for this purpose.
2. Select the reference temperature, Tr. Without loss of generality, pick
Tr = 22.8 C.
3. For each test temperature, Tk:
a. Provide an initial value of logaTk 0
b. Add the data: logti logaTk :
c. Plot the data: logti logaTk ; Pi : At this time, logaTk 0; so that the
plots correspond with the original data: logti ; Pi :
At this point, the shifted times corresponding to scheme should look as in
Table 7.2 and the corresponding plots, as in Fig. 7.1, given earlier.
4. Starting with the temperature that is closest to Tron either sideguess a
value of logaTk and observe the shift of the curve relative to curve chosen as
reference:
Table 7.2 Sample relaxation modulus test time data for a natural rubber before temperature
shifting
Temperature
i
Tk
log(aTk) =
t (sec)
-17.8
-1.11
22.8
48.9
82.2
log(t/aTk)
1
0.1
-1.00000
-1.00000
-1.00000
-1.00000
-1.00000
0.2
-0.69897
-0.69897
-0.69897
-0.69897
-0.69897
0.3
-0.52288
-0.52288
-0.52288
-0.52288
-0.52288
0.4
-0.39794
-0.39794
-0.39794
-0.39794
-0.39794
7.5
181
a. If the guessed value shifts the curve away from the reference, change the
sign of the guessed value.
b. Otherwise, fine-tune the guess until the shifted curve aligns satisfactorily
with the reference curveor its extension.
5. Repeat step 4 until the data for all available temperatures have been shifted.
This completes the shifting process.
At this stage, the shifted times should look as indicated in Table 7.3, and the
shifted data should look as shown in Fig. 7.5.
Although the pairs of temperature shift values Tk ; logaTk fully characterize
the shift function, it is sometimes possible and convenient to fit an analytical
expression, such as the WLF function to the data. This can be done rather easily by
rewriting (6.3) as
logaT T; Tr T Tr C1 T Tr C2 logaT T; Tr
7:32
Table 7.3 Sample relaxation modulus test time for a natural rubber shifted to construct master
curve
Temperature
i
Tk
log(aTk) =
t (sec)
-17.8
-1.11
22.8
48.9
82.2
6.5
2.6
-1.2
-2.5
log(t/aTk)
1
0.1
-7.50000
-3.60000
-1.00000
0.20000
1.50000
0.2
-7.19897
-3.29897
-0.69897
0.50103
1.80103
0.3
-7.02288
-3.12288
-0.52288
0.67712
1.97712
0.4
-6.89794
-2.99794
-0.39794
0.80206
2.10206
Fig. 7.5 Tensile relaxation

rubber shifted to construct the
master curve at 22.8 C
16
14
Modulus (MPa)
12
10
8
6
4
2
0
-8.0
-6.0
-4.0
-2.0
0.0
2.0
4.0
6.0
log time (sec)

-17.8C
-1.11C
22.8C
48.9C
82.2C
8.0
182
Fig. 7.6 Curve fit of the

WLF shift function to tensile
relaxation modulus data for a
natural rubber compound

8.0
6.0
Curve Fit
4.0
log aT
Test Data
2.0
0.0
C1 5.50
- 2.0
C2 75.0 C
- 4.0
-40
-20
20
40
60
80
100
Temperature ( C)
Introducing the change in variables

z logaT T; Tr T Tr ;
x1 T Tr ;
x2 logaT T; Tr
7:33
leads to the following expression in the standard form (7.28):

z C1 x1 C2 x2
7:34
A linear least-square fit of this expression to the test data provides the coefficients C1 and C2 of the WLF equation. Using (7.31a, b, c) leads to the system:
" P
Np 2
x
PNpi 1i
x
1i x2i
i
# ( P
)
P Np
Np
C
x
x
x
z
1
1i
2i
1i
i
i
PiNp
P
Np 2
C2
i x2i
i x2i zi
7:35
The solution, C1 & 5.499 and C2 = 74.96, at Tr = 22.8 C, is depicted in

Fig. 7.6.
7.5.2
Numerical Characterization of Modulus and Compliance
The analytical expressions used to represent the relaxation modulus, be they of

Dirichlet-Prony series or power-law type, are of the same form as for the creep
compliance. This can be seen by comparing expression (7.4) with (7.8), (7.12) with
(7.13), and (7.14) with (7.15). Without loss of generality, we demonstrate how to
numerically characterize either of these material property functions using
expressions listed for the relaxation modulus only. We address both Prony series
and power law forms.
7.5
183
(a) Functions in PronySeries Form

The Dirichlet-Prony series used to model relaxation and creep compliance,
listed in (7.4) and (7.8), respectively, are of the same following general form:
P t Pe
Nc
X
Pk et=ak
7:36
k1
A practical procedure to fit this expression to test data, when the time parameters are known, is as follows:
1. Shift the test data in accordance with the procedure described in Sect. 7.5.1 and
plot them as in Fig. 7.5. In this case, use the data in Table 7.1, with the shift
values listed in Table 7.2 in the row marked logaTk :
2. Use the plotted data to guess the equilibrium value, Pe. Even though this is not
strictly necessary, it is physically meaningful and convenient to do so.
a. If you make this assumption, rewrite (7.36) as:
P t P e
Nc
X
Pk et=ak
7:37a
k1
and introduce the following change of variables:

z t P t Pe ;
a k Pk ;
xk et=ak ;
k 1; . . .NC
7:37b
A good initial estimate of Pe for the present case is Pe = 0.30 MPa.

b. Otherwise, introduce the following change of variables:
zt Pt;
a0 Pe ;
ak P k ;
xk et=ak ;
k 1; . . .NC
7:37c
With these changes, expression (7.36) is put in the standard form given in (7.28).
3. Apply the least-squares procedure to (7.37a, b, c) and obtain the unknown
coefficients as the solution of the resulting system, which is of the form in
(7.31a, b, c).
Regarding the time parameters, ak, one can either assume, more or less arbitrarily, the NC values required to span the range of the data or make a more
meaningful estimate of their values by trial and error. Although it is easy to do,
arbitrarily picking the time parameters in the range of the test data will often lead
to undesirable oscillations of the predicted property, which are absent from the
data, but are due to the fact that each exponential dominates in the neighborhood
of its time parameter [4]. An easy alternative to minimize this effect is to estimate
the time parameters from a trial-and-error fit of the test data. This can be done per
the following procedure:
184
1. Perform steps 1 and 2 of the previous procedure.

a. Plot the analytical expression with initially arbitrarily selected values of the
time parameters and coefficients, including Pe.
b. Adjust Pe so that the prediction matches the long-term test data.
2. Sequentially guess a time parameter and a corresponding coefficient and adjust
their values trying to bring the predicted curve close to the test data.
This trial-and-error process may be somewhat lengthy, the first couple of times
it is tried. However, the end results are usually ready to use even if optimization of
the coefficients through least squares is not carried out. In addition, unlike the
unconstrained least-squares procedure presented earlier, which would yield the
coefficients that best fit the data, irrespective of their sign, the user has control over
the sign of the coefficientswhich should always be positive, as is required by the
fading memory hypothesis [c.f. Chap. 2]. In the present case, nine exponential
terms and the equilibrium value were used to obtain a reasonably good fit to the
test data, as indicated in Fig. 7.7. The figure also shows each of the exponential
terms, used in the curve fit. The figure clearly shows the region of dominance of
each exponential term.
The curve fit shown in the figure was obtained with the coefficients listed in
Table 7.4 together with the originally suggested equilibrium value Pe = 0.30.
(b) Functions in PowerLaw Form.
As with the Prony series expressions, the power law forms used to model
relaxation and creep compliance listed in (7.12) and (7.13) are of the following
general form:

p
Pt Pe Pt 1 t=h
7:38
100.0
Curve Fit
S#1
S#2
log M (t/a T )
10.0
S#3
S#4
S#5
S#6
1.0
S#7
S#8
S#9
Test Data
0.1
-10
-8
-6
-4
-2
log (t/a T)
Fig. 7.7 DirichletProny series fit to relaxation modulus data of a natural rubber compound
7.5
185
Table 7.4 Trial-and-error curve fit of Prony series to relaxation modulus data of a natural
rubber compound
Pk 16.0
4.3
0.8
2.0
0.75
0.45
0.09
0.15
0.10
ak 3.333.10-8 1.538.10-6 5.000.10-5 2.000.10-5 4.000.10-4 2.857.10-2 6.667.10-1 3.333.100 2.500.103
A practical procedure to fit this expression to test data is as follows:

1. Shift the test data in accordance with the procedure described in Sect. 7.5.1 and
plot them as in Fig. 7.5. The data listed in Table 7.1 are used, after shifting, to
illustrate the procedure.
2. Assume that the equilibrium value, Pe, the transient value, Pt : Pg - Pe, and
the time parameter h are known and recast Eq. (7.38) in the form:

p
Pt Pe Pt 1 t=h
7:39
3. Take the common logarithm of this expression and introduce the notation:
zt logPt Pe ;
x1 1;

x2 log 1 t=h ;
a1 logPt ;
a2 p
7:40a
to transform the original Eq. (7.38) into the standard form (7.28), as
z t a1 x 1 a 2 x 2
7:41a
In practice, it is more advantageous to include the parameter Pt in the trial-anderror process and use the following notation, instead
zt logPt Pe logPt ;

x1 log 1 t=h ;
a1 p
7:40b
to transform the original equation into the standard form as

zt a1 x1
7:41b
4. Make an initial estimate of Pe (and Pt, if needed) from the equilibrium and glassy
values that you think the data would extrapolate to. Take the time parameter h, as
a value near where the data would show the steepest change. These two
(alternatively, three) values will be adjusted during the fitting exercise.
5. Plot Eq. (7.38) with the initial guesses, and least-squares fit the data by means
of (7.31a, b, c). This will yield the values of p and log (Pt) if (7.40a) is used or
p, if (7.41b) is used instead.
6. Iteratively adjust the values of Pe, h, and Pt, if needed, and perform the leastsquares fit until the predictions match the data to your satisfaction.
186
Fig. 7.8 Trial-and-error

least-squares fit of a power
law form to relaxation
modulus data of a natural
rubber compound

16
Relaxation Modulus (MPa)
14
12
Shifted data
Prediction
10
P(t) = Pe +Pt (1+ t/ ) p
Pe = 0.315 MPa , Pt = 24.6 MPa,
p = -0.236,
= 1.8510 -9
4
2
0
-10
-8
-6
-4
-2
log(t/aT) (seconds)
An acceptable solution, such as shown in Fig. 7.8, is obtained in just a couple of

minutes. In the present case, the trial values used in the curve fit shown were
Pe = 0.315 MPa, Pt = 24.6 MPa, and h = 1.8510-9 s; the final least-square
value of the time exponent was p = -0. 236.
Standard linear regression may be used to fit power law functions of the forms
listed in (7.14) and (7.15) to test data, using the following general form:
P t P t t p
7:42
In this case, taking common logarithms and introducing the change in variables
zt log Pt;
a1 logPt ;
x1 1;
a2 p;
x2 logt
7:43
transform the original expression (7.38) into the standard form (7.28) as
z t a1 a2 x 2
7:44
The system of equations presented earlier as (7.35), for the case of two variables, is directly applicable here and takes the following explicit form:
2
6 Np
6
6 Np
4P
x2i
i
9
8
Np
>
>
P
>
>
>

x2i 7
zi >
=
<
a
7 1
i
i
7
Np
Np
>
>
5 a2
P
P
>
>
>
x22i
;
: x2i zi >
Np
P
7:45
7.6
Source Data Files and Computer Application
7.6
187
Source Data Files and Computer Application
Accompanying this text are several files containing the source data used in some of
the examples, as well as a computer application called Prony-Inverse, specifically
developed to obtain the exact convolution inverse of a DirichletProny series.
Both the files and the computer application can be accessed at extras.springer.com. Here, we list the names of the data files, identify their contents,
and provide a guide to using Prony-Inverse.
7.6.1
Source Data Files
The files available at extras.springer.com for this book are as follows:

(1) NR and PBAN data for engineering viscoelasticity.
This is an Excel file containing three worksheets:
(a) Modulus data for NR. This worksheet contains the relaxation modulus data
for a natural rubber compound, shown in Table 7.1 of the text.
(b) C and M for NR. This worksheet contains the Prony coefficients of the
compliance and relaxation functionsin that orderof the natural rubber
compound of Problem P.7.5.
(c) M and C for PBAN. This worksheet contains the Prony coefficients for the
relaxation modulus and creep compliance of the solid propellant of Problem
P.7.4.
(2) PBAN-Modulus.dat. This file contains the information in the form required by
Prony-Inverse to invert the relaxation modulus of the solid propellant of
Problem P.7.4.
(3) NR-Compliance.dat. This file contains the information in the form required by
Prony-Inverse to invert the creep compliance of the natural rubber compound
of Problem P.7.5.
(4) PBAN-Compliance.dat. This file contains the information in the form required
by Prony-Inverse to invert the creep compliance of the solid propellant of
Problem P.7.6.
(5) NR-Modulus.dat. This file contains the information in the form required by
Prony-Inverse to invert the relaxation modulus of the natural rubber compound
of Problem P.7.6.
7.6.2
Computer Application
Prony-Inverse is a computer application developed exclusively to obtain the

convolution inverse of DirichletProny series, by the exact method of inversion
discussed in Sect. 7.4. Prony-Inverse is an interactive code which explains, at run
time, what information is necessary to use it. The application assumes the Prony
series of the source function to be given as in Eq. (7.4):
188
S t Se
N
X
Si et=si
7:46a
i1
in case it is a relaxation modulus and as in Eq. (7.8):

S t Se
N
X
Si et=si
7:46b
i1
if the source is a creep compliance function.

As will be learned during execution of Prony-Inverse, the following information is asked for
1. Name of a data file on which the results will be stored
2. Input option:
(a) 1, if input is manual (through screen)
(b) 2, if input is through a data file
3. If input is manual:
(a) Type of source function (1, if modulus; -1, if compliance)
(b) Number of transient terms, N, and equilibrium value, Se
(c) N pairs of Prony coefficients and time constants, (Si, si)
The Si and si are always positive. The type of source function indicator (1 or 1) is used to assign correct sign during execution.
The sis represent relaxation times if the source is a relaxation modulus and
And retardation times if the source is a creep compliance. In either case, these
time constants must be provided in ascending order of numerical value, to
prevent erroneous solutions.
4. If input is through file:
(a) Name of the file containing the source function data (the data file structure
is as described before)
The accuracy of the inversion procedure hinges on that of the bisection method
used to obtain the roots of the N-degree polynomialeither U(b) = 0 or
W (a) = 0. This may give rise to a slight discrepancy between a source function
that is first used to obtain its inverse and the source function that is obtained by
using as source the inverse that was initially calculated. Bearing this in mind,
provide as accurate a set of values of the parameters of the source function as
possible. This is illustrated in the problems at the end of the section.
7.7
Problems
P.7.1 Making the simple assumption that the relaxation and retardation times are
the same, obtain the creep compliance of a viscoelastic material whose
relaxation modulus in Prony series form is M (t) = 1.00 ? 0.50e-2t.
Answer : Ct 1:00 1=3 e2t
7.7
Problems
189
Hint: Take the creep compliance function in the form C(t) = Ce - C1e-2t, as
per Eq. (7.8), and calculate Mg : M(0) = 1.50, and Me = 1.00. Then, use (2.22):
Mg = 1/Cg and (2.23): Me = 1/Ce, as well as (7.10): Cg = Ce - C1 to obtain
Ce = 1/Me = 1.00, Cg = 1/Mg = 1/1.50, and C1 = 1/3.
P.7.2 Solve Problem P.7.1 using the computer application Prony-Inverse, and
double-check your answer using the results of Example 7.1.
Answer:
Ce = 1.00 kPa,
C1 = 0.333 kPa,
k = 0.75 s,
or
C t 1:00 1=3 e4t=3
Hint: Double-click on the Prony-Inverse application and follow the directions
on the screen, noting that the code expects relaxation times for modulus and
retardation times for creep compliance. Hence, use N = 1, Se = 1.00, S1 = 0.5,
s1 = 1/2 = 0.5, when entering the data. Using that Prony-Inverse returns relaxation or retardation times, leads to the result presented.
P.7.3 Use the answer to Problem P.7.2 as the source input to Prony-Inverse to
obtain the (initial) relaxation modulus.
Answer: Me = 1.00 kPa, M1 = 0.499 kPa, s = 0.500 s
Hint: Proceed as in Problem 7.2, this time using the creep compliance data. The
discrepancy between the initial relaxation modulus data and the solution check is
due to the fact that while the solution to P.7.2 was C1 = 1/3, the approximation
C1 = 0.333 was used for this problem.
P.7.4 The equilibrium modulus of a solid propellant is 81.02, and the coefficients
and time constants of the DirichletProny series form of its relaxation modulus
are as follows:
Mk 2,057.4
sk
1,014
458.4
212.8
98.00
44.88
21.50
7.860
36.78
3.820
1.00.10-5 1.00.10-4 1.00.10-3 1.00.10-2 1.00.10-1 1.00.100 1.00.101 1.00.102 1.00.103 1.00.104
Use Prony-Inverse to obtain the coefficients and time constants of the creep
compliance.
Answer: Ce = 1.2343E-02; with Prony coefficients and time constants given
by
Ck 2.018.10-4 4.562.10-4 8.630.10-4 1.400E.10-3 1.765.10-3
kk
1.923.10
-5
-4
2.034.10
-3
1.929.10
1.757.10
-2
-1
1.528E.10
1.607.10-3 1.152E.10-3 4.844.10-4 3.555E.10-3 6.115E.10-4

1.311.100
1.171.101
1.064.102
1.434.103
1.050.104
Hint: To avoid typographical errors, use a text editor to prepare a data file with
the information given. This file should have the following structure:
1 Source function type (1 = modulus, -1 = compliance)
10, 81.02 Number of transient terms and equilibrium value (Se). Other lines:
(Si, ai)
190
2057.4
1.0E-5
1014.0
1.0E-4
458.4
1.0E-3
212.8
1.0E-2
98.0
1.0E-1
44.88
1.0E+0
21.50
1.0E+1
7.86
1.0E+2
36.78
1.0E+3
3.820
1.0E+4
It is important to note that annotations are only permissible on each data line,
but only following the expected data for the line.
Run Prony-Inverse and select file input mode when prompted. The result should
be as stated.
P.7.5 The equilibrium compliance of a natural rubber is 3.333, and the coefficients
and time constants of the DirichletProny series form of its creep compliance
are as follows:
Ck 6.741.10-2 9.126.10-2 6.979.10-2 2.193.10-1 4.200.10-1 6.357E.10-1 2.295E.10-1 7.263.10-1 8.341E.10-1
kk
9.116.10-8 2.902.10-6 3.072.10-5 8.107.10-5 7.085.10-4 4.862.10-2
7.706.10-1
4.641.100
3.334.103
Use Prony-Inverse to obtain the coefficients and time constants of the relaxation
modulus.
Answer: Me = 0.3000; with Prony coefficients and time constants given by
Mk 6.741.10-2 9.126.10-2 6.979.10-2 2.193.10-1 4.200.10-1 6.357.10-1 2.295.10-1 7.263.10-1 8.341.10-1
sk
9.116.10-8 2.902.10-6 3.072.10-5 8.107.10-5 7.085.10-4 4.862E02
7.706.10-1 4.641.100
3.334.103
Hint: To avoid typographical errors, use a text editor to prepare a data file with
the information given. This file should have the following structure:
-1 Source function type (1 = modulus, 0 = compliance)
7.7
Problems
191
9, 3.3333 Number of transient terms and equilibrium value (Se). Other lines:
(Si, si)
6.7406E-02
9.1159E-08
9.1256E-02
2.9020E-06
6.9792E-02
3.0715E-05
2.1931E-01
8.1073E-05
4.1998E-01
7.0845E-04
6.3568E-01
4.8618E-02
2.2947E-01
7.7063E-01
7.2629E-01
4.6410E+00
8.3405E-01
3.3338E+03
P.7.6 Proceed as in Problem P.7.3 and use Prony-Inverse to double-check the

results obtained in problems P.7.4 and P.7.5.
Hint: For each problem in turn, prepare a Prony-Inverse input file and run the
application; then, compare the results obtained for the inverse to the initial input
given in each case.
References
1. B.E. Read, G.D. Dean, The determination of dynamic properties of polymer composites
(Wiley, London, 1978) pp. 112
2. W.G. Knauss, I. Emri, Volume change and the nonlinearly thermo-viscoelastic constitution of
polymers. Polymer Eng. Sci. 27, 86100 (1987)
3. R.A. Schapery, Approximate methods of transform inversion for viscoelastic stress analysis, in
Proceedings of 4th U.S. national congress of Appl. Mech., 1075 (1962)
4. D. Gutierrez-Lemini, Exact inversion of viscoelastic property functions of exponential type,
JANNAF, JSF, San Diego, CA (2005)
Three-Dimensional Constitutive
Equations
Abstract
This chapter generalizes to three dimensions the one-dimensional viscoelastic

constitutive equations derived in earlier chapters. The concepts of homogeneity,
isotropy, and anisotropy are introduced and the principle of superposition is
used to construct three-dimensional constitutive equations for general anisotropic, orthotropic, and isotropic viscoelastic materials. So-called Poissons
ratios are introduced, and it is shown that uniaxial tensile and shear relaxation
and creep tests suffice to characterize orthotropic viscoelastic solids. A rigorous
treatment extends applicability of the Laplace and Fourier transforms to threedimensional conditions, and constitutive equations in both hereditary integral
form and differential form for compressible and incompressible isotropic solids
are developed and discussed in detailed.
Keywords
Anisotropic
Summation
8.1
Orthotropic Isotropic Symmetry Tensor Matrix

Indicial Transform Spherical Deviatoric Incompressible
Introduction
The one-dimensional viscoelastic constitutive equations derived up to this point in

the book are generalized here to three dimensions. For ease of reference, Sect. 8.2
gives an overview of the notation used, regarding the representation of stresses,
strains, and constitutive properties. The subject matter is then developed in Sects.
8.3 to 8.8.
Section 8.3 introduces the concepts of material homogeneity, as well as isotropy and anisotropy and uses the principle of superposition to construct threedimensional constitutive equations for general anisotropic viscoelastic solids.

193
194
Three-Dimensional Constitutive Equations
Section 8.4 introduces material symmetry planes, on which the direct or normal
stresses decouple from shear strains and vice versa, leading to orthotropic materials and their constitutive equations. Contraction or Poissons ratios are introduced also, and it is shown that uniaxial tensile and shear relaxation and creep tests
suffice to characterize orthotropic viscoelastic solids. Using the argument that the
constitutive tensors defining the three-dimensional stressstrain equations are
ordered arrays of scalar relaxation or creep functions, Sect. 8.5 extends the
applicability of the Laplace and Fourier transforms to three-dimensional
conditions.
Sections 8.6, 8.7 and 8.8 are devoted to constitutive equations for isotropic
viscoelastic solids. Sections 8.6 and 8.7 develop constitutive equations in hereditary integral form, applicable to compressible and incompressible solids,
respectively. Section 8.8, on the other hand, treats compressible and incompressible isotropic viscoelastic solids of differential type.
8.2
Background and Notation
Three types of notation are used in this book: symbolic notation, indicial tensor
notation, and matrix notation. In symbolic notation, the same symbols used in the
one-dimensional case are used to represent the corresponding tensors. Therefore,
the one-, two-, and three-dimensional equationsconstitutive and otherwise
look exactly alike, and this applies to the Laplace and Fourier transforms too.
In indicial tensor notation, tensors, as ordered arrays of quantities, are denoted
by subscripted symbols, such as Aijk or Br. The number of non-repeated indices,
also called free indices, denotes the order of the tensor, which is related to the
number of elements it has. In three-dimensional Euclidean space, a tensor with
p free indices has 3p elements and is said to be a tensor of order p. For instance,
Mijkl would represent a tensor of fourth order, having 34 = 81 components, while
Dij would represent a tensor of second order, with 32 = 9 components; Aii, having
no free indices, would represent a tensor of order zero, also called a scalar
(c.f. Appendix B).
Just like matrices, tensors of the same orderwith the same number of free
indicesmay be added together, term by term, to yield a tensor of the same order
as the addends. Thus, for instance, Dij ? rij = Aij, etc. Tensors can also feature
repeated indices. However, no index can be repeated more than once in a tensor.
Thus, Aijj, Mijkk, Cii, and Miikk are valid tensors, but Ajjj and Mikkk are not.
Also like matrices, tensors of the same order can be multiplied together. This is
done according to the summation convention: when an index appears twice in a
tensor, a summation is implied over all the terms that are obtained by letting that
index assume all its possible values, unless explicitly stated otherwise [1]. This
convention is suspended if an index appears three or more times in a term or a
tensor. According to the summation convention then, the explicit meaning of the
tensor expressions AiBi and Aijxj, for instance, is
8.2
Background and Notation
Ai Bi A1 B1 A2 B2 A3 B3 ; Aij xj Ai1 x1 Ai2 x2 Ai3 x3 ;
195
j 1; 2; 3
As a consequence of the summation convention, the presence of a repeated

index reduces the order of a tensor by two. For instance, setting k = l in the fourthorder tensor Mijkl leads to Mijkk = Mij11 ? Mij22 ? Mij33, which is a sum of three
second-order tensors.
The symbol dij is reserved for a special second-order tensor, with the properties
of the identity matrix: dij = 1, if i = j; and dij = 0, if i = j. This tensor is also
usually called the Kronecker delta and, on occasion, the substitution operator. The
latter name is due to its property that, for example, Aijdjk = Ai1d1k ? Ai2d2k ? Ai3d3k, which, based on the definition of the Kronecker delta, evaluates to
either Ai1, Ai2, or Ai3, depending on the value of k. Thus, Aijdjk = Aij, where
dummy index j has been substituted with free index k.
All quantities entering the one-dimensional constitutive equations, that is the
stress, the strain, and the material property functions relating themrelaxation
modulus and creep complianceare considered scalar quantities, each fully
described by a single entity. In three dimensions, the state of stress at a point
requires an ordered set of nine functions to be fully specified and similarly for the
state of strain at a point. The ordered collection of nine functions forms the stress or
the strain tensor. The nine components of the stress and strain tensors are arranged
in 3 9 3 arrays, or matrices with entries rij(t) and eij(t), respectively, as follows:
2
r11 t
rij t 4 r21 t
r31 t
3
2
r12 t r13 t
e11 t e12 t
r22 t r23 t 5 eij t 4 e21 t e22 t
e31 t e32 t
r32 t r33 t
3
e13 t
e23 t 5
e33 t
In this notation, the components of the stress tensor that are listed on anyone of
its three rows denote the components of a traction vector acting on a surface whose
outward unit normally points in the direction of the axis represented by the index
of the rowthat is, by the first index of the tensor component. The second index
indicates the axis or direction in which that stress component acts. The sign
convention for the components of the stress tensors is presented in Fig. 8.1.
According to it, a stress component may act on either a positive plane or a negative
plane at a material point. A stress tensor component on a positive plane is positive
if it acts in the positive direction of a coordinate axis. A stress component on a
negative plane is positive if it acts in a negative coordinate direction.
In somewhat similar fashion, the components of the strain tensor, which are
listed on the main diagonal of the matrix of the strain tensor, measure the change
in length of a material element lying along the direction of the axis represented by
the first index, per unit length of the material element along the axis represented by
the second index which, being on the main diagonal, is the same as the first. The
off-diagonal elements in the strain tensor measure the decrease in angle of a
material plane with edges parallel to the axis represented by the two indices of the
strain tensor component [2].
196
Fig. 8.1 Sign convention for

the components of the stress
tensor
X3
Positive face 3
33
31
32
23
22
13
12
X2
21
11
Positive face 2
X1
Positive face 1
Explicit matrix notation is very similar to indicial tensor notation, including the
use of the summation convention. The difference is that the stress and strain
tensors are represented as extended (9-by-1) vectors, and the elements of the
constitutive tensors are arranged in 9-by-9 square matrices.
The reader is assumed familiar with the usual rules of operation with matrices:
addition, multiplication, transposition, and inversion. Regarding differentiation
and integration, which enter viscoelastic expressions, suffice it to say that they are
defined in the same fashion for matrices and tensors, so that the integral of a
matrix is the matrix of the integrals of its elements and the derivative of a matrix
is the matrix of the derivatives of its elements.
8.3
Constitutive Equations for Anisotropic Materials
Material property functions, such as relaxation modulus or creep coefficient of

thermal expansion, may depend on the orientation the test specimen had before it
was extracted from the parent material and even on the position it occupied in the
bulk. If the value of a material property function is the same, within acceptable
material variability, when the test specimen is extracted in the same orientation,
but from different positions, such as (a1) to (a3) in Fig. 8.2, the material is said to
be homogeneous with respect to the property measured. If that is not the case, the
material is said to be inhomogeneous. Additionally, if the value of a material
property function does not depend on the orientation that the test article had in the
bulksuch as (b) or (c) in Fig. 8.2, the material is said to be isotropic with respect
to the property function being measured. Otherwise, the material is said to be
anisotropic. Only homogeneous materials are considered here.
8.3
197
Fig. 8.2 Dependence of

material properties on
position and orientation
(a2)
(b)
(a1)
(c)
(a3)
The approach followed to derive the three-dimensional constitutive equations

for general anisotropic materials is essentially identical to the one-dimensional
case, except that for anisotropic materials, attention has to be paid to the orientation of the stresses and strains relative to the parent material itself.
The principle of superposition, which applies to any linear material, requires
that each component of stress must be linearly related to each of the strain components and vice versa. Letting rij tekl represent the stress rij due to strain ekl
acting alone; similarly for eij t , the principle of superposition may be expressed
rkl
in either of the following forms:

rij t
3 X
3
X
rij tekl ;
i; j 1; 3
8:1
eij trkl ;
i; j 1; 3
8:2
k1 l1
eij t
3 X
3
X
k1 l1
Each term in these expressions relates one single stress or strain component to
one single strain or stress component. In other words, each term in each linear sum
in (8.1) and (8.2) is a one-dimensional stressstrain (strainstress) viscoelastic
relationship. Denoting by Mijkl(t), the scalar modulus-type function giving the
stress response rij to a strain input ekl and Cijkl(t) to represent the scalar compliance-like function for strain response eij to a stress input rkl allows each term in
(8.1) and (8.2) to be represented by mutually independent one-dimensional
expressions such as r11 je12 M1112 de12 or e33 jr11 C3311 dr11 , .
Using indicial tensor notation, the sums (8.1) and (8.2), which represent the
three-dimensional constitutive equations of a general anisotropic viscoelastic
material, may be cast as
rij t Mijkl t s dekl s
Zt
0
eij t Cijkl t s drkl s
Zt
0
d
ekl sds
ds
8:3
d
rkl sds
ds
8:4
Mijkl t s
Cijkl t s
198
Before proceeding, it must be pointed out that although each material property
function in the sets Mijkl and Cijkl is a scalar function relating a single component
of stress (strain) to a single component of strain (stress), the individual functions in
M and C are generally not the same as the moduli or compliances of the onedimensional theory. The relationships between the two sets of functions will be
made clear later on.
Since each of the four indices in these expressions ranges from 1 to 3, Mijkl(t) and
Cijkl(t) each contain 3333 = 34 = 81 scalar relaxation and compliance functions,
respectively. These two ordered sets of 81 scalar property functions each are the
fourth-order material property tensors for so-called polar anisotropic viscoelastic
materials, for which the stress and strain tensors are non-symmetric [5]. For
non-polar materialswhich are the typical engineering materials considered in this
textthe stress and strain tensors are symmetric and each contains only 6 independent components. Hence, the constitutive tensors M and C for general non-polar
anisotropic materials are symmetric in their first and second pairs of indices,
respectively, which indicate that the number of material property functions of a
general anisotropic material is, at most, 36.1
Switching to matrix notation for simplicity, the relaxation moduli and the creep
compliances for anisotropic viscoelastic materials become the 6-by-6 matrices
[Mij] and [Cij], or [M] and [C], for short. Expressions (8.3) and (8.4) then take the
equivalent forms [3]:
ri Mij t s dej s;
i j 1; 6
8:5
ei Cij t s drj s;
i j 1; 6
8:6
In these expressions, the 6 independent components of the stress and strain tensors
are arranged in the 6-by-1 extended column vectors2:{ri}T : {r}T : {r11, r22,
r33, r12, r13, r23} and {ei}T : {e}T : {e11, e22, e33, e12, e13}. And clearly, as in
the one-dimensional case, combining (8.5) and (8.6) proves that the moduli and
compliance matrices are convolution inverses of each other3:
Mik t s dCkj s Cik t s dMkj s Htdij
8:7
Now, the work per unit volume is ri dei Mij dej dei dei Mij dej . After
interchanging the dummy indices, i $ j, on the right-hand side and reordering,
1
For each pair of indices, ij and kl, ranging from 1 to 3, the number of independent components
is 34/2 = 6.
2
In matrix notation, the superscript T is used to denote the transpose of the matrix it is
appended to.
3
The equivalent tensor expression, derived combining (8.3) and (8.4), is Mijkl dCklpq Htdip djq
8.3
199
leads to Mij dej dei Mji dei dej Mji dej dei , which implies Mij Mji .
In other words, the 6-by-6 material property matrix, M, is symmetric, requiring
only 6 9 7/2 = 21, independent coefficients. The same can be said of the matrix
of compliances, Cij. From all these arguments follow that the number of independent material property functions of a general non-polar anisotropic viscoelastic
material is 21.
8.4
Constitutive Equations for Orthotropic Materials
By definition, the mechanical response of anisotropic materials is always coupled

in that their response to any action will generally involve stretching, contraction,
and shearing. Accordingly, 21 independent tests would be required to characterize
linear anisotropic materials, viscoelastic or otherwise.
A material symmetry plane is one defined by two perpendicular directions
along which the direct stresses decouple from shearing in the plane and vice versa.
The directions along which decoupling occurs are termed material principal
directions.4 By definition, then, a direct pull in a principal material direction
produces a stretch in that direction and, quite generally, a contraction in the
material principal directions that are perpendicular to the direction of pull. This
deformation, however, occurs without shearing in the material plane of symmetry.
An orthotropic material is defined as one with three material symmetry planes,
meaning that in these planes, the three direct stresses (strains) decouple from the
three shear strains (stresses). This reduces the number of independent constitutive
coefficients by 6. Also, since shear stresses (strains) applied on a material principal
plane decouple from other shear strains (stresses), the constitutive equations for
shear involve only shear terms of the same type and decouple completely from the
rest. This reduces the number of independent constitutive functions by an additional 3 per shear component, to 9 overall. In explicit matrix notation, omitting the
time arguments, for conciseness, the constitutive equations for general orthotropic
viscoelastic materials take the form:
8 9 2
M11
< r1 =
r2 4 M12
: ;
8 r3 9 2 M13
M44
< r4 =
r5 4 0
: ;
0
r6
M12
M22
M23
0
M55
0
3 8 9
M13
< e1 =
M23 5 d e2 ;
: ;
M33
e
3 8 39
0
< e4 =
0 5 d e5
: ;
M66
e6
8:8
Material principal directions are not to be confused with the principal directions of stress or the
principal directions of strain, the latter so-named because along them the stress and, respectively,
the strain attain their extreme numerical values.
200
8 9 2
C11
< e1 =
e2 4 C12
: ;
8 e3 9 2 C13
C44
< e4 =
e5 4 0
: ;
0
e6
C12
C22
C23
0
C55
0
3 8 9
C13
< r1 =
C23 5 d r2 ;
: ;
C33 3 8 r3 9
0
< r4 =
0 5 d r5
: ;
C66
r6
8:9
As pointed out previously, the relaxation functions Mij are not in general the
one-dimensional relaxation moduli in the indicated directions, except for shear
response. In this case, for instance, r12 : r4 = M44*de4 : 2G12*de12 and so
M44 = 2G12. In similar fashion, M55 = 2G13 and M66 = 2G23. The diagonal form
of the constitutive equations for shear, together with (8.7), implies that the shear
compliances C44 to C66 are convolution inverses of M44 to M66, respectively.
Clearly, standard shear relaxation or shear creep tests performed in material
principal planes would suffice to establish the three material property functions in
shear.
The uniaxial relaxation modulus in each material principal direction may be
related to the components of the creep compliance matrix for direct strainstress
response. This is done using (8.9) to evaluate the strain response to each of three
separate uniaxial states of stress: {r1 = 0, r2 = r3 = 0}, {r2 = 0,
r1 = r3 = 0}, and {r3 = 0, r1 = r2 = 0} and noting the analogy with the
uniaxial conditions.
Using the first state of stress, (8.9) produces e1 = C11*dr1, e2 = C12*dr1,
1
de1 E11 de1 , where
e3 = C13*dr1. The first of these relations gives r1 C11
1
E11 is the uniaxial (tensile) relaxation modulus. In other words, E11 C11
.
Inserting r1 = E11*de1 in the second and third relationships produces
e2 = C12*dE11* de1 and e3 = C13*dE11* de1. The negative of the functions
C12*dE11 and C13*dE11 are given the special symbols v12 and v13, respectively.
Then, e2 and e3 are simply written as e2(t) = C12*dE11*de1 : -v12*de1 and
e3(t) = C13*dE11*de1 : -v13*de1. From these expressions follow that C12
1
1
E11
dv12 and C13 E11
dv13 . These newly introduced functions are called
contraction ratios or, more generally, Poissons ratios.
Exactly the same arguments can be followed with the other two states of direct
stress listed before. Hence, collecting the findings for the pure shear and pure
direct states of stress, the following relationships are obtained:
1
1
1
t; C22 t E22
t; C33 t E33
t
C11 t E11
8:10a
1
1
1
dv12 ; C13 E11
dv13 ; C23 E22
dv23
C12 E11
8:10b
1
1
1
C44 t G1
t; C55 t G1
t; C66 t G1
t
2 12
2 13
2 23
8:10c
8.4
Constitutive Equations for OrthotropicOrthotropic Materials
Fig. 8.3 Expected time

dependence of Poissons ratio
for orthotropic viscoelastic
solids
201
vij(t)
log(t)
Also, had the functions C21, C31, and C32 been used in the derivations, instead of
C12, C13, and C23, the following would have resulted:
1
1
1
C21 E22
dv21 ; C31 E33
dv31 ; C32 E33
dv32
8:11
Because of the symmetry of the compliance matrix: Cij = Cji, (8.11) and
1
1
dv12 E22
dv21 or, equivalently, due to the commu(8.10b) imply that E11
tative property of the Stieltjes convolution (see Appendix A)
v12 dE22 v21 dE11 . Since the same is true for the other Poissons ratios, we
see that just as for orthotropic elastic materials, suspending the summation convention,5 these relations can be expressed as [4]:
vij dEjj vji dEii ; no sum on i or j;
i; j 1; 3; i 6 j
8:12
Using (8.10a) to (8.12) with (8.9) leads to the following matrix of compliances for
orthotropic viscoelastic solids in terms of tensile and shear relaxation moduli and
Poissons ratios:
8 9 2
1
E11
< e11 =
1
4
e22 E11 dv12
: ;
1
e33
E11
dv13
1
E11
dv12
1
E22
1
E22
dv23
9
3 8
1
E11
dv13
< r11 =
1
E22
dv23 5 d r22
;
:
1
r33
E33
8 9 8 1 1
9
< e12 = < 2 G12 dr12 =
1 G1 dr13
e
: 13 ; : 21 13
;
1
e23
2 G23 dr23
8:13a
8:13b
From their definitions, each Poissons ratio functions, v12, v13 and v23, relates a
strain response to a strain caused by a single stress component. Their functional
behavior should therefore be like that of a compliance function, as indicated in
Fig. 8.3.
In indicial tensor notation, the summation convention is typically suspended by adding an

underscore to the indices excluded from the summation.
202
In practical applications, the Poissons functions, vij(t) of an orthotropic viscoelastic material are established by subjecting a slender test specimen to a step
strain ei(t) = eoiH(t) and by monitoring the strain response, ej(t), in material
principal direction j; according to its definition
ej t vij dei
Zt
vij t s
o
d
ei sds
ds
8:14
Inserting the step strain history input and rearranging yield the practical
expression:
vij t
8.5
ej t
eoi
8:15
Constitutive Equations in Integral Transform Space
Because the constitutive tensors and associated matrices are ordered collections of
scalar functions, any linear transformation of them will yield tensors or matrices of
transforms. This allows writing the Laplace and Fourier transforms of the constitutive equations of anisotropic viscoelastic materials directly, as the tensors or
matrices of the corresponding transforms of the individual material property
functions. Thus, for instance, expressions (8.3) and (8.4) in Laplace-transformed
space become
ijkl sekl s
ij s sM
r
8:16
ijkl s
eij s sC
rkl s
8:17
In like fashion, the Fourier transform of the same expressions read

ijkl jxekl jx
ij jx jx M
r
8:18a
ijkl jx
eij jx jx C
rkl jx
8:19a
Equivalently, in terms of the complex stress, strain, modulus, and compliance, as is

more customary in dealing with steady-state conditions (see Chap. 4)

rij jx Mijkl
jxekl jx
8:18b

jxrkl jx
eij jx Cijkl
8:19b
According to the derivations in this section, the results in (8.16) to (8.19b) apply to
each term of a modulus or compliance matrix.
8.5
Constitutive Equations in Integral TransformTransform Space
203
1
Example 8.1 Using the quantity e11 tjr22 E11
dv12 dr2 , appearing as one
of the addends in the strainstress constitutive equation of an orthotropic viscoelastic solid, according to (8.13a) demonstrate by direct evaluation of the con
volution integrals involved, that for steady-state loading, e11 jxtr
22

1
E11
jx v12 jx r2 jxt; in accordance with the developments of this
section.
Solution:
1
dv12 dr2 to the cyclic
The steady-state strain response e11 tjr22 E11
stress r2 t ro2 sinxt is obtained as the imaginary component of the complex
stress r2 jxt ro2 ejxt , as explained in Chap. 4. Hence, integrating the given
strain component of the constitutive relation from -? to t, using the complex
input, leads to
e11 jxtr
22
Zt
1
E11
t s
s1
Zt
jxro2
d Zs
d
v12 s s ro2 ejxs ds
ds s1
ds
1
E11
t s
s1
d Zs
v12 s sejxs ds
ds s1
Introducing the change in variable u = s - s, using ds = -du, to invert the

resulting limits of integration and taking the derivative of the inner integral with
respect to s:
e11 jxtr jxro2

22
jxro2
Zt
s1
Zt
1
E11
t s
d jxs 1
Z
e
v12 uejxu du
ds
u0
1
E11
t sejxs jx
s1
Defining v12 jx jx
1
R
1
Z
v12 uejxu du
u0
v12 uejxu du, changing variables of integration on the
s0
outer integral, to z = t - s, and inverting the limits of integration, as before, yield

"
1
E11
dv12 dr2 jx
1
Z
#
1
E11
zejxz dz
v12 jxro2 ejxt
z0
1

Recognizing that the quantity in brackets is the complex compliance E11
jx

1
and r jx ro2 ejxt leads to e jxt E jx v jx r jxt. By
22
11
r22
11
12
22
the arbitrariness of the term selected, the same is true of any other component of
the compliance matrix and similarly for the matrix of moduli.
204
8.6
Integral Constitutive Equations for Isotropic Materials
There is a very important and large class of materials, which are called isotropic,
whose material property functions are the same irrespective of the orientation in
which the test specimens used to establish them are extracted from the bulk. The
mechanical response of isotropic materials is fully described by means of two
independent material property functions. This is so for the following reasons [5]:
1. The most general isotropic representation of a fourth-order tensor, such as M or
C, can be expressed in terms of only three independent scalar property
functions
Mijkl t kt dij dkl dik djl lt dik djl dil djk
mt dil djk dik djl
8:20
2. The stress and strain tensors are symmetric: rij = rji and eij = eji. Using this
and that the repeated indices k and l are dummy and thus interchangeable
omitting the time dependence for conciseness:
rij Mijkl dekl Mijlk delk rji Mjilk delk
This implies that

Mijkl Mijlk Mjilk
The rate of work, W rij deij , is a scalar quantity, so that using the associative property of convolution integrals, interchanging dummy indices, and
invoking the commutative property of convolutions, in that order:
W Mijkl dekl deij Mklij dekl deij
From these two expressions follow the symmetry relationships satisfied by the
relaxation and compliance material property tensors, M and C:
Mijkl t Mijlk t Mjikl t Mklij t
8:21
Cijkl t Cijlk t Cjikl t Cklij t
8:22
Using (8.21) in (8.20), together with the properties of the identity tensor,6 adding,
in turn, the subscripts M and C to the root symbols k and l to distinguish between
modulus and compliance material property functions, there result
Mijkl t kM tdij dkl dik djl lM tdik djl dil djk
dij = 1 if i = j, dij = 0 if i = j; dii = 3; Aikdkj = Ai1d1j ? Ai2d2j ? Ai3d3j = Aij.
8:23
8.6
Integral Constitutive Equations
205
Cijkl t kC tdij dkl dik djl lC tdik djl dil djk
8:24
This leads to the following form of the constitutive equation for isotropic
viscoelastic materials:
rij t
Zt
0
eij t
Zt
0
o
Zt
o
ekk sdij ds 2 lM t s eij t sds
os
os
0
8:25
o
Zt
o
rkk sdij ds 2 lC t s rij t sds
os
os

0
8:26
kM t s
kC t s
More simply, using convolution notation, and omitting the time dependence, for
brevity:
rij t kM dekk dij 2lM deij
8:27
eij t kC drkk dij 2lC drij
8:28
Likewise, introducing the vector {e}T : {1,1,1,0,0,0} and combining convolution

and matrix notations:
fri tg fei g kM dekk 2lM fdei g
8:29
fei tg fei g kC drkk 2lC fdri g
8:30
The manner in which the ks and ls are related to the uniaxial direct and shear
property functions may be established from the observation of the possible forms
of the entries in the strainstress matrix equations (8.28) and from the use of the
correspondence between the Carson transforms of elastic properties and the
transforms of the homologous viscoelastic material property functions.
Since the response of an isotropic solid is independent of orientation, it follows
that the uniaxial moduli should all be equal: E11 = E22 = E33 : E. By the same
token, the three Poissons ratios must also be equal to each other:
v12 = v13 = v23 : v; the same has to be true of the three shear moduli:
G12 = G13 = G23 = G. Since only two of the material property functions are
supposed to be independent, E, G, and v must be related. The relationships among
these properties are established by means of the elastic relationship
E = 2(1 ? v)G and the elasticviscoelastic correspondence. This leads first to the
cancelling out the common factor, s, and taking the
21 svsG;
relation sE
inverse Laplace transform yield the result:
Et 21 v dGs 2Gt 2
Zt
0
v t s
d
Gsds
ds
8:31
206
In similar fashion, the relationship between kM and the uniaxial property

mE
and the
functions E and v is established using the elastic relation k 1m12m
correspondence between elastic and viscoelastic constitutive relations. Using this
correspondence, the previous expression may be cast as 1 sv1 2sv
Cancelling out the common term s, taking the inverse Laplace transs
k svsE.
form, and rearranging yield the result:
m dE kt m dk 2m dm dk
8:32
As will become apparent later shortly, the constitutive equations for isotropic
viscoelastic materials can be expressed by using material property functions other
than the ks and ls, or E, v, and G. The relations among the corresponding elastic
properties listed in Appendix B may be used together with the elasticviscoelastic
correspondence to establish the viscoelastic expressions.
For isotropic materials, it is sometimes convenient to separate the stress and
strain tensors into their spherical (rSij, eSij) and deviatoric (rDij, eDij) parts and to
split the constitutive equations accordingly. This is done as follows.
The spherical part, AS, of the 3 9 3 square matrix of a second-order tensor, A, is
defined as the diagonal matrix with non-zero entries equal to the average of the
diagonal elements of the given matrix. Thus, ASij : (Akk/3)dij = 1/
3(A11 ? A22 ? A33)dij. The deviatoric part, AD, is what is left over, that is, ADij : Aij - ASjj. In explicit matrix notation, A AS AD or more simply
A AS AD
2
1
Akk
A11 A22 A33 4
0
dij
AS
3
3
0
d
3
0 0
1 05
0 1
2
3
A11 As
A12
A13
Akk
5
dij 4
AD Aij
A21
A22 As
A23
3
A31
A32
A33 As
By definition then, the spherical part of a second-order tensor is obtained by

making its two free indices the samea tensor operation called contractionand
dividing the result by 3. Applying this operation to (8.27) produces

2
rkk t 3kM 2lM dekk 3 kM lM dekk
3

2
rDij t kM dekk dij 2lM deij kM lM dekk
3
and
8.6
207
Grouping terms and using the definition of spherical and deviatoric strains, (8.27)
and (8.28), may be cast as
rS t 3kM 2=3lM deS
8:33a
rDij t 2lM deDij
8:33b
eS t 3kC 2=3lc drS
8:34a
eDij t 2lC drDij
8:34b
By its definition, the spherical stress is the average of the three direct stresses
that act at a point in a body in any three mutually perpendicular directions such as
(x, y, z) or (r, h, z). For this reason, the spherical stress is also called the pressure
stress. When the strains are infinitesimal, the spherical strain measures the volume
strain, that is, the change in volume at a point after deformation, relative to the
volume before application of the loads (c.f. Appendix B). Based on these observations, the function (kM ? 2/3 lM) entering (8.33a) relates pressure and volume
strain; hence, it represents a volumetric or bulk relaxation modulus. Since (8.33a)
accounts for all changes in volume, (8.33b) must account for all changes in shape.
Hence, the material property function lM represents the uniaxial shear relaxation
modulus, G, so that lM(t) = G(t) and lC(t-s)*dG(s) = H(t).
Expressions (8.33a) and (8.34a) for bulk response are also often written in terms
of the bulk relaxation modulus, KM, and the bulk creep compliance, KC, as
rkk t 3kM 2=3lM dekk 3KM dekk
8:35a
rDij t 2lM deDij
8:35b
ekk t 3kC 2=3lc drS 3KC drkk
8:36a
rDij t 2lM deDij
8:36b
in which the bulk modulus and bulk compliance have been defined as KM and KC,
through
KM t 3kM 2=3lM
8:37a
KC t 3kC 2=3lc
8:37b
Example 8.1 Write the explicit integral form of the stressstrain equations for an
isotropic viscoelastic solid with shear relaxation modulus, G, and bulk relaxation
modulus K.
208
Solution:
Insert (8.37a) into (8.35a) and use the definition of convolution to get
rkk t 3
Zt
K t s
0
o
ekk sds
os
Likewise take the convolution of (8.33b) to write

rDij t 2
Zt
G t s
0
o
eDij t sds:
os
There are also many viscoelastic materials of practical interest whose spherical
response is elastic or approximately so. For such materials, (8.35a, b) take the
form:
rkk t 3Kekk t; rDij t 2
Zt
0
lt s
o
eDij t sds
os
8:38a
Or, equivalently, but more simply

rkk t 3Kekk t; rDij t 2l deDij
8:38b
Example 8.2 Write the matrix form of the plane-stress constitutive equations of a
viscoelastic material of hereditary integral type, whose volumetric response is
elastic with bulk modulus K and shear relaxation modulus G.
Solution: In plane stress, all stresses associated with one reference direction, the
3-axis, say, are zero. Specifically, r31 = r32 = r33 = 0. Therefore, from the
definitions listed in (d) and (e): rkk = r11 ? r22, rD11 = 2/3r11- 1/3r22,
rD22 = -1/3r11 +2/3r22 and hence rkk = (r11 ? r22)/3 and eD12 = e12. Using
these with (8.37a, b) and simplifying give one form of the result, as
2
6K
1 4
3K
9K
0
3K
6K
0
8
9 8
9
9
38
2
0 < r11 = < 9K
G dr11 =
< G de11 =
2
0 5 r22 9K
G dr22 2 G de22
:
:
; :
;
;
0
r12
G de12
r12
The constitutive property functions as well as the components of the stress and
strain are assumed to be piecewise smooth, bounded, and with, at most, a finite
number of finite amplitude discontinuities. Under these conditions, their Laplace
and Fourier transforms are well defined (c.f. Appendix A). Hence, either transform
may be applied to the constitutive equations listed in this section. For instance, the
Laplace- and Fourier-transformed stressstrain relations corresponding to (8.27) are
8.6
209
ij s skM s ekk s dij 2s

lM eij s
r
8:39
rij jx kM jx ekk jx dij 2lM jx eij jx
8:40
Example 8.3 Compute the steady-state response of a slender bar subjected to the

harmonic loading fegT eo11 ; eo22 ; eo33 ; 0; eo13 ; 0 sinxo t. Assume the materials
volumetric response to occur at a constant bulk modulus, K, and its deviatoric
response to be that of Kelvin solid with modulus, E, and viscosity g.
Solution:
To apply the constitutive equations, we need the complex form of the applied

strain, which is fe jxo tgT eo11 ; eo22 ; eo33 ; 0; eo13 ; 0 ejxo t , but expressed in terms of
its spherical and deviatoric components, using (e) and (f); thus, eS 1=3
eo11 eo22 eo33 ejxo t
2 o
2e eo22 eo33
1 4 11
0
eD
3
eo
13
3
0
eo13
5ejxo t
2eo22 eo33 eo11
0
o
o
o
0
2e33 e11 e22
From the statement of the problem, K* = K, while G* must be derived for a Kelvin
solid. To do this, use the stress-strain equations for a Kelvin solid (3.16a) and the
applied strain e*(jxt) = eo ejxt to write r jxt ro ejxtd E jxgeo ejxt .
After regrouping, using Eq. (4.2) and dividing by eo, get: G jx E jxg.
Inserting K* and G*evaluated at the forcing frequencytogether with the
applied strain, the following is obtained:
rS jxo t Kfeo11 eo22 eo33 gejxo t
rD jxo t
2 o
3
2e eo22 eo33
0
eo13
1 4 11
o
o
o
5ejxo t
0
2e22 e33 e11
0
2E jxo g
3
0
2eo33 eo11 eo22
eo13
Since the applied strain is the imaginary part of the complex strain used, the shear
stress response will correspond to the imaginary part of the complex stress and the
volumetric response, being elastic, will be proportional to the applied strain. That is,
rS t Keo11 eo22 eo33 sinxo t
2 o
2e11 eo22 eo33
0
2
0
2eo22 eo33 eo11
rD t E sinxo t xo g cosxo t4
3
eo
0
13
3
eo13
5
0
2eo33 eo11 eo22
210
8.7
Constitutive Equations for Isotropic Incompressible

Materials
An important class of viscoelastic materials also existssuch as most rubbe compoundswhich are very nearly incompressible, in that they respond to loading by
changing their shape but preserving their volume. Incompressible materials preserve
volume throughout deformation. In geometrically linear problems, volume preservation requires that the volume strain, ekk, be identically zero throughout the loading
process. Incompressibility acts as a constraint on the equations of motion. This means
that one can add a spherical stressuniform in all directionsof any magnitude to
the acting stress field, without altering the strains. That is, for incompressible
materials, the stress tensor is determined from the strains, only up to a spherical
stress. The spherical stress is usually called the pressure stress, p. The stressstrain
constitutive equations for isotropic incompressible materials take the form:
Zt
rij t ptdij 2lt s

0
o
eij t sds; ekk t 0
os
8:41a
Or, in convolution notation, omitting the time arguments of the convolution

integrals:
rij t ptdij 2l deij ; ekk t 0
8:41b
Example 8.4 Strip biaxial test. A short, thin, and very wide sheet of rubber is
stretched along its height direction, x2, as depicted in Fig. 8.4. Because of its large
aspect ratio, the sheet does not contract appreciably in the wide direction. As a
consequence, e11 = 0 in its central region. Using this fact, determine the stress in
the direction of pull, in the center of the specimen, assuming that the shear
relaxation modulus of the material is l(t).
Solution
In this case, e11 = 0 and r33 = 0. The incompressibility condition, ekk = 0,
requires that e33 = -e22, while, according to (8.41b), r33 = 0 = p(t) ? 2l*de33 : -p(t) - 2l*de22. Using this to evaluate r22 with (8.41b) again
and simplifying lead to r22 = 4l*de22.
x2
(t)
x1
8.7
211
All material property functions as well as the components of the stress and
strain tensors are assumed well behaved for their Laplace and Fourier transform to
exist. Under these conditions, the Laplace and Fourier transforms of the stress
strain relations listed in (8.41b) are:
ij s psdij 2s
r
lseij s; ekk s 0
8:42
ij jx pjxdij 2jx

ljxeij jx; ekk jx 0
r
8:43a
l ; p p ; r
ij rij ; eij eij , as is customary, (8.43a)
Using the notation jx l
may be expressed alternatively as
r jx p jxdij 2l jxe jx; ekk jx 0
8.8
8:43b
Differential Constitutive Equations for Isotropic Materials
As with integral models, generalization of the constitutive equations to threedimensional conditions is straightforward. Using indicial notation and the summation convention,7 for instance, the most general constitutive equation in differential operator form becomes
Prij t Qeij t
8:44
For homogeneous isotropic materials, which require only two property functions, the viscoelastic constitutive equations in differential operator formin
analogy with the hereditary integral forms in (8.35a) and (8.36a)read
PD rDij 2QD eDij ; PS rkk 3QS ekk
8:45
In general, then, four differential operators would be required to define the constitutive equations of a viscoelastic material in differential form. On the other
hand, for materials whose volumetric response may be idealized as elastic, the
stress strain Eq. (8.38a) takes the form:
PD rDij 2QD eDij ; rkk 3Kekk
8:46
Following the same reasoning, the constitutive equations for isotropic incompressible viscoelastic materials of differential type are written as
1
PD rDij 2QD eDij ; pt rkk ; ekk 0
3
7
8:47
According to the summation convention, terms in an expression are summed over the range of
their repeated indices, so that, for instance, Aikk = Ai11 ? Ai22 ? Ai33, for i = 1, 2, 3, etc.
212
Example 8.5 A slender bar made of a Kelvin material having constant bulk
modulus K and spring and dashpot parameters G and g is subjected to a uniaxial
stress r11(t) = roH(t). Determine the strain e11(t) and its long-term value.
Solution:
Use (8.46) to get PD rD11 2QD eD11 and rkk 3Kekk . For a Kelvin material,
1
rkk Ht. Or, in
P = 1 and Q = G ? gqt so that rD11 2G got eD11 ; ekk 3K
strainstress form, introducing the retardation time k = g/G of the differential
1
1
rD11 t; ekk 3K
rkk Ht.
model dtd eD11 t 1k eD11 t 2g
In this case, the only non-zero stress is r11. Hence, rkk = roH(t);
rD11(t) = r11(t) - rkk/3, that is, rD11(t) = (2/3)roH(t) and eD11(t) = e11(t) - ekk/
3 : e11(t) - roH(t)/(9 K). Taking these results into the previous expression,
cancelling the common factor, and reordering, the following is obtained:

d
1
1
G
g
g
1
e11 e11 ro H t
ro dt; k ; ekk
ro Ht
dt
k
3g 9Kg
9K
G
3K
Integrating between 0- and t, using the integrating factor et/k with the initial con

g t=k
t=k
e
ro .
ro 9K
dition e11(0-) = 0, and rearranging produce e11 t 3KG
9KG 1 e
3KG
The long-term response is obtained as the limit t ? ?: e11 1 9KG ro rEo .
Where the expression 9 KG/(3 K ? G) = E, for the elastic, Youngs modulus was
borrowed from Appendix B.
Physically realistic differential constitutive equations admit Laplace and Fourier integral transformations. As with one-dimensional models, the integral
transform of a differential operator of order n turns out to be a polynomial of
degree n in the transform variable (s or jx), as seen in (3.13). Using this, and
letting v stand for either s or jx, the Laplace and Fourier transforms of (8.45) to
(8.47) can be succinctly written as
D veDij v; P
S vekk v
D v
S v
P
rDij v 2Q
rkk v 3Q
8:48
D veDij ; r
D v
kk v 3Kekk v
P
rDij v 2Q
8:49
D veDij v;
D vrDij v 2Q
P
8:50a
1
kk v; ekk v 0
pv r
3
8:50b
The relationships between the various property functions may be easily established
by means of the Laplace transformation. Taking the Laplace transform of the
corresponding relationships (8.35a) and (8.45) to do this, for instance, and
equating the results, the following is obtained:
8.8
Differential Constitutive Equations
213
D s

Q
M s QS s
M s
;K
l
S s
sPD s
sP
8.9
8:51
Problems
P.8.1 Write the convolution integral form of the two-dimensional constitutive

equation for an isotropic incompressible viscoelastic solid of shear relaxation
modulus G.
8
8
9
9
8 9
< rxx t =
< exx s =
<1=
ryy t pt 1 2Gt s d eyy s
Answer :
:
:
;
;
: ;
exy s
0
8rxy t
9
< pt 2G dexx =
pt 2G deyy
:
;
2G dexy
Hint: Use the extended vector form Aij ? {Aij}T : {Axx, Ayy, Axy} of the stress
and strain matrices to rewrite (8.41b): rij t ptdij 2G deij ; ekk t 0.
P.8.2 Use matrix notation to extend to three dimensions the concept of effective
elastic modulus for constant rate loading.
1
Answer : Eeff t
t
Zt
Mudu
Hint. Note that, in this case, even though the prescribed boundary conditions may
be changing at a constant rate, the induced strains may vary not only from point to
point, but also from coordinate direction to coordinate direction at every point in
the body. Using this, take ei (t) = Rit for each coordinate direction, use matrix
algebra, and proceed as in Example 2.1 to arrive at the matrix form.
P.8.3 Use the Laplace transform to establish the tensile relaxation modulus of an
elastomer for which a uniaxial Maxwell model is available with spring and dashpot
parameters E and g, respectively.
Hint: Write the uniaxial constitutive equation of the model in Laplace-transformed space for both the differential form Pr = Qe and the convolution form
r = M*de. Then, apply the Laplace transform to both expressions and equate them
Egs,
Q . Now, insert the transformed operators, P
E sg and Q
to get M
sP
Egs
E

and arrive at M sEsg 1=ss ; s g=E. The inverse transform of this expression [see Appendix] yields the desired result Mt Eet=s .
P.8.4 Write the matrix form of the plane-stress constitutive equations of a viscoelastic material of differential type, whose volumetric response is elastic, with bulk
modulus K.
214
2
6KPD 2QD
1 4
Answer :
2QD 3KPD
9K
8 90
< e11 =
2QD e22
: ;
e12
2QD 3KPD
6KPD 2QD
0
9
38
0
< r11 =
0 5 r22
:
;
9KPD
r12
Hint:
In plane stress, all stresses associated with one reference direction, the 3-axis,
say, are zero. Specifically, r31 = r32 = r33 = 0. Therefore, from the definitions
given in (e) and (f) of Sect. 8.6: rkk = r11 ? r22, rD11 = 2/3r11 -1/3r22, rD22 = 1/3r11 +2/3r22, and also ekk = (r11 ? r22)/3. Use these facts with (8.46) and
perform the indicated operations to arrive at the result.
P.8.5 Solve Example 8.5 assuming that the Kelvin material is incompressible.
Hint:
The constitutive applicable constitutive equation in this case is (8.47). Because
of incompressibility, ekk(t) = 0, so that eD11(t) = e11(t). In addition, p(t) = roH(t)/3. Hence, rD11 becomes rD11(t) = r11(t) ? p(t) : (2/3)roH(t).
Taking this into (8.47), recalling that P = 1 and Q = G ? gqt yields
d
1
1
dt e11 t k e11 t g ro Ht. Integrating this equation between 0 and t, using the
t/k
integrating factor e and the initial condition e11(0 ) = 0, and rearranging pro

duce e11 t G1 1 et=k ro . Evaluating this expression in the limit as t ? ?
gives the long-term solution as e11 1 rGo .
P. 8.6 Determine the differential operator expression for the Poissons ratio of a
linear isotropic viscoelastic material having constant bulk modulus.
Answer : v
3KP 2Q
6KP 2Q
Hint: Here, the target is an operator expression for the ratio v = -e22/e11, resulting
from a stress r11, when all other stresses are zero. Therefore, select a onedimensional tension test, for which the only non-zero stress would be r11. Because
of symmetry, e22 = e33. Hence, rkk = (1/3)r11, ekk = (e11 ? 2 e22)/3, together
with rD11 = (2/3)r11 and eD11 = (2/3)(e11 - e22). Use expression (8.46) to obtain
Pr11 = 2Qe11- 2Qe22 and r11 = 3 K(e11 ? 2e22). Combine these relationships
and perform the necessary algebra to obtain the differential operator form of the
Poissons ratio.
P.8.7 Under two-dimensional conditions, the direction of the maximum principal
stress is given by the relation: tan2hrmax 2rxy =rxx ryy ; similarly, the
direction of the maximum principal strain is given by tan2hemax 2exy =exx eyy .
Use these facts to demonstrate that, in general, for linear isotropic viscoelastic
materials, the directions of maximum principal stress and strain do not coincide.
8.9
Problems
215
Answer : hrmax
2l dexy
2exy
6
hemax
l dexx eyy exx eyy
Hint: Evaluate hrmax by inserting any one of the general constitutive equations
for linear isotropic viscoelastic solids, such as (8.27) or (8.41a), cancel the common terms, and arrive at the proof. This problem shows that in viscoelastic
materials, the directions of maximum principal stress and maximum principal
strain do not coincide in general. As may be seen from the answer to this problem,
the directions of principal stress and strain in a viscoelastic solid would coincide if
the stress response were elastic like, that is, simply proportional to the strain, to
allow cancelling the material property function. Conditions under which this
happens are discussed in Chap. 9.
P. 8.8 A rigid cylindrical container is filled with a Maxwell material having elastic
bulk modulus, K, and spring and dashpot parameters G and g, respectively.
Determine the axial strain if a rigid plunger is used to apply a pressure poH(t) at its
open end, as indicated in Fig. 8.5.

1
1
1
3
t=s
Answer : ezz t po

e
;s
K
K 2Gg K 4G 3K

4G 3K
g
3GK

Hint: Write the three-dimensional constitutive equations for a Maxwell material

with elastic bulk response: PrDij = 2QeDij and rkk = 3 Kekk, in cylindrical coordinates (r,h,z). In this case, err = ehh = 0, ekk = ezz, rkk = 3 Kezz, eDzz = (2/3)ezz,
and rDzz = -poH(t) - Kezz lead to G got po H t Kezz 2Ggot 23 ezz . Col3po
lect terms and arrive at the equation dedtzz 1s ezz g4G3K
GH t gdt. Solve
po
this equation using the initial condition ezz 0 2GgK , which can be obtained from
PrDij = 2QeDij, following the procedure discussed in Chap. 3 (see Sect. 3.3.4).
P.8.9 Prove that the Poissons ratio of an incompressible viscoelastic material is 1/2.
Hint: Consider a uniaxial tensile test on a general incompressible material.
Invoke its incompressibility ekk = 0 and symmetry of strain response (e22 = e33) to
establish that e22 = -(1/2)e11. Use this result and the definition of Poissons ratio
that v = -e22/e11 to complete the proof.
zz(t)
= -poH(t)
216
P.8.10 Write the stressstrain constitutive equations applicable to plane-stress

steady-state conditions for an orthotropic viscoelastic solid.

Answer :
r11
r22

1
E11

1 m12 m21 m21 E11

m12 E22

E22

e11
; r12 2G12 e12
e22
Hint: Write the

constitutive
complex
form of the strain
Eq. (8.13a, b), as discussed

m21 =E22
1=E11
r11
e11
; e12 r12 =2G12 . Invert

in Sect. 8.5:

e22
m12 =E11
1=E22
r22
the matrix and the shear equation and use the complex form of the symmetry
relations (8.12) to arrive at the result.
P.8.11 A rectangular plate made of a linear isotropic viscoelastic solid is supported
along its long sides while subjected to a strain e1(t) = 0.05cos(0.25 t) on its short
edges, as shown in Fig. 8.6. Determine the reaction stress at the support, for a
material with relaxation modulus E(t) = 100 ? 1,000e-0.5t MPa and Poissons
ratio m(t) = 0.400.15e-0.25t.
Answer : r2 t 1:540 MPa
Hint: Note that for an isotropic
solid, m12 = m21 = m
and E11 = E22 = E and

r11

e11
1 m
use the result of P. 8.10, to write r 1Em 2
; r12 2G e12 . Since
e
m
22
22
the plate is restrained in the 2-direction but free to extend in the 1-direction, no
shearing would develop in it, the strain field would be e11 eo11 cos0:25t and
e22 0 e12 , and as a consequence, the stress in the 2-direction becomes

r22 1m Em 2 e11 . Next, evaluate the complex modulus and the complex Poissons
ratio using the results of Example 4.1. For E*, use Ee = 100, E1 = 1,000, and
s = 2; and for m, me = 0.40, m1 = -0.15, and s = 4. So that E 0:25j 100
1;0000:2522
10:2522
j 0:15:254
10:2542
j 1;000:252
300 200j
10:2522
and
m 0:25j 0:4 0:150:254

10:2542
0:325 0:075j. Take these values and the applied strain field into
the expression for r22 and perform the complex algebra and since the actual
1(t)
= 0.05cos(0.25t)
1(t)
= 0.05cos(0.25t)
References
217
applied strain is the real part of the complex strain e22 , take the real part of the
result to arrive at the answer sought.
References
1. A.E. Green, W. Zerna, Theoretical Elasticity, 2nd edn. (Dover, NY, 1968), pp. 13
2. L.E. Malvern, Introduction to the Mechanics of a Continuous Medium (Prentice-Hall,
Englewood Cliffs, 1963), pp. 6480
3. B.E. Read, G.D. Dean, The Determination of Dynamic Properties of Polymer Composites
(Wiley, New York, 1978), pp. 1217
4. R.M. Christensen, Mechanics of Composite Materials (Wiley, New York, 1979), pp. 150160
5. W.R. Little, Elasticity (Prentice Hall, Englewood Cliffs, 1973), pp. 6263. 1617
Isothermal Boundary-Value Problems
Abstract
This chapter contains a comprehensive discussion of the types of boundaryvalue problems encountered in linear viscoelasticity. The chapter presents
detailed solution methods for compressible and incompressible solids, including
materials with synchronous moduli, whose property functions are assumed to
have the same time dependence. The method of separation of variables in the
time domain and frequency domains is also described in full, as is the use of the
Laplace and Fourier transformations. The elasticviscoelastic correspondence
principle, which allows viscoelastic solutions to be constructed from equivalent
elastic ones and as a consequence of the applicability of integral transforms, is
also developed and examined in detail.
Keywords
Boundary-value Balance Conservation Momentum Energy Equilibrium

Motion Quasi-static Compatibility Synchronous Separation of variables
Correspondence principle
9.1

Introduction
This chapter introduces several methods of solving boundary-value problems of

isotropic linear viscoelastic continua for which the temperature is assumed constant in space and time. Thus, all required material property functions are assumed
known at the arbitrary but constant temperature of the discussions.
The chapter begins by listing the differential equations of motion, which
embody the principles of balance of linear and angular momenta and are postulated to be obeyed by all materials in bulk. The boundary-value problems of
viscoelasticity are presented next, restricting attention to quasi-static conditions,
219
220
for which the acceleration, or inertia terms are negligible. Specific methods of
solution of boundary-value problems are discussed after that. The methods presented include the separation of variables and integral transformations, as well as
the special case of materials with synchronous moduli, for which the viscoelastic
property functions in shear and bulk are assumed to have the same dependence on
time.
9.2
Differential Equations of Motion
There are certain laws of physics which are obeyed by all substances in the bulk,
be they elastic, viscous or viscoelastic, irrespective of whether their response to
external stimuli is linear or non-linear [1]. These laws proclaim the conservation of
mass, linear momentum, angular momentum, and energy. All the mentioned laws
are discussed in Appendix B, but for continuity with the topic of this chapter, the
balance laws of linear and angular momenta are presented here as well.
9.2.1
Balance of Linear Momentum
Newtons second law of motion requires a balance between the external resultant
load acting on a system and the rate of change of its linear momentum. The
integral version of this law is due to Cauchy and gives rise to the equations of
motion, which are valid for all materials in the bulk. The derivation of the motion
equations is straightforward; and, as shown in Appendix B, for instance, in
unabridged notation, the x-component of the equations of motion for a body with
density q and body force qbx, reads:
o
o
o
o2
rxx rxy rxz qbx q 2 tx
ox
oy
oz
ot
The y- and z-component equations are written with appropriate permutations of

the coordinates. Thus, using indicial tensor notation, where terms with repeated
indices are summed over their full range, the equations of balance of linear
momentum take the form:
o
o2
rij qbi q 2 ui ;
oxj
ot
i; j 1; 3
9:1
When inertia terms are zero, as in static problems, or can be neglected, which
leads to so-called quasi-static problems, the acceleration term on the right-hand
side of the equations is dropped. This leads to the equations of static equilibrium:
o
rij qbi 0i
oxj
9:2
9.2
Differential Equations of Motion
221
Where differentiation is understood with respect to coordinates in the undistorted state. The comma notation: rij,k : qrij/qxk, is also frequently used to
denote differentiation with respect to the coordinate(s) following the comma.
9.2.2
Balance of Angular Momentum
Non-polar materials are defined as those without intrinsic body couples or spin,
and for which the resultant internal moment on the surface of any infinitesimal
material element is zero. For such materials, the principle of conservation of
angular momentumthat the resultant external moment on a body is equal to the
time rate of change of its angular momentumyields the simple requirement that
the stress tensor, rij, be symmetric:
rij rji
9:3a
Using unabridged notation and the standard x, y, z Cartesian coordinates, these

expressions take the form:
rxy ryx ;
rxz rzx ;
ryz rzy
9:3b
The implication of the balance of angular momentum for non-polar materials is

that the stress tensor has only 6 independent components. This is so because a
symmetric N-by-N matrix has N N 1=2 independent components (6, when
N = 3).
9.3
General Boundary-Value Problem
The isothermal boundary-value problem of linear isotropic viscoelasticity consists

of the three equations of motion (9.1) together with their initial and boundary
conditions. Taking the initial configuration of the body to be free of stress and at
rest prior to the application of the loading (all field variables are identically zero
for t \ 0), the equations of motion and the associated boundary conditions are as
follows1:
o
oxj
ui t
nj rji t
1
rij t qbi t q oto 2 ui t;

uoi t;
Tio t;
in
on Su ;
V
t0
9:4
on ST ; t 0
The components of the displacement vector, the stress and strain tensors, the material density,
the boundary values, and the normal to the surface, will in general depend on position. For clarity
of exposition, however, dependence on position is omitted most of the time, but shall be
understood.
222
In these expressions, ni represents the unit outward normal to the boundary, S,

formed of Su and ST, where displacements and tractions, respectively, are prescribed. Here, ni is not a function of time. In addition, in a well-posed problem, Su
and ST do not intersect (that is: Su \ ST U). This last requirement means that one
cannot prescribe different types of boundary conditions at the same point and in the
same direction.
The system of equations in (9.4) contains 9 unknown quantities3 displacement
components and 6 independent stress components. For a solution of this system to
exist, the system has to be complemented by other relations. The displacements ui,
are related to the strains, eij through 6 relationships (c.f. Appendix B):
eij t

1 o
o
ui t
uj t
2 oxj
oxi
9:5
Finally, the 6 constitutive equations relate 6 additional strains to the 6 components of the stress tensor. The constitutive equations may be taken in hereditary
integral form, in differential form, or, for steady-state problems, in terms of
complex quantities. As indicated in Chap. 8, the constitutive equations of isotropic
materials are completely defined by any two of five interrelated relaxation functions: the tensile, shear, and bulk moduli, E(t), G(t), K(t), respectively; the Poissons or contraction ratio, v(t); and Lames function k(t). As discussed in Chap. 8,
the relationships among these functions may be derived using their elastic counterparts (as listed in [2]) and the elasticviscoelastic correspondence. For the sake
of presentation, the constitutive equations developed as (8.27) and (8.41) for
compressible and incompressible materials, respectively, and listed here:
rij t kM t s dekk sdij 2lM t s deij s
9:6
rij t ptdij 2lM t s deij s
9:7
Equations (9.4), (9.5), and any of the sets of 6 constitutive equations derived in
Chap. 8, such as either (9.6) or (9.7), represent 15 equations in 15 unknowns: 3
equilibrium equations, 6 straindisplacement relations, and 6 stressstrain equations; and 3 unknown displacements, 6 unknown stresses and 6 unknown strains.
These are the field equations of isothermal viscoelasticity, which will possess a
unique solution if the initial values of the material relaxation functions are nonnegative [3].
The straindisplacement equations listed in (9.5) relate the three components of
the displacement field to the six components of strain. These expressions result in a
unique set of strains for any prescribed set of displacements, but in general, do not
suffice to produce a unique displacement field from an arbitrarily prescribed set of
strains. The system of equations in the latter case is over-determined, as it has six
equations in three unknowns. This prevents the six components of strain to be
prescribed arbitrarily. The additional conditions that the strain tensor must satisfy
9.3
General Boundary-Value Problem
223
to allow a unique displacement field upon integration of the straindisplacement

relations are known as the integrability or compatibility conditions.
The equations of compatibility are obtained by differentiating the straindisplacement relations twice and permuting indices (c.f. Appendix B). This process
yields the following 81 equations, of which only six are independent [4]:
eij;kl ekl;ij eik;jl ejl;ik
9:8
Using the standard Cartesian coordinates x, y, z, for subscripts 1,2,3, respectively, the six independent compatibility equations in unabridged notation are of
the following form:

o2
o2
o2
o2
o
oeyz oezx oexy

e
e
2
;
yy
xx
xy
xx
ox
ox2
oy2
oxoy
oyoz
ox
oy
oz
Similar permutations of x, y, and z produce the other four independent relations.

In the case of two-dimensional problems in the xy plane, the only non-trivially
satisfied relation is the first one listed above.
On occasion, the compatibility conditions are required in terms of stresses.
Since stresses and strains are connected through the constitutive equations, the
integrability conditions in terms of stresses depend on material properties. These
equations may be obtained by direct substitution of the strainstress constitutive
equations into (9.8). For a viscoelastic material of hereditary type, the result is the
convolution form of the so-called Beltrami-Michell relations of the theory of
elasticity [5].
9.4
Quasi-Static Approximation
In what follows, inertia effects are ignored but it is recognized that all field
variables, ui, rij, and eij will generally change with time. This is the quasi-static
approximation, in which problems remain time-dependent, because the stress
response depends, through the strains, on the complete history of displacements,
and not only on the current state, even when the boundary conditions remain
constant. In the latter case, the step functions that define the boundary displacements and tractions produce disturbances that propagate into the body at high
speed. Stress relaxation and heat conduction in viscoelastic materials, however,
tend to damp out this wave motion rather quickly. In using the quasi-static
approximation, it is assumed that the wave motion generated by any sudden
changes in the boundary conditions is damped out more quickly than the boundary
data change afterward. The quasi-static approximation, then, assumes that at any
instant the body is in equilibrium with the concurrent boundary data [6].
224
9.5
Classification of Boundary-Value Problems
The differential equations of motion (9.4) and associated boundary conditions

suggest that the boundary conditions may be of one of the three types. The first
class corresponds to problems where only tractions are prescribed. A second class
involves pure displacement boundary values, while a third class allows displacements and tractions to be prescribed on the boundary. The three types of boundaryvalue problems are examined next; and for simplicity, only the quasi-static
approximation is considered, so that the inertia terms are set to zero.
9.5.1
Traction Boundary-Value Problem
In this case, (9.4) takes the quasi-static form:

o
oxj
rij t qbi t 0; in V
on S
nj rji t Tio t;
9:9
It is important to note that in this case the differential equations of equilibrium

and the boundary data involve only stresses. Since this system does not involve
material properties, its primary solution, that is the stress field, has to be independent of material constitution. In other words:
The stresses in a linear viscoelastic bodyirrespective of whether it is isotropic or not
subjected only to tractions on its boundary, are exactly the same as those which the same
body would experience if its material were replaced with any other linear material.
The strain and displacement solutions, however, will in general depend on the
material in question, and the equations of compatibility need to be employed to
ensure that the strains, calculated from the stress field through the constitutive
equations, will yield single-value displacements in simply connected domains
when the body has no internal holes.
9.5.2
Displacement Boundary-Value Problem
In this case, only displacements are prescribed on the surface of the body, and the
governing equations and boundary conditions become:
o
rij t qbi t 0;
oxj
ui t uoi t;
on S
in V
9:10
9.5
Classification of Boundary-Value Problems
225
Here, since strains are determined uniquely by the displacement field, strain
compatibility is automatically satisfied. Also, because stresses and displacements
are connected by the straindisplacement relations and constitutive equations, the
equations of equilibrium may be cast in terms of displacements only.
Unlike the traction-only problem, no savings can be realized in the present case
for viscoelastic materials in general. However, as with the traction-only problem,
the viscoelastic solutions for incompressible isotropic materials, and for isotropic
materials of constant Poissons ratio can be constructed from those of elastic
solids, as shown in Sect. 9.6.
9.5.3
Mixed Boundary-Value Problem
This is the most general type of boundary-value problem in that both displacements and tractions are specified on the bounding surface. Just as the traction-only
problem, the constitutive equations and compatibility conditions are required. In
addition, however, the straindisplacement relations are needed for the problem to
be mathematically well posed. The field equations for this case take the form:
o
oxj
rij t qbi t 0;
nj rji t Tio t;
ui t uoi t;
in V
on ST
on Su
9:11
No savings in solution effort are available in general for the mixed boundaryvalue case. Also, as with the other two types of problem, viscoelastic solutions can
be constructed from elastic solutions for isotropic solids of constant Poissons
ratio, which include as a special case, incompressible materials.
Example 9.1 A disk of diameter D, made of a viscoelastic material is subjected to
a pair of diametrically opposing concentrated loads, P(t), as indicated in Fig. 9.1.
Find the stresses along the load line, if the corresponding values for a linear elastic
solid are [2]:
2Pt
;
pD

2P
2
2
1
rey x 0; y
; rexy x 0; y 0;
p D 2y D 2y D
rex x 0; y
Solution:
This is a traction-only boundary-value problem. Consequently, the stress
picture is independent of material properties, and the viscoelastic solution must be
equal to the elastic one given.
226

P(t)
y
x
P(t)
9.6
Incompressible Materials
The mechanical response of isotropic incompressible materials is fully described

by one single material property function, as indicated by (9.7). The solution of a
displacement boundary-value problem for this material type can be constructed
from the solution of the same problem posed for an isotropic incompressible
elastic material of arbitrary shear modulus. The method followed here to
accomplish this, normalizes the shear relaxation modulus, lM, as:
lM t lR :mN t
9:12
This normalization introduces an elastic material of shear modulus lR that is

meaningful to the problem at hand. At the same time, through (2.20) or (2.21), it
defines the creep compliance, Cl and its normalized counterpart cN, and establishes the relationship between the normalized functions mN and cN, as:
Cl t CR cN t
1
cN t
lR
9:13
mN t s dcN s cN t s dmN s
9:14
The method also introduces auxiliary displacement, uVi, in terms of the actual
displacements, ui, by means of the definition:
uVi t mN t s dui s
Zt
0
m N t s
o
ui sds
os
9:15
According to (9.5), the strains, eVij, corresponding to this auxiliary field would
be given by:
9.6
Incompressible Materials
227

1 o
o
1
eVij t
uVi
uVj uVi;j uVj;i
2 oxj
oxi
2
9:16
and, in particular:

lM deij lR mN d ui;j uj;i lR mN dui;j mN duj;i lR eVij t;
Using this result in (9.7), yields
rij xk ; t pxk ; tdij 2lR eVij xk ; t rVij xk ; t
9:17
That rij = rVij, follows from the fact that, considered as a function of the
strains, r(e) = -p ? 2le, implies that r(eV) = -p ? 2leV : rV. From this, follows that the equations of equilibrium in (9.10) are identically satisfied by the
auxiliary elastic stress (9.17). To satisfy the displacement boundary conditions
also, it is necessary to convolve them with the normalized shear modulus mN, and
use (9.15) to convert them into the auxiliary elastic form: m dui uVi m duoi .
This transforms the displacement boundary-value problem (9.10) for an incompressible viscoelastic material to the auxiliary elastic form:
o
rVij qbi 0; inV;
oxj
uVi m duoi ;
on S
9:18
Once the auxiliary elastic displacements are obtained, the actual displacements
are determined inverting (9.15)convolving cN with uVi. The strains follow from
the straindisplacement relationships.
9.7
Materials with Synchronous Moduli
The mechanical response of viscoelastic materials for which the relaxation functions in bulk and shear have the same time dependence are said to have synchronous moduli. The proportionality between the shear and bulk relaxation
moduli implies that the stressstrain behavior for this type of material is fully
described by a single relaxation or compliance function. Synchronous moduli are
possible if the Poissons ratio is constant. In this case, several options are available
regarding the time-dependent material properties used: kM and lM : G; G and v;
the uniaxial tensile modulus, E, and Poissons ratio, v; and so on.
Since materials with synchronous moduli have only one viscoelastic property
function, the same approach can be followed as for incompressible materials to
transform a displacement boundary-value problem of a viscoelastic material to that
of an auxiliary elastic one. The difference lies in the specific form of the constitutive equation and the choice of the material property functions used.
228
The approach used introduces the normalized relaxationor creepfunctions

entering the problem, as follows:
kM t kR mN t; lM t lR mN t; Et ER mN t
9:19
1
1
1
cN t; Cl t
cN t; CE t
m N t
kR
lR
ER
9:20
C k t
This normalization, for which (9.14): mN t s dcN s cN t s dmN s,

still holds, and the constancy of the Poissons ratio imply that the normalizing
moduli kR, GR, and ER, are related elastically. That is (c.f. Appendix B):
kR
m
2m
; ER
GR
1 m1 2m
1 2m
9:21
From this point on, the method is the same as for incompressible materials. It
uses the auxiliary elastic displacements, uVi, and the corresponding strains, eVij,
listed in (9.15) and (9.16), introduces (9.19) into the selected constitutive equations
and arrives at the same result as for incompressible materials, that the displacement boundary-value problem (9.10) of viscoelasticity is converted to that of an
auxiliary elastic solid, as in (9.18).
Example 9.2 Write the explicit form of the auxiliary elastic stress field for use in a
displacement boundary-value problem of viscoelasticity, for a viscoelastic solid
with synchronous moduli whose constitutive equation is given in the form:
rij t kt s dekk sdij 2lt s deij s.
Solution:
Use the normalization suggested (9.19) with the selected constitutive equation
to write that: rij t kR mN dekk dij 2lR mN deij . Then, interchange the order
of differentiation and integration and invoke the definitions of the auxiliary elastic
displacement and strain fields to arrive at the form:
rij xk ; t kR eVkk xk ; tdij 2lR eVij xk ; t rVij xk ; t:
9.8
Separation of Variables in the Time Domain
The method of separation of variables is applied to construct solutions to viscoelastic boundary-value problems for isotropic materials of constant Poissons ratio.
The basis of the method is the assumption that the solution variables have the same
dependence on time as their boundary data counterparts. That is, the stresses and
body forces have the same time variation as the prescribed surface tractions and
the displacements and strains, the same time variation as the boundary displacements. As will be seen, the method of separation of variablesvalid only for
9.8
Separation of Variables in the Time Domain
229
materials with constant Poissons ratioapplies to any of the three types of

boundary-value problems. Specifically, it is assumed that to boundary data2:

Tio xk ; tST Toio xk fT t;

uoi xk ; tSu uooi xk fu t
9:22
There correspond solution variables in the separable forms:

rij xk ; t roi xk fT t;
bi xk ; t boi xk fT t
9:23
ui xk ; t uoi xk fu t;
eij xk ; t eoij xk fu t
9:24
Where the spatial part of the strain, eoij, is computed from the spatial part of the
displacement field through the straindisplacement relations (9.5), as:
eoij

1 o
o
uoi
uoj
2 oxj
oxi
9:25
Using (9.22) to (9.24) with the equations of equilibrium under mixed boundary
data (9.11), canceling the common factors fT(t) and fu(t), results in a mixed
boundary-value problem for the spatial parts roij, uoi, and eoij of the viscoelastic
solution fields. Explicitly:
o
roij xk boi xk 0; in V
oxj
nj xl roji xk Toio xk ;
uoi xk uooi xk ;
9:26
on ST
on Su
The viscoelastic solution is retrieved from the elastic one through (9.23) and
(9.24). Since the stress and strain fields are related through the constitutive
equations, the functions fT and fu cannot be prescribed independently. The relation
between them is established by means of the constitutive equations. Taking the
strain field (9.24) into the constitutive equation, such as (9.6), using (9.23) and
(9.25), and separating the spatial and temporal parts, there results:

rij t roij fT t kR eokk dij 2lR eoij mN t s dfu s
9:27
Since the quantity inside the braces is the spatial part, roij, of the stress tensor,
this expression leads to the relation sought between the time functions fu and fT:
fT t mN t s dfu s
Zt
0
m N t s
d
fu sds
ds
9:28
The reader is reminded that all boundary data as well as the displacement, stress and strain
fields are dependent on position but that, sometimes, such dependence has been omitted for
clarity.
230
The same approach may be applied when the boundary conditions are either
pure traction or pure displacement. The difference is that in problems with
boundary data of a single type only one of the time functions is given, and the
other has to be determined from (9.28). The approach described applies equally
well to isotropic incompressible viscoelastic materials, because their constitutive
equations involve a single material property function.
Example 9.3 A small hole of radius, ri, in a very large block of a viscoelastic
material of constant Poissons ratio, v, is subjected to an internal pressure P f t,
as indicated in Fig. 9.2. Determine the stresses and displacements in the block, if
the solution for the cavity in an infinite elastic medium under a pressure of
magnitude, Pi, is given by the relations [2]:
rer r; h
r 2
i
P;
r
reh r; h
r 2
i
P;
and
uer r; h
1 v ri2
P
E
r
Solution:
If the cavity in question is small compared with the dimensions of the viscoelastic solid, the elastic boundary-value problem approximates the viscoelastic
one, and its solution may be used to construct the one to the problem posed.
Although only tractions are prescribed on the boundary of the hole, the remote
displacement field is zero. That is, this problem is a mixed boundary-value
problem. Hence, according to (9.23) the elastic and viscoelastic stresses at each
point and time are the same; that is:
rr r; h; t
r 2
r 2
i
i
Pf t; rh r; h; t
Pf t
r
r
The actual displacements are obtained from (9.24): uio(t) = uifu(t), and the
convolution inverse of (9.28): fu t m1
N dfT s, using the normalization E t
ER mN t: This gives the solution ur r; h; t
ri2
r
P 1v
ER
Rt
0
m1
N t s df sds.
( r, )
ri
9.9
9.9
Integral-Transform Correspondence Principles
231
Elasticviscoelastic correspondence principles are a convenient means of constructing solutions to quasi-static viscoelastic boundary-value problems from
elastic solutions. The steady-state case falls under this general definition.
Inasmuch as the Laplace and Fourier transformations are mathematically
related, the elasticviscoelastic correspondence principle presented here may be
interpreted as derived from either integral transform. For definiteness, we let
s denote the transform parameter and understand it to mean s in case of the Laplace
transform, or jx if the Fourier transform is meant. In the latter case, j stands for the
pffiffiffiffiffiffi
imaginary unit: j 1. Also, the same symbol is used to represent a quantity
and its transform, distinguishing the latter with an over-bar.
Viscoelastic boundary-value problems in transformed space can be derived by
applying the corresponding transformLaplace or Fourierto the field, compatibility, and constitutive equations of linear viscoelasticity. Proceeding thus,
interchanging the order of integration and differentiation, implied, respectively, by
the integral transforms of the field equations, there result:
(a) For the equilibrium equations and boundary conditions (9.4):
o
i s 0;
ij s B
r
oxj

ji sS Tio s;
nj r
T
ui sjSu uoi s
9:29
(b) For the general stressstrain constitutive equations, such as (9.6) or (9.7):
ij s skM sekk sdij 2s
r
lM seij s
9:30
ij s psdij 2s

r
lM seij s
9:31
(c) For the strain-displacement relationships (9.5):

eij s

1 o
o
ui s
uj s
2 oxj
oxi
9:32
(d) For the strain-compatibility conditions (9.8):
eij;kl s ekl;ij s eik;jl s ejl;ik s
9:33
232
As is apparent from these expressions, the transformed viscoelastic field

equations are identical to those of linear elasticity, if the transforms of all viscoelastic variables are associated with their elastic counterparts, and if the products
of the transforms of each material property function and the transform variable
s or jxis associated with the corresponding elastic constant. From this analogy
follows that the transformed solution to a viscoelastic boundary-value problem
may be obtained directly from the solution of the same problem posed for an
elastic material, by replacing the boundary data with their transforms, and each
elastic constant with the transform-parameter multiplied transform of the corresponding viscoelastic property function. The solution to the original problem is
then obtained by inverting the transformed solution back to the time domain.
In closing, it is remarked that the present approach is applicable to displacement, traction and mixed boundary-value problems, and that, unlike the methods
presented for incompressible materials and for materials with synchronous moduli,
or the separation of variables method, the correspondence principle does not
require the Poissons ratio to be a constant. One limitation of elasticviscoelastic
correspondence principles that are based on integral transforms, however, is that
the prescribed conditions must not change type during the deformation process.3
This excludes from application cases such as contact, where the data prescribed on
some parts of the boundary change from displacement to traction and vice versa,
during deformation.
Example 9.4 A cantilever beam of length l and constant cross-sectional area,
made of a viscoelastic material, is subjected to an oscillating load
P(t) = Posin(xt) applied at its free end, as indicated in Fig. 9.3. Use the correspondence principle for steady-state conditions to obtain an expression for the
steady-state displacement of the tip of the beam, if the materials relaxation
modulus is E(t).
Po sin(t)
l, I, E (t)
Solution:
Following the correspondence principle, first write the elastic solution:
Po l3
Do 3E
. Next, replace the amplitudes of the displacement, load, and modulus by
oI
3
P l
l
their complex forms, using that E*C* = 1. This produces: D 3E
I C P 3I .
The viscoelastic solution is obtained by noting that the load actually applied
This limitation is implicit also in the methods presented in Sects. 9.6 and 9.7, for displacementonly boundary conditions. .
9.9
233
corresponds to the imaginary part of the complex load P* = Po{cos(xt)

?
n
o j
l3
sin(xt)}. The solution sought is, therefore, Dt ImfD g Im C P 3I
Po l3
3I
ImfC0 jC 00 cosxt j sinxtg. Or, carrying out the indicated algebra:

3
Dt P3Io l C 0 x sinxt C 00 x cosxt.

Example 9.5 The stress and displacement in the direction of pull in a very wide,
centrally cracked plate of an elastic material, in the vicinity of each crack tip are

pffiffiffi
given by the following relations [7]: (Fig. 9.4) re22 t ro t 2ra cos h2

1 sin h2 sin 3 h2 and

u2 t
ro t
Ge
rffiffiffiffi

ar
h
h
sin
2 2ve cos2
:
2
2
2
o (t )
o (t )
In these expressions, r and h represent the radial and angular coordinates

measured from a polar system centered at the crack tip, and Ge stands for the
materials shear modulus. Using the correspondence principle obtain the solution
for a plate of a linear viscoelastic solid with shear relaxation modulus G(t) and
constant Poissons ratio v.
Solution:
According to the correspondence principle, the transformed solution takes the
form:
rffiffiffiffi

a
h
h
h
22 s r
o s
r
cos
1 sin
sin 3
2r
2
2
2
rffiffiffiffi

o s ar
r
h
h

u2 s
sin
2 2svs cos2
2
2
2
sG
rffiffiffiffi

ar
h
h
G s
ro s
2 2svs cos2
sin
sC
2
2
2
234
Where sG sC1G has been used to recast the expression for the transformed
displacement.
Since Poissons ratio is a constant, noting that the s-multiplied value of the
transform of a constant is the constant itself, the viscoelastic solution becomes:
rffiffiffiffi

a
h
h
h
cos
1 sin
sin 3
2r
2
2
2
rffiffiffiffi
Z t
ar
h
h
o
u2 t
CG t s ro sds
sin
2 2v cos2
2
2
2
os
r22 t ro t
0
As expected of a traction boundary-value problem, the stress solution is independent of material constitution, but the displacement and strain fields are not. In
the context of the mechanics of fracture, the solution is commonly stated in terms

ffiIffiffiffi cos h2
22 s pKffi2pr
of the stress intensity factor; so that, for instance: r

1 sin h2 sin 3 h2 . According to this, the stress intensity factor, KI, which represents the amplitude of the stress field at a crack tip, must be independent of
material properties.
9.10
The Poissons Ratio of Isotropic Viscoelastic Solids
As discussed in Sects. 9.7 and 9.8, the assumption of a constant Poissons ratio
allows any type of viscoelastic boundary-value problem to be solved directly from
the solution of the same problem posed for an isotropic elastic solid. For most
viscoelastic materials, however, the assumption of a constant Poissons ratio is not
supported by experimental evidence and is justified solely by the simplicity it
confers upon the solution process. A more realistic assumption is that the volumetric response of many polymers, particularly rubbers, is very nearly elastic; and
it is the bulk modulus which may be taken as constant in time. This observation is
used here to examine the nature of the viscoelastic Poissons ratio of an isotropic
material.
Applying the correspondence principle
to1the elastic relation K = E/{3(1-2v)}
Inverting this expression results
and rearranging, produces that: m 12 1s 6K
E.
in the following relationship between the uniaxial tensile modulus, E(t), and the
Poissons ratio function, v(t):
1
1
m t H t
Et
2
6K
9:34
Since E(t2) B E(t1), whenever t2 C t1, [c.f. Chap. 2], expression (9.34) shows
that v(t) is an increasingor at least, a non-decreasingfunction of t. In other
words, its glassy value, vg, is smaller than its long term, or equilibrium value, v?.
9.10
The Poissons Ratio of Isotropic Viscoelastic Solids
235
This expression also shows that the long-term Poissons ratio of a viscoelastic fluid
is equal to since for such materials E? = 0. In addition, since the long-term
modulus of elastic solids never vanishes, relation (9.34) also shows that the
Poisson ratio of any linear isotropic viscoelastic solid is never larger than . These
observations reinforce the notion that the Poisson ratio of isotropic viscoelastic
materials has the time dependence shown in Fig. 8.3. Expression (9.34) also shows
that the Poissons ratio of incompressible materials, which by definition should
posses an infinite bulk modulus, is also . The same would be approximately true,
as well, for materials which have a bulk modulus orders of magnitude larger than
the uniaxial tensile or shear relaxation moduli. Such materials, like many rubber
compounds, are usually referred to as nearly incompressible.
9.11
Problems
P.9.1 Use the correspondence principle to derive the equations governing the
bending response of straight viscoelastic beams from the three equations for elastic
Mx;t
o
d
d2
Mx; t, qx; t dx
Vx; t, and dx
beams: V x; t ox
2 yx; t
EI ; where,
V, M, and q are the shearing force, bending moment, and load per unit length on
the beam, respectively, and y is the deflection.
Answer:
V x; t
o
d
d2
1
M x; t; qx; t V x; t; 2 yx; t CE t s dMx; s
ox
dx
I
dx
Hint:
Apply the correspondence principle to the elastic expressions and write the
1=sCE . Invert these
transformed equations for viscoelastic beams using thatsE
expressions, to arrive at the viscoelastic forms.
P.9.2 Write the auxiliary elastic constitutive equation in terms of the uniaxial
tensile relaxation modulus for use in a displacement boundary-value problem of
viscoelasticity.
h
i
ER
Answer: rVij t 1m
m
12m eVkk dij
eVij .
Hint:
Proceed as in Example 9.2 using the normalization (9.19) and the elastic form
(9.21).
P.9.3 In Problem P8.7, we showed that the directions of principal stress and
principal strain in a viscoelastic material do not coincide in general. Under what
conditions would these directions coincide, if ever?
Answer: The directions of maximum principal stress and maximum principal
strain will coincide for viscoelastic materials with synchronous moduli, including
incompressible materials, when the boundary data are specified as proportional
loading; that is, either as roij xfT t, or as uoi xfu t.
236
hrmax
2eoxy l df
2exyo
hemax

exxo eyyo l df
exxo eyyo
2r
xy
Hint: Evaluate hrmax rxx r
inserting the terms from constitutive Eq. (9.27):
yy

rij t roij fT t kR eokk dij 2lR eoij mN t s dfu s, cancel the common
2e
m df
oxy N
u
terms to arrive at hrmax eoxx e

oyy mN dfu
2e
2e
f t
oxy
oxy u

hemax .
eoxx eoyy eoxx eoyy fu t
P.9.4 How would you use the finite element method to obtain the viscoelastic
solution to a traction boundary-value problem of a material with a constant
Poissons ratio, v, if the traction data were prescribed in the form Fo(xi)f(t)?
Answer:
The finite element analysis would have to be carried out using the Poissons
ratio, v, of the viscoelastic solid, and an arbitrary shear modulus GR, computed
such that ER = 2(1 ? v)GR. The boundary data would have to correspond to
Fo(xi), or be proportional to it. This leaves the temporal part as fT(t) = f(t).
Assuming the finite element solution is given by uFEi, rFEij, and eFEij, one would
use (9.23) and (9.24), to write the viscoelastic solution as:
rij x; y; z; t rFEij x; y; zfT t

1
ui x; y; z; t uFEi x; y; zfu t uFEi x; y; z
lR
eij x; y; z; t eFEi x; y; zfu t
1
eFEi x; y; z
lR
Zt
0
Zt
0
m1
N t s
m1
N t s
o
fT sds
os
o
fT sds
os
P.9.5 The radial stress in a semi-infinite elastic half space subjected to

a concentrated load, P, at the origin of coordinates, as indicated in Fig. 9.5) is
e
e
given by [5]: rerr t Pt
2p 1 2v f r; z gr; z in which v is the Poissons
ratio of the elastic material, and f and g are known functions of position only. (a)
What would the radial stress be in a viscoelastic half space subjected to the same
boundary values? (b) What difference would there be if the viscoelastic half space
had a constant Poissons ratio?
P(t)
r
z
9.11
Problems
237
Answer:
(a) First assume the Poissons ratio function for the half space is v(t); then, invoke
the correspondence principle to write the transformed solution and invert to
arrive at the expression:
(
)
1
Zt
d
Ptf r; z gr; z 2f r; z mt s Psds
rrr t
2p
ds
0
(b) If the half space had a constant Poissons ratio, the viscoelastic solution would
revert to an elastic-like solution, since v would pull out from under the integral
in (a).
P.9.6 Formulate the method of separation of variables for the mixed boundaryvalue problem in which the surface tractions and the body forces have different
dependence on time.
Answer:
(a) Solution at zero body force

o T

rij 0; nj rTji t Toio fTT t; uTi tSu uooi fuT t
oxj
ST
rTij t rToij fTT t; uTi t uToi fuT t; eTij t eToij fuT t
with: fuT t l1
Rt
0
o T
m1
N t s os fT sds
(b) Solution at zero boundary data

o B

rij Bi 0; nj rBji t 0; uBi tSu 0
oxj
ST
rBij t rBoij fTB t; uBi t uBoi fuB t; eBij t eBoij fuB t
with: fuB t l1
Rt
0
o B
m1
N t s os fT sds
Hint:
Split the solution fields into two parts: one satisfying the prescribed boundary
data at zero body force, and another for which the only external force is the body
force and the boundary values are identically zero. The viscoelastic solution will
be the sum of the two solutions. Specifically, using the superscripts B and T,
respectively, for the solutions to the problem with and without body force, write:
rij rBij rTij , ui uBi uTi , and eij eBij eTij . Split the original problem into the
two parts and separate variables to arrive at the desired result.
238
References
1. D.C. Leigh, Nonlinear Continuum Mechanics, (McGraw-Hill, NY, 1968) pp. 117138
2. E. Volterra, J.H. Gaines, Advanced Strength of Materials, (Prentice-Hall, NJ, 1971), pp. 1219,
pp. 177186
3. R.M. Christensen, Theory of Viscoelasticity, 2nd Ed. (Dover, 1982), pp. 3741
4. Y.C. Fung, Foundations of Solid Mechanics, (Prentice Hall, NJ, 1965) pp. 99103
5. I.S. Sokolnikoff, Mathematical Theory of Elasticity, (McGraw-Hill, US, 1956), pp. 7179
6. A.C. Pipkin, Lectures on Viscoelasticity theory, (Springer, Berlin, 1956), 2nd Ed. pp. 7779
7. T.L. Anderson, Fracture Mechanics, Fundamentals and Applications, 2nd Ed. CRC, CS,
1956), pp. 5155
10
Wave Propagation
Abstract
This chapter examines the propagation of harmonic and shock waves in

viscoelastic materials of integral and differential type. For simplicity, the
different topics are introduce in one dimension, presenting the balance of linear
momentum across the shock front, and the jump equations in stress, strain and
velocity without obscuring the subject matter. As must be expected, harmonic
waves in viscoelastic media are always damped. Also shown is that shock
waves travel at the glassy sonic speed of the viscoelastic material in which they
occur; that is, at the speed of sound in an elastic solid with modulus of elasticity
equal to the glassy modulus of the viscoelastic material at hand.
Keywords
Wave Harmonic Cyclic Shock

Attenuation Momentum Balance
10.1
Speed Frequency Front Jump
Introduction
This chapter examines in some detail the propagation of harmonic waves and
shock waves in viscoelastic materials. To keep the presentation simple, only onedimensional conditions are assumed in what follows. Harmonic waves are discussed in Sect. 10.2, and shock waves in Sect. 10.3. In both cases, the treatment
covers materials of integral and differential types.
Waves develop in a body as a result of the conditions imposed on its boundaries. Harmonic waves result from cyclic boundary conditions, while shock waves
are due to discontinuous boundary values. As must be expected, harmonic waves
in viscoelastic media are always damped. Also, shock waves travel at the glassy
sonic speed of the viscoelastic material in which they occur; in other words, at the
239
240
10 Wave Propagation
speed of sound in an elastic solid having modulus of elasticity equal to the impact
or glassy modulus of the viscoelastic material at hand.
10.2
Harmonic Waves
In this section, we examine the propagation of harmonic waves in a viscoelastic

bar. Such waves result from cyclic boundary conditions and their speed thus
depends on the forcing frequency. The treatment is restricted to longitudinal
vibrations, but the same principles apply to lateral and torsional vibrations.
In longitudinal vibrations, it is assumed that bar elements can only extend or
contract along the original axis of the bar; that plane cross sectionsperpendicular
to the original axis of the barremain plane during motion; and that the inertia of
transverse motion is disregarded. Under these conditions, as can be inferred from
Fig. 10.1, balance of linear momentum leads to the single partial differential
equation:
o
o2
rx; t q 2 ux; t
ox
ot
10:1
Where r is the axial stress, u, the axial displacement, and q stands for the
density of the material, which may depend on position, but, for simplicity, is
assumed constant in time.
Consistent with the balance laws [c.f. Appendix B], the relationship in (10.1) is
valid irrespective of material constitution. The type of material at hand enters
through the constitutive equation, as discussed later on.
The purpose here is to examine solutions of Eq. (10.1) associated with either
displacement or traction boundary conditions which vary cyclically in time. Thus,
aside from fixing one end of the bar, when needed, boundary conditions at the
other end, x = 0, say, will be of one of the following forms:
Fig. 10.1 Balance of linear

momentum in a onedimensional bar
dx
u(x,t )
x( x , t ) +
x ( x , t)
dx
x ( x , t ) dx
10.2
Harmonic Waves
241
8
< U cosxo t
or :
u0; t
:
U sinxo t
10:2
if a harmonic displacement of amplitude U is prescribed, or:

8
< S cosxo t
r0; t
or :
:
S sinxo t
10:3
when a harmonic surface traction of intensity S is applied.

In any event, since a frequency-domain analysis requires that both loading and
response be specified in complex formwhich includes cosine and sine terms
instead of (10.2) we would use:
u 0; t Uejxo t
10:4
whereas (10.3) would be replaced with:

r 0; t Sejxo tu
10:5
Clearly, the solution corresponding to a cosine boundary condition should be

real part of the complex solutionthe cosine component of ejxt = cosxt ? jsinxt;
and, for the same reason, the imaginary part of the complex solution would
represent the solution to a sine forcing function [c.f. Chap. 4].
10.2.1 Materials of Integral Type

In this case, using the strain-displacement relation exx ou=ox, converts (10.1)
into:
Et s o
h o2
i
o2
ux;
s
q
ux; t
ox2
ot2
10:6
Since the boundary condition is a circular function, the solution will be a

combination of circular functionsthat is, cosines as well as sinesof the same
frequency. In addition, one would expect the intensity of the forcing function to be
less at sections farther from the point of application. That is, on physical grounds,
one would expect the solution to (10.6) to be a decreasing function of position.
Hence, we seek a solution in the form:
u x; t Aejxo tkx
10:7
242
10 Wave Propagation
In this expression, xo is the forcing frequency and, k, as will be seen subsequently, is a complex quantity whose components represent the rate of amplitude
decay and the wave speed. Taking (10.7) into (10.6) leads to:
k2 E jxo u x; jxo t x2o qu x; jxo t
From which k can be obtained after simplification, as:

k xo
r
q

E jxo
10:8
This shows that there are two values of kone being the negative of the
otherwhich are complex because E* is a complex number. In what follows, we
reserve the name k for one of the roots of (10.8), the other beingk. With that:
0
k1 = k = k ? j k00 , k2 = -k, and:
00
00
u x; jxo t A1 ek x ejxo tk x A2 ek x ejxo tk x
10:9
The physical interpretation of this form is that under a cyclic forcing function
there will always be two waves running in opposite directions along the bar. The
first term in (10.9) represents a wave running in the direction of increasing x, and
the second term is a wave moving in the opposite direction. The solution (10.9)
also shows that each wave decreases exponentially as it advances; the imaginary
part, k00 , of k is responsible for the waves rate of decay; and, by dimensional
homogeneity, k0 must be inversely proportional to the wave speed, as will be
shown later. Finally, in the case of an infinitely-long bar, the solution can contain
only the first term, since the second term must be omitted for regularity, to keep
the solution from blowing up.

In this case, the stressstrain law is introduced by applying the differential operator
P to the equation of motion (10.1), thus:
P
ho
i
h o2
i
rxx x; t P q 2 ux; t
ox
ot
Using that P(r) = Q(e), and exx ou=ox, and assuming the materials density
to be time independent, transforms the dynamic equation into:
h o2
i
h o2
i
Q
ux;
t
qP
ux;
t
ox2
ot2
10:10
10.2
Harmonic Waves
243
On the same arguments as before, the solution of (10.10) is sought in the form
(10.7); so that:
k2 Qjxo u x; jxo t x2o qPjxo u x; jxo t
Or, since E*(jxo) = Q(jxo)/P(jxo) = Q*/P* [c.f. Chap. 4]:

s
r
Pjxo
q
xo
k xo q
Qjxo
E jxo
10:11
This expression is exactly the same as that in (10.8), which was derived for a
material with constitutive equation of hereditary integral type. Hence, irrespective
of the form of the materials stressstrain law, once the complex modulus E*(jxo)
is made available, k can be evaluated from either of the equivalent forms (10.8) or
(10.11).
In particular, since E*(jxo) = E0 (xo) ? jE00 (xo), with tand = E00 /E0 [c.f.
Chap. 4], multiplying and dividing the quantity under the radical sign in (10.11) by
the complex conjugate of E* leads, after some algebraic manipulations, to the
following form, which is more adequate for numerical computations1:
k xo
r
q
xo
ejd=2
ejd2kp=2 q
jjE jxo jj
jjE jxo jj
10:12
In this expression, the quantity under the radical sign on the far right has the
dimensions of speed. In analogy with the elastic case [1], the complex sonic speed
is defined as follows:
s
E jxo
c jxo
jjc jxo jjejd=2
q
10:13
This gives the amplitude ||c*(jxo)|| of the complex sonic speed, as:
s
jjE jxo jj
jjc jxo jj
q
10:14
With the definitions introduced in (10.12) and (10.13), k, in (10.11), may be put
in the form:
k
xo

jjc jx
o jj
ejd=2
10:15
The root of index n of a complex number, z rejh2kp , is given by: z1=n r 1=n ejh2kp=n ,
as shown in Appendix A.
244
10 Wave Propagation
which serves to express k0 , and k00 as:

0
xo
cosd=2;
jjc jxo jj
00
xo
sind=2
jjc jxo jj
10:16
Two limiting cases of the complex sonic speed amplitude, (10.14), which are of
interest, pertain to forcing functions of very low and very high frequencies. Using
the limit properties of E* [c.f. Chap. 4], the following results are obtained:
s s s
0
E jxo
E1
E1

lim
c1
xo !0
q
q
q
10:17
and:
s s s
E jxo
Eo0
Eg
cg

lim
xo !1
q
q
q
10:18
These expressions indicate that waves induced by harmonic forcing functions

with very low or very high frequencies propagate as if the material were linearly
elastic. Harmonic waves of very low forcing frequency will travel at the long-term
sonic speed of the viscoelastic material at hand, while those of very high frequency
will travel with the materials glassy sonic speed.
The preceding discussion shows that, irrespective of whether the constitutive
equations are of integral- or differential-operator type, harmonic waves in a viscoelastic substance will always be a superposition of decaying sine and cosine
waves of the form (10.7)or in more detail, (10.9) and that the wave speed and
the rate of decay depend on the materials density and on the forcing frequency
through the complex modulus.
Example 10.1 Find the magnitude of the complex sonic speed of an elastomer
whose complex modulus at the frequency of interest is E*(jx) = 10 ? 3.5 j MPa,
if its mass density is 1,660 kg/m3.
Solution: The sonic speed depends on the magnitude of the complex modulus,
||E*|| & 10.6 Mpa, and the mass density: q = 1,660 kg/m3; thus, using (10.14),
it follows that: ||c*|| & 79.9 m/s.
Example 10.2 A bar of a viscoelastic material with tensile relaxation modulus
E(t) = 70 ? 35e-t/0.1 kPa, and mass density = 1,800 kg/m3 is subjected to the
following boundary conditions(2):
u0; t U cosxo t;
uL; t 0
xo 25 rad/s
10.2
Harmonic Waves
245
Determine the bars response at xot = 0, p/4, p/2, 3p/4, p, if the length of the bar
is such that L = 7p/k0 (&6.5 m).
Solution: First, the boundary conditions are expressed in the complex form
Uejxo t . Noting that the actual condition, U cos(xot), corresponds to the real
part of U eejxo t , the solution will be given by the real part of (10.9). Putting
the boundary conditions in it leads to a system of equations in the unknown
amplitudes, A1 and A2:
U A1 A2
00
00
0 A1 ek L A2 ek
Solving for A1 and A2 and plugging the results back into 10.9), leads, after some
manipulation, to the complex form [cf. Appendix A]:
u x; t U
sin kL x jxt
e
sin kL
Since k is a complex number, to decipher this expression requires the circular

functions of a complex variable [c.f. Appendix A]: sinz = sin(x+jy) = sin(x)
cosh(y) ? jcos(x) sinh(y) and similarly for cos(z). After some complex and
trigonometric algebra, there results:
u x; t
U
h cos k0 x cosh k00 2L x
0
cosh 2k L cos 2k L
cosh k00 x cos k0 2L x cosxo t
sin k0 x sinh k00 2l x sinh k00 x sin k0 2L x sinxo t
00
jfcos k0 x cosh k00 2L x cosh k00 x cos k0 2L x sinxo t

sin k0 x sinh k00 2l x sinh k00 x sin k0 2L x cosxo tgi
The solution corresponding to the boundary values of the present problem
(which pertain to a cosine forcing function) is the real part of the previous
expression. The parameters of this problem: E* = 100.17 ? 12.07 j MPa;
d & 0.120 rad; ||c*|| & 7.487 m/s; k0 & 3.333; and k00 & -0.200, yield the
solution, which is displayed in Fig. 10.2 at a few selected times.
Also shown by this solution is that although the wave amplitudes decrease from
the left to right, they increase upon first entering the bar. This indicates that the bar
is at or near resonance. In fact, an elastic bar of the same density and modulus
M = E0 (xo) would be in resonance under the prescribed boundary conditions and
would exhibit standing waves of infinite amplitude [2].
246
10 Wave Propagation

harmonic waves in given
viscoelastic bar
1.5
u ( x, t ) / Uo
1.0
0.5
0.0
-0.5
-1.0
= 0
= /4
= /2
= 3 /4
-1.5
10.3
Shock Waves
This section examines the propagation of shock waves in a bar made of a linearly
viscoelastic material. When the governing equationsthe equations of motion in
the present caseare of hyperbolic type, any abruptness in the boundary values
show up as discontinuities in the first derivatives of the displacement field. These
discontinuities travel along surfaces, which are known as the characteristics of the
differential equation, and which are in themselves smooth functions of the spatial
coordinates and time. In continuum mechanics, these discontinuities are commonly referred to as shock waves.
A rigorous theory admits the propagation of shock waves only in unbounded
media. In a finite body, multiple reflections of waves from the boundaries give rise
to dispersion and blur the shock front. To avoid these complications and still use a
one-dimensional model in the study of shock waves, we assume a bar in which
plane sections remain plane, there is no transverse inertia, nor lateral motion of any
sort, not even due to Poissons effect; and most of all, such that waves do not
bounce off its lateral surfaces. These assumptions allow the study of shock waves
in an idealized bar in virtually the same way as in an infinite medium.
Per the previous definitions, a shock wave through the one-dimensional bar
model will be a smooth function, x = Y(t), in the (x,t) plane. The first partial
derivative of the shock wave with respect to time, which exists by the smoothness
assumption, represents the wave speed:
d
d
x Yt ct
dt
dt
10:19
As long as the bar does not come apart, the displacement field, u(x,t), will be
continuous throughout. However, the first partial derivatives of the displacement
field with respect to time and space will exhibit jump discontinuities across the
shock front.
10.3
Shock Waves
247
The jump of a function, f, at a point, is denoted by the symbol [f] and defined as
the difference between the values of the function at points infinitely close to the
point in question, but situated on opposite sides of it. More explicitly, the magnitude of the jump of a function f at x = y is the difference between the right- and
left-hand side limits of the function at x = y:
f f f
10:20
where the one-sided limits are defined as follows [4]:

f lim f x;
x!y
x\y
f lim f x
10:21
x!y
x[y
Once in motion, a shock wave travels forward from one end of the bar to the
other; and upon reaching it, reflects and travels back to the point of origin. The
wave fronts or characteristics of the differential equation of motion can be represented by straight lines in the x - t plane:
x
nt ;
c
gt
x
c
10:22
The equation n = constant, represents a wave moving in the direction of

increasing x, and g = constant corresponds to a wave front moving backwards,
toward the origin, as illustrated in Fig. 10.3.
As seen in Fig. 10.3, on the forward wave n = constant, dn = 0; which implies
that dt = dx/c. Likewise, on the return wave g = constant, dg = 0, and so, dt =dx/c. Hence:
8
<
o
ox :
1o
c ot ;
1c oto ;
on n const
10:23
on g const
of
Since the total differential of a function f (x,t) is df ox
dx ofot dt, (10.23)
implies that:
Fig. 10.3 Graphical

representations of shock
waves in the x-t plane
(a)
(b)
t
dx
dt=-dx/c
dt=dx/c
dx
248
10 Wave Propagation
8 of
1 of
Df < ox c ot ;
Dx : of 1 of
ox c ot ;
on n const
10:24
on g const
And also that:

8 of
of
Df < ot c ox ;
Dt : of
of
ot c ox ;
on n const
10:25
on g const
Additionally, (10.24) and (10.25) lead to:

8 D
c ; on n const
D < Dx
Dt :
D
c Dx
; on g const
10:26
The differential equation of motion (10.1) applies on either side of the front, but
not on the front itself. For ease of reference that equation is repeated here, with a
change in notation from displacement to velocity: q2u\qt2 : qv\qt; thus:
o
o
rx; t q vx; t
ox
ot
10:27
To see what happens at the front, the balance of momentum equation is used in
its global form. To this end, consider a portion of the bar which straddles the wave
front, as shown in Fig. 10.4, and let the size of the element of bar approach zero.
Eq. (10.27) then leads to:
2
ArX1 t; t ArX2 t; t A
o6
4
ot
YZ t
ZX2
qvxt; tdx
X1
3
7
qvxt; tdx5 a
Y t
To cast this equation in a more useful form, cancel out the common factor A,
and carry out the integration using Leibnitz rule [c.f. Appendix A]. Let x1 ? Yfrom the left, and x2 ? Y+ from the right, and, recognizing the continuity of the
Fig. 10.4 Balance of linear

momentum across the wave
front
x = Y(t)
wave front
(x2,t)
(x1,t)
x = x1
x = x2
10.3
Shock Waves
249
shock wave (which requires that: dY-/dt = dY+/dt = c), re-order terms and arrive
at the balance of linear momentum across the front:
r qcv
10:28
The continuity condition at the shock front is obtained by differentiating the

jump of the displacement. Since the displacement is continuous across the front,
[u] = 0 there; and so must be, its total derivative, Du/D. Using (10.24):
Du
ou
1 ou
Dx 0 ox c ot , to evaluate the jump leads to:
ou
1 ou
1
; or : e v
ox
c ot
c
10:29
Combining the last two expressions leads to a relation between the jumps in
stress and strain across the front, which seems independent of the constitutive
equation.
r qc2 e
10:30
Expressions (10.2810.30) relate the jumps in stress, strain, and velocity.

Material type or constitution enters directly through the density, per (10.28), and
the stressstrain relations, through (10.30).
10.3.1 Materials of Integral Type

Without loss of generality, the stress-strain law in this case may be represented in
the integral form [c.f. Chap. 2]:
rx; t Eg ex; t
Zt
oEt s
ex; sds
os
10:31
Since the relaxation function E() is continuous across the front, and the strain
history, e(x,s), is continuous for all s except s = t, the integral vanishes across the
jump. Thus:
r Eg e
10:32
This, together with (10.30) yields the expression for the speed of the shock
wave, which is of the same form as for shock waves in a linearly elastic bar [1]:
s
Eg
c
q
10:33
250
10 Wave Propagation
As indicated before, shock waves in a viscoelastic bar travel with the materials
glassy sonic speed.
Example 10.3 Obtain the speed of longitudinal shock waves in a bar of viscoelastic material with tensile relaxation modulus E(t) = 4.2 ? 39.3e-t/0.25 MPa
and density 1,800 kg/m3.
Solution:The speed of the shock front is given by (10.33). The glassy modulus of
the material in this case is 43.5 MPathe sum of the equilibrium and the
transient components of the tensile relaxation modulus. Hence, c = (43.5106/
1,800)1/2 & 155.5 m/s.
A shock wave moving in a viscoelastic material would change its amplitude
both as the wave progresses through the medium, and as time goes by. This may be
proven by constructing a differential equationin the spatial coordinate or in time,
respectivelyout of the stress-velocity jump equation and examining its solution,
as follows.
Differentiating the stress-velocity jump equation (10.28) across the front and
using (10.24), leads to:
D
D
r qc
v
Dx
Dx

or
1 or
D
v
qc
ox
c ot
Dx
Or:
The second term on the left-hand side of this expression is evaluated from the
stress-strain law, (10.31), using Leibnitz rule to differentiate under the integral
sign, to get:
8
9
2 t
3

Z 2
<
=

1 or
1
oe
o Et s
oEt s
Eg
ex; sds5
ex; t
c
4

;
c ot
c:
ot
otos
os
st
0
Since the integrand in the second term on the right is continuous for all s,
except s = t, its jump is zero. Therefore:

1 or
1
oe
1 oEt s
ex; t
Eg

c ot
c
ot
c
os st
Upon using that: oe=ot ov=ox, and (10.33): Eg = qc2, on the first term of the
right-hand side of the last expression; together with (10.29): [e] = -[v]/c, on the
second term, there results:
10.3
Shock Waves
251

1 or
ov
1 oEt s
1
qc
v

c ot
ox
c
os
st c
Taking this result into (b), using the equation of motion (10.27) to replace
qr/qx, collecting terms and recalling the definition (10.24) of the derivative, D/Dx,
across the front, leads to:
D
1
v C0v 0;
Dx
2c
C0
1 o
1 d
Et sjst
Etjt0 10:34a
Eg os
Eg dt
This is a linear differential equation of first order with constant coefficients. It

thus has an integrating factor eC(0)x/(2c)[c.f. Appendix A]. This allows casting it as
d [v eC(0)x/(2c)] = 0; from which follows the solution:
C0
v Ae 2c x
10:35a
Since C(0) [ 0, this relationship shows that the jump in velocity decreases as the
wave moves in the direction of increasing x.
Example 10.4 Prove that C(0), as defined in (10.34a) is always positive for
relaxation functions expressed in Dirichlet-Prony series, as sums of exponentials.
o
Solution: Use (10.34a): Ct s 1=Eg os
Et s, and the Prony-series form
N
P
d
of the relaxation modulus:Et Eg Ei 1 et=si to compute: ds
Et s
i1
N
P
Ei
i1
si
ets=si and arrive at the result: Cs t C0 E1g
N
P
Ei
i1
si .
Use that since
all the Ei, and si [ 0, it follows that C(0) [ 0.

The attenuation of shock waves with time may be examined using Eq. (10.26):
D/Dx = (1/c)D/Dt; to write (10.34a) as an equation in time:
D
1
v C0v 0;
Dt
2

1 d
1 d
C0
Et sjst
Et
10:34b
Eg ds
Eg dt
s0
Proceeding as before, the solution in time becomes:

v Be
C0
2 t
10:35b
This result indicates that the amplitude of a shock wave decreases with time, as
expected it should in viscoelastic materials. Using (10.28) with either of Eq.
(10.35a, b) establishes the stress jump:
252
10 Wave Propagation
8
1
< qcBe2C0t
r
or
:
1
qcAe2cC0x
10:36
Similarly, (10.29) and (10.35a, b) produce the strain jump:

8
1
< 1c Be2C0t
e
or
: 1 2c1 C0x
c Ae
10:37
A wave front may also move from right to left, on the characteristic
g = constant. In that case, the wave velocity will be negative, and the previous
equations apply as well.
Additional information about the response of the bar can be gained solving
(10.27). To this end, insert into it the constitutive equation rt Et s
o
d ox
us [c.f. Chap. 2] writing:
Z
Et s

o o2
o2
u
ds
q
u
os ox2
ot2
Now take the Laplace transform, assuming at-rest initial conditions to get [c.f.
Appendix A]:
Essu00 x; s qs2 ux; s 0
Or:
u00 x; s
qs
ux; s 0
Es
10:38
This is a partial differential equation in x for the transformed function u(x, s).
Because only derivatives with respect to x enter this equation, its solution may be
obtained by methods for ordinary differential equations. Employing differential
operators leads to a characteristic equation with roots [c.f. Appendix A]:
k
r
r
qs
q
s
sEs
Es
10:39
These roots imply a general solution of the form:

ux; s A1 sekx A2 sekx
10:40a
Inserting the Laplace-transformed strain, ex; s oux; s=ox, leads to the

transformed stress:
10.3
Shock Waves
253
x; s kA1 seksx A2 seksx

r
10:41a
From here, the solution in physical space-time would be obtained, at least in

principle, by Laplace-transform inversion. The mathematical manipulations
required to accomplish this, however, become intractable for all but the simplest of
constitutive models.
Lastly, for a semi-infinite bar, the root with positive sign in (10.40a) has to be
discarded.2 It represents a wave that starts traveling from the right at time zero.
This is inconsistent with at-rest initial conditions. Thus, the solution for a semiinfinite bar in transformed space is simply:
p
ux; s Aseksx Ases q=sEsx
10:40b
x; s ksAseksx
r
10:41b
Example 10.5 Find the Laplace-transformed response of a semi-infinite bar of a

viscoelastic material subjected to a discontinuous velocity of amplitude V at its
end.
Solution:First, note that v = qu/qt, so that: vx; s sux; s. The transform of
the boundary value v(0,t) = VH(t) is v0; s V=s; u0; s V=s2 . Evaluating
(10.41b) at x = 0 establishes A(s) = V/s2. Thus, the Laplace-transformed
solution is ux; s sV2 ekx . Note that, per (10.39), k is a function of the
transformed modulus, so the inverse transform of ux; s is not simply the
transform of 1/s2 times Ve-kx.
Example 10.6 Find the Laplace transform stress response of the semi-infinite bar
of Example 10.5 if the bars material is of Maxwell type, with modulus Eg and
viscosity g.
Solution:Using that r sE e sE ou=ox together with the results of
Example 10.5, leads to the transformed stress: r E Vs kekx . The relaxation
modulus for the Maxwell model is given by [c.f. Chap. 3]: Et Eg et=s , with
Eg
and hence, from (10.39), using that
relaxation time s = g/Eg, so that: E s1=s
q
p
q
2
1
Eg = qc , k Eg ss 1=s c ss 1=s. And the transformed stress
p p
becomes:3: r qVc ss 1=se ss1=sx=c .
2
In mathematical terms, this has to be so because every Laplace transform must vanish at s??,
per the limit theorems [c.f. Appendix A].
3
In this case, the solution in physical space can be established from a table of Laplace transforms

and it is [3] rx; t qcVet=2s Io 2s1 t2 x2 =c2 Ht x=c, where Io () is the modified
Bessel function of order zero.
254
10 Wave Propagation

The general form of the stressstrain law in this case may be expressed as: P(r) =
Q(e) [Cf. Chap. 3], where P and Q are differential operators of order m, and n,
respectively, such that either m = n - 1, or m = n:
m
X
pk
k1
n
dk r X
dl e
q
l
dtk
dtl
l1
10:42
Where, for a homogeneous material, all the pks and qls are constant.
The differential equation (10.42) can be integrated with respect to time, t,
starting from a fixed but arbitrary point x = X1, located before the shock front, and
ending at a variable point x = X2(t), before or after the front. After the first
integration, the generic term on either side of (10.42) becomes:
ZX2
ok
dt
otk
X1
X2 t
Z
X1
X2 t
o ok1
ok1
dt
ot otk1
otk1 X1
With this, any term of the sum in (10.42) becomes a function of the upper limit
and can be integrated again. Repeating this integration a total of n times produces an
integral whose limits lie on both sides of the shock front. At this stage, most of the
terms that result from the integrations would be continuous functions of time, in such
a manner that their jumps across the front would be zero. The exception being the
two terms with the highest order derivative. If we now let X1 and X2 approach each
other and the shock front at x = X*, the following is obtained after n integrations:
(
lim
X1 !X2 t
X1 X t
pn
X
R t
X1
)
o
ot rdt
lim
X2 !X1
X2 [ X t
pn
XR
2 t
X t
)
o
ot rdt
and similarly:
lim
X1 !X2 t
qn feX eX g qn feX eX g qn jej
These expressions lead to the stress-strain law in jump form for a viscoelastic
substance of the differential-operator type, as:
r Eg e;
Eg
qn
pn
10:43
10.3
Shock Waves
255
This expression is the exact analog of (10.32). Also, since (10.2810.30) apply
irrespective of the form of the stressstrain law, combining (10.30) and (10.43)
yields the same equation for the wave speed, c, as Eq. (10.33):
s
Eg
qn
c
; Eg
q
pn
10:44
Hence, the jump equations: (10.2810.30) and (10.32) or (10.43), as well as

(10.33) or (10.44), for the sonic speed, are the same for all viscoelastic substances.
Example 10.7 Obtain the speed of longitudinal shock waves in a bar made of a
Maxwell material with modulus Eo, viscosityg, and density q.
Solution:The constitutive equation for this Maxwell material is [c.f. Chap. 3]:
p
de
r Ego dr
Eo =q.
dt g dt . Using (10.44) with n = 1, pn= g/Eo, qn= g yields: c
The differential equations governing the jumps in strain, stress, and velocity for
the materials of differential type are obtained in the same way as for viscoelastic
materials of integral type: by taking the total derivative, on the shock front, of
the stress-velocity jump equation (10.28), evaluating the partial derivative of the
stressstrain law with respect to time, and using expressions (10.2810.30) in the
process. Specifically, taking the derivative of (10.28) on the shock front leads to:
D
D
r qc v
Dt
Dt
Using (10.25) on the left-hand side of this expression converts it to:

or
or
D
c
qc v
ot
ox
Dt
To evaluate the first term on the left-hand side of (l), first differentiate the stress
strain law once with respect to time. Considering the case m = n only, this leads to:
p0
or
on r
on1 r
oe
on e
on1 e
pn1 n pn n1 q0 qn1 n qn n1
ot
ot
ot
ot
ot
ot
Integrating this expression n times with respect to t, following the same approach
that led to Eq. (10.43)starting at a fixed but arbitrary point below the shock front
and ending at a variable point above it and then taking the limit as the two points
approach the frontproduces the jump equation:

or
qn oe
qn1
pn1
e
r
ot
pn ot
pn
pn
256
10 Wave Propagation
Using that qn/pn = Eg = qc2, together with [qe/qt] = [|qv/qx|], [e] = -[v]/c, and
[r] = -qc[v], converts this expression into the following:

or
ov
qn1
pn1
qcv
qcv
qc2

ot
ox
pn
pn
Inserting (o) and the jump of the equation of motion across the front: [qr/
qx] = q[qv/qt], into (l), yields, after collecting and regrouping terms:
qc

ov
ov
pn1 qn1
D

c
qcv qc v
ot
ox
Dt
pn
qn
Since, by virtue of (10.25), the expression in braces is the total time derivative,
D[v]/Dt, of [v] across the front, the final result is:

D
1 pn1 qn1
v

v 0
Dt
2 pn
qn
10:45
This first-order differential equation is the exact analog of (10.34b), derived for
materials of relaxation integral type. The solution of this equation can therefore be
cast in the same form as (10.35b):
1
jvj Ae2C0t ; C0

pn1 qn1

pn
qn
10:46a
As was the case with materials of integral type, using (10.8): D/Dt = cD/
Dx converts (10.45) into an equation in x with solution:
v Ae
1
2c
C0x

pn1 qn1
; C0

pn
qn
10:46b
This seems to be the first time the general expression for C(0) listed in
(10.46a, b) appears in the literature. In lieu of this expression, the procedure to
estimate the rate of decay of shock waves in viscoelastic substances of differential
type seems to have use (10.34a), which requires the relaxation modulusa more
cumbersome process.
Example 10.8 Determine the rate of decay of shock waves in a bar made of the
standard linear solid with the material properties shown in Fig 10.5.
Solution:The rate of decay of shock waves in viscoelastic materials of differential type is given by (10.46a, 10.46b). To compute it requires the coefficients
of the constitutive equation for this material. Using [c.f. Chap. 3]: E0gE1 r
1
or
ot
Eo E1
Eo E1
Eg0 E1 e E0 oe
ot , identifies po g , p1 1, qo g , and q1 Eo . Taking
1
10.3
Shock Waves
257
Fig. 10.5 Standard solid of

example 10.8
E1
E0
this into (10.46a, 10.46b) givesthe rate of

of shock
decay

waves
for the
pn1
qn1
E0 E1
E1
standard linear solid as: C0 pn qn
g Eg 0 .
g
1
10.4
Problems
P.10.1 Determine the complex sonic speed at x = 50 rad/s for the three-parameter
solid of Example. 10.6 if Eo = 140 kPa, E1 = 140 kPa, g = 276 kPa-s,
q = 1,600 kg/m3.
Answer: c* = 9.35e-0.0101j = 9.353 ? 0.0474 j
Hint: Use the constitutive polynomials P and Q [c.f. Chap. 3] to compute the
complex modulus E*(jx) = Q*(jx)/P*(jx) & 139.97 ? 1.42 j kPa. From it,
00
0
determine the loss angle d = tan-1(E /E ) & 0.0101 rad, and use Eq. (10.13) to
obtain c*.
P.10.2 The equilibrium modulus of an elastomeric compound is 145 psi; its mass
density is 0.065 lbm/in3; and the transient part of its tensile relaxation modulus, in
Prony-series form, has the following coefficients and time constants at the temperature of interest:
Ek (MPa)
9.665
20.69
3.448
5.517
1.379
1.379
0.2410
0.2070
sk (sec)
4.510-5
6.510-4
5.010-3
2.010-2
2.510-1
3.5100
1.5101
4.0102
Determine the complex sonic speed of the material at a forcing frequency of

50 Hz.
Answer: c* & 74.2 ? 13.0 j m/s.
Hint: Compute E* per Chap. 3, as: E*& 9.64 ? 3.345 j MPa, together with its
amplitude, ||E*|| & 10.20 MPa, and phase angle d = tan-1(3.345/9.64) &
0.347 rad. Using ||E*|| and q = 1,600 kg, in (10.14), gives ||c*|| & 75.3 m/s. The
complex sonic speed at the given frequency is obtained from (10.13).
P.10.3 Determine the response of a 30 inch long bar of the same material and
under the same conditions as the bar in Example 10.2. What would the response of
00
the bar be if its material were such that E 0?
Answer: The solution is displayed in Fig. 10.6 for the viscoelastic bar and in
00
Fig. 10.7, for the bar without damping (E 0), which would then be elastic. As
seen in the figures, the viscoelastic and elastic bars respond quite similarly.
258
10 Wave Propagation

viscoelastic bar
2.0
= 0
= /4
= /2
= 3/4
=
1.5
u ( x, t ) / Uo
1.0
0.5
0.0
-0.5
-0
-1.0
-1
-1.5
-1
-2.0
-2
Hint: In this case, E*, x, and k, are the same as for the bar in Example 10.2.
Hence, the form of the solution will be the same, and the difference is due to the
change in length. Note the symmetry of the response for the case with no damping
(E00 = 0).
E0 0=Eg
2Etg cdE
dt jt0
A e 2c t

Hint: Use that: C0 E1g dsd Et s E1g qEq
P.10.4 Prove that: jvj A e
t0
q0
EE0
, to put
g
(10.35b) into the desired form.

P.10.5 Find the stress response in transformed space, of a semi-infinite bar of
Maxwell material subjected to a stress of amplitude R, suddenly applied at its end.
p
Answer: rx; s Rs ex=c ss1=s
Hint: First, obtain the transform of r(0,t) = RH(t), as: r0; s R=s. Now use
the stress-strain law (3.23a) in transformed space, to get: AS A R=ks.
x; s.
The result follows from this and the expression for r
P.10.6 Find the Laplace-transformed solution of a bar of length L, fixed at its end
x = L and subjected to a discontinuous change in velocity of amplitude V at x = 0.

00
elastic bar E 0
2.0
1.5
u ( x, t ) / Uo
1.0
0.5
0.0
-0.5
-1.0
-1.5
-2.0
= 0
= /4
= /2
= 3/4
=
10.4
Problems
259
sinh ksx
Answer: vx; s Vs eksL sinh
ksL
Hint: Use that v(x,t) = qu(x,t)/qt to write: vx; s sux; s, and thus:
v0; s V=s. Use this, and the fact that v(L,t) = 0, vx; s D1 ekx D2 ekx to
evaluate the functions D1(s) and D2(s) and arrive at the result.
P.10.7 Determine the rate of decay of shock waves in the standard linear solid
of Example 10.5, using the materials relaxation modulus and expression (10.34a).
Answer: C(0) = E0/g1
Hint: The rate of decay of shock waves in terms of the relaxation modulus given
in (10.34a) requires the materials glassy modulus as well as the first derivative of
the relaxation modulus at t = 0. The glassy modulus of this materials is Eg = E 0;

E0 E1 t
g1
E1
E0 E1
and its relaxation modulus [c.f. Chap. 3]: Et EE00E
E

.
0
E0 E1 e
1

2
E
From this, it follows that dtd Ett0 g 0 ; and, from (10.34a): C(0) = E0/g1; the
1
same as in Ex. 10.8, which used Eqs. (10.46a, b), pertinent for materials of differential type.
References
1.
2.
3.
4.
H. Kolsky, Stress Waves in Solids (Dover Publications, NY, 1963), pp. 4153
W. Flgge, Viscoelasticity, 2nd edn. (Springer, Berlin, 1975), pp. 121140
R.M. Christensen, Theory of Viscoelasticity, 2nd edn. (Dover, NY, 1982), pp. 110116
N. Cristescu, I. Suliciu, Viscoplasticity (Martinus Nijhoff, The Hague, 1982)
11
Variational Principles and Energy

Theorems
Abstract
This chapter introduces the subject of the variation of a functional and develops
variational principles of instantaneous type which are the equivalent of
Castiglianos theorems of elasticity for computing the generalized force
associated with a generalized displacement and vice versa, by means of partial
derivatives of the potential energy and the complementary potential energy
functionals, respectively. A natural consequence of the variational principle of
instantaneous type is that the constitutive potentials of viscoelastic materials are
not unique. Any dissipative term can be added to them without changing the
stress strain law. The viscoelastic versions of the unit load theorem of elasticity,
and the theorems of Betti and Maxwell for elastic bodies, are also developed in
detail.
Keywords

Variation Variational Functional Potential Stationary Admissible Field

Castigliano Unit load Reciprocal
11.1
Introduction
The calculus of variations is concerned with the development of methods to

establish the maxima and minima of a special class of functions, called functionals.
Loosely speaking, a functional is a function whose arguments are themselves
functions. Functionals occur naturally in viscoelasticity through the constitutive
equations, in which the arguments are the position coordinates, the current stress or
strain, and the history of the stress or strain.

261
262
11
Variational Principles and Energy Theorems
Many variational theorems have been developed for the theory of elasticity;
among them are the principles of minimum potential energy and minimum complementary potential energy, the principle of virtual work or virtual displacements,
and the theorems of HuWashizu, HellingerReiner, Castigliano, and Lagrange
[1]. Not surprisingly, variational theorems also exist which are the viscoelastic
counterparts of those for elasticity.
Variational principles provide a rigorous means of establishing the field
equations of the theory of viscoelasticity without recourse to simplifying
assumptions. The purpose of the present chapter, however, is to use variational
principles to develop computational tools that facilitate solving certain types of
viscoelastic boundary-value problems. To do this, the subject matter is developed
in five parts, as follows.
Section 2 treats viscoelastic functionals as functions of two time-dependent
arguments: the present values and the past values of the stress or strain. Depending
on whether the current values or the past values of the arguments are varied, two
types of variations are identified. Instantaneous variations result when only the
current values of the arguments (stress or strain) are varied, and history variations
are concerned only with variations in the past values of the arguments. This is used
in Sect. 3 to develop two variational principles of instantaneous type, which are the
equivalent of Castigliano theorems of elasticity for computing the generalized force
associated with a generalized displacement and vice versa, by means of derivatives
of the potential energy and the complementary potential energy, respectively. The
viscoelastic version of the unit load theorem of elastic solids, which is used to
obtain the deflection at an arbitrary point in a structure, is developed in Sect. 4. The
reciprocal theorems of viscoelasticity which correspond to the theorems of Betti
and Maxwell for elastic bodies [2] are developed in Sect. 5.
11.2
Variation of a Functional
Aside from the position coordinates, xk, two types of time-dependent arguments
occur in viscoelastic functionals. To account for memory effects, the response is
made to depend on all past values of the stress or strain. To account for instantaneous response, the current values of the stress or strain are also included. Simply
put then, a viscoelastic functional may be considered as a function of two variables:
a history variable and a variable that describes the current state. Therefore, in
examining the variation of a given viscoelastic functional, one is at liberty to
consider variations in the current values of its arguments or to consider variations in
the past history of the arguments. The former type of variation leads to instantaneous variational principles and the latter to hereditary variational principles. Both
types of variations produce equivalent principles, but only variational principles of
instantaneous type are discussed in the present chapter.1
1
Viscoelastic variational principles concerned with variations of the histories of the arguments
will be included in subsequent editions of the text.
11.2
Variation of a Functional
263
With this, a functional, F, of the past history, e(t-s), and current value, e(t), of a
variable may be represented as follows [3]:
1
F w eij t s; eij t
11:1
s0
Assuming that all pertinent smoothness and continuity requirements are satisfied by all the functions involved and considering the past history eij(t-s) fixed, the
first instantaneous variation, or total differential, dF, of the functional is defined as
the limit2:
1
dF lim
h!0 h

w eij t s; eij t h deij t w eij t s; eij t
s0
s0
a
h0
As for ordinary scalar-valued scalar functions, this expression is the derivative

of w with respect to h, evaluated at h = 0 [c.f. Appendix A]. Therefore, the first
variation of the functional F with respect to the current value of its arguments,
which is also called first instantaneous variation of F, is given by
dF

d 1
w eij t s; eij t h deij t
dh s0
h0
11:2
Using the chain rule of differentiation to evaluate it, results in

dF
o 1
weij t s; eij t dekl t
oekl t 0
11:3
Clearly, entirely similar expressions would be obtained if the roles of the stress
and strain tensors were interchanged.
11.3
Variational Principles of Instantaneous Type
These principles assume variation in the current, instantaneous values of the

variables rij and eij, while their histories determine the state of the material at the
current time. In this case, as indicated by 11.3, the first variation of a functional is
obtained by taking its partial derivative with respect to its instantaneous argument,
regarding the hereditary component as constant.
In keeping with standard practice, in this chapter only, we use the Greek letter, d, to denote
variation of a function or functional; and as such, is not to be mistaken for the unit impulse or
Dirac Delta function, nor for the Kronecker delta, used to represent the unit tensor.
264
11
For the present treatment, the three-dimensional forms of the constitutive laws
presented in 2.34a and b will be used.3 Thus, as discussed in Chap. 8, for a general
anisotropic material
rij t Mgijkl ekl t
eij t Cgijkl ekl t
t
0
o
Mijkl t sekl sds
os
11:4a
t
0
o
Cijkl t srkl sds
os
11:5a
which, according to 11.3, have the following instantaneous variations:

drij t Mgijkl dekl t
11:4b
deij t Cgijkl drkl t
11:5b
Different viscoelastic potential functionals and associated variational principles

can be constructed, which have their counterparts in the theory of elasticity.
Prominent among these are a HellingerReissner-type functional and a Castigliano-type functional that is similar to the potential energy functional of elasticity.
The Castigliano-type functional is equivalent to the elastic complementary
potential energy functional. Like their elastic counterparts, these viscoelastic
functionals acquire stationary values and give rise to principles which can be used
to derive the field equations of viscoelasticity and various computational methods.
11.3.1 First Castigliano-Type Principle

This is the instantaneous variational counterpart of the theorem of minimum
potential energy of elasticity. The corresponding functional which, for reference
only, is called instantaneous viscoelastic potential energy, /o, in this text, is
defined for all kinematically admissible displacement fields, u0i t. A displacement
field u0i is said to be kinetically admissible if it is continuously differentiable up to
third order4 inside the region occupied by the body in question and, in addition,
identically satisfies both the equations of strain compatibility and the displacement
boundary conditions on the part of the boundary of the body where they are
prescribed. Clearly, to a kinematically admissible displacement field, there correspond kinematically admissible strains e0ij 12 u0i;j u0j;i , defined by the strain
As in previous chapters, the subscript g stands for glassy, to indicate the value of the
corresponding material property function at t = 0.
4
The requirement of continuity up to the third order derivatives of the displacement field stems
from the fact that the equations of compatibility involve second derivatives of the strains, which
are defined in terms of the first derivatives of the displacement field.
11.3
265
displacement relationships (9.5), and stresses, r0ij , defined through constitutive

equations (11.4a).
Similar to elastic solids, the instantaneous viscoelastic potential function, /o, is
defined for all kinematically admissible displacement fields, u0i t, including the
actual field, ui t, in the form:

Z
Z
uo u0i Uo e0ij dV Fi tu0i t dV
V
Tio tu0i t ds
11:6
ST
In this expression, the functional Uo is of the same form as the internal energy
of elasticity theory and, just like it, is assumed to be an (instantaneous) internal
energy potential, from which the stressstrain relations of viscoelasticity can be
obtained by differentiation [c.f. Appendix B]. Similar to the elastic strain energy
density, and using the actual fields rij and eij, for generality, the instantaneous
potential energy density functional, Uo, is defined as
rij
o
Uo emn
oeij
11:7
As a potential function, Uo (e) can be formally constructed from 11.7 by

integration. To do this, the stressstrain constitutive Eq. (11.4a) is put on the lefthand side of 11.7 and integration is carried out holding the hereditary part constant, in accordance with the present type of variation. This leads to
1
Uo emn t eij tMgijkl ekl t eij t
2
Zt
1
o
Mijkl t sekl sds D eij t s
os
s0
In this expression, D [(eij(t-s)] is a functional of the strain history and represents a purely dissipative contribution. The physical interpretation of this is that
The constitutive potentials of viscoelastic substances are not unique. Any dissipative
term can be added to them without changing the stress strain law.
Thus, without any loss of generality, the functional D is assumed to be identically zero, so that:
1
Uo emn eij tMgijkl ekl t eij t
2
t
0
o
Mijkl t sekl sds
os
11:8
By its construction, as the integral of 11.7 with respect to the current strains, it
should be clear that the first instantaneous variation in this expression produces the
constitutive equations.
Example 11.1 Check that the first instantaneous variation in Uo with respect to
current strains yields the constitutive equations of viscoelasticity in stressstrain
form.
266
11
Solution:
Take the first instantaneous variation
in Uo as
dUo 12 deij Mgijkl ekl eij Mgijkl dekl and use the symmetry of the material
property tensor: Mgijkl Mgklij to collect terms and arrive at constitutive Eq.
(11.4a).
For future reference, Uo is split into an elastic part, Ugo, and a hereditary part,
Uto:
Uo Ugo Uto
11:9a
1
Ugo eij tMgijkl ekl t
2
11:9b
Uto eij t
t
0
o
Mijkl t sekl sds
os
11:9c
Multiplying (11.4) by eij(t), and using the previous definitions, yields the
expression:
rij teij t 2Ugo Uto
11:10
Returning now to the instantaneous potential /o, introduced in 11.6, the other
integrals in it correspond to the instantaneous work of the body forces, Fi(t), and
the specified surface tractions, Toi (t). Also, the stress and strain fields will in
general be functions of position, but such spatial dependence is omitted for clarity.
The conditions that 11.6 must fulfill at its stationary points are obtained by
setting its first variation to zero.
duo u0i
dUo e0ij
dV
Fi tdu0i t
Adding and subtracting
duo u0i
Z
V
dUo e0ij
dV
Tio tdu0i t dS 0
ST
Ti du0i dS
ST
dV
nj rji dui 0 :
ST
Z
V
Fi du0i
dV
Z
ST
Ti
Tio du0i
dS
nj rji du0i dS
ST
0
c
Extending the integral on the far right to the whole surface of the body ST ? Su,
which is valid because du0 0 on Su, and invoking Gauss theorem to convert that
integral into a volume integral [c.f. Appendix A], use the straindisplacement
relations, and collect terms to write:
11.3
duo u0i
o
rij
Uo e0kl deij dV
oeij
rij;j
267
Fi du0i
dV
Ti Tio du0i dS
ST
d
So that, for arbitrary variations in the kinematically admissible displacement
field and its associated strain field, the first variation of the instantaneous potential
is stationary if
oUo ekl
Mgijkl ekl t
rij t
oeij t
Zt
o
Mijkl sekl t s ds
os
11:11
rij;j t Fi t 0; in V
Ti t Tio t; on ST
Conversely, if these conditions are fulfilled, one can construct the potential /o.
For this reason, the first variation in the instantaneous potential /o has a stationary
value, if and only if the constitutive and equilibrium equations as well as the
traction boundary conditions are satisfied. Clearly, by definition of the admissible
fields, the displacement boundary conditions are also identically satisfied. These
are the exact same requirements on the potential energy functional of elasticity.
That the stationary value of the instantaneous potential /o corresponds to a
minimum may be proven in simple terms. Indeed, the symmetry of the stress and
strain tensors implies that the glassy modulus tensor, Mgijkl is symmetric. Because
of this, the glassy term of the instantaneous potential /o, which is quadratic in
eij(t), must be positive definite. As in ordinary differential calculus, the second
variationor second derivativedetermines the character of the stationary points
of any given functional. Also, both the current and hereditary terms of the
instantaneous potential, /o, and the external work are linear in eij(t). Hence, it is
the quadratic term in eij(t) that defines the second variation. By the positive definiteness of this term, the instantaneous functional /o will attain a minimum for all
kinematically and statically admissible displacement fields.
The theorem of minimum instantaneous potential /o may be used to develop a
method to determine the generalized force required to maintain a set of displacements in a body. This is so when the instantaneous work of the external
forces can be expressed as a linear combination of the work of a set of generalized
forces, Pr(t), as they move through their corresponding generalized displacements,
dr0 t, that is, if
Z
F tu0i t dV
V i
ST
Tio tu0i t dS
X
r
Pr dr0
11:12
268
11
In this expression, the Prs are known generalized forces, while the dr0 s are
kinematically admissible, but otherwise arbitrary generalized displacements.
The displacement field, u0i t, and with it the corresponding strain field, e0ij t,
can then be expressed as linear combinations of the dr0 . Therefore, the instantaneous potential, Uo, and through it, /o, become quadratic functions of the dr, as
uo
U d 0 dV
V o q
X
r
Pr dr0
11:13
Invoking the stationarity of /o, and recalling the definition

of instantaneous

R
P
Uo dq0 dV Pr dr0 dds0 0; which yields
variation, one can write duo odo 0
s
the relationship sought:

Ps
o Z
Uo dq0 dV
ods0 V
11:14
11.3.2 Second Castigliano-Type Principle

This is the instantaneous viscoelastic variational counterpart of the theorem of
minimum complementary potential energy of elasticity. The viscoelastic functional,
which may be thought of as an instantaneous complementary potential energy
density, wo , is defined for all statically admissible displacement fields u00 . A displacement field, u00 , with strains e00ij 12 u00i;j u00j;i , and stresses, r00ij Mijkl de00kl , is
statically admissible if it is continuous and continuously differentiable inside the
region occupied by the body and identically satisfies the equations of equilibrium
inside the body, as well as the traction boundary conditions, where prescribed.
In analogy with elastic materials, an instantaneous complementary potential
functional, wo , is defined for all statically admissible fields u00i , e00ij , and r00ij which
include the actual field ui tin the form:
Z
Z

wo r00ij
Yo r00ij dV
Ti00 tuoi tds
V
11:15
Su
In this expression, the functional Yo is of the same form as the complementary

potential energy density of elasticity theory and, just like it, is assumed to be an
(instantaneous) energy potential, from which the strainstress relations of viscoelasticity can be obtained by differentiation [c.f. Appendix B]. Similar to the
elastic case, and using the actual fields rij and eij, for generality, the instantaneous
complementary potential energy functional, Yo, is assumed such that
eij
o
Yo rkl
orij
11:16
11.3
269
As a potential function, Yo (r) can be formally constructed from 11.16 by

integration. To do this, the strainstress constitutive Eq. (11.5a) is put on the lefthand side of 11.16 and integration is carried out holding the hereditary argument
constant, in accordance with the present type of variation. This leads to
1
Yo rmn t rij tCgijkl rkl t rij t
2
Zt
1
o
Cijkl t srkl sds Q rmn t s
os
s0
11:17
This expression has the following instantaneous variation:
2
dYo rmn drmn t4Cgijkl rkl t
Zt
o
Cijkl t srkl sds5
os
In general, although the term Q, above, may be a functional of the stress history,
as far as the instantaneous variation is concerned, it does not contribute at all to the
strainstress equations and may thus be taken as zero without loss of generality. In
other words, the complementary energy potential functional for a viscoelastic
substance is not uniquely defined, since one can add to it an arbitrary functional of
the stress history without altering the strainstress relations. In other words, the
functional Q represents a purely dissipative contribution. Hence, in the sequel, Yo
is taken simply, as:
1
Z
o
Yo rmn rij tCgijkl rkl t rij t t0 Cijkl t srkl sds
2
os
11:18
Also, for future reference, Yo is separated into an instantaneous component, Ygo,

and a transient part, Yto, as follows5:
Yo Ygo Yto
11:19a
1
Ygo rij tCgijkl rkl t
2
11:19b
Yto rij t
t
0
o
Cijkl t srkl s ds
os
11:19c
Premultiplying (11.5a) by rij(t), using the above decomposition of Yo, and

collecting terms produce:
rij eij 2Ygo Yto
5
The negative sign to define Yto is only used for mathematical convenience.
11:20
270
11
In addition, replacing Yto with the aid of 11.19a and rearranging

Yo rij eij Ygo
11:21
Returning to the instantaneous potential wo, introduced in 11.5a and b, in terms

of the arbitrary but statically admissible fields u00i ; e00ij ; andr00ij , the second integral in
it is the work of the arbitrary but statically admissible surfaced tractions, T 00 , acting
on the actual displacement field uoi (t).
The stationarity conditions of wo are derived by setting the first variation in wo
to zero:
dwo r00i
dYo r00ij dV
dTi00 tuoi tdS 0
Su
Su
ST
Adding and subtracting dTi00 ui dS and using
dTi00 ui dS
nj dr00ij ui dS result
Su
in:
dwo r00i
dYo r00ij dV
00
dTi
uoi
u00i dS
nj dr00ij ui S 0 c
ST Su
Su
Before proceeding, the third integral on the right of this expression is converted
into a volume integral by means of the Gauss theorem [c.f. Appendix A]:
Z
nj dr00ij ui dS
dr00ij;j ui dV
dr00ij ui:j dV
The first term on the right vanishes because the equations of equilibrium are
identically satisfied by the statically admissible field, and the instantaneous variation in the actual body forces is identically zero. Now, by the symmetry of the
stress tensor, the second integral may be expressed in the form:
Z
V
dr00ij ui:j dV
1
ui;j uj;i dr00ij dV
2
Using these results and rearranging, dwo becomes

dwo r00mn
Z
V
o
1
Yo r00kl ui;j uj;i drij dV
or00ij
2
u00i uoi dTi00 dS 0 f
Su
So that, for arbitrary variations in the statically admissible fields, the first
variation of the instantaneous complementary energy potential is stationary if
11.3
e00ij t
271

1 00
ui;j u00j;i ; inV
2
11:22
u00i t uoi ; on Su
Clearly, if these conditions are fulfilled, one can also construct the potential wo.
For this reason, the first variation of the instantaneous complementary potential
has a stationary value, if and only if the straindisplacement relationships and the
displacement boundary conditions are satisfied. By the static admissibility of the
fields involved, the equations of equilibrium are satisfied as well. As stated before,
these are precisely the requirements put on the potential energy functional of
elasticity.
The proof that the stationary value of the instantaneous complementary
potential energy corresponds to a minimum follows from the simple fact that wo is
quadratic in the instantaneous stresses and only linear in its hereditary part. By the
symmetry of the tensor of glassy compliances, Cgijkl, the quadratic term of wo is
positive definite. Since the character of a stationary point is determined by the
second variation in a functional, it being positive for wo due to the positive definiteness of its quadratic term, the stationary points of wo must correspond to a
minimum.
The theorem of minimum instantaneous complementary potential wo may be
applied to derive the viscoelastic counterpart of a second Castigliano-type theorem
of elasticity. This principle is used to determine the generalized displacement
required to maintain a set of specified tractions in a structure. In so doing, it is
assumed that the work term appearing in (11.15) may be expressed as discrete sum
00
of products of arbitrary but statically admissible generalized forces,6 Pr , and actual
generalized displacements dr as:
Z
Ti00 tuoi tdS
P00r dr
11:23
Su
In this case, the stress field may be expressed as a linear combination of the
generalized forces; consequently, the instantaneous complementary potential
energy functional, Yo, is quadratic in the P00r . Using this and (11.23) allow one to
express (11.15) as:
wo
Z
V
Yo P00q dV
P00r dr
11:24
The term generalized force is used to denote either a concentrated force or a concentrated
moment, while the term generalized displacement denotes either a linear or an angular displacement. A generalized force and a generalized displacement are work-conjugate if the work done
by the former acting on the later can be correctly calculated from their product.
272
11
Stationarity of wo requires that:

2
3
Z
X
o 4
dwo 00
Yo P00q dV
P00r dr 5dP}r 0
oPs
r
That is
ds
o
oP00s
Yo P00q dV
11:25
This expression shows that the generalized displacement corresponding to a

generalized force may be obtained as the first partial derivative of the instantaneous complementary potential function with respect to that force. This is the
analog of Castiglianos second theorem for linear elastic structures.
Example 11.2 A beam of a viscoelastic material with tensile relaxation modulus
E(t) is subjected to a concentrated load P(t), as shown Fig 11.1. The length of the
beam is L and its cross-section have second moment of area of magnitude I. Find
the deflection of the beam under the point of load application .
Solution:
The deflection, D, of the beam under the load P may be readily obtained by
means of Castiglianos theorem (11.25). Hence, the following expression needs to
R
o
be evaluated: D oP
Yo PdV. For the beam in question [c.f. Chap. 5],
V
My
I
Ptxy
2I .
Using the strainstress relation
et Cg rt os Ct s rs Cg rt Kt s rs
and Eq. (11.21): Yo re Ygo leads to the following expression for the complementary potential function:
Z

Z
1
rtCg rt rtKt s rs dV
Yo dv
2
P2 tCg L3 PtCg Kt s Ps
48I
96I
P(t)
L, E(t), I
11.3
273
R
PtC L3
L3 C KtsPs
o
Applying the theorem leads to Dt oP
Yo dm 48Ig g 48I
; or:
V
L
Dt 48I
Cg Pt
Rt
0
o
os Ct
sPsds. This is the viscoelastic analog of the
elastic form PL /(48EgI) and could, of course, have been obtained by the methods
of Chap. 5.
Castiglianos theorem may appear to be of limited value because its form
(11.25) seems to imply that one can determine generalized displacements only at
the exact locations where generalized forces act. The unit theorem presented
subsequently is derived from Castiglianos theorem and extends its applicability to
locations in a structure where no actual generalized forces need to be acting.
11.3.3 Unit Load Theorem

The unit load theorem for viscoelastic structures is derived from Castiglianos
principle in a simple fashion, as follows. A generalized fictitious load P is applied
to the structure at the location and in the direction in which the deflection is
desired.7 The stresses and strains resulting from P are then determined and used to
construct the complementary potential functional Yo(P). The generalized displacement, d, is obtained by applying Castiglianos theorem (11.25), differentiating Yo with respect to P and, since P is fictitious, evaluating the result at P = 0.
Indeed, let rij(t) be the stress field induced by the loads acting on the structure
and ruij(t) that due to a generalized load of unit magnitude. With this, the stress
field induced in a linear viscoelastic structure by a generalized load of magnitude
P will be Pruij(t), and the combined action of the actual loads and P will yield the
field: rij(t) ? Pruij(t). Under this system, using 11.19a, b, and c results in the
following instantaneous potential functional, Yo:
1
Yo rij t Pruij tCgijkl frkl t Prukl tg
2
Zt
o
rij t Pruij t Ctijkl t sfrkl s Prukl sgds
os
11:26
Applying the second Castigliano-type theorem, (11.25) produces the expression

sought:
For simplicity of exposition, the prime notation used to distinguish statically admissible fields
is dropped.
274
11
o
Yo j
oP P0
2 t
3
Z
Z
Z
o
Ctijkl t srkl s ds5 dV
ruij tCgijkl rkl t dV ruij t4
os
ujP
11:27
This form is easily applied to obtain the deflection of linearly viscoelastic
bodies made of structural members such as beams, columns, and bars, for which
the most general state of stress, r, is made up of a normal component and two
shear stress components. The direct or normal stress is due to the normal force and
bending momentpossibly two of themacting on the cross-section. The shear
stresses are induced by the shear forces and, if present, the torsional moment. As it
turns out, the use of principal centroidal axes in (11.27) eliminates all terms
containing mixed mechanical elements.8
Example 11.3 Derive the viscoelastic unit load theorem for prismatic structural
members accounting only for bending moment effects.
Solution:
y
Start with the normal stress rt Mx;t
produced in a viscoelastic beam by a
I
y
bending moment M(x,t) [c.f. Chap. 5] and note that ru t Mu x
would be the
I
stress induced by a generalized load of unit magnitude acting along the generalized
direction at the desired location. Then, insert these two expressions on the righthand side of (11.27) to obtain
ujP
Z Z
Mu x; t
Mx; t
yCg
y dA dx
I
I
L A
2 t
3
Z Z
Z
Mu x; t 4 o
Mx; s

y
Ct s
yds5 dA dx
I
os
I
L
Carry out the integral over the cross-sectional area, using that I
y2 dA and
simplify to get:
ujP
Z
L
Mu x; tCg Mx; t
dx
I
8
9
Zt
Z <
=
Mu x; t oCt s
Mx; sds dx
: I
;
os
L
11:28
8
This is so, because mixed terms involve integrals of the form:
R
A
by the assumption that the reference axes are centroidal.
ydA; which are identically zero
11.3
275
This is the analog of the linear elastic expression: ujP
R
L
Mu tMt
dx,
EI
only when
bending effects are included. For convenience, Eq. (11.28) will be simplified by
^
introducing an auxiliary operator, C:
^ Bt AtCg Bt At
A t C
o
C t sBsds
os
0
Zt
11:29
So, that, operationally

t
^ Cg Z o Ct s
C
os
0
11:30
With this definition, Eq. (11.28) takes the symbolic form:

ujP
^ Mx; t
Mu x; t C
dx
I
11:31
Example 11.3 Determine the mid-span deflection of the uniformly loaded, simply
supported viscoelastic beam shown in Fig. 11.2, where q, L, I, and E represent,
respectively, the uniform load on the beam, reckoned per unit length of the beam,
the beams length, the second moment of area of the beams cross-section, and the
tensile relaxation modulus.
Solution:
The solution is obtained by evaluating Eq. (11.31), in which M is the moment
induced by the actual loading q and Mu is the moment due to a unit concentrated
load acting downward.9 To do this, obtain the required mechanical elements by
establishing equilibrium of the beam under the unit, generalized load and the
actual load.
To establish equilibrium of the beam under a unit load, use Fig. 11.3:
q(t)
L, E, I
The sense of the generalized unit load is arbitrarily chosen; but if the computed deflection is
negative, its sense will be opposite that assumed for the generalized unit load.
276
11

Equilibrium under a unit load
1/2
Equilibrium under actual load
x
x/ 2
qx/2
x
qL/2
From it, write that:

1
Mu
2
L
2
x; 0 x L2
2x ; L2 \x L
For equilibrium under the actual load, use Fig. 11.4 to write
M
qL
q
x x2 ;
2
2
0xL
L=2
R
Insert these expressions into Eq. (11.31) to obtain ux L2 2I
0
1
^ E qL x q x2 dx
x

C
2
2
2
Here, use has been made of the symmetry of Mu and M with respect to the
center of the beam to perform the integration over one half of the beam only.
^ is independent of x because the material is
Integrating this relation, noting that C
5L4 ^
homogeneous and using the Cg = 1/Eg yields: ux L2 384I
C qt
t
R
4
4
oCE ts
5L
5qtL
qs ds; in which the first part on far right is the deflection
384Eg I 384I
os
0
at the center of an elastic beam of Young modulus Eg.

Expressions (11.31) can be generalized to cover the case of multiple mechanical elementsaxial force Nz, shear forces Vx and Vy, bending moments Mx and
My, and torsional moment Mzacting simultaneously on a structural member.
Indeed, let the mechanical elements in a structural member be defined in a
coordinate system with its z-axis passing through the centroid of the members
11.3
Fig. 11.5 Vectorial sign

convention for structural
mechanical elements
277
My
Vy
y
Mx
Vx
x
z
Nz
Mz
cross-sections and located along the tangent to the principal centroidal rectangular
coordinate system.10 For definiteness, the vector sign convention is used.
According to this convention, a mechanical element is positive if as a vector it acts
in the positive direction of the corresponding coordinate axis, as illustrated in
Fig. 11.5.
Under these assumptions, the most general state of stress in any given structural
member would include the following components [4]:
rxx t rxy t ryy t 0
11:32a
rzx t
Vx Qx Mz tx

Jz
Ix b y
rzy t
Vy Qy Mz ty
Jz
Ix b x
11:32b
Nz t Mx ty My tx

A
Ix
Iy
11:32c
rzz t
In these expressions, A is the area of the cross-section; Ix and Iy are the second
rectangular moments of area of the cross-section with respect to principal and
10
A centroidal and principal coordinate system is located at the center of area or centroid of a
cross section, with its axes coinciding with the cross sections principal axes of inertia. In such a
system, the following relations hold:
Z
A
xdA
Z
A
ydA
Z
A
rdA 0; Ixy
Z
A
xydA 0; Ix
Z
A
y2 dA; Iy
Z
A
x2 dA; Jz
Z
A
r 2 dA
278
11
centroidal x and y axes, respectively; and Jz = Ix ? Iy is the polar moment of area

of the cross-section with respect to the z-axis. Also, bx(y) is the length of the fiber
at which the shear stress is being evaluated, which is located at a distance y from
the centroidal x-axis, and Qx(y) is the first moment of area with respect to x of that
portion of the cross-section delimited by this fiber and the outermost fiber of the
cross-section; by(x) and Qy(x) have analogous interpretations.
Using Eq. (11.32a, b, c) and the corresponding expressions for the P-multiplied
stress field with Eq. (11.27) leads to:
3
o
ruzz tCg rzz t dV
ruzz t4
Ct srzz sds5 dV
os
L A
L A
0
2 t
3
Z Z
Z
Z Z
o
Ct srzx sds5 dV
ruzx tCg rzx t dV
ruzx t4
os
L A
L A
0
2 t
3
Z Z
Z
Z Z
o
ruzy tCg rzy t dV
ruzy t4
Ct srzy sds5 dV
os
o
ujP
Yo j
oP P0
Z Z
Z Z
Zt
11:33
Example 11.4 Derive the expression for the unit load theorem for a prismatic
beam loaded only on its principal plane of bending z-y, by axial and transverse
forces, disregarding the effect the shear forces might have on the deflection of the
beam.
Solution:
In this case, My = Mz = Vx = 0; and rxx = rxy = ryy = rzx = 0, and
VQ
rzz NAz MIxx y. Also, although rzy Iyybyy , its effect on the deflection will be
neglected. Inserting this field in 11.33 together with the corresponding stresses
ruzz NAuz MIuxx y, due to the unit load, and using (11.29) to simplify notation,
produce the
expression:
R N z M x y
R Nuz Mux y
^
C

ujP
A
A Ix dA dx. Carrying out the operations and
Ix
L R
A
noting that ydA 0, because the reference axes are centroidal, and also that
A
Z
A
y dA Ix lead to ujP
Z
L
^ Nz
Nuz C
dx
A
^ Mx
Mux C
dx:
Ix
11:34
Just as for elastic systems, the unit load theorem can also be used to determine
reactions in statically indeterminate structures. This and other related topics will be
taken up in a future edition of the text.
11.4
Reciprocal Theorems
11.4
279
Reciprocal Theorems
Consider a viscoelastic body under the action of two separate sets of body forces,
1
1
2
2
surface tractions and boundary displacements, fF ; T o ; u1o g and fF ; T o ; u2o g, to
which correspond stress and strain solution fields fr1 ; e1 g and fr2 ; e2 g, respectively. It can then be shown that under certain additional restrictions, any two such
systems enjoy several reciprocity relations.
11.4.1 Static Conditions

Here, viscoelastic bodies are considered which have been at rest prior to the start
of the observation and are loaded so that their accelerations remain negligible.
Such systems have boundary tractions and displacements that satisfy a reciprocity
relation analogous to the theorem of Betti for elastostatics [2].
Z
S
Ti1
du2i
dS
Z
V
Fi1
du2i
dV
Z
S
Ti2
du1i
dS
Z
V
Fi2 du1i dV
11:34a
Or written out in full:

Z Z
ST
t
0
Toi1 t s
o 2
u s ds dS
os i
Z Z
V
t
0
Fi1 t s
o 2
u s ds dV 2 j1 11:34b
os i
where dependence on the position coordinates is omitted for clarity. Also, to avoid
repetition, the symbol p jq is meant to indicate that the expression on the right-hand
side is identical to that on the left-hand side, but with superscript p taking the role
of superscript q and vice versa [1].
The present reciprocity relation is proven by showing that starting with one side
of it, say the left, one can produce its other side. To do this, first transform the
surface integral on the left-hand side of (11.34a, b) to a volume integral, replacing
the surface traction in it by the scalar product of the surface normal and stress
tensor and applying Gauss divergence theorem [c.f. Appendix A]:
Z
S
Ti1 du2i dS
Z
ZS
nj r1ij du2i dS
du2i
r1ij du2i ;j dV
du2i;j dS
Z
Z
1
2
F dui dV r1ij du2i;j dV
r1ij;j
r1ij
280
11
In the last step, the equilibrium equations, r1ij;j Fi1 0; were invoked.
Operating now on the second integral on the right-hand side of (a), using the
symmetry of the stress tensor
Z
V
r1ij
du2i;j
dV
Z
V
r1ij
1
du2i;j u2j;i dV
2
Z
V
r1ij de2ij dV
Use of (a) and (b) transforms the left-hand side of (11.34a) into
Z
S
Ti1 du2i dS
Z
V
Fi1 du2i dV
r1ij de2ij dV
r2ij de1ij dV
Consequently, also
Z
S
Ti2
du1i
dS
Z
V
Fi2
du1i
dV
Now, in view of the stressstrain constitutive equations, r1ij Mijkl de1kl ,

together with the associative and commutative properties of the Stieltjes convolution [c.f. Appendix A]:
Z
r1ij
de2ij
dV
Z
V
Mijkl
de1kl
de2ij
dV
Mijkl de2kl de1ij dV
Z
V

r2ij de1ij dV
Expressions (c) and (e) prove the reciprocity relations (11.34a, b).
11.4.2 Dynamic Conditions

Using the Laplace transformation and assuming at-rest initial conditions, a
reciprocity relationship for dynamics may be derived, which is applicable to both
homogeneous and inhomogeneous elastic and viscoelastic materials of arbitrary
density [1].
The reciprocity relations for dynamics, which are entirely analogous to those of
the quasi-static caseincluding that no time derivatives appear in the convolution
integralstake the form:
Z
S
Ti1 u2i dS
Z
V
Fi1 u2i dV
Z
S
Ti2 u1i dS
Z
V
Fi2 u1i dV
11:35a
11.4
Reciprocal Theorems
281
Or, explicitly
Z Z
ST
Toi1 t su2i s ds dS
Z Z
V
Fi1 t su2i s ds dV 2 j1
11:35b
Indeed, starting with the equations of motion for the body under the first loading
system r1ij;j Fi1 qui , convolving them with the dot product of the displacement
vector of the second system, u2i , and integrating over the volume, manipulating the
spatial derivative:
Z
V
r1ij;j Fi1 u2i dV
r1ij u2i ;j dV
ZV
q
V
d 2 u1i
u2i dV
dt2
Z
V
r1ij u2i;j dV
Z
V
Fi1 u2i dS
Applying now the divergence theorem to the first integral on the right of the
identity sign and using the straindisplacement relations and symmetry of the
stress and strain tensors in the second integral; and rearranging:
Z
Ti1 u2i dS
Z
V
Fi1 u2i dV
Z
V
r1ij e2ij dV
Z
q
V
d2 u1i 2
u dV
dt2 i
On applying the Laplace transform to this expression, assuming at-rest initial

conditions:
Z
T i u2i dS
F i u2i dV
r1ij e2ij dV
qs2 u1i u2i dV
And, upon inserting the Laplace transform, r1ij sM ijkl e1kl ; of the constitutive
relations:
Z
S
T i u2i dS
F i u2i dV
Z
V
sM ijkl e1kl e2ij dV
Z
V
qs2 u1i u2i dV
qs2 u2i u1i dV
Similarly, reversing the roles of solution fields 1 and 2:

Z
S
2
T i u1i
dS
Z
V
2
F i u1i
dV
Z
V
sM ijkl e2kl e1ij
dV
Z
V
Equating (a) and (b) yields:

Z
S
T i u2i dS
F i u2i dV
Z
S
T i u1i dS
F i u1i dV
282
11
The inverse Laplace transform of this expression recovers the general

reciprocity relation for elastokinetics presented in Eqs. (11.35a, b).
11.5
Problems
P.11.1 derives explicit expressions for the instantaneous potentials Ugo and Uto, of
a linear isotropic viscoelastic solid of constant Poissons ratio, v, and uniaxial
tensile relaxation modulus E(t).
Answer:
o
Eg n
v
eij teij t
ekk t2
21 v
1 2v

1
1
Z
o
Z
o
ekk t t0 Et sekk sds eij t t0 Et seij sds
Uto
1 v 1 2v
os
os
Ugo
Hint:
Use constitutive Eq. (8.26) and relations (9.21) for isotropic viscoelastic
materials with constant Poissons ratio, together with Eq. (11.9b, c), and carry out
the indicated operations.
P.11.2 Use the unit load theorem to determine the rotation at the left support of the
beam in Example 11.3.
h
i
2
L
Answer: hx 0; t 16I
Pt
Eg
R t0 oso CE t sPsds
Hint:
Proceed as in Example 11.3, using a unit moment as generalized load, applied
at the left support, where the rotation, as generalized deflection is wanted. Thus, as
seen in Fig. 11.6, equilibrium under the unit generalized load leads to
q 2
Mu 1 x=L. Insert this and M qL
2 x 2 x ; for 0 x L; into (11.31) and
carry out the operations to arrive at the result.
1/L
References
1. Y.C. Fung, Foundations of Solid Mechanics, Prentice Hall, Inc., p 429433 (1965)
2. A. Ghali, A.M. Neville, Structural Analysis. A Unified Classical and Matrix Approach,
International Textbook Co., p. 90110 (1972)
3. R.M. Christensen, Theory of Viscoelasticity, 2nd Ed., Dover, p. 39 (1982)
4. T. Oden, Mechanics of Elastic Structures, McGraw-Hill, pp. 7678, 8993 (1967)
Errata to: Engineering Viscoelasticity

12
Errata to:
D. Gutierrez-Lemini, Engineering Viscoelasticity,
DOI 10.1007/978-1-4614-8139-3
In the below table, text content given in Reads column should be corrected, and
the corrected text content have been given in Should read as column:
The online version of the original book can be found under 10.1007/978-1-4614-8139-3
D. Gutierrez-Lemini (&)
Special Products Division, Oil States Industries, Inc., Commercial Blvd. N. 1031,
Arlington, TX 76001, USA
e-mail: [email protected]

E1
Equtaion (l)
11
Last sentence
Equation (c)
Last paragraph
Second
paragraph
Last sentence of
first paragraph
Last paragraph
Last sentence of
first paragraph
First paragraph,
second line
First paragraph,
fourth line
First paragraph,
after Solution
24
26
26
27
34
39
41
43
43
43
Chapter 2
Section 1.3,
second paragraph
Location
Chapter 1
Page
numbers
The phrase: then invert it should be deleted
rDij 2G deDij
rDij 2G deDij
According to (2.31), the creep compliance....then

invert it...
rDij 2sGeDij
rDij 2sGeDij
and the derivative of the applied action
/ and dw=dt
/ and w
and the applied action
time scale
The phrase: and using the properties of the

Heaviside step function, should be deleted
tk
timescale
and using the properties of the Heaviside

step function,
tk+1
This topic is of great

(
0;
for t \t1
rt
M t t1 e0 ; for t t1
rt gR
rt R
This topic of great

(
0;
for s\s/
rt
M t t1 e0 ; for s s/
and let r stand
Should read as
and let r stands
Reads
(continued)
E2
D. Gutierrez-Lemini
First paragraph,
after Solution
Equation (2.39)
First paragraph
Answer to P.2.5
43
45
47
50
Midway down
Abstract
First paragraph
First paragraph
Equation (3.6a)
Last paragraph
First line, second

bullet point
Second bullet
point
Equation (h)
Last paragraph
53
59
59
59
59
60
62
66
66
Chapter 3
Location
Page
numbers
(continued)
f t r0
f t f t e0
When stress is
(a) When stress is
M 1
q0

p0

pn1 r t0
(continued)
Ee(t) + gqte(t) = [E + gqt]e(t) = [gqt + E]e(t) = gqte(t) +Ee(t);
t0
Rt
C t t0 ert0 1g ds ttg 0 H t t0
Ee(t) + gqte(t) = [E + gqt]e(t) = [gqt

+ E]e(t) = gqte(t) +Ee(t);
p0
M 1
q0

n1 r t0
t t0
H t t0
g
(3.1a) and (3.4b)
t0
e0 ds
(3.1a, b)
C t t0 1g
Rt
, performing the required integration, and dividing by ro:
the constitutive
and performing the required integration:
serel t; sds
given as e = u/l, and r = F/A, we cast

h
i
E1 xs2
E1 xs
Answer : rt Ee 1xs
e sinxt
2 eo cosxt
1xs2 o
it turns out that the constitutive
h
i
E1 xs2
E1 xs
Answer : rt Ee 1xs
e cosxt
2 eo cosxt
1xs2 o
given e = u/l, we cast
o
os mt

Rt
rt Mg erel t; 0

Rt
rt Mg erel t

serel sds
there results: 1 s2 M s
there results: s2 M s
o
os mt
Should read as
Reads

E3
Location
Equation (l),
third term on
right hand side
First paragraph
Equation (m),
third term on
right hand side
Third paragraph
Item 2
First paragraph
Equation (c), first

term on right
hand side
Paragraph after
Eq. (k)
Equation (3.27a)
Paragraph after
Eq. (3.27a)
Equation (3.27B)
Page
numbers
66
67
67
73
75
76
76
78
80
80
80
(continued)
qn d
e t
pn dt 0
qn d 2
e t
pn dt2 0
to
Rt
s
Eesr
df
ds. . .
dt0
. . .H t to
to
Rt
Eesr
2. The overall
R
u e dt=sr et=sr
df
ds. . .
ds
i0 j0
j6i
Ei gi ot r
i0
n
Q
Ei gi ot e
i0 j0
j6i
j0
n

Q
Ej gj ot r
Ej gj ot e
n Q
n
P
ei
n Q
n
P
i0
n
P
Kelvin unit, and switching
ei
Kelvin unit, switching
i1
n
P
r and (t) = roH(t - to) should not be split, but should be r(t) = roH(t - to)
. . .H t to
2. he overall
R dt=sr
ue
et=sr
setting f(t) = r(t), and to = 0,
setting to = 0,
pn d 2
r t0
qn dt2
...
(b) Similarly, when

Should read as
pn d
r t
...
qn dt 0
Similarly, when

Reads
(continued)
E4
D. Gutierrez-Lemini
End of list item

(a)
Second
paragraph
Equation (f)
Equation (j), first

term on right
hand side
Second
paragraph
Section 3.7.2,
second paragraph
Problem P3.6
83
85
85
86
87
88
90
Example 4.2,
Solution
Example 4.2,
Solution
98
98
Chapter 4
List item 2
End of first
paragraph
82
Paragraph after
Eq. (3.29)
81
83
Location
Page
numbers
(continued)
Mo q1 =p1 Eo Mg . . .00
. . .Mo q1 =p1 Mg . . .00
jxt
xt2 jxt
E
1 jxt
1 2
and (4.7a), as:
M E
Fig. 3.19 adjacent sketch, show
equation of the Maxwell
and strain becomes
The coefficient ro should be deleted from the term
jxt
xt2 jxt
E
1 jxt
1 xt2
and (4.7c), as:
M E
Fig. 3.19, show
equation of the generalized Maxwell
and stress becomes
(continued)
Two occurrences of ets should read as et =s also, ds should read dt0 and f(t) should be replaced with f(t0 )
will always exhibit
equation (3.28) reveals
we write these expressions in stress
Should read as
will always exhibits
equation (3.31a) reveals
we express it in stress
si, 1/ Ei should not be split, but should be (si ,1/Ei )
Reads

E5
Location
First paragraph
after Eq. (4.11)
Second
paragraph
Equation (4.19a)
Equation (4.19b)
First paragraph
after Eq. (4.22)
Last paragraph
Example 4.5,
Solution
Example 4.5,
Solution
End of second
paragraph
Second
paragraph
Figure 4.5,
caption
Phrase after Eq.

(4.43)
Page
numbers
98
99
99
99
100
100
104
104
105
107
108
109
(continued)
dr 1
d
r E et
dt sr
dt
dr 1
d
r E f t
dt sr
dt
M 00 1 1; C 0 0 Ce ; . . .
. . . M 00 1 1C 0 0 Ce ; . . .
The correct expression should be rt r0 sinxt d; and it should not be split
whereas those of the three-parameter fluid
by parts, using u = M(t) and dv ejxt dt:
whereas those of a fluid
by parts:
Replace all occurrences of the Greek letter s in the expression for C* with the subscripted Greek letter sr
equation (2.1b).
equations (2.32).
such as constitutive equations.
C jx kC kejd
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kC k C 0 2 C 00 2
kC kjx kC kejd
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kC k C 0 2 c00 2
such as constitutive equation.
(4.7a) and (4.14)
this into (2.33b) and
Should read as
(4.7a) and (4.13)
this into (2.33a) and
Reads
(continued)
E6
D. Gutierrez-Lemini
Hint to Problem
P.4.9
Problem 4.10,
first line
112
112
Figure 5.1,
caption
End of first
paragraph
Second line of
second paragraph
Figure 5.3, part

(a)
Example 5.1,
statement
Equation (5.17)
left hand side
Paragraph, after
Eq. (5.18)
115
116
117
119
119
121
121
Chapter 5
Last line
Problem P4.5
109
First line of last

paragraph
109
111
Location
Page
numbers
(continued)
kC kro
in (4.19d):
substituting eo kC kro
Should read as
s(r,x,t) =
s(r, xt) =
at station x.
in Fig. 5.3
in Fig. 5.2
at station, z.
P(t) = Fo f(t)
in the limit as Dx ? 0, yields:
d Force resulants at x
for k = 1,2,
ek Ck rk
P(t) = Po
neglecting higher order terms yields:
The term tan u . . . should not be split
d Force resulant at x
fork=1,2,
ek Ck ek
Question mark appearing at end of problem statement should be removed
in (4.19):
substituting
Reads
(continued)

E7
Location
Equation (5.19a)
Second
paragraph
First paragraph
of 5.3.2
First paragraph
of 5.3.3
Second
paragraph of
5.3.3
Paragraph after
Eq. (5.25)
Second
paragraph. after
Eq. (5.32)
Paragraph. after
Eq. (5.42b)
Equation (g)
Equation (5.46a)
Second line, last

paragraph
Page
numbers
121
122
122
123
123
125
126
133
135
135
137
(continued)
elastic relationships
K T sws
viscoelastic relationships
T s
JQs=Ps
ws
L
K T sws
JQs=Ps
ws
L
T s
time, as in Chap. 2
is as follows:
this takes the form:
time of as in Chap. 2
is that as follows:
this take the form:
T t JGt s d
(5.19a,b)
(5.19a)
T t JGt s d
(5.13). Thus
ow
x; s
ox
o
wx; t . . .
ox
(5.17)
hx; t
Should read as
(5.15). Thus
dw
x; s
dx
o
wx; t . . .
oz
(5.18)
hx; t
Reads
(continued)
E8
D. Gutierrez-Lemini
Fourth paragraph
Bullet point b
Footnote
Paragraph after
Eq. (5.48)
Fifth paragraph
Paragraph after
Eq. (5.59)
Problem P.5.1,
Answer
Problem P.5.3,
Answer
Problem P.5.9,
Hint
138
138
138
139
141
144
145
146
147
Last paragraph
Second
paragraph
Last paragraph
Second
paragraph after
Eq. (6.2b)
150
152
153
154
Chapter 6
Location
Page
numbers
(continued)
to t/aT(T,Tr) units
time t, and T,
where it is not for
dealt within this
Apply (2.37)
0L
QvL; t L3I PF t F3I
P F t
to t/aT(T,Tr) units
time t, and temperature T,
were it not for
dealt with in this
Apply (2.45)
0L
QvL; t L3I PF t F3I
P f t
Chap. 4, developed
Chap. 4 developed
= -q(x,t)
the magnitude of F* is
the magnitude, F* is
= q(x,t)
the components K 0 and
the component K and
the complex, controlled variable
(continued)
,
, or torsional acceleration, w
bending acceleration, u
, or rotational
angular acceleration, w
,
acceleration, w
the complex-controlled variable
except as follows:
Should read as
except that as follows:
Reads

E9
Last paragraph
Second line of
footnote
Fourth
paragraph,
second line
Fourth
paragraph, third
line
Fourth
paragraph, fourth
line
Last line of
fourth paragraph
Last paragraph
Last reference
154
159
160
160
160
160
161
163
Abstract, second
to last line
Third paragraph
165
166
Chapter 7
Location
Page
numbers
(continued)
using Eq. (2.32a)
function in a Prony
using Eq. (2.1b)
functions in Prony
7. W.G. Knauss, I. Emri
Tr(F) = 9/5 Tr(F) ? 32_73.0 F
W.G. Knauss, I. Emri
C2(F) = 9/5 C2(C) & 134.9 F

Tr(F) = 9/5 Tr(C) ? 32 & 73.0 F
C2(F) = 9/5 C2(C)_134.9 F
M dC H t
M dC H t
with (2.28) yields
eM t C t sdrs
eM t C t s drs
with (2.20) or (2.21) yields
inserting (6.18) into
when Tr = Tg.
curve at t/aT(T,Tr)
Should read as
inserting into
when Tr = Tg.
curve at t/aT(Tr)
Reads
(continued)
E10
D. Gutierrez-Lemini
Location
First paragraph
of 7.2.2
Last paragraph,
3rd line
Last paragraph
Equation (7.24)
Equation (7.25)
Equation (7.26)
Third paragraph
Equation (7.27)
Paragraph after
Eq. (7.27)
Page
numbers
169
169
169
176
176
177
177
177
177
(continued)
i1
N
P
Cr br
br ai
i1
Mi br
br ai
0;
Cr ai
br ai
Mi ai
br ai
i1
bai
N
P
Mi ai
0;
0;
i 1; . . .; N
r 1; . . .; N
are shorter than the corresponding relaxation

times
(continued)
are longer than the corresponding relaxation times
b1 \a1 ; ai \bi1 \ai1 ; for i 1; N 1
Ub Mg
r1
N
P
i1
N
P
ai \bi \ai1 ; for i 1; N 1; and bN [ aN
Ce
Mg
These considerations reveal that U tends to Me and Mg as b

approaches 0 and ?, respectively. In addition, U tends to +? and
-? as b approaches ai from above
Mi b
bai
i 1; . . .; N
r 1; . . .; N
temperature. By calculating
to the loading
per Eq. (2.2b)
Should read as
These considerations reveal that U tends to Mg and

Me as b approaches 0 and ?, respectively. In
addition, U tends to -? and +? as b approaches ai
from above
i1
N
P
0;
Ub Mg
Ce
Mg
N
P
temperature, by calculating
to the loading is
per Eq. (2.2)
Reads

E11
Location
Figure 7.3
Figure 7.4
Page
numbers
177
178
(continued)
i+1
i +1
Fig. 7.4 Behavior of the polynomial equation in the

relaxation times. C is the source function, and M, the
target
Cg
Ce
()
Fig. 7.3 Behavior of the polynomial equation in the

retardation times. M is the source function and C,
the target.
Reads
Fig. 7.4 Behavior of the polynomial equation in the reciprocal

relaxation times. C is the source function, and M, the target.
(continued)
Fig. 7.3 Behavior of the polynomial equation in the retardation

times. M is the source function and C, the target.
Should read as
E12
D. Gutierrez-Lemini
First line
Middle of page
First bullet (),

second line
183
184
188
Last paragraph
Second
paragraph
(expression split
at index)
194
195
Chapter 8
Last paragraph
Table 7.3
heading
181
190
Paragraph before
Eq. (7.32)
181
Second bullet ()
Bullet point 4.
180
First paragraph
Paragraph after
bullet c.
180
188
Second
paragraph
178
190
Location
Page
numbers
(continued)
of temperature and shift values
relative to the curve
. . .Aij djk Ai1 d1k Ai2 d2
k Ai3 d3k . . .
tensors of the same order can be multiplied
(1 = modulus, 0 = compliance)
annotations are only permissible
and And
. . .Aij djk Ai1 d1k Ai2 d2k Ai3 d3k
tensors can be multiplied
(1 = modulus, -1 = compliance)
annotations are permissible
and
7.5.2.2
to assign the correct sign
(b)
7.5.2.1
to assign correct sign
(a)
Column headings i and t (sec) should appear at same level as column heading logt=aTK
of temperature shift values
relative to curve
to the scheme
(continued)
Wa 0 Ce and Wa 1 Cg , respectively. Also, W tends

to -? and +? as a approaches
Wa 0 Cg and Wa 1 Ce , respectively.
Also, W tends to +? and -? as a approaches
to scheme
Should read as
Reads

E13
Location
Last paragraph
Paragraph after
Eg. (8.6)
Section 8.4, 1st

paragraph
Last paragraph
Example 8.1,
Solution
Equation (8.20)
Equation (8.23)
Equation (8.24)
First paragraph
Second
paragraph
Fourth paragraph
Fourth paragraph
Equation (8.36b)
Third line of text
Page
numbers
195
198
199
201
203
204
204
205
206
206
206
206
207
208
(continued)
Likewise take the convolutions of (8.33b) to write
rDij t 2lM deDij

Likewise, from (8.33b)
eDij t 2lC drDij
ADi j
Asij Akk =3dij

1=3A11 A22 A33 dij
. . .Asij Akk =3dij 1=

3A11 A22 A33 dij . . .
As will become apparent shortly
As will become apparent later shortly
ADi

1 sv1 2svs
k svsE
Cijkl t kC tdij dkl lC t . . .
Mijkl t kM t dij dkl lM t . . .
Mijkl t kt dij dkl lt . . .
of the response to the complex stress
each Poissons ratio function
linear orthotropic
fei gT fegT fe11 ; e22 ; e33 ; e12 ; e13 ; e23 g
parallel to the axes
Should read as
. . .1 sv1 2sv
svsE.
. .
sk

Mijkl t kt dij dkl dil djk lt . . .

Mijkl t kM t dij dkl dik djl lM t . . .

Cijkl t kC t dij dkl dik djl lC t . . .
of the complex stress
each Poissons ratio functions
linear anisotropic
fei g feg fe11 ; e22 ; e33 ; e12 ; e13 g
parallel to the axis
Reads
(continued)
E14
D. Gutierrez-Lemini
Equation (8.40)
Second
paragraph
Paragraph in first
Hint
Problem P.8.5,
Hint
Problem P.8.6,
Hint
Problem P.8.8,
Answer
Figure 8.6,
caption
209
211
214
214
214
215
216
Abstract, last
sentence
Third paragraph
219
222
Chapter 9
Location
Page
numbers
(continued)
rD22
1=3r11 2=3r22
ekk = (r11 ? r22)/3K
rD22
1=3r11 2=3r22
ekk = (r11 ? r22)/3
(as listed in [2])
elastic ones and as a consequence
(as listed in Appendix B)
elastic ones and is a consequence
ezz t po . . .;

4G 3K
sg
3KG
rkk = r11, ekk = (e11 ? 2 e22)
rkk = (1/3)r11, ekk = (e11 ? 2 e22)/3
ezz t po . . .; s

4G 3K
g
3KG
The applicable constitutive

. . .In addition;
pt ro H t=3. . .
The constitutive applicable constitutive
. . .In addition;
pt ro H t=3. . .
(continued)
The steady-state response rij and p* to the complex strain eij may
obtained from (8.41a), proceeding as in Chap. 4, using the notation
jx l l as:
l ; p p ; rij
Using the notation: jx l

rij ; eij eij ; as is customary, (43-a) may expressed
alternatively as
kl jx jx
kM jx emm jx dkl
r
2jx
lM jx ekl jx
eij jx
rij jx kM jx ekk jx dij
2lM jx
Should read as
Reads

E15
Third paragraph
last line
Paragraph after
Eq. (9.14)
Equation (9.21)
Equation (9.30)
222
226
228
231
Abstract, third
line
Paragraph after
Eq. (10.8)
Equation (10.11)
Last paragraph
Second
paragraph
Third paragraph
Last paragraph
Equation (f)
239
242
243
244
245
245
245
252
Chapter 10
Location
Page
numbers
(continued)
decrease from the left to right

Rt
o o2
E t s
u
ds
os ox2
o
back into 10.9)
decrease from left to right

Rt
o o2
Et s
u
ds
os ox2
o
back into (10.9)
(continued)
. . .real part of U ejxo t ; the. . .
qffiffiffiffiffiffiffiffiffiffiffiffiffi
qffiffiffiffiffiffiffiffiffiffiffi
Pjxo
q
q Qjx
x
o
E jxo

o
. . .real part of U eejxo t ; the. . .
(2)
k xo
other being -k
are introduced in
A parametric treatment of a problem of this type, for a

three-parameter solid, may be found in W. Flgge, Viscoelasticity,
pp. 135140, Springer-Verlag, 1975.
qffiffiffiffiffiffiffiffiffiffiffiffiffi
qffiffiffiffiffiffiffiffiffiffiffi
Pjxo
q
q Qjx

x
o
E jxo

o
v
2v
ER
GR
1 v1 2v
1 2v
rij s skM sekk sdij . . .
kR
auxiliary displacements, uVi ,
respectively, are listed here:
Should read as
Footnote 2 is missing
k xo
other being k
are introduce in
kR
v
2v
; ER
GR
1 v1 2v
1 2v

rij s skM sekk sdij . . .
auxiliary displacement uVi ,
respectively, and listed here:
Reads
E16
D. Gutierrez-Lemini
P.10.4, equation
in Hint
P.10.5, Hint,
second line
P.10.5, Hint,
second line
P.10.7, Statement
258
258
258
259
Equation
(10.46b)
256
P.10.2, Hint
Paragraph after
Eq. (10.46a)
256
257
Paragraph after
Eq. (10.46a)
256
P.10.1
Second
paragraph
256
P.10.2, Statement
Example 10.6,
Solution
253
257
Example 10.5,
Solution
253
257
Location
Page
numbers
(continued)
ffiffiffiffiffiffiffiffiffiffiffiffi x
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiss1=s
c
ss 1=se
pffiffiffiffiffiffiffiffiffiffiffiffiffi x pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r qVce ss1=sc = ss 1=s
(10.40b)
Should read as
using (10.26)
of Example 10.8
A(s) =
A(S) A =
of Example 10.5
(10.41b)
C0 E1g dsd Et sjts E1g dtd Etjt0
C0 E1g dsd Et sjt0 E1g qEqjq0 EEg0
(3.23a)
q = 1600 kg/m3
1 MPa and mass density 1600 kg/m3
q = 1600 kg
Example (10.8)
Example. 10.6)
v Be . . .
145 psi and mass density 0.065 lbm/in3
v Ae . . .
The term D=Dt in expression D=Dt cD=Dx should not be split
using (10.8)
The term or=ox in expression or=ox pov=ox should not be split
r qVc
(10.41b)
Reads
(continued)

E17
Problem P.10.7,
Hint
259
Paragraph after
Eq. (11.9c)
Second
paragraph
Section 11.3.2,
First paragraph
Second
paragraph
Line after
Equation (c)
Equation (c)
Equation (11.22)
266
268
268
270
270
270
271
Chapter 11
Location
Page
numbers
(continued)
Su
dTi
00
uoi
u00i dS
ST Su
dYo r00ij dV
Su

1
00
ui;j u00j;i ; in V
e00ij t
2
dwo r00ij
ST
and using
R
R
00
00
dTi ui dS nj drij ui dS
nj dr00ij ui S
0
Su
00

1
ui;j uj;i ; in V
2
dYo r00ij dV
ST Su
dTi uoi ui dS
e00ij t
dwo r00ij
ST Su
ST Su
also subtracting
R
00
dTi ui dS 0, and using
ST
R
R
00
00
dTi ui dS
nj drij ui dS
introduced in 11.15
fields u00 . A
displacement field, u00
fields u. A
displacement field, u
introduced in 11.5
of the d0 r
E2
g0
Multiplying by (11.4a)
d
dt Etjt0
Should read as
of the dr
Multiplying (11.4)
E02
d
dt Etjt0 g1
Reads
nj dr00ij ui dS 0
(continued)
E18
D. Gutierrez-Lemini
Figure 11.4,
inset
Next-to-last
paragraph
Example 11.4,
Solution, last
equation
Equation
(11.34b)
Paragraph after
Eq. (d)
276
276
278
279
280
First line
273
First line
Example 11.2,
Solution
272
Example 11.3,
Solution
Example 11.2,
Statement
272
275
Last paragraph
271
275
Location
Page
numbers
(continued)
Vo dv
T0i1 t s
o 2
u sdsdS . . .
os i
2
j1
^ Nz
^ Mx
R Mux C
R Nuz C
dx
dxs
A
Ix
L
L
r1IJ Mijkl de1kl
ST 0
R Rt
ujP
the first part on far right is
xL=2
by establish equilibrium
only when bending effects are included
R
V
Yo dV
Ti1 t s
o 2
u sdsdS . . .
os i
j1
^ Nz
^ Mx
R Mux C
R Nuz C
dx
dx
A
Ix
L
L
r1ij Mijkl de1kl
ST
R Rt
ujP
the first part on the far right is
qL=2
by establishing equilibrium
when only bending effects are included
leads to Dt
and Eq. 11.21: Yo re Ygo
and Eq. 11.21: co re cgo
leads to Dt
its cross section has second
express 11.15 as:
Should read as
its cross section have second
express 11.23 as:
Reads
(continued)

E19
Location
Equation
(11.35b)
Paragraph after
Eq. (11.35b)
Page
numbers
281
281
(continued)
j1
ST 0
system r1ij;j Fi1 q

u1i
ST 0
ui
system r1ij;j Fi1 q
Ti1 t su2i sdsdS
R Rt
R Rt
T0i1 t su2i sdsdS
Should read as
Reads
j1
E20
D. Gutierrez-Lemini
Appendix A
Mathematical Background
A.1 Average Value of a Function

Simply put, an integrable function is one whose integral exists. The average value
of an integrable function, f, over a finite interval [a,b] is defined as the quotient of
the value of the integral divided by the amplitude of the integration interval.1
1
A f
b a
Zb
f sds
A:1
f sds b aA f
A:2
Stated another way:

Zb
a
In other words, the average value A(f) of the function f may be thought of as the
height of a rectangle with base equal to the length of the interval over which the
average is taken. This is indicated schematically in Fig. A.1.
When f is continuous, its average value is equal to the value of f at some point in
the interval [a,b]. This is the mean value theorem for integrals, proven next.
A.2 Mean Value Theorem for Integrals

If f is continuous on [a,b], then for some c e [a,b]:
Zb
f sds b a f c
A:3a
Reference material for sections A.1 to A.5 may be found in T.M. Apostol, Calculus, 2nd
Edition, Xerox College Publishing, (1967), pp. 154, 184186.
D. Gutierrez-Lemini, Engineering Viscoelasticity, DOI: 10.1007/978-1-4614-8139-3,

283
284
Appendix A: Mathematical Background
Fig. A.1 Geometric interpretation of mean value

theorem
f (x)
A( f )
x
a
Proof Let m and M denote, respectively, the minimum and maximum values of
f (x) on [a,b]. Then, m B f (x) B M. Integrating this system of inequalities in the
given interval, dividing through by (b-a), and using the average value of a
function, there results
1
m
b a
Zb
f sds Af M
In addition, the intermediate value theorem for continuous functions tells us that
f takes on every value between f (a) and f (b), somewhere in the interval [a,b].
Thus, A (f) = f (c), for some c e [a,b]; completing the proof. The theorem is
alternatively expressed as
Zb
f sds b a f a kb a;
0\k\1
A:3b
A.3 Weighted Mean Value Theorem for Integrals

Let f and g be continuous on [a,b]. If g never changes sign on [a,b], then for some
c in [a,b]:
Zb
a
f sgsds f c
Zb
gsds
A:4
Proof The proof proceeds as that of the mean value theorem for integrals. Since
g never changes sign in [a,b], it is always nonnegative or always non-positive.
Assuming g is nonnegative, we multiply the system of inequalities m B f (x) B
M by g(x), to yield m g(x) B f (x)g(x) B M g(x), and integrate it between the limits
of the given interval, so that
Zb
m
gsds
285
Zb
a
f sgsds M
Zb
gsds
If the integral of g is zero, the theorem is trivially satisfied; since in that event, the
integral of f g is also zero, and both members of the theorem are zero, for any c in
[a,b]. Otherwise, the integral of g would be positive. This would allow division of
the above system of inequalities by the integral of g and apply the intermediate
value theoremas we did for the mean value theoremto complete the proof.
This theorem is useful to get an estimate of the integral of a product of two
functions, especially when one of the functions is easy to integrate.
R 1 x9
1
Example A.1 Prove the inequality: 101p2 0 p
dx 10
1x
Solution
p
First note that both, x9 and 1= 1 x are continuous and nonnegative in the
interval of integration. Then, apply the weighted mean value theorem for integrals,
R 1 x9
R1 9
1
1
to write 0 p
dx p
x dx 10p1
; n 2 0; 1. Now, since p12 p
1;
1n 0
1n
1n
1x
R
1 x9
1p
1
for all n in [0,1], it follows that 10 2 0 p
dx 10:
1x
A.4 Rolles Theorem

For any function, f, which is continuous in [a,b], and differentiable in (a,b), and
such that f(a) f(b), there exist at least one point c e (a,b) where its derivative
vanishes f 0 c 0:
Proof This theorem is proven assuming that f 0 x 6 0 everywhere in (a,b), and
noting that such assumption leads to a contradiction, which then implies that
f 0 x 0 for at least one x in (a,b). Indeed, because f is continuous in [a,b], it
acquires its minimum and maximum values, m, and M, respectively, somewhere in
[a,b]. Also, since a necessary condition for the existence of an extreme value
(maximum or minimum) of a function is that its derivative vanishes at the critical
point, it follows that f 0 x 0 for some x in [a,b]. However, by assumption,
f 0 x 6 0 everywhere in the interval (a,b). This leaves x a and x b as the only
critical points of f in [a,b]. That is f 0 a 0 and f 0 b 0. This, and the property
that f(a) f(b), would require that m M; in which case, f would be constant in
[a,b], and f 0 x 0; everywhere in [a,b]. This contradicts the hypothesis that
f 0 x 6 0 everywhere in (a,b), and hence, f 0 x 0 somewhere in (a,b);
completing the proof.
286
A.5 Mean Value Theorem for Derivatives

If f is continuous in [a, b], and differentiable everywhere in (a,b), then there is a
point c in (a,b) for which
f 0 c
f b f a
b a
A:5
In other words, as is depicted in Fig. A.2, there isat leasta point in the
interval (a, b), where the tangent to the curve has the same slope as the segment
joining the end points of the interval.
Proof The proof is based on Rolles theorem, which applies to the same type of
function as the present theorem, but with the added requirement that f (a) f (b).
To apply Rolles theorem, we construct the function: hx
f x b a x f b f a. By construction, h(x) is continuous in [a,b],
differentiable in (a,b), and such that h(a) h(b). Under these conditions, Rolles
theorem applies to h(x); thus, h0 (c) 0 for some c in (a,b). Hence, h0 c 0
f 0 c b a 1 b a; proving the theorem.
A.6 The Total Derivative

If y f(x) represents a real-valued function of a real variable, its derivative, f 0 x;
is defined by
1
f 0 x lim f x h f x
h!0 h
A:6
Now consider the following limit, denoted by f 0 x; z; or df x; z and called the

total derivative or first variation of f with increment z2:
1
d
df x; z lim f x hz f x f x hzjh0
h!0 h
dh
A:7
As will be shown subsequently, this limit is linear in z, so that

df x; z f 0 x; z f 0 xz
A:8
This establishes the relation between the derivative f 0 x and its total derivative
or first variation, f 0 x; z; ordf x; z:
The linearity of df x; z on z follows from its homogeneity and additivity. To
prove that df x; z is homogeneous of degree one in z, consider the variation
df x; kz; and using its definition in (A.7), multiply and divide its right-hand side
by k and invoke that the product of the limit is the limit of the product, to write
2
D.C. Leigh, Nonlinear Continuum Mechanics, McGraw-Hill (1968), pp. 4649.

Fig. A.2 Mean value theorem for derivatives
287
(c, f (c))
1
k
df x; kz lim f x hkz f x lim f x hkz f x
h!0 h
h!0 kh
c
a
Now set l kh and arrive at the following expression, which proves

homogeneity:
1
df x; kz k lim f x lz f x kdf x; z
l!0 l
To prove additivity, evaluate df x; z w using (A.7). Then, split the argument

x + h(z + w) into (x + hz) and hw, and add and subtract the function f(x + hz) to
express the result as
1
df x; z w lim f fx hz wg f fxg
h!0 h
1
lim f fx hz hwg f x hz
h!0 h
1
lim f fx hzg f f xg
h!0 h
df x; z df x; w
A.7 Differentiation Under the Integral Sign

Differentiation under the integral sign is carried out by means of Leibnitzs rule,
that
If u1(a) and u2(a) are differentiable functions in the closed interval [a,b], and
f(x,a) and qf(x,a)/qa are continuous in the region of the x-a plane delimited by a B
a B b, u1(a) B x B u2(a), then
288

uZ2 a
d
da
f x; adx
u1 a
uZ2 a
o
f x; adx
oa
u1 a
f u2 a; a
d
d
u2 a f u1 a; a u1 a
da
da
A:9
R u a
Proof Letting: Ua u12a f x; adx; we proceed to compute its derivative using
the four-step rule. We thus evaluate U(a + Da), DU, DU/Da and
limDa!0 DU=Da:
Z u2 a
Z u2 aDa
f x; a Dadx
f x; adx a
DUa Ua Da Ua
u1 aDa
u1 a
We now split the limits of integration of the first integral on the right-hand side
into three subintervals; the first, from u1(a + Da) to u1(a); the second, from u1(a) to
u2(a); and the third, from u2(a) to u2(a + Da). We then combine the second integral
on the right-hand side of the above expression with the split integral of the same
limits, to arrive at
uZ1 a
DUa
f x; a Dadx
u1 aDa
u2 ZaDa
uZ2 a
f x; a Da f x; adx
u1 a
f x; a Dadx
u2 a
Rewrite the first integral in this expression by switching its limits of

integrationwhich picks up a negative sign:
uZ1 a
U 1 a
u1 aDa
f x; a Dadx
u1 ZaDa
f x; a Dadx
u1 a
Apply the mean value theorem for integrals to this expression to get
U1 a f n1 ; a Da u1 a Da u1 a
U1 a f n1 ; a Da Du1 a; u1 a \n1 \u1 a Da
That is
In similar fashion, invoking the mean value theorem for integrals, the third
integral in the expression for U may be cast in the form:
U2 a
u2 ZaDa
u2 a
f x; a Dadx f n2 ; a Da u2 a Da u2 a
289
In other words,
U2 a f n2 ; a Da Du2 a; u2 a \n2 \u2 a Da
Using the expressions for U1 and U2 to replace the corresponding integrals

appearing in DU, leads to
DUa

Da
uZ2 a
f x; a Da f x; a
Du1 a
dx f n1 ; a Da
Da
Da
u1 a
f n2 ; a Da
Du2 a
Da
Taking then limit as a?0, noting that n1 ? u1(a) and n2 ? u2(a) as a?0, and
using the definition of the partial derivative of a function, completes the proof.
Example A.2 Obtain the derivative with respect to t, of the function
R t ts
qt 0 e a psds
Solution
Direct application of Leibnitzs rule yields
d
qt
dt
Zt
o ts
dt
d0
tt
t0
e a psds e a pt e a p0
ot
dt
dt
That is
d
1
qt
dt
a
Zt
1
ts
e a psds pt qt pt
a
R t ts
Expressions of the form dtd qt 1a 0 e a psds pt 1a qt pt;
occur naturally in viscoelasticity and are solved for the state variable q (stress or
strain) in terms of the source function, p (strain or stress, respectively) using the
procedure described in the next section.
A.8 Linear Differential Equation of First Order

The general linear ordinary differential equation of first order may be written as
d
yx p x y x qx
dx
A:10
290
The solution to this equation is obtained by rewriting it in terms of differentials,

and multiplying it by a continuous function, u(x)3:
ux dyx ux p x y x dx ux qx dx
With this, the left-hand side of the equation becomes an exact differential of the
product u(x) y(x). That is, look for a (non-zero) function u(x) such that
du x y x u x qx dx
The solution, y(x), of the general linear differential equation posed, is obtained
integrating this expression. This can be accomplished either by indefinite or by
definite integration.
Using indefinite integration and the fact that the indefinite integral of the total
differential of a function is the function itself, results in:
Z
u x y x us qs ds C
A:11
x
The dummy variable of integration of the indefinite integral on the right-hand

side was changed for clarity. The constant of integration, C, is established from the
condition (xo,yo) at the end x xo of the interval of integration.
Alternatively, if definite integration between the limits xo and x is employed
u x y x uxo yxo
Zx
us qs ds
A:12
xo
The integrating function u(x) may be established from the requirement that
multiplying the original equation by it should turn its left-hand side into an exact
differential. That is
du x y x ux dyx ux p x y x dx
Evaluating the left-hand side of this expression, canceling like terms, and
R
x
regrouping, produces that du
pxdx;
ux p x dx; with integral lnfu xg
which leads to the integrating factor:
R
u x e pxdx
A:13
Example A.3 Obtain the general solution of the differential equation dtd qt
1
s qt pt; if the solution passes through the point (to, qo). That is, if q(to) qo.
Elementary Differential Equations, E.D. Rainville, P.E. Bedient, 4th Edition, Collier
Macmillan, London (1969), pp. 3440.
291
Solution
Expressions of this type occur naturally in viscoelasticity and are used to
determine the state variable, q, in terms of the source function, p. By the present
method, an integrating factor for this equation is obtained from (A.13) as u(t) et=s ;
and noting that t is the independent variable, its general solution, in accordance to
(A.11) or (A.12), is
q t e
at
Zt
ea psds e
tto
a
qo
to
Example A.4 The stress, r, in viscoelastic system may be expressed in terms of

the applied strain, e (t), in the form: dtd r a r dtd e; a 2 R. Determine r (t) if the
initial stress is r(0) ro and the applied strain is e (t) R t.
Solution
The given stressstrain relation is a first-order ordinary linear differential
equationR in r (t). An integrating factor for it may be found from (A.13), as
ut e adt eat

R
This, and (A.11) or (A.12), produce that rt eat r0 eas Rsds .
Integrating this expression by parts, simplifying, and
using the initial
condition on
r, produces the result sought: rt eat r0 Ra t 1a 1 eat .
A.9 Differential Operators

Differentiation may be considered as an abstract operation, which when applied to
a given function returns another function. With this idea in mind, the first
derivative, df
dt ; of a function, f, with respect to some variable, t, may be regarded as
the result of applying the derivative operator of first order, dtd ; to the function f. The
derivative operator of first order is represented by any one of several symbols, as
d
dt ; ot ; D; Dt . Also, note that this operator is linear, because the operation of
differentiation is homogeneous and additive; that is
By homogeneity is meant that for any function k which does not depend on t:
d
d
kf k f
dt
dt
By additivity is meant that, for any two suitable functions, f1 and f2:
d
d
d
f1 f2 f1 f2
dt
dt
dt
Invoking the property of differentiation that the derivative of the first derivative
of a function is the second derivative of the function; and so on, we can construct
292
derivative operators of order higher than one. Thus,

n
d d n1
f
dtd n f .
n1
dt dt
d d
dt dt f
dtd 2 f ; . . .;
We denote the derivative operator of order n by dtd n ; ont ; Dn ; Dnt ; and, for
practical purposes, generalize itto include
the derivative operator of order zero, to

signify no derivative is taken:
d0
dt0
1 f f.
The foregoing serves to define the general linear differential operator of order
n with constant coefficients, or with coefficients which do not depend on the
differentiation variableas a sum of derivative operators of orders up to and
including order n:
P p0
Xn
d0
d1
d2
dn
di

p

p
1
2
n
i
i0
dt0
dt1
dt2
dtn
dti
A:14
where the pis are any suitably continuous functions which do not depend on t;
and, as required by the order of the operator, pn = 0.
By definition, linear differential operators can be added together (by additivity),
or factored by constants or functions that do not involve the variable of
differentiation (by homogeneity). Specifically, given any two linear operators, P,
and Q, a suitably smooth function, f, of the differentiation variable, and any
constant, k, the following relation holds:
P Qkf Pkf Qkf kP f kQ f
In addition, linear differential operators are both commutative and associative.

These properties result from the fact that the mth derivative of the nth derivative
is equal to the nth derivative of the mth derivative. Because of their linearity,
these operators are distributive as well. Thus, for linear operators P, Q and R, and
function f1 and f2:
Commutative property
PQf1 PQf1 QPf1 QPf1
Associative property
PQRf1 PQRf1 PQRf1
Distributive property
PQf1 f2 PQf1 PQf2
293
A.10 Linear Differential Equations of Higher Order

A differential equation is one involving an unknown function and some of its
derivatives. If the derivatives appearing in the equation are ordinary derivatives,
the equation is called ordinary; and if the derivatives are partial derivatives, the
equation is a partial differential equation. The order of the highest-ordered
derivative in the equation represents the order of the differential equation. Also, a
differential equation is said to be linear if it is linear in the unknown function and
its derivatives.
Using differential operator notation, an ordinary linear differential equation of
order n may be expressed in the following symbolic form:
f yx Rx
A:15
In which, as in the previous section, the operator (f) is defined by

f a0 D0 a1 D1 a2 D2 an Dn
A:16
Clearly, for the equation to be linear, none of its coefficients can depend on the
unknown, function y.
Only linear equations with constant coefficients are considered here. Also, if the
function R(x) is identically zero in the interval of interest, the differential equation
is called homogeneous; otherwise, the equation is non-homogeneous.
A.10.1
Homogeneous Differential Equations
By definition, a homogeneous ordinary linear differential equation has the simple

form:
f yx 0
A:17
Solutions to this type of differential equation are found by inspection. Indeed,

noting that Dk emx mk emx , we try the function y(x) Aemx as a solution to (A. 17)
and arrive at

a
f Aemx a0 a1 m a2 m2 an m Aemx f m Aemx 0
After canceling out the factor Aemx, this produces the auxiliary equation:

f m a0 a1 m a2 m2 an mn 0
b
The auxiliary equation (b) is a polynomial equation of degree n in the exponent
m of the trial solution and has, in general, n roots, mi. Once these roots are
obtained, the solution of the homogeneous differential equation should be given,
by its linearity, as the sum of the n independent solutions, Ai emi x . Two cases are
distinguished: non-repeated and repeated roots, as follows.
294
Non-repeated roots
In this case, the complete solution to (A.17) is given by the sum:
Xn
y x
A em i x
i1 i
A:18
The coefficients Ai are established from the boundary (or initial) values of the
function y(x) and its derivatives up to order n - 1.
Repeated roots
When the auxiliary equation f (m) 0 has repeated roots, the operator (f) has repeated
factors. Assume, for definiteness that the auxiliary equation has p equal roots mi b.
Then, the operator (f) must have a factor (D - b)p, and Eq. (A.17) can be cast as
f y x gD bp y x 0
where the operator (g) contains all the factors other than (D - b)p. Clearly, any
solution of
D bp y x 0
is also a solution of Eq. (c), and therefore of the original Eq. (A.17). In summary,
we need to find p independent solutions of Eq. (d). This is easily done, noting that
the functions ykxkebx, k 0, 1,, p - 1, are linearly independent because the
xks are linearly independent, and consequently, so are the proposed functions, yk.
Using this, we construct the general solution of (A.17), for the case of p repeated
roots mkb, of its auxiliary equation, as
Xnp
Xp1
y x
A em i x
B xk ebx
A:19
i1 i
ko k
The same ideas and procedure are used when the auxiliary equations has
multiple sets of repeated roots.
Example A.5 Find the general solution of the homogeneous equation
2 dy
dx 4 0:
d2 y
dx2
Solution
The operator form of the given equation is D2 2D 1y x 0: This leads
to the auxiliary equation m2 2m 1 0; with roots m1;2 1; 1; for which the
solution is y x Aex Bxex .
A.10.2
Non-Homogeneous Differential Equations
The general solution of the non-homogeneous ordinary linear differential equation

in (A.15) may be expressed as the sum:
295
y x yh x yp x
A:20
of the complementary solution, yh, to homogeneous Eq. (A.17), and some

particular solution, yp, to the original Eq. (A.15).
The solution of the homogeneous equation is obtained as described before, from
the roots, mi, of the auxiliary equation f m 0: In what follows, a particular
solution is sought using the method of undetermined coefficients. This method is
applicable only to cases where the function R(x) is itself a solution of some linear
homogeneous differential equation, say: gyp x Rx: Indeed, in such a case,
the roots, m0k ; of the auxiliary homogeneous equation gm0 0 are obtained by
inspection, from R(x). In addition, the differential equation:
gf y x 0
has an auxiliary equation whose roots are the roots, mi, of the auxiliary equation
f (m) 0, together with the roots m0k of the auxiliary equation gm0 0: This
means that the general solution of (e) contains the particular solution yp, referred to
in (A.20) and so, is of the form: y x yh x yr x: Then, f yr x Rx;
because f yh x 0: Removing yh from the general solution of (e) leaves the
function yr, which for some specific numerical values of its coefficients, must
satisfy the original non-homogeneous equation (A.15). The required coefficients of
yr are established from the requirement that yr yp.
Example A.6 Find the general solution of the equation
d2y
dx2
2x
2 dy
dx 1 2e .
Solution
The roots of the complementary homogeneous equation were obtained in
Example A.5, as m1;2 1; 1: The function e2x on the right-hand side of the given
equation corresponds to an auxiliary equation with root: m01 2: Therefore, the
roots of the general solution y yh yp ; have to be: m1;2;3 1; 1; 2; which
requires that y A1 ex A2 xex Be2x : The constant B is established from the
requirement that (f)y(x) 2e2x, which produces B 2. Hence, the general solution
of the equation posed is A1 ex A2 xex 2e2x :
A.11 The Divergence Theorem

This theorem is an example of an integral transformation. It concerns a function
f(x,y,z) which, together with its first partial derivatives, qf/qx, qf/qy, qf/qz, is singlevalued and continuous inside a simply connected volume Vthat is, one without
holes bounded by a close surface, S. Then, using the notation: x1 $ x, x2 $ y,
x3 $ z, there follows:
Z
Z
o
o
o
n1 f1 n2 f2 n3 f3 dS
f1 f2
f3
dV
A:21
ox1
ox2
ox3
S
296
Fig. A.3 Domain of definition of function f, in GreenGauss theorem
g
E
Here, ni is the ith direction cosine or component of the outward unit normal to
the surface S that encloses the volume, V, and always points away from the surface.
Clearly, the ni are thus functions of position on S. The surface S is required to be
piecewise smooth, and such that it unambiguously defines and inside and an
outside. The theorem is based on the following relationship involving the function f:
Z
Z
o
ni fdS
fdV; i 1; 2; 3
A:22
oxi
V
Proof To prove the Green-Gauss theorem, consider a function f (x, y) defined on a

plane area S, and having continuous first partial derivative with respect to x, and let
C be a closed curve surrounding the plane area S, having the property that any
straight line parallel to a coordinate axis cuts C in at most two points, as shown in
Fig. A.3.
In addition, let yb (x) and yt (x) represent, respectively, the equations of the bottom
and top curves, AEB and AFB, respectively. From (A.22), choosing i 2, for
R
R o
R
R
definiteness, one can write that z dz A oy
f x; ydA z dz C ny x; yf x; ydS;
R
which, after canceling the common term z dz; results in the expression to be proven:
3
2
Zb
Z
Zyt x
Z
o
o
7
6
f x; ydA
f x; ydy5dx
ny fdS
4
oy
oy
C
xa
Zb
yt x
f x; yjyy
dx
b x
xa
Zb
xa
I
C
yyb x
Zb
f x; yt f x; yb dx
xa
f x; yt dx
Zb
f x; yb dx
xa
f x; ydx
Zb
xa
f x; y
f x; yb dx
Za
f x; yt dx
xb
dx
ds
ds
As seen in Fig. A.4, the normal, ^n; at any point on a smooth plane curve C, may
be expressed as ^
n înx ^jny îsinh ^jcosh; whereby
297
Fig. A.4 Unit normal to a

smooth plane curve
ds
dy
dx
nx
dy
dx
; ny
ds
ds
Using (b) on the right-hand side of (a) proves that

Z
Z
o
fdV
ny fdS
oy
A
The expressions for i 1 and i 3 in (A.22) can be proven similarly. Also,

although for simplicity the proof was carried out assuming the closed curve
C could be cut no more than at two points by any lines parallel to the coordinate
axes, the theorem is valid for cases in which lines parallel to the axes meet the
bounding curve, C, in more than two points. It is also easy to show that the GreenGauss theorem applies just as well to multiply connected regions.4 Because the
quantity oxo1 f1 oxo2 f2 f3 oxo3 under the volume integral (surface integral, in two
dimensions) is the divergence of the function f, the theorem is also known as the
divergence theorem of Gauss, or simply, divergence theorem.
A.12 The Unit Impulse Function

The unit impulse function, also known as the delta function and the Dirac delta
function, was developed by the French mathematician, Jean-Baptiste-Joseph
Fourier, but was first put to practical use by the engineer and theoretical physicist
Paul Dirac, in the 1920s. There are several ways to define the unit impulse function,
but it is always represented by the Greek letter d and a real argument, such as x and t.
The delta function is usually defined as a special function whose value is
infinite when its argument is zero and is identically zero everywhere else on the
real axis; but such that its integral over the entire real axis equals 1. That is,
M.R. Spiegel, Theory and Problems of Advanced Calculus, Schaums Outline Series,
McGraw-Hill (1963), pp. 202205.
298

d t
Z1
dtdt
1
Ze
0; t 6 0
1; t 0
dtdt 1;
A:23
e [ 0:
A:24
e
The delta function may also be defined as a symbolic or generalized function, in

terms of its integral properties alone, as they affect a so-called testing function. A
testing function is an arbitrary continuous function that vanishes identically
outside some finite interval. For any such testing function, u(t), the d-function is
defined by the relation:
Z1
dtutdt utjt0 u0
A:25
1
Unlike the definition in (A.23) and (A.24), no value is assigned to the

d-function as a symbolic function. The integral itself has no meaning as an
ordinary integral either. The integral and the d-function are defined by the value
u(0) assigned to the testing function (H.P. Hsu 1967). The following are some
practical properties of the d-function:
Z1
dt to utdt ut to jt0 uto
1
t
1
u ju0 u0
jcj jcj
jcj
where f is continuous at t 0
1
Z1
dc tutdt
1
f tdt f 0dt;
Proofs Properties (a) and (b) are proven through simple changes of variables. To
prove (a) introduce the change of variable u t - to and use (A.25) to obtain the
value of the testing function. The proof of (b) requires the change of variable u
ct, once for c [ 0, and then again for c \ 0. Use of (A.25), and the definition of
absolute value, completes the proof. To prove (c), select a testing function u(t),
and since f (t) is continuous, write:
Z1
f tdtutdt
1
f 0
Z1
dtf tutdt
1
Z1
1
f tutjt0 f 0u0
Z1
dtutdt
f 0dtutdt
1
299
Since the testing function u(t) is arbitrary, the integrals on the left- and right-hand
sides are valid for all choices of u(t), which implies that the quantities under the
integral signs be identical, proving (c).
A.13 The Unit Step Function

The unit step function, also known as the Heaviside step function, is typically
denoted by H. It is defined to be identically equal to zero for negative values of its
argument and identically equal to one when its argument is non-negative, but is
undefined at the origin:

0; t \ 0
Ht
A:26
1; t [ 0
The graphical representation of the unit step function is shown in Fig. A.5.
From the figure, note that the derivative of the unit step function is everywhere
zero, except at the origin.
As it turns out, the unit step function is a symbolic function which may be
alternatively defined in terms of a testing function u(t) by means of the relation:
Z1
H tutdt
1
Z1
utdt
A:27
The unit step function and the d-function are related through the generalized
derivative, which is defined for any differentiable function f (t) and testing function
u(t), by means of the relation:
Z1
f tutdt
1
Z1
f tu0tdt
A:28
1
R1
R1
Indeed, setting f(t) H(t), this yields 1 H 0 tutdt 1 Htu0 tdt;
R1 0
R1
which, according to (A.27), becomes: 1 H tutdt 0 u0 tdt
u1 u0 u0
Fig. A.5 Geometric representation of the unit step

function
H(t)
1.0
t
300
R1
Since: u0 1 dtutdt; in accordance with the definition of the
d-function given in (A. 25), it follows that
Z1
Z1
H tutdt
1
dtutdt
1
Consequently, by the arbitrariness of the testing function, u(t):

H 0 t
d
H t dt
dt
A:29
Note, however, that the derivative of the unit step function is zero both for
negative and positive values of its argument.
A.14 Stieltjes Convolution

The Stieltjes convolution of two functions: u(t), which is continuous in [0,?), and
w(t), which vanishes at -?, is defined as
Z t
d
A:30a
/t s dws
/t s wsds / dw
ds
1
where the form on the far right is used when the argument is understood. Under the
additional assumption that u(t) 0 for t \ 0, one may split the interval of
integration from (-?, t) into (-?, 0-)U(0-, t) and write an alternate form for the
convolution:
Z t
d
A:30b
/t s dws
/t s wsds
ds
0
If, in addition w(t) 0 for t \ 0, then
/t s dws
Zt
d
/t s wsds /tw0
ds
0
Zt
/t s
d
wsds
ds
A:31
-
which is obtained by expressing the integration interval (0 , t) as (0 , 0 )U(0+, t),

and by integrating by parts the integral defined over (0+, t). In this case, as will be
shown, the convolution integral is commutative, associative and distributive. Thus,
u dw w d/
Commutivity
u dw dh / dw dh / dw dh
u dw h / dw / dh
Associativity
Distributivity
301
Proof of Commutivity Starting from the defining expression, integrating it by

parts, and shifting the argument of the derivative operator, s; to t s; following
this by a change of variable of integration from t s to s; and finally, changing
the dummy variable s to s again:
Zt
d
t
/ dw /tw0 /t sws0
ws /t sds
ds
0
wt/0
Z0

ws

d
/t s dt s
dt s
wt/0
Zt

w t s

d
/s ds
ds
w d/
Proof of Associativity Setting f w dh and g u dw; the associative law
may be written as
u dw dh / df / dw dh g dh
Now:
/ df
0
d
/t s f sds
ds
0
d
/t s
ds
ws s
0
d
hsdsds
ds
c
Differentiating the nested integral and effecting the change s s u leads to

u df
Z1
/t s u
0
Z1
Z1
0
0
Z1
d
d
wu hsdsdu
du
ds
0
d
d
/t s u wu hsdsdu
du
ds
In similar fashion:
g dh
Z1
0
d
gt s hsds
ds
Z1 Z1
0
/t s u
d
d
wu hsdsdu
du
ds
0
Hence, u dw dh u dw dh u dw dh; as was to be proven.

Proof of Distributivity The distributive property of the convolution integral
emanates directly from the linearity of integration. Thus,
302
u dw h
Zt
/t s
d
ws hsds
ds
Zt
d
/t s wsds
ds
Zt
/t s
d
sds
ds
u dw u dh
A.15 The Laplace Transform

The Laplace transform of a function f(t), denoted by L{f(t)} or f s; is defined as
Lff tg f s
Z1
est f tdt; s a jb
A:32
for some c 2 R; c [ 0
A:33
Provided
Z1
jf tject dt\1;
In expression (A.32), s is a complex number, and for the Laplace transform of a

piecewise continuous function f(t) to exist, the function itself has to be of
exponential order (that is, bounded by an exponential function of a real variable).
According to its definition, the Laplace transform converts a function from one
space to another; in the present case, from the t-domain to the s-domain. The
conversion from the transformed space back to the original space may be carried out
by means of the inverse transformation, which is definedfrom Fourier transform
theoryas
1
L ff tg f t
2pi
1
ki1
Z
estf sds
A:34
ki1
Without going into any details, the evaluation of the line integral in (A.34) is
carried out using analytic function theory. In practice, however, the definition given
in (A.32) is used to construct a table of transforms and the table is then used in
reverse to obtain the inverse transforms. This is possible because the Laplace
transform of a function and the corresponding inverse transform are uniqueexcept
possibly for a so-called zero function. Either way, the important thing to bear in
mind is that one can go from one space to the other and back, directly or indirectly.
303
As will be evident subsequently, the Laplace transform converts differential and

integral operators into algebraic ones. The practical implication of this being that
by applying the transform one can convert differential or integral equations into
algebraic ones, which are easier to solve. Applying the inverse transform to the
transformed solution will then lead to the solution in the original, physical space.
A.15.1
Properties of the Laplace Transform
A list of some properties of the Laplace transform, which are pertinent to the study
of viscoelasticity, is provided next. For completeness of presentation, the proofs
are given in each case.
Linearity
Lfa1 f1 t a2 f2 tg a1 Lff1 tg a2 Lff2 tg a1f1 s a2f2 s
A:35
The proof follows from (A.32) and the linearity of the integration operator:
Lfa1 f1 t a2 f2 tg
Zt
0
a1
est a1 f1 t a2 f2 tdt
Zt
e
st
f1 tdt a2
Zt
est f2 tdt
Transform of Derivatives
Lff n tg snf s sn1 f 0 sn2 f 0 0 . . . f n1 0
A:36
The proof is carried out by induction, starting with the transform of the firstorder derivative, using integration by parts:
0
Lff tg
Z1
e
0
st 0
f tdt f 0 s
Zt
est f tdt sf s f 0
Having proven the formula valid for k 1, we presume it valid for k n-1,
that is
Lff n1 tg sn1f s sn2 f 0 sn3 f 0 0 sn3 f 00 s . . . f n2 0 c
And proceed to prove that the formula is valid for k n. To this end, take the
Laplace transform of f (n), introducing the notation g f (n-1), and using the
transform of its first derivative: Lff n tg Lfdtd f n1 tg Lfg0 tg sgs g0:
The proof is completed inserting the transform of f (n-1), and g(0) f (n-1)(0).
304
Initial Value Theorem

lim f t lim sf s
A:37
s!1
t!0
This theorem is proven by taking the limit as s ? ? of the Laplace transform

of the first derivative of f, using the integral form of it and its value, sf s f 0:
Z1
lim
s!1
est f 0 tdt lim sf s f 0
s!1
From this expression follows that 0 lim sf s f 0; which completes the
s!1
proof, because the continuity of f assures that f 0 lim f t:

t!0
Final Value theorem

lim f t lim sf s
t!1
A:38
s!0
This statement is proven by taking the limit as s ? 0 of the transform of the first
derivative, in manner similar to the initial value theorem.
lim Lff tg limsf s f 0 lim
0
s!0
s!0
Z1
st 0
s!0
f tdt
Z1
f 0 tdt f 1 f 0
e
Implying the theorem, since, from the continuity of f: f 1 lim f t:
t!1
Transform of Integrals
9
8 t
= 1
<Z
L
f tdt f s
; s
:
A:39
Rt
To prove this, we let gt 0 f tdt: In this manner, g0 (t) f(t), g(0) 0, and
R t

Lff tg Lfg0 tg sgs g0 sL 0 f tdt :; as asserted by (A.39).
Likewise, mathematical induction would prove that for integrals iterated n times:
8 t t
9
Zt
<Z Z
= 1
L
f udu n f s
A:40
:
; s
0
Substitution
f s a Lfeat f tg
A:41
This property is proven from the definition of the transform, by replacing the
transform variable, s, with the shifted variable, s - a; thus
f s a
Z1
e
sat
305
f tdt
Zt
est feat f tgdt Lfeat f tg
Translation
A:42
LfHt af t ag easf s; a [ 0
R 1 st
R 1 st
Indeed, LfHt af t a 0 e Ht af t adt a e f t adt:
The proof is concluded introducing the change of variable: s t-a, and carrying
out the integration:
LfHt af t a
Z1
e
ss
sa
sa
f sds e
Z1
ess f sds esaf s
Convolution
L
8 t
<Z
:
f t sgsds
9
=
;
Lff tg Lfgtg f sgs
A:43
This property, that the Laplace transform of the convolution of two functions is
equal to the product of the transforms of the functions is also referred to as Borels
theorem. The proof uses the translation property of the Laplace transform to write
f s
gs f s
Z1
e
ss
80
Z1 < Z1
s0
ts
gsds
Z1
fess f sggsds
09
=
est f t sdt gsds
;
Interchanging the order of integration completes the proof:

f s
gs
Z1
e
t0
A.15.2
st
Zt
f t sgsdsgdt Lf
s0
Zt
f t sgsdsg
Brief List of Laplace Transforms
Dirac delta function:

Lfdtg 1
Lfdt ag e
j
as
306
Unit step function:

LfHtg
1
s
LfHt ag
eas
s
l
m
Exponential function:
Lfeat g
1
sa
Lftn g
n!
sn1
Power function:
Sine function:
Lfsinatg
s2
a
a2
s2
s
a2
Cosine function:
Lfcosatg
A.15.3
Partial Fraction Expansion
As indicated before, a simpleand often effectivemeans to find the inverse

transform of a function is to use a table of transformations in reverse. When the
function that is to be transformed back to the physical plane appears as a quotient
of two polynomials for which the inverse transform is not already available, it is
usually advantageous to cast the given quotient of polynomials in term of its
partial fractions.
Every rational function q(x) can be expressed as the quotient of two
polynomials in x. The division algorithm allows expressing q(x) in the form:
q x u x
gx
hx
A:44
where u(x), g(x) and h(x) are polynomials in x, and the degree of the numerator,
g(x), is lower than that of the denominator, h(x). If h(x) can be expressed as the
product of its irreducible factors, then it is possible to write q(x) as the sum of
u(x) and a given number of so-called partial fractions.
With the above in mind, let the known transform of a function f (t) be given as
1
2
n

f s Ps p0 p1 s p2 s pn s
1
2

Qs q0 q1 s q2 s qm sm
A:45
307
To expand it into its partial fractions requires the m zeroes, ri, of the
denominator, so that
f s A1 A2 Am
s r1 s r2
s rm
A:46

Which applies when all the roots of Qs
0 are distinct. A different expansion
applies in the case of repeated roots. For instance, if r1 occurs k times, the
appropriate expansion becomes:
f s
A11
s r1
A12
s r1

A1k
s r1
A2
A3
Amk

s r2 s r3
s rmk
A:47
In any case, the coefficients Ai correspond to the limiting process:

Ps
Ai lim s ri
s!ri Qs
A:48
In practice, the partial fraction expansion of the ratio of two given polynomials
is developed by first writing down the expansion based on the roots of the
denominator, as in (A.46) or (A.47), multiplying by the polynomial in the
denominator, equating coefficients of like powers of the transformed variable, and
solving the resulting linear system in the coefficients. Also, when it is
advantageous to leave the denominator of a partial fraction as a second-degree
polynomial, its numerator must be a linear polynomial, in accordance with the
requirements of partial fraction expansion indicated before. Thus, in analogy with
px
(A.47), the partial fraction expansion of the rational polynomial f x ax2 bxc
n
would be given by the following decomposition:
px
A1 x B1
A 2 x B2
A n x Bn

ax2 bx cn ax2 bx c ax2 bx c2
ax2 bx cn
A:49
Example A.7 As an illustration, we obtain the inverse transform of the polynomial

fraction:
f s 4s 7
s 12
Solution
In this case, the denominator has two equal zeroes: r -1, -1. Therefore, we
A1
A2
4s7
seek the partial fraction expansion in the form: s1
: Multiplying
2 s1
s12
by the minimum common multiple of the expansion on the right-hand side and
308
canceling the denominators on both sides of the resulting form: 4s 7

A1 s 1 A2 : This yields a linear system in A1 and A2, whose solution is A1 4,
4s7
4
3
: The inverse Laplace transform of this
and A2 3; and thus, s1
2 s1
s12
expression can be easily obtained using the substitution property f(s-a) L{eatf(t)}
and the table of transforms in reverse.
4
s1
1
4 s1
4f s 1; f s 1s
n o

4
) L1 s1
4fet L1 1s g 4et Ht
3
s12
Therefore: L1
1
3 s1
f s s12
2 3f s 1;
n
o

3
) L1 s1
3fet L1 s12 g 3et t
2
4s7
s12
L1
4
s1
L1
3
s12
Example A.8 Find the partial fraction expansion of
4 3tet .
x5 1
x2 42
Solution
In this case, two facts need to be noted. Firstly, the given fraction is improper,
as the degree of the numerator is larger than the degree of its denominator.
Secondly, the denominator of the given expression is a quadratic polynomial and is
repeated twice. Hence, according to (A. 44), we need to expand the denominator as
x4 8x2 16; and use it to divide the numerator x5 1 by it. The division
5
3
1
x xx16x1
. We use (A.49) and seek
algorithm then leads to the result: xx2 4
2 42
2
3
CxD
x AxB
the partial fraction expansion of the second term, as xx16x1
2 42
x2 4 x2 42 :
Writing the partial fraction expansion in terms of the common denominator

2
x2 4 and collecting terms leads to x3 16x 1 Ax3 Bx2
4A C x 4B D: Equating coefficients of like powers of x produces: A
-8, B 0, C 16, D 1. These values and the result of the division yield the
5
1
x x28x4 16x1
:
partial fraction expansion: xx2 4
2
x2 42
A.15.4
Laplace Transform of Differential Equations
Any initial value problem which involves an ordinary linear differential equation
of order n with constant coefficients may be converted into one in transformed
space. Indeed, let the general differential equation in question be expressed as

n
X
ak
k0
0
309
dk
yt gt
dtk
d
yt 1
dt0

dk
yt y0;
dtk
A:50
k 1; . . .; n 1
t0
Applying the Laplace transform, this equation takes the form:

(
)
n
k
X
X
k
kr r1
ak s ys
s y
0 gs
A:51
r1
k0
Pulling ys outside the summation sign and rearranging:

gs
ys
n
P
k1
ak
k
P
skr yrk 0
r1
n
P
A:52
ak s k
k0
The solution function, y(t), may be obtained by Laplace transform inversion.
A.15.5
Laplace Transform of Integral Equations
An integral equation is one in which the unknown function appears inside an

integral sign and, possibly, outside of it too. The general form of an integral
equation in v(t) is
Zsb
at
Kt; svsds btvt ct
A:53
sa
where a, b, and c are arbitrary functions of the independent variable, t; a = 0; and

the lower limit of integration, a, is a constant. If both limits of integration are
constant, the integral equation is called a Helmholtz integral equation; and when the
upper limit is the independent variable, t, the equation is a Volterra integral
equation. Equations for which the function c (t) 0 are called homogeneous; while
equations in which the unknown appears only inside the integral (i.e., b (t) 0) are
called integral equations of the first kind; and integral equations of the second kind,
otherwise.
310
In what follows, we concentrate on non-homogeneous Volterra integral

equations of the second kind. There are two reasons for this: first, this type of
equation always admits a solution (at least in principle); secondly, they occur
naturally in viscoelasticity.
Without loss of generality, a Volterra integral equation of the second kind may
be cast as
Zt
ct vt Kt; svsds
A:54
a
Where we have divided (A.53) throughout by b(t), absorbed -a(t)/b(t) into K,

and retained the notation c(t) for c(t)/b(t). Of particular interest is the case when
a 0, and the kernel K(t,s) of the integral is a so-called difference kernel, when
K(t,s) may be expressed as K(t,s) K(t - s), such as occur for non-aging
viscoelastic materials. The integral in (A.54) is then a convolution integral, so that,
applying the Laplace transform, collecting terms, and re-arranging:
vs
cs

1 Ks
A:55
The solution to the original integral equation follows from the inverse transform
of the expression in (A.55), as

cs
vt L1
A:56

1 Ks
Example A.9 The relaxation modulus, M, and the creep compliance, C, of linear
Rt
d
Csds: Assuming
viscoelastic materials are related by MtC0 0 Mt s ds
as happens to be the casethat C(0) 1/M(0), use the Laplace transform to solve this
equation for the function C(t), if M(t) M1e-t/s.
Solution
Direct application of the Laplace transformation, noting that the integral is the
convolution of M and the time derivative of C, and using the pertinent properties of the

transform yields MsC0
MssCs
C0 1s . Collecting terms and

s1=s

rearranging: Cs
2 1 1=M1 2 1 1=M1 1 1=s1 12 . Using the list
s Ms
of Laplace transforms in reverse: Ct M11 Ht t=s1 . As a matter of curiosity, the

material in question represents a fluid, among other things, because C(?) ? ?. Also,
C(0) 1/M1 is called the glassy compliance, and M1s1 is the materials viscosity.
311
A.15.6 Relationship Between the Laplace and Fourier

Transforms
In general, the Laplace and Fourier transforms of an acceptable function5 differ,

but their definitions show considerable similarity:
Lff tg
Z1
f test dt; s a jb
Z1
A:57
jf tject dt\1; c 2 R; c [ 0
Fff tg
Z1
1
Z1
f tejxt dt
A:58
jf tjdt\1
1
The question then arises if the two transforms can be made equivalent; and if
so, under what conditions? The answer is that the two integral transforms will be
equivalent for all functions which are absolutely integrable and vanish identically
for all negative values of their argument, so that
Z1
1
jf tjdt
Z1
jf tjdt\1
Since most functions in engineering, physics and viscoelasticity in particular,

are causal, they vanish for all negative values of their arguments. The Fourier
transform of any causal function, is given as
In both instances, an acceptable function is one that is piecewise continuous and has a finite
number of finite discontinuities on its interval of integration. As can be seen from their
definitions, a function has to be of exponential order (bounded by an exponential function) on the
positive real axis for its Laplace transform to exist, and absolutely integrable on the entire real
axis for it to have a Fourier transform.
312
Fff tg
Z1
f tejxt dt
1
Z0
1
Z1
f te
jxt
dt
f tejxt dt
Z1
0
Z1
0
f tejxt dt

st
f te dt

A:59
sjx
Hence,
Fff tg Lff tgjsjx ;
for :
f t 0;
8t\0;
Z1
and :
jf tjdt\1
A:60
This shows that the Laplace and Fourier transforms are equivalent for all
absolutely integrable causal functions.
Example A.10 Using (60), find the Fourier transform of the exponential function
defined by
at
e ; t [ 0; a [ 0
f t
0; t\0
Solution
To apply (A.60), we must first make sure the given function satisfies the
required conditions. By definition, the function f satisfies the causality condition
that f(t) : R0, t \ 0. The function
f is also absolutely integrable on the positive real
R1
1
axis, since 0 jeat jdt 0 eat dt 1a \1. Hence, the Fourier transform of f can
be obtained from its Laplace transform. From the list of Laplace
transforms:

at
1
at
at
1
1
Lfe g sa. This and (A.60) yield F fe g Lfe gjsjx sa sjx jxa
Example A.11 Demonstrate that the Fourier transform of the unit step function
cannot be obtained from its Laplace transform.
Solution
The unit step function satisfies
R 1 the causality
R 1 condition, but the integral of its
absolute value is infinite: 0 jutjdt 0 dt 1. Since this violates the
Dirichlet conditions, the Fourier transform of the unit step function cannot be
derived from its Laplace transform.
313
A.16 Orthogonal Functions

A set of functions {fk (t)} is said to be orthogonal on an open interval a \ t \ b if
for any two function fi (t) and fj (t)}in the set, the following relation holds
Zb

fi tfj tdt
0 for i 6 j
mi for i j
A:61
The set of sinusoidal functions f1; cosxt; cos2xt; . . .; cosnxt; . . .;

sinxt; sin2xt; . . .; sinnxt; . . .g is an orthogonal set of functions on an open
interval of total amplitude p 2p/x. As may be shown by elementary calculus, the
functions in this set are such that
Ztp
sinkxtdt 0 for all k
A:62a
Ztp
coskxtdt 0
for k 6 0
A:62b
Ztp
sinkxtcoslxtdt 0
for all k and l
A:62c
Ztp
(
sinkxtsinlxtdt
for k 6 0
p=2
Ztp
t
(
coskxtcoslxtdt
0
p=2
for k l 6 0
for k 6 0
for k l 6 0
A:62d
A:62e
The proofs of (A.62-a) and (A.62-b) are straightforward, as their integrals are
proportional to a cos(kxt) and sin(kxt), respectively; and in addition, these
functions are periodic, meaning that coskxt p coskxt; and
sinkxt p sinkxt. Hence,
Ztp
1
sinkxtdt fcoskxt p coskxtg 0 for all k
k
1
coskxtdt fsinkxt p sinkxtg 0 for all k 6 0
k
Ztp
t
314
To prove (A.62-c), we use the identity sinA cosB 1=2sinA + B)

sinA B); so that
Ztp
1
sinkcoslxtdt
2
Ztp
sink lxt sink lxtdt
1 1
1 1
cosk lxttp
cosk lxttp
t
t
2 k lx
2 k lx
0 if k 6 l
c:1
In the case k l, we use the trigonometric identity 2sin(A)cos(A) sin(2A), to
write
Ztp
sinkxtcoslxtdt
Ztp
sinkxtcoskxtdt
Ztp
sin2kxtdt
c:2
1
cos2kxttp
t
4kx
0
The
proof
of
(A.62-d)
uses
1=2cosA B cosA + B. Thus,
Ztp
t
1
sinkxtsinlxtdt
2
Ztp
the
identity:
sinA sinB
cosk lxt cosk lxtdt
1 1
1 1
sink lxttp
sink lxttp
t
t
2 k lx
2 k lx
0 if k 6 l
d:1
In the case k l, we use the trigonometric identity sin2 A 1=2 1=2cos2A;
to write
Ztp
315
sinkxtsinlxtdt
Ztp
sin2 kxtdt
Ztp
1 cos2kxtdt
d:2
1
1
sin2kxttp
tttp
t
2
4kx
p
:
2
The proof of (A.62-e) uses the identity: cosAcosB) = 1=2cosA B)

cosA B. Thus,
Ztp
coskxtcoslxtdt
1
2
Ztp
cosk lxt cosk lxtdt
1 1
sink lxttp
t
2 k lx
1 1
sink lxtttp
2 k lx
e:1
0 if k 6 l
In the case k l, we use the trigonometric identity cos2(A) + cos(2A), to

write
Ztp
coskxtcoslxtdt
Ztp
cos2 kxtdt
1
2
Ztp
1 cos2kxtdt
e:2
1 tp
1
t t
sin2kxtttp
2
4kx
p
A.17 Complex Numbers

A complex number, z, is an ordered pair of real numbers, x and y, which, for this
reason, is usually written in the form z (x, y); just like a point or vector in two
dimensions.
The collection of all ordered pairs of real numbers forms the set, C; of complex
numbers. The first component of the ordered pair (x, y) is termed the real part of
the complex number and is denoted by Re{z}. For reasons which will become
316
logical later on, the second number in the pair is referred to as the imaginary part
or component, Im{z}, of the complex number. Using this notation, then
z x; y;
x Refzg;
y Imfzg
A:63
The complex conjugate, z; of a complex number z, is defined by the relation:

z x; y
A.17.1
A:64
Elementary Operations with Complex Numbers
The arithmetic operations of addition or subtraction multiplication, and division

of complex numbers are defined in terms of the corresponding operations for real
numbers, as described next.
Addition
The addition of complex numbers is defined component-wise, by the simple rule:
z1 z2 x1 ; y1 x2 ; y2 x1 x2 ; y1 y2
A:65
From this definition follows that the complex number (0,0) : 0 is the identity
for complex addition; that is
z 0; 0 x; y 0; 0 x 0; y 0
0 x; 0 y
0; 0 x; y
A:66
0; 0 z
Another consequence of the definition of addition of complex numbers is that
any complex number may be expressed uniquely as the sum two complex
numbers; one with zero imaginary part, and the other with zero real part:
z x; y x; 0 0; y
A:67
Using this decomposition, z and z may be combined to express the real and the
imaginary parts of a complex number, as
x Refzg
zz
;
2
y Imfzg
zz
2
A:68
The operation of addition or subtraction enjoys the same properties as the

addition of real numbers. That is, the addition of complex numbers is associative,
distributive, and commutative. This statement is proven using the definition of
addition of complex numbers and the properties of addition of real numbers.
Example A.12 Prove that the addition of complex numbers is commutative.
317
Solution
To prove that the addition of complex numbers is commutative is equivalent to
showing that the order of the addends in the sum of tow complex numbers can be
interchanged without changing the result. To do this, examine the sum z1+z2 of two
complex numbers:
z1 z2 x1 ; y1 x2 ; y2 x1 x2 ; y1 y2
x2 x1 ; y2 y1 x2 ; y2 x1 ; y1
z2 z1
Example A.13 Prove that the addition of complex numbers is associative.
Solution
To prove associativity, consider the sum (z1+z2) + z3:
z1 z2 z3 x1 x2 ; y1 y2 x3 ; y3
x1 x2 x3 ; y1 y2 y3
x1 ; y1 x2 x3 ; y2 y3 z1 z2 z3
For simplicity, the following representation of a complex number is used:
z x; y x; 0 0; y x1; 0 y0; 1 x j y
A:69
Multiplication
The product of two complex numbers z1 and z2 is defined by the following rule:
z1 z2 x1 ; y1 x2 ; y2 x1 x2 y1 y2 ; x1 y2 y1 x2
A:70
From this follows that the complex number (1,0) : 1 is the identity element
for complex multiplication. Indeed,
z 1; 0 x; y 1; 0 x 1 y 0; y 1 x 0 x; y z
A:71
1; 0 z 1; 0 x; y 1 x 0 y; 1 y 0 x x; y z
A:72
and
In this notation, the letter j is used to represent the unit, (0,1), of the
imaginary numbers,6 which, in accordance with the definition of the product of
complex numbers, is such that
j2 0; 1 0; 1 1; 0 1
A:73
In other words,
j
p
1
A:74
More commonly, the letter i is used to denote the complex unit, but we reserve this letter for
use as a subscript.
318
Multiplication of two complex numbers enjoys the same properties as that of real
numbers. For instance, that multiplication of complex numbers is commutative and
associative results from the corresponding rules for real numbers.
z1 z2 x1 ; y1 x2 ; y2 x1 x2 y1 y2 ; x1 y2 y1 x2
x2 x1 y2 y1 ; x2 y1 y2 x1
z2 z1
z1 z2 z3 x1 x2 y1 y2 ; x1 y2 y1 x2 x3 ; y3
x1 x2 x3 y1 y2 x3 x1 y2 y3 y1 x2 y3 ; x1 x2 y3 y1 y2 y3 x1 y2 x3 y1 x2 x3
x1 x2 x3 y2 y3 y1 y2 x3 x2 y3 ; x1 y2 x3 x2 y3 y1 x2 x3 y2 y3
x1 ; y1 x2 x3 y2 y3 ; y2 x3 x2 y3 x1 ; y1 x2 ; y2 x3 ; y3
z1 z2 z3
b
Additionally, from the definition of multiplication of complex numbers, one has
x x 1 x 1; 0 x; 0 1; 0 x; 0
y 0; 1 y 1; 0 0; 1 y; 0 1; 0 0; 1 y; 0 0; 1 0; y
d
Combining these results with the definition of addition given earlier, there
results
z x; y x; 0 0; y x1; 0 y0; 1
This expression shows that a complex number may be interpreted as the sum of
two points or vectors in a two-dimensional Euclidean space. Indeed, the Argand
plane is a two-dimensional plane in which the real part, x, of a complex number, z,
is measured along the horizontal axis or real axis, and its imaginary part, y, is
measured along the vertical or imaginary axis, as shown in Fig. A.6.
From its graphical representation, it can be seen that a complex number can be
represented in so-called polar form, as
p
z x; y r cosh j r sinh; r x2 y2
A:75
p
z x; y r cosh j r sinh; r x2 y2
A:76
In these expressions, r, as the positive square root of nonnegative real numbers,
x2 and y2, is a real number. The number r is called the modulus, or magnitude of
the complex number z and is usually denoted as jzj. From the previous expressions,
it is apparent that the modulus of any complex number is the same as the modulus
of its conjugate; that is
p
r jzj x2 y2 jzj; 8z x; y
A:77
319
Fig. A.6 Graphical representation of a complex number, z; and its conjugate, z; in
the complex, Argand plane
y
x
x
y
Division
The rule for the division z1/z2 of two complex numbers z1 and z2 may be readily
developed from the concept of the inverse, z-1, of a non-zero complex number, z:
z z1 1; 0 1
A:78
At the same time, using the rule for multiplication of complex numbers and the
definition of complex conjugate, it follows that
z z x; y x; y

x2 y2 ; xy xy

x2 y2 ; 0
A:79
r 2 ; 0 jzj2
In other words,
zz
j zj 2
1 1; 0
A:80
Combining this with the definition of the complex multiplication inverse yields
z z1 1; 0
zz
j zj 2
A:81
Canceling out the common term z, and using that jzj2 z z, leads to
z1
z
jzj
1
z
A:82
Not only does this relation prove that the inverse of a non-zero complex number
is, as for real numbers, equal to its inverse, but also provides the rule: the inverse
of a complex number is equal to its complex conjugate divided by it squared
modulus.
Example A.14 Find the inverse of the complex number z 3-j.
320
Solution
Apply the rule to obtain the inverse of a complex number: z1 jzzj2 to get
1
323j
z1 3j
12 0:3 0:1j
The same process yields the rule that to obtain the quotient z1/z2 of two
complex numbers, multiply the numerator and denominator of the fraction by the
conjugate of the denominator. This is so, because
z1
z2
z1 z2
z1 z1
2 z1
2
z2
z2 z2
j z2 j
Example A.15 Find the quotient: (2-3j)/(1+2j)
Solution
Apply the rule for division to get
A.17.2
23j
12j
12j
23j
0:8 1:4j
12 22
Eulers Formula
A useful representation of a complex number is by means of the so-called Euler

formula. This formula expresses the exponential function of a complex argument
in terms of the sine and cosine of the argument. Eulers formula may be easily
derived by developing the function eih in Taylor series, as follows:
jh jh2 jh3 jh4

1!
2!
3!
4!

2
4
3
h
h
h h

1 j
1! 3!
2! 4!
ejh 1
A:83
cosh j sinh
Combining this, with the polar representation z rcosh j sinh; gives
z x; y rcosh j rsinh rejh
A:84
In using the polar or the exponential forms of a complex number, it must be

borne in mind that the cosine and sine functions are periodic functions with period
2p. Hence,
z rejh rejh2pk ;
A.17.3
k 0; 1; 2; . . .
A:85
Powers, Roots, and Logarithms of Complex Numbers
Powers, logarithms, and integral roots of complex numbers are easily calculated
using their exponential representation and the corresponding rules for real
321
numbers. Letting z rejh be an arbitrary complex number, the following

expressions follow directly from the properties of real numbers:
Integral Powers
The nth power of a complex number is easily established from the polar form, and
the rules of exponentiation of real numbers, as
n
n
A:86
zn rejh rejh2pk r n ejnh2pk ; k 0; 1; 2; . . .; n 1
Radicals
The root of integral index n of a complex number is easily obtained from its polar
form, and the rules of exponentiation of real numbers, as

1=n
1=n
jh2pk
z1=n rejh2pk
rejh2pk
r 1=n e n ; k 0; 1; 2; . . .; n 1
A:87
Logarithms
The natural logarithm, loge(z) : ln(z), of a complex number, z, can be established
from the polar form of the complex number and the properties of logarithms of real
numbers, as

loge z ln rejh2pk lnr jh 2pk; k 0; 1; 2; . . .; n
A:88
The logarithm, logb(z), of a complex number, z, in an arbitrary real base, b, may
also be established from the properties of logarithms of real numbers, using that
for any two bases, b1 and b2: logb1 f logb1 b2 logb2 f ; as

logb z logb rejh logb e ln rejh2pk
A:89
logb eln r jh 2pk; k 0; 1; 2; . . .; n
Trigonometric Functions
The circular functions sin(z) and cos(z) of a complex argument result from Eulers
formula with complex argument, ejz cosz j sinz. The sine function is
obtained by subtracting e-jz from e jz, and the cosine function results from adding
them. Thus,
sinz

1 jz
e ejz ;
2i
cosz

1 jz
e ejz
2
A:90
These relations may be recast in terms of the real and imaginary parts, x and y,
of the complex number z x + jy. Taking the sine function, for instance, leads to
322

1 jxy
e
ejxy
2j
1
ey cosx j sinx ey cosx jsinx
2j

1
ey ey
ey ey
j sinx
cosx
j
2
2
sinz sinx jy
sinx coshy j cosx sinhy

So that
sinx jy sinx coshy j cosx sinhy
A:91
In this expression, sinhu 12 eu eu and coshu 12 eu eu ; are the

usual hyperbolic functions of real analysis. In an entirely similar fashion, the
cosine function leads to
cosx jy cosx coshy j sinx sinhy
A:92
Example A.16 Calculate cos (1+2j).

Solution:
Use: cosx jy cosx coshy j sinx sinhy; with x 1 and y 2 rad, to
obtain:
cos1 2j cos1 cosh2 j sin1 sinh2
0:54033:7622 j0:84153:6269
2:0327 3:0520j
The inverse trigonometric functions sin-1z and cos-1z are also readily defined.
For instance, if w sin1 z; then taking the sine of this expression leads to
z sinw

1 jw
e ejw
2j
Multiplying throughout by 2j, to remove the denominator, and multiplying the

result by ejw ; and collecting terms, there results e2jw 2jzejw 1 0; which is a
p
quadratic equation in ejw ; with roots ejw jz 1 z2 . The natural logarithm of
this relation, using that w sin1 z; yields

p
sin1 z j ln jz 1 z2
A:93
Note that the function sin-1z is multi-valued because the natural logarithm is
multi-valued. In an entirely similar way,

p
cos1 z j ln z z2 1
A:94
j
jz
j
1 jz
tan1 z ln
ln
2
jz
2
1 jz
323
A:95
The form in the far right of (A.95) results from the property that j2 -1, and
the fact that
j z j x jy j1 jx jy j1 jz
j z j x jy j1 jx jy j1 jz
and similarly,
Appendix B
Elements of Solid Mechanics
B.1 Scalars, Vectors, and Tensors

A tensor is an ordered system of objects, which are typically numbers or functions.
As such, tensors may be represented by matrices. In standard three-dimensional
space, the number of components of a tensor of order n is simply 3n. Accordingly,
a tensor of order zero has only one component (30 1), arranged in a 1-by-1
matrix. In Euclidean 3-space, scalars or tensors of order zero are quantities, such as
temperature, density, volume, which are fully described by one value. The matrix
representation of a tensor of order one, also called a first-order tensor, has 3
components (31 3). These tensors are the well-known vectors of Euclidean
geometry: 3-by-1 or 1-by-3 matrices. Continuing thus, the matrix representation of
a second-order tensor has 32 9 entries, corresponding to a 3-by-3 matrix. Tensors
of fourth order have 34 81 components.
It is common practice to use the terms scalar and vector, respectively, to refer to
tensors of order zero and one; and to reserve the term tensor exclusively for
tensors of order 2 and higher. Also, symbolically, a vector is represented placing a
bar or an arrow over the stem letter, or by the use of bold type: a;~
a; a:
The components of a tensor may be designated when the tensor is referred to a
coordinate system. In Cartesian coordinates, for instance, the components of a
vector, A, may be expressed as Ax, Ay, Az; or, as A1, A2, A3, if the axes are
numbered. The vector itself may be represented in several forms:
(a) As a point in space, A A1 ; A2 ; A3 :
(b) In explicit form, displaying the unit vectors, along the coordinate axes, using
P
Gibbs notation:7A A1ê1 A2ê2 A3ê3 3i1 Aiêi :
(c) In component indicial form, combining the summation conventionthat a
term is developed by summing over the range of its repeated indices:
A Aiêi ; i 1; 2; 3:
p
The magnitude of the vector A is the real number: jAj A21 A22 A23 :
7
We typically represent unit vector with a hat, êi ; to emphasize their magnitude is 1. However,
they are also represented in bold face type, such as, ei :

325
326
Appendix B: Elements of Solid Mechanics
Vector addition is defined component-wise. Therefore: A + B C is such that

C1 A1 + B1, C2 A2 + B2, and C3 A3 + B3. Therefore, vector addition is
commutative.
There are three types of multiplication of vectors: scalar or dot product, vector
or cross product, and tensor product.
Scalar or Dot Product
The scalar product A B of two vectors A and B is the scalar, a, defined by the
relation:
A B A1 B1 A2 B2 A3 B3 Ai Bi a
B:1
where the summation convention is used in the expression on the far right. The
scalar may be viewed pictorially as the product of the magnitudes of the two
vectors, and the cosine of the smallest angle between them:
A B A1 B1 A2 B2 A3 B3 Ai Bi a jAjjBjcosA; B
B:2
Using this definition, it is clear that the dot product ei ej of any two unit base
vectors of a rectangular Cartesian coordinate system is such that ei ej 1; if i j;
and ei ej 0; if i=j. This gives rise to a special symbol called the Kronecker
delta, dij, which is equivalent to the identity 3 9 3 matrix, and hence, to the secondorder identity tensor:

1; if i j
dij
B:3
0; if i 6 j
Using this definition and Gibbs notation, the dot product of two vectors, A and B,
may be computed as
e j Ai Bj dij Ai Bj dji Ai Bi
A B Ai b
e i Bj b
e j Ai Bj b
ei b
B:4
This shows that, in particular, the scalar product of two vectors is commutative.
The dot product is also associative and distributive with respect to vector addition.
Vector or Cross Product
The cross product A B of two vectors A and B is a third vector, C, that is
perpendicular to the original vectors. The magnitude of the resultant vector may be
viewed pictorially as the product of the magnitudes of the two vectors, and the sine
of thesmallestangle between them:
jA Bj jCj jAjjBjsinA; B
B:5
Using this definition, the vector or cross product, ei ej of any two unit base
vectors of a rectangular Cartesian coordinate system, is such that ei ej ek ; if
the indices i, j, k are an even permutation of the coordinate directions, 1, 2, 3; and
ei ej o when i, j, k are an odd permutation of 1, 2, 3. This can be expressed
more succinctly by means of the alternating symbol eijk, defined such that
eijk
8
>
< 1; when ijk are an even premutation of 123
1; when ijk are an odd premutation of 123
>
:
0; when any two subscripts are equal
327
B:6
Using this definition and Gibbs notation, the cross product of two vectors,
A and B, may be computed as
A B Ai b
e j Ai Bj eijk b
e i Bj be j Ai Bj b
ei b
e k Ck b
ek C
B:7
Hence, in indicial tensor notation, the vector C, resulting from the cross product
A B; is given by Ck eijk Ai Bj . The alternating symbol, having three free
indices and being an ordered system of numbers, is a tensor of third order, called
the alternating tensor. From the definition of the alternating tensor, follows directly
that the cross product of two vectors is not commutative, as
A B Ai Bj eijk b
e k Ai Bj ejik b
e k Bj Ai ejik b
e k B A
B:8
Tensor Product
The operation of lumping vectors in juxtaposition is called tensor product. The
tensor product is also called the dyadic product; and the factors in the product,
dyads. Second-order tensors result when two vectors are multiplied in this fashion.
For instance, the tensor product AB of two vectors, A, and B, is the second-order
tensor:
AB A b
C
e j Ai Bj b
eib
ej
i e i Bj b
B:9
By this definition, the second-order identity tensor, represent by the letter I with
a double over-bar, I; is the continued dyadic product:
I b
eib
ei
B:10
It is also clear that tensor multiplication is not commutative, as changing the

order of the vectors changes the order of the components of the resulting tensor
indicated by the dyad b
eib
e j : The tensor BA Ai Bj b
ejb
e i is called the transpose of the
tensor AB. By this definition, AB is itself the transpose of the tensor BA. The
transpose of a given tensor is indicted by appending the superscript T to the stem
symbol representing the tensor. Thus, the transpose of (AB) is (AB)T. Higher-order
tensors are defined by means of continued dyadic products. For instance, a third
b
eib
order tensor may be written as C
ejb
e k Cijk :
Tensor or dyadic multiplication of any order is assumed to obey the following
laws of operation:
Associative law. The factors of a tensor product may be lumped in any manner
without changing the result: A B C A B C A B C:
Distributive law. If B and C are tensors of the same rank, and A is any continued

dyadic product, then A B C A B A C:
328
Scalar multiplication law. A scalar, a, may be placed anywhere in a continued

dyadic product without changing the result: aA B C A a B C A B a C
A B C a:
The dot and cross products of dyads are an extension of the definitions of these
operations between two vectors. Thus,
AB CD AB CD B CAD
B:11
AB CD AB CD
B:12
Two types of double dot and cross products are defined. In the first type of
double product, the operations are resolved from right to left:
AB CDAB C DB CA DA DB C
AB CD B CA D
B:13
B:14
In the second type of double product, the operations are resolved from left to
right:
AB : CDA CB D
B:15
AB
CD A CB D
B:16
Mixed dot and cross products are correspondingly defined. Thus, for instance:
AB CD A CB D
B.1.1
B:18
Coordinate Transformations
There are two types of coordinate transformations between reference systems

having the same origin: rotation and reflection. A rotational transformation of a
coordinate system preserves its right- or left-handedness. A reflection of a
coordinate system about one of its coordinate planes, which is called an inversion,
changes a right-handed system into a left-handed one, and vice-versa. The
components of tensors have meaning relative to a reference system; hence, it is
important to study the effect that coordinate transformations have on tensor
components. We do this by considering how the mathematical description of
tensors of various orders change in going from one reference system, (X1, X2, X3),
say, to another X10 ; X20 ; X30 ; which is obtained from the first by an arbitrary rigid
rotation, as indicated in Fig. B.1.
Rotation of Scalars
By definition, a scalar, or tensor of order zero, is a quantity that has the same value
and hence mathematical description, in all coordinate systems. The value of a
329
Fig. B.1 Rigid rotation of

two coordinate systems
X2
X1
X3
scalar, a, like mass or temperature, will only change if the units of measurement
are changed. Thus, its values in the primed and unprimed systems are equal:
a0 a
B:19
Rotation of Vectors
As a mathematical entity, a vector exists irrespective of the coordinate systems
used to represent it. The components of the same vector, however, change from
one coordinate system to another. This is represented in Fig. B.2.
Using the summation convention with Gibbs notation, an arbitrary vector v,
may be expressed in terms of the unit vectors of each of the two coordinate
systems, and its corresponding components, as
v v0j b
e 0j vi be i
B:20
The relationship between the components in the primed and unprimed systems
may be obtained post-multiplying this relation by b
e 0k and using the definition of the
unit tensor:
e 0k v0j d0jk v0j djk v0k vi b
e 0k
e 0j b
ei b
vb
e 0k v0j b
From this result, first note thatfrom the commutative property of the dot
product:
vb
e 0k be 0k v v0k
B:21
Also:

0

v0k b
ei b
e 0k vi b
ek b
e i vi
B:22

e 0k cosXi ; X0k cosX0k ; Xi
where by definition of the dot product b
ei b
0

0

be k b
ek b
e i . Hence, the array b
e i ; which is a second-order tensor because it is
described
by
two
free
indices,
is
the rotation matrix. We denoted it simply by
0

ak 0 i b
ek b
e i . In this notation, the rotational transformation for vectors may be
expressed as
0

v0k b
ek b
e i vi ak 0 i vi
B:23
330
Fig. B.2 Graphical representation of a vector in two

coordinate systems differing
by a rigid rotation
X2
X2
X1
X1
X3
X3
Had we chosen to post-multiply v by b

e k ; in the previous development, we
would have obtained, respectively that
v e k vk
B:24

0
0
0
vk b
ek b
e i vi aki0 vi
B:25
and
Using this and the corresponding expression for v0k ; the following results

vk b
ek b
e 0i v0i aki0 ai0 l vl
B:26
Because the left- and right-hand sides of this equation represent the same
component, it follows that the rotation matrix is such that
aki0 ai0 l dkl
B:27
To simplify, we drop the prime notation and adopt the convention that the first
index in the rotation tensor, akl ; always represents the resultant vector, and the
second index, the vector being rotated. We then rewrite the previous relation as
aki ail dkl
B:27a
The left-hand side of this expression is the indicial form of the product of a
matrix and its transpose; and the right-hand side is the indicial form of the identity
matrix. Hence, in matrix notation,
A AT I AT A
B:27b
By definition of the inverse of a non-singular square matrix, it follows that

A1 AT
A square matrix whose inverse is equal to its transpose is called orthogonal.

Hence, the rotation matrix is orthogonal.
331
Rotation of Tensors
The effect of rotation of the coordinate system on a tensor is easily established
from that of vectors, expressing the tensor in continued dyadic form using Gibbs
representation. For instance, let T Tij be i b
e j represents a second-order tensor.
Then, since the tensor is the same in all coordinate systems and the only thing that
might change is its mathematical description in terms of components and unit
vectors along the coordinate axes:
T Tij0 b
e 0i b
e 0j Tij b
eib
ej
B:28
Proceeding as before, we dot-multiply this expression by b

e 0k and b
e 0l and arrive at
Tij0 aik ajl Tkl
B:29
e 0i ; b
e k ; and ajl be 0j ; b
e l :
where to emphasize, aik b
el:
In similar fashion, multiplying by b
e k and b
Tij aki alj Tkl0
B:30
ei; b
e 0k ; and ajl b
ej; b
e 0l .
In this case, according to our convention, aik b
Rotation of tensors of higher order is defined in an entirely similar fashion.
B.1.2
Isotropic Tensors
An isotropic tensor is one whose components are the same in every orthogonal
coordinate system. By this definition,
(a)
(b)
(c)
(d)
(e)
Every tensor of order zero, that is every scalar, is isotropic.

There are no isotropic tensors of order one. That is no vector is isotropic.
The identity tensor, dij ; is the only isotropic tensor of second order.
The alternating tensor, eijk ; is the only isotropic tensor of third order.
The product of isotropic tensors is isotropic. Therefore, the tensors dij dkl and
eijk dkl are isotropic tensors of fourth order.
(f) The most general isotropic tensor of order four may be expressed as a linear
combination of fourth-order isotropic tensors, as
Aijkl adij dkl bdik djl cdil djk
B:31
It may be written as the sum of symmetric and an anti-symmetric tensors, as

Aijkl kdij dkl l dik djl dil djk m dik djl dil djk
B:32
332
B.1.3
Tensors of Second Order
Second-order tensors enjoy the following properties:

The double dot product, T I; of any second-order tensor T with the identity
tensor, I; is a scalar, tr{T}, called the trace of the tensor:
B:33
ekb
e j Tij dij Tii trfTg
eib
e j b
e k Tij b
e i djk b
e k Tij b
ei b
T I Tij b
A second-order tensor T Tij b
eib
e j is said to be symmetric, if it is equal to its
transpose; that is, if
T
T Tij b
eib
e j T Tji b
ejb
ei
B:34
is symmetric if its components satisfy the relations:

That is, a tensor, T;
Tij Tji
B:35
may be expressed uniquely as the sum T T T

A second-order tensor, T;
S
A
of a symmetric and an anti-symmetric tensor, T S and T A ; respectively. Where
TSij

1
Tij Tji
2
B:36a
and

1
Tij Tji
B:36b
2
n o
As may be seen from these definitions: tr T A TAii 0. Also, when T is
symmetric, this decomposition yields a zero antisymmetric tensor: T A 0:
may be expressed uniquely as the sum T T T
A second-order tensor, T;
S
D

of a symmetric and an anti-symmetric tensor, T and T ; respectively; where the
TAij
symmetric tensor is chosen to be diagonal, and the antisymmetric tensor is

defined as the tensor that is left over, so that
1
TSij Tkk dij
3
B:37a
1
TDij Tij Tkk dij
3
B:37b
In this decomposition, the symmetric tensor, T S ; is called the spherical tensor,

and the antisymmetric tensor, T D ; is the deviatoric tensor. Also, even if the
original tensor is symmetric, its deviatoric component is not zero: T D 6 0:
B.1.4
333
Principal Value Problem
The dot product of a tensor T Tij b

eib
e j with a vector u ui be j ; produces another
vector, v vi b
e j ; for: T v Tij b
eib
e j uk b
e k Tij uk b
e i djk Tij uj b
e i vi b
ei
Thus,
Tij uj vi
B:38
If the resulting vector, v, is parallel to the original vector, u, then since v ku,
one has
Tij uj kui
This expression may be written in the form:

Tij kdij uj 0
B:39a
B:39b
This represents a homogenous system of linear equations in the components uj

of u. This system will possess a unique solution, only if its determinant is zero:

Tij kdij 0
B:40
Since the matrix of a second-order tensor is a 3 9 3 square matrix, this equation
represents a cubic equation in the parameter k. It is customary to write this
equation in the form:
k3 IT k2 IIT k IIIT 0
B:41
This equation is called the characteristic equation of the problem, and its roots
are called the characteristic roots. The quantities IT, IIT, and IIIT are called first,
second, and third (principal) invariants of the tensor. They are given by the
following relations:

IT T11 T22 T33 Tii trfTg
2
2
2
IIT T11 T22 T22 T33 T33 T11 T23
T31
T12

1
Tii Tjj Tij Tij
2
IIIT eijk Ti1 Tj2 Tk3 detTij
B:42
The characteristic values are also called principal values, proper values, and
quite frequently, eigenvalues, borrowing the German word, eigen, for proper.
By definition, to each eigenvalue, k(m), there will correspond an eigenvector, u(m).
If the tensor is real and symmetric, its eigenvalues and associated eigenvectors will
be real. The magnitude of the eigenvectors is indeterminate, because the resulting
system for each eigenvalue k(m)
334
m
Tij km dij uj 0;
no sum on m
B:43
is homogeneous; and for a non-trivial solution to exist, its rank is less than the
number of equations it represents. For this reason, the eigenvectors are usually
normalized as unit vectors. Taking a set of coordinate axes (that is, a primed
coordinate system) along the eigenvectors, allows expressing each eigenvector,
u(m), as
m
um ui b
e i ami b
ei
B:44
m
This defines the components of the eigenvector u(m) as ui ami . We use this
to write the eigenvalue problem for eigenvector u(m) in the form:
Tij amj km ami ;
no sum on m
B:45
Multiplying both members of this expression by api ; using the orthogonality of

0
we arrive at the relation:
the rotation tensor, and noting that api Tij amj Tpm
0
Tpm km dpm ; no sum on m
B:46
This expression proves that the transformation to the coordinate system

represented by the eigenvector triad diagonalizes the original tensor, and that the
elements along the main diagonal of the matrix representation of the tensor are the
principal values of the tensor. That is, in the principal system, the matrix of any
second-order symmetric tensor takes the diagonal form:
2
3
0
0
k1
0 5
T 4 0 k2
B:47
0
0 k3
B.2 Bodies and Motion

A body B may be defined as a collection of material points. The simultaneous
positions of the particles or material points of a body constitute a configuration of
the body. For reference purposes, it is convenient to choose one configuration of
the body to which all other configurations are compared. This special configuration
is called the reference or initial configuration of the body. We use a rectangular
coordinate system to describe the configurations of the body B and denote the
particles by their position vectors. We use capital letter X to indicate a particle with
!
position vector X Xi ei X1 ; X2 ; X3 in the original configuration and lower
ei
case letter, x, to refer to the same particle with position vector !
x xi b
x1 ; x2 ; x3 in another configuration. Figure B.3 shows this convention.
335
Fig. B.3 Position of a particle in the reference and

another configuration
X
x
X
x
A configuration may be represented by a one-to-one (invertible) mapping, !

v ; as
!
!
x !
v X ; xi vi X1 ; X2 ; X3 ; i 1; 2; 3
B:48
With this, a motion of a body is a continuous sequence of configurations in time
and is denoted by
!
!
x !
v X ; t ; xi vi X1 ; X2 ; X3 ; t; i 1; 2; 3
B:49
According to the previous definitions, the initial configuration is that occupied
by the body at time t 0. Hence, in particular,
!
!
x !
v X ; 0 ; xi vi X1 ; X2 ; X3 ; 0; i 1; 2; 3;
B:50
As a matter of terminology, the configuration of the body at the present time is
called current configuration. The displacement of a particle at the current time may
then be defined as the difference between its spatial coordinates at that time, xi, and
its initial coordinates, Xi, as8
ui X1 ; X2 ; X3 ; t xi X1 ; X2 ; X3 ; t Xi
B:51
B.3 Analysis of Strain

Consider any two particles, X 1 and X 2 ; which in the reference configuration are
2
1
an infinitesimal distance apart, dXi Xi Xi . After deformation, these two
2
particles will be an elemental distance dxi xi xi

displacements
1
ui
and
2
ui
apart, and their
will also differ by a differential amount; that is
In keeping with standard practice when using indicial notation, reference to the range of the
indices is omitted but shall be understood to run from 1 to the number of dimensions of the
Euclidean space of the discussions.
336

2
dui ui ui . Using these facts and the displacement relationships ui xi Xi ;

introduced earlier, it follows that
2
dxi Xi
Xi
ui ui
dXi dui
B:52
Recalling that the ui are functions of the original position coordinates, Xk, it
follows that
dxi dXi dui
oui
dXj
oXj
B:53
Consequently,
dxi dXi oui
dXj
oXj
B:54
Because of its two free indices, this expression represents a second-order tensor.
Also, since the indices take values over the same range, the left-hand side of the
relation measures the change in length in the coordinate direction indicated by the
index of the numerator, per unit of original length along the coordinate indicated
by its denominator. When the coordinates of numerator and denominator are the
same, the ratio corresponds to the so-called direct strain in the selected coordinate
direction. When the two indices are different, the ratio yields the shear strain, as
measured by the change in displacement in one direction as one moves in another
direction. This definition of shear strain gives rise to two different shear strains for
each pair of coordinate directions, such as qu1/qX2 and qu2/qX1; and such shear
strains would depend on rigid body motion. Only the sum (qu1/qX2 + qu2/qX1) is
independent of rigid body motion. For this reason, the latter definition of shear
strain is used, and the infinitesimal strain tensor, eij, is defined as

1 oui ouj
eij
B:55a
2 oXj oXi
Or, in explicit matrix form:

2 ou
ou
1
6
e 6
4
oX1
1
2
ou2
oX2 oX1
1
ou2
oX2
Sym:
1
2
1
2

ou1
oX3
ou3
oX
1
ou3
ou2
oX3 oX2
ou3
oX3
3
7
7
5
B:55b
As a second-order symmetric tensor, the strain tensor may be decomposed

uniquely into spherical and deviatoric components, has three principal values and
three principal directions, and becomes diagonal in the coordinate system defined
by its principal directions.
337
B.4 Analysis of Stress

The forces acting on a body may be classified as internal and external. Internal
forces are those originating inside the body. External forces originate outside the
body. Forces may be further classified as body forces and surface forces. Body
forces are external forces that act upon each volume element of a body. Surface
forces may be external or internal. External surface forces act on the bounding
surfaces of a body. Internal surface forces are due to interactions between material
particles and thus act on the surfaces of internal volume elements. Surface forces
are reckoned as force per unit surface area.
The body force per unit mass acting on an infinitesimal element dV of the body
!
is denoted by the vector b bi b
e i . With this, the body force on the element dV is
qbi b
e i dV; and the vector sum of all the body forces acting on a body of finite
volume V is given by the volume integral:
Z
Z
Z
Z
Z
qb
bdV qbi b
e 1 qb1 dV b
e 2 qb2 dV b
e 3 qb3 dV B:56
e i dV b
V
The surface force per unit area, also called surface traction, stress traction, or
!
simply traction, is denoted by the vector t ti b
e i . The force acting on an
!
elemental surface of area dS is t dS ti be i dS; and the vector sum of the tractions
across a finite surface S will be given by the surface integral:
Z
Z
Z
Z
Z
btds ti b
e 1 t1 dS be 2 t2 dS b
e 3 t3 dS
B:57
e i dS b
S
Now assume the body in Fig. B.4 is subjected to some prescribed external
surface tractions and body forces, which are functions of the position coordinates
X1, X2, X3.
At any point within the body, an elemental surface DS has a normal unit vector b
n.
!n
The stress traction, t ; at a point (X1, X2, X3) is defined in terms of the surface
!
force F that acts on DS, as the limit:
!
!
DF d F
!n
lim
t
DS!0 DS
dS
B:57
where limit is assumed to exist. In this notation, the superscript n is appended to

!
the surface traction, t ; to indicate the normal to the surface on which the traction
acts. Of particular interest are surface traction vectors with normal parallel to the
coordinate axes. Such traction vectors may be expressed as linear combinations of
unit vectors along the coordinate axes, as
!i
i
i
i
i
t tj b
e j t1 be 1 t2 b
e 2 t3 b
e3
B:58
338
Fig. B.4 Surface tractions in

a body
X2
S
X1
X3
i
!i
where by definition, tj represents the component of traction vector t along the
i
positive direction of coordinate axis Xj. It is customary to write rij tj ; and thus
to express the traction vector as
!i
t rij be j ri1 b
e 1 ri2 b
e 2 ri3 b
e3
B:59
By convention then, the first subscript of rij indicates the direction of the
outward unit normal to the elemental area; and the second index represents the
direction of the component of the traction vector. This sign convention is shown in
Fig. B.5. Since the symbol rij has two free indices, it represents a tensor of second
order, which is called the stress tensor at the point in question.
Once the components of the stress tensor are available at a point, the traction
vector acting on any other plane passing through the same point may be
established. Indeed, using the continued dyadic form of the stress tensor and the
properties of the Kronecker delta, expression (B.59) may be written as the inner or
dot product of the unit vector ei and the stress tensor:
!i
t b
ei b
e k rkj be j dik b
e k rkj be j rij b
ej
B:60
By the same token, the surface traction vector acting at a point on a plane with a
normal b
n ni b
e i ; would be given by
!n
t
ei b
e k rkj b
e j ni rij b
ej
tjn be j b
n r ni b
e k rkj be j ni dik b
From which, the components of the surface traction vector associated with a
plane with normal ^n are
n
tj
ni rij
B:61
For simplicity, we usually omit reference to the surface normal and simply
express the traction vector as
tj ni rij
B:62

Fig. B.5 Sign convention for
the components of the stress
tensor
339
X3
33
31
13
32
Positive face 3
23
22
12
11
21
X2
Positive face 2
X1
Positive face 1
As a second-order symmetric tensor, the stress tensor may be decomposed

uniquely into spherical and deviatoric tensors, has three principal values and three
principal directions, and becomes diagonal in the coordinate system defined by the
principal directions.
B.5
The constitutive equations of linear elasticity are nine equations relating the nine
components of the stress tensor to the nine components of the strain tensor, in the
form:
rij Mijkl ekl
B:63
These nine equations contain 81 material constants, but since the stress and
strain tensors are both symmetric, Mijkl Mjikl Mijlk , the number of independent
constants reduces to 36, because for each pair of indices ranging from 1 to 3, there
are 3(3+1)/2 6 distinct combinations, accounting for the 66 36 independent
material constants. This allows for a convenient representation of (B.63) as a matrix
equation relating a six-by-one column matrix of stresses to a six-by-one column
matrix of strains, though a six-by-six matrix of elastic constants, as follows:
9 2
8
38 9
M1111 M1122 M1133 M1144 M1155 M1166 >
r11 >
>
>
>
>
> e11 >
>
>
>
>
>
> e22 >
6 M2211 M2222 M2233 M2244 M2255 M2266 7>
r
>
>
>
22
>
>
>
>
7>
= 6
=
<
<
7
6
r33
M
M
M
M
M
M
e
3311
3322
3333
3344
3355
3366
33
7
6
7> e12 > B:64
> 6
> r12 >
>
>
6 M4411 M4422 M4433 M4444 M4455 M4466 7>
>
>
>
>
>
4 M5511 M5522 M5533 M5544 M5555 M5566 5>
r13 >
e13 >
>
>
>
>
>
>
>
;
;
:
: >
r23
M6611 M6622 M6633 M6644 M6655 M6666
e23
340
For elastic materials, the number of independent constants is further reduced

from 36 to 21 due to the existence of a strain energy functiondiscussed
subsequently. This extends the symmetry of the constitutive tensor to
Mijkl Mjikl Mijlk Mklij .
The specific coefficients entering the stressstrain relationships will depend on
the directions of the reference axis. A material that allows certain types of
transformation of reference axes without changing its elastic properties is said to
possess elastic symmetry. If the material properties are independent of rotation
about an axis, that axis is an axis of symmetry. If the properties remain the same
under reflection in a plane, that plane is a plane of symmetry.
An elastic material with three orthogonal planes of symmetry called
orthotropic symmetry requires only 9 independent constitutive coefficients. The
corresponding constitutive equations take the form:
9 2
8
38 9
M1111 M1122 M1133
r11 >
0
0
0
>
>
>
>
>
> e11 >
>
>
>
>
>
> e22 >
7>
6
M
r
M
0
0
0
>
>
>
22
2222
2233
>
>
>
>
7>
= 6
=
<
<
7
6
M
r33
0
0
0
e
3333
33
7
6
B:65
6
M4444
0
0 7
>
> 6
>
> r12 >
> e12 >
7>
>
>
>
>
>
Sym:
M5555
r >
0 5>
e >
> 4
>
>
>
>
>
;
;
: 13 >
: 13 >
M6666
r23
e23
An elastic material for which every plane is a plane of symmetry is said to be
isotropic. In this case, the fourth-order constitutive tensor, Mijkl, is isotropic. As
such, it may be represented as [see Sect. B.1.2]:
Mijkl kdij dkl ldik djl dil djk mdik djl dil djk
B:66
By the symmetry requirement, the last term drops out, so that

Mijkl kdij dkl ldik djl dil djk
B:67
In other words, only two material property constants are needed to fully
characterize the constitutive equations of isotropic materials. Combining this
expression with (B.63), using the properties of the identity tensor, and simplifying,
yields the constitutive equations for isotropic elastic materials, as
rij kekk dij 2leij
B:68
In this expression, l is the materials shear modulus; and both k and l are called
Lames constants. Also, as shown in a later section, when strains are small, ekke11
+ e22 + e33 : eV represents the volumetric strain.
Using the decomposition of the stress and strain tensors into their spherical and
deviatoric parts, the constitutive equations may be cast in the following equivalent
form:

2
rkk
ekk
r S 3 k l eS ; r S
; eS
B:69a
3
3
3
rDij 2leDij ;
341
rDij rij rS dij ;
eDij eij eS dij
B:69b
Note that in this form, the six constitutive equations of linear elasticity decouple
into 7 one-dimensional constitutive equations, connected through their spherical
and deviatoric components. Equation (B.69-a) between the spherical stress and
spherical strain is usually presented in the following alternate forms:
rS 3KeS
p KeV ;
p
B:70a
rkk
; eV ekk
3
B:70b
The material constant K k 23 l; relating the hydrostatic pressure p r3kk ;

to the volumetric strain eV ekk is properly termed volumetric or bulk modulus.
In practice, the tensile relaxation modulus, E, and the so-called Poissons ratio,
v, are easier to determine than k, l, or K. When E and v are used, the constitutive
equations take on the following form:

E
v
rij
eij
ekk dij
B:71
1 v
1 v
Only two independent constants are needed to fully characterize a linear elastic
material, and there are several material constants available to do thisk, l, K, E,
and v. The form of the constitutive equations will depend on which properties are
chosen to express them. The elastic constants E, v, and K, are related to k and l, as
follows:
E 21 ml;
k
;
2k l
2
K k l
3
B:72a
With these, the following interrelationships may be verified:

l
E
21 m
B:72b
2
2lm
mE
lE 2l
kK l
3
1 2m 1 m1 2m
3l E
B:73c
2
2l1 m
E
El
K k l
3
31 2m 31 2m 33l E
B:72d
l3k 2l
9Kl
2l1 m
kl
3K l
B:72e
k
3K 2l
E
1
2k l 6K 2l 2l
B:72f
There are isotropic elastic materials whose bulk modulus is orders of magnitude
larger that their shear modulus. The practical implication of this is that their
342
response to mechanical loads takes place by shearing only, with virtually no

change in their volume. Such materials are called incompressible. As will be seen
later on, volume preservation in linear materials requires that the volume strain, eV
ekk, be identically zero throughout the loading process. Hence, incompressibility
enters as a constraint on the equations of motion. This means that one can add a
spherical stressuniform in all directionsof any magnitude to the stress field,
without altering the strains. For incompressible materials, then, the stress tensor is
determined from the strains, only up to a spherical stress. Since the spherical
stress, rS : (rkk)/3, is the average of the normal stresses, its negative, -rS, is
called the hydrostatic pressure. Thus, -p : rS : (rkk)/3. The constitutive
equations for linear elastic isotropic incompressible materials may be written
directly from Eq. (B.69-b), using rS :-p, noting that eDij eij, because ekk : 0,
and expressing rij in terms of rDij. Hence,
rij pdij 2leij
B:73
B.6 Conservation Principles

There are laws of physics which are obeyed by all substances in the bulk, be they
elastic, viscous or viscoelastic, irrespective of whether their constitutive equations
are linear or non-linear.9 These laws proclaim the conservation of mass, linear and
angular momenta, and energy. The expressions used to represent conservation
principles are known as conservation or balance laws.
B.6.1
Conservation of Mass
The total mass of a system, which does not exchange mass with its surroundings,
remains the same at all times. In particular, the total mass, M(t), of a body at any
time t, must be equal to its original mass, Mo. Using the materials density,
q(x) which measures mass per unit volume and letting the subscript o denote the
original value of a quantity, we write the equation of mass conservation, as10
Z
Z
qo X dVo
qxdV
B:74
Vo
D.C. Leigh 1968.

For simplicity of exposition, here, and in derivations that follow, we use X and x to denote a
material particle as well as its position vector before and after deformation, respectively.
10
343
Changing the volume of integration from the current, deformed configuration at

time t, to the original configuration:
Z
Z
Z
qo X dVo q xdV
J qxdVo
a
Vo
Vo
Noting that the resulting expression must be valid for an arbitrary initial
volume, we cancel the integral sign and write the balance of mass equation:
qo X qxJx
B:75
For clarity, we have omitted the indices that identify the coordinates, Xi, and xj,
and used instead the particle representation, X, and x. The additional implication
that the motion of the particle X depends on time; that is, that x v(X, t) must also
be kept in mind.
The Jacobian determinant, J, represents the ratio of the change in volume per
unit original volume. Now, by the definition of small axial strain, as change in
length per unit original length, a cube of sides of initial lengths lxo, lyo, lzo subjected
to strains exx, eyy, ezz along its edges, will change its volume so that

V Lxo 1 exx Lyo 1 eyy Lzo 1 ezz

Lxo Lyo Lzo 1 exx eyy ezz O e2
b
2
Vo 1 exx eyy ezz O e
where the symbol O(e2) indicates terms of second order and higher. That is, to first
order, the balance of mass may be alternatively expressed in terms of the direct
strains, as
J
V
1 exx eyy ezz
Vo
B:76
Two important consequences of this expression are that

(a) The change in volume per unit original volume, also known as volume strain,
is equal to the sum of the direct strains in any three perpendicular directions:
evol
V Vo
exx eyy ezz e11 e22 e33
Vo
B:77
(b) Incompressible materials deform without changing volume: J V/Vo 1, and

thus, for such materials, the sum of the direct strains in any three mutually
perpendicular directions is zero:
evol exx eyy ezz e11 e22 e33 0
B:78
344
B.6.2
Conservation of Linear Momentum
Newtons second law of motion requires a balance between the external resultant
load on a body and the rate of change of its linear momentum. The integral form of
this law is due to Cauchy and gives rise to the equations of motion, valid for all
materials in bulk. In terms of the Cartesian components of the stress tensor, rij, and
velocity field, vi, these equations take the following form:
o
o
rij qbi q vi
oXj
ot
B:79
When inertia terms are zero, as in static problems, or can be neglected, as under
steady-state conditions, the acceleration term on the right-hand side is dropped:
o
rij qbi 0
oXj
B:80
Here, differentiation is understood with respect to the original coordinates, Xi,

so that, for instance, the X-component of the equations of motion in unabridged
form reads:
o
o
o
o
rxx
ryx
rzx qbx q vx
oX
oY
oZ
ot
B:81
The y- and z-component equations are written with corresponding permutations

of the coordinates.
B.6.3
Conservation of Angular Momentum
Non-polar materials are defined as those for which the resultant internal moment
on the surface of any volume element is zero. For such materials, the principle of
conservation of angular momentumthe resultant external moment on a body is
equal to the time rate of change of its angular momentumyields the requirement
that the stress tensor, rij, be symmetric:
rij rji
B:82
Using unabridged notation, these expressions take the form:

rxy ryx ;
rxz rzx ;
ryz rzy
B:83
The balance of angular momentum reduces the number of independent

components of the stress tensor from 9 to 6.
B.6.4
345
Conservation of Energy
The experimental fact that the total energy in a thermodynamically closed system
remains constant represents the first law of thermodynamics. According to this
experimental law, there is always a balance between all energy put into a system
and all internal energy. The external energy consists of mechanical and thermal
work, while the internal energy is separated into macroscopically observable
kinetic energy and intrinsic energy. The rate form of the first law of
thermodynamics, that qEint/qt qEext/qt, stipulates a balance between the
intrinsic energy, on one side, and the stress and thermal power on the other:
q
oe
o
oqi
rij eij qr
ot
ot
oXi
B:84
In this expression, the first term on the right-hand side is the sum total of the
products of the stress components and the corresponding strain rates and is thus
called stress power. The other two terms on the right-hand side represent the
thermal power: r is due to the heat sources and the qi are the components of the
heat flux vector.
B.7 Boundary Value Problems

The isothermal boundary value problem of linear isotropic elasticity consists of
the three equations of motion with their initial and boundary conditions. Taking
the initial configuration of the body to be free of stress and at rest prior to the
application of the loadingall field variables are identically zero for t \ 0the
equations of motion and boundary conditions are
o
o2
rji qbi q 2 ui ; Xk 2 V
oXj
ot

ui ~
X; t uoi ~
X; t ; Xk 2 Su ; t
0

X rij ~
X; t ; Xk 2 ST ; t
0
X; t tio ~
nj ~
B:85
In these expressions, ni represents the unit outward normal to the boundary, S,

formed of Su and ST, where displacements and tractions, respectively, are prescribed.
Here, ni is not a function of time, which is consistent with the assumptions of small
displacements and strains. Also, in a well-posed problem, Su and ST do not intersect
Su \ ST U; which means that one cannot prescribe different types of boundary
conditions at the same point and in the same direction, at the same time.
The boundary value problem of linear elasticity, as posed in (a) consists of three
equations in nine unknownsthree components of the displacement vector and six
independent stress components. For a solution to exist, the system in (B.85) has to
be complemented by other relations. The additional relationships are as follows:
346
The displacements, ui, are related to the strains, eij, through the six relationships
that were used earlier to define the strain tensor:

1 oui ouj
eij
B:86
2 oXj oXi
We now have three equations of equilibrium, six straindisplacement relations,
and six stressstrain equations, for fifteen equations; and there are three
displacements, six stresses and six strains, for fifteen unknowns. These are the
field equations of isothermal linear elasticity. The uniqueness of solution of the
boundary value problem posed is due to the fact that linear elastic materials have a
positive definite strain energy function.
B.8 Compatibility Conditions

The straindisplacement relations relate the three components of the displacement
field to the six components of strain. These expressions result in a unique set of
strains for any prescribed set of displacements, but do not suffice in general to
produce a unique set of displacements from an arbitrarily prescribed set of strains.
The system of equations in the latter case, with six equations in three unknowns, is
over-determined, thus preventing the six components of the strain tensor to be
prescribed arbitrarily. The conditions that the strain tensor must satisfy to allow a
unique displacement field upon integration of the straindisplacement relations are
known as the integrability or compatibility conditions.
The equations of compatibility are obtained differentiating the strain
displacement relations twice and permuting indices. This process yields the
following 81 equations, of which only six are independent:
eij;kl ekl;ij eik;jl ejl;ik
B:87
Using the standard Cartesian coordinates x, y, z, for subscripts 1, 2, and 3,

respectively, the equations of compatibility take the following unabridged form:
o2
o2
o2
exy
eyy 2 exx 2
2
ox
oy
oxoy
o2
o
o
o
o
exx eyz ezx exy
ox ox
oy
oz
oyoz
B:88
Similar permutations of x, y, and z produce the other four independent relations.

For two-dimensional problems in the x-y plane, the only non-trivially satisfied
relation is the first one listed above.
On occasion, the compatibility conditions are required in terms of stresses.
Since stresses and strains are connected through the constitutive equations, the
integrability conditions in terms of stresses depend on material properties. They
may be obtained by substituting the strainstress constitutive equations in the
compatibility equations. The resulting expressions are known as the BeltramiMichell relations.
347
B.9 Energy Principles

Energy principles play an important role in the theory of elasticity. One reason for
this is that the work performed on an elastic solid by external agents is stored in the
solid in the form of internal energy, and this internal energy is fully recoverable11
upon removal of the external agents. Because of this, energy principles are
frequently used to derive methods of solution to elastic boundary value problems.
Among the various energy principles that are available for elastic solids, we
present the principle of virtual work, and the important theorems of minimum
potential energy and minimum complementary potential energy.12 Before
discussing these principles, we introduce the concepts of statically admissible
stress fields and kinematically admissible displacement fields.
Statically admissible field
A stress distribution or field is called statically admissible if it is continuously
differentiable in the domain, V, occupied by the body, identically satisfies the
equations of equilibrium inside V, and takes on the values assigned to the surface
tractions on the portion, ST, of the body, where tractions are prescribed.
Kinematically admissible field
A displacement field is called kinematically admissible if it is three times
continuously differentiable13 inside the domain, V, occupied by a body, identically
satisfies compatibility in V, and takes on the values of the displacement field on the
portion, Su, of the body, where the displacements are prescribed.
B.9.1
Principle of Virtual Work
The principle of virtual wok, also referred to in the literature as the principle of
virtual displacements, states that
Given any statically admissible stress field: fr0ij ; tj0 ni r0ij ; Fi qbi g; and any
kinematically admissible displacement field, u00i then
Z
Z
Z
ti0 u00i dS Fi u00i dS r0ij e00ij dV
B:89
S
11
This means that if the external agents that put work into an elastic solid were completely
removed, the internal energy stored in the solid could be used to perform an amount of work
equal to the work that was put into the solid in the first place.
12
The principle of virtual work (or its generalization to virtual velocities), as well as the
theorems of minimum potential energy, and minimum complementary potential energy have been
applied to derive finite element methods.
13
A kinematically admissible displacement field has to be thrice continuously differentiable
because the compatibility conditions involve the second derivatives of the strains, and the strains
are defined in terms of the first derivatives of the displacement field.
348
To prove the principle, we first transform the surface integral into a volume one
using the Green-Gauss theorem [c.f. Appendix A]. Afterward, we invoke the
equations of equilibrium: r0ji;j Fi 0; the straindisplacement relations: e00ij
00
1=2ui;j u00j;i ; and the symmetry of the stress tensor, that rij, rji, on account of
which r0ji u00i;j r0ji e00ij . Since the proof of the principle does not involve constitutive
equations, the principle is applicable to all materials in bulk.
B.9.2
Principle of Minimum Potential Energy
An elastic material may be defined as one for which there exists a single-valued
positive definite potential function of the strains, W(eij), such that
Weij
Zeij
rkl dekl
B:90
The function W is also required to be path independent and convex in the

strains; the latter requirement is in the sense that, for any two strain fields, eij and
eaij the following relationship must be satisfied:

oW
a
a

B:91
W eij W eij
eij eij
oekl eij
For W to be path independent and a function of the final strains only, its total
differential must be a perfect differential. This condition is satisfied if
o2 W
o2 W
oeij oekl oekl oeij
B:92
orkl orij
oeij
oekl
B:93
or, equivalently,
oW
dekl rkl dekl is a
This results follow directly from (B.90), since dW eij oe
kl
sum of differentials, and thus, a generalization of the two-dimensional form df
Pdx Qdy; for which the proposition is easily proven14. The total differential of
W also defines the explicit relationship between W and the stress field, as
14
of
of
If df Pdx Qdy; is a perfect differential, then df ox
dx oy
dy; which implies that
of
of
P ox
and Q oy
. By the theorem of Schwartz for mixed partial derivatives,
oP
oy
o f
o f
oyox
oxoy
oQ
ox .
rij
349
oW
oeij
B:94
This relationship is what gives W its character as a potential function, since the
stresses are calculated from it as the derivatives of the function with respect to the
corresponding strains. Also, because W is measured per unit volume, it is referred
to as the strain energy density.
The potential energy, U, of an elastic body subjected to conservative body
forces, Fi, and surface tractions tio ; is defined for any kinematically admissible
displacement field, u00i , as
Z
Z
Z
00
00
o 00
U ui Weij dV ti ui dV Fi dV
B:95
V
In this expression, e00ij 1=2u00i;j u00j;i . If rij ; eij ; and ui represent the actual
stresses, strains, and displacements, respectively, it is an easy matter to prove that,
because of the convexity of the strain energy density, W:

U u00i
Uui
B:96
This equation represents the theorem of minimum potential energy, that among
all the kinematically admissible displacement fields, the actual displacement field,
which is also statically admissible by definition, minimizes the potential energy.
B.9.3
Principle of Minimum Complementary Potential Energy
In the previous section, we defined an elastic material as one for which a strain
energy density, W, can be found, such that the stressstrain constitutive equations
can be obtained from it by differentiation according to rij qW/qeij. The existence
of a unique inverse of these constitutive equations, together with the symmetry
condition
oekl
oeij
orij orkl
B:97
guarantees there is a complementary strain energy density Y, defined by the

relation:
Y rij eij W
B:98
Since dY drij eij rij deij dW rij deij ; the complementary strain energy
density, Y, defines the strainstress constitutive equations, through the relations:
eij
oY
orij
B:99
350
The complementary potential energy, W, of an elastic body is defined for any

statically admissible field r0ij ; by the relation:
Z
Z
0
0
W ri
Yrij dV
ti0 ui dS
B:100
V
Su
Where the statically admissible surface tractions, tj0 ; are computed from the
statically admissible stress field, according to tj0 ni r0ij . If rij ; eij ; and ui represent
the actual stress, strain, and displacement fields, then it can be proven, using the
convexity of the complementary strain energy density, Y, that:

W r0ij
W rij
B:101
This embodies the principle of minimum complementary potential energy that
of all the statically admissible stress fields, the actual field, which is also
kinematically admissible, minimizes the complementary potential energy.
Index
A
Admissible, 264, 267, 268, 270
Aging, 3, 20, 21
Amorphous, 2, 3, 16, 17
Amplification, 140, 142
Amplitude, 94, 96, 97, 99, 111
Analytical, 166, 167, 170
Anisotropic, 193, 196199
Argand, 96
Attenuation, 251
B
Balance, 219221, 240, 248, 249
Bending, 114, 115, 119, 122, 129, 132
Boltzmanns, 24, 25, 27, 44
Boundaryvalue, 219, 221, 224227, 229, 232
Buckling, 114, 124
C
Castigliano, 262, 264, 268, 271273
Characterization, 166, 170, 179, 180, 182
Compatibility, 223, 224, 231
Complex, 9496, 98, 99, 101, 102, 105, 170
Compliance, 2, 9, 1618, 20, 21, 56, 59, 62,
7274, 78, 81, 91, 94, 99, 104107, 110,
166, 167, 170172, 174, 175, 182
Conservation, 220, 221
Constitutive, 24, 4043, 45, 93, 98, 100, 101,
103, 166, 170
Convolution, 38, 39, 41, 45, 166, 172174, 187
Correspondence, 136
Correspondence principle, 231234
Creep, 2, 8, 9, 16, 17, 21, 24, 30, 31, 33, 37,
38, 41, 56, 59, 64, 70, 7274, 78, 85, 91,
121, 124, 166, 167, 170172, 174, 175, 177
Critical, 124, 126, 127
Cross-linked, 3, 16
Crystalline, 3
Cycle, 24, 34, 37, 51
Cyclic, 239, 240, 242
D
Damper, 54, 57, 60, 61
Dashpot, 54, 5760, 69, 70, 75, 88
Delta function, 5
Deviatoric, 206, 209
Differential, 93, 103, 114, 118, 119, 122, 127
Dirac, 5
Dirichlet, 100, 111, 166, 170, 178, 179, 182,
183, 187
E
Elastic, 1, 2, 57, 1214, 1619
Energy, 2, 12, 14, 18, 94, 109, 220
Equation, 166, 167, 174, 176, 177, 179
Equilibrium, 2, 16, 17, 19, 33, 34, 37, 39, 45,
49, 220, 224, 225, 227, 229, 231
Euler, 94, 97, 126
Excitation, 94, 95, 100, 102, 109
F
Fading, 24, 30, 34, 36, 49
Field, 262, 264, 265, 267, 268, 271, 273
Fluid, 1, 2, 4, 5, 10, 54, 62, 69, 74, 88
Forced, 114, 139, 145
Fourier, 94, 95, 97, 100102
Free, 114, 129, 144
Frequency, 94, 95, 97, 150, 151, 155, 241,
242, 244
Front, 246250, 252, 254256
Functionals, 261263, 265, 268, 273
G
Glass, 2, 16
Glassy, 16, 18, 20, 37, 45
H
Harmonic, 93, 95, 102, 103, 240, 241, 244
Heaviside, 5

351
352
Index
Hereditary, 28, 35, 117, 119, 121, 125, 171

Hooke, 5, 8
Hygrothermal, 160, 161
O
Operator, 45
Orthotropic, 193, 199, 201
I
Imaginary, 9698, 102, 112
Incompressible, 209, 213
Indicial, 194, 197, 211
Integral, 9395, 97, 114, 117, 121, 125
Inversion, 166, 172, 173, 175
Isothermal, 24
Isotropic, 197, 203206, 209, 211
P
Parameter, 170, 174, 183
Polymer, 2, 3, 17
Polynomial, 171, 177, 188
Potential, 264, 265, 267, 269
Power, 166, 170, 171, 179, 182, 184
Pressure, 150, 159
Principle, 24, 25, 27, 30, 34, 36, 38
Prony, 170, 171, 174, 175, 179, 182, 183, 187
Property function, 166, 169175, 179
J
Jump, 246, 247, 249251, 255, 256
Q
Quasi-static, 219, 220, 223, 231
K
Kelvin, 69, 70, 73, 74, 79, 88
Kernel, 166
L
Laplace, 24, 4143, 63, 90, 94, 100, 172, 173
Law, 166, 170, 171, 179, 182, 184
Least squares, 179, 184
Logarithmic, 166, 167
Long-term, 2, 34, 45
Loss, 97, 98, 107, 109, 110
M
Mass, 114, 138, 140, 143, 144
Master, 152, 154, 155
Matrix, 195, 198, 199, 201, 206
Maxwell, 69, 74, 75, 81, 88
Memory, 24, 30, 3336, 38, 49
Model, 54, 55, 63, 67, 69, 79, 81, 83
Modulus, 2, 4, 6, 7, 17, 18, 20, 55, 56, 62, 65,
85, 90, 91, 94, 9698, 100, 102, 107
Moisture, 160, 161
Momentum, 220, 221, 240, 248, 249
Motion, 219, 220, 223, 224
N
Navier, 115
Newton, 5, 101
Numerical, 166, 179, 180, 182
R
Real, 94, 9698, 102, 107
Reciprocal, 262, 279
Recovery, 56, 59, 70, 73, 85, 87
Relaxation, 2, 6, 7, 16, 17, 21, 24, 25, 28, 33,
49, 51, 56, 58, 64, 69, 77, 90, 91, 166, 167,
171, 172, 174, 175, 177, 179, 182, 183, 188
Retardation, 73, 81, 174, 175, 178, 188
Roots, 177, 188
S
Scale, 150, 151, 154
Separation of variables, 220, 228, 232, 237
Shift, 152155, 159
Shiftfunction, 167, 179, 180
Shock, 239, 246, 247, 249251, 254, 255
Solid, 54, 62, 69, 84
Spectrum, 81, 83
Speed, 239, 243, 244, 246, 249
Spherical, 206210
Spring, 5456, 59, 69, 84, 90, 114, 128, 129,
131, 132, 134
Static, 119, 124, 128, 129, 141, 144
Stationary, 266, 267, 271
Steady-state, 103, 119, 123
Storage, 94, 97, 98
Strain, 2, 4, 5, 6, 8, 10, 11, 54, 55, 57, 58, 61,
63, 66, 68, 69, 71, 75, 85
Stress, 2, 7, 55, 56, 58, 64, 71, 87
Index
Summation, 194, 195
Symbol, 35, 41
Symmetry, 193, 199, 200, 204
Synchronous, 220, 227, 235
T
Temperature, 2, 3, 6, 10, 16, 17, 150, 151, 153,
154, 156159
Tensor, 194, 195, 198, 204, 206
Test, 166, 167, 169, 179, 180, 182, 183, 185
Thermal, 150, 157159, 161
Thermorheologically simple, 155, 156
Time, 149, 150, 153, 155, 160, 161, 166, 174,
177, 178, 183, 188
Time-temperature superposition, 150152,
154, 155, 160
Torsion, 114, 120, 122, 134
Transform, 94, 95, 100, 102, 173, 175, 202,
205, 208, 212
Transition, 2, 1618, 20, 150, 157, 158,
172, 159
Transmissibility, 141
353
U
Unit load, 262, 273, 279
V
Variation, 262, 263, 266, 267, 269271
Variational, 262, 264
Vibration, 114, 138, 139, 144
Viscoelastic, 113, 16, 20
Viscous, 25, 10, 11, 1315
Volterra, 175
Volume, 157, 158
W
Wave, 239, 240, 242, 246, 247, 250, 252
WLF, 154, 159

Engineering of Viscoelasticity

Uploaded by

Copyright:

Available Formats

Engineering of Viscoelasticity

Uploaded by

Document Information

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

Engineering of Viscoelasticity

Uploaded by

Copyright:

Available Formats

Danton Gutierrez-Lemini

Additional material to this book can be downloaded from http://extras.springer.com/

Springer New York Heidelberg Dordrecht London

To Lolita, the love of my life; whose support,

type of productcalled convolutionbetween a function which represents the

I would like to express my deep appreciation to Michael Luby, Engineering Editor

Fundamental Aspects of Viscoelastic Response . . . . . . . . .

Constitutive Equations in Hereditary Integral Form

Constitutive Equations in Differential Operator Form

Constitutive Equations for Steady-State Oscillations . . . .

Material Property Functions and Their Characterization. . . . .

Three-Dimensional Constitutive Equations . . . . . . . . . . . . . .

Isothermal Boundary-Value Problems . . . . . . . . . . . . . .

11 Variational Principles and Energy Theorems. . .

12 Errata to: Engineering Viscoelasticity . . . . . . . . . . . . . . . . . . . . .

Appendix A: Mathematical Background. . . . . . . . . . . . . . . . . . . . . . .

Appendix B: Elements of Solid Mechanics . . . . . . . . . . . . . . . . . . . . .

Fundamental Aspects of Viscoelastic

This chapter describes the molecular structure of amorphous polymers, whose

Aging Amorphous Compliance Creep Cross-linked Crystalline Delta

1 Fundamental Aspects of Viscoelastic Response

Finally, it will be observed that viscoelastic materials may undergo aging,

The Nature of Amorphous Polymers

1 Fundamental Aspects of Viscoelastic Response

Mechanical Response of Viscoelastic Materials

According to the previous discussion, a material may be classified as viscoelastic if

Mechanical Response of Viscoelastic Materials

On the previous definitions and assumptions, we introduce the one-dimensional

Alternatively, Newtons law of viscosity is used to define the constitutive

Material Response to Step-Strain Loading

We consider the response to a constant strain history of a one-dimensional article

According to (1.1), the response of a straight bar of a linearly elastic solid to a

In words, a uniaxial bar of an elastic solid would respond instantaneously to the

1 Fundamental Aspects of Viscoelastic Response

Per our elementary definition of viscoelastic behavior, we would expect the

According to the previous discussion, M(t) must be a decreasing function of

Mechanical Response of Viscoelastic Materials

Fig. 1.4 Qualitative time

Specifically, to establish the relaxation modulus at a constant temperature:

Material Response to Step-Stress Loading

By definition, the relaxation modulus of linear viscoelastic substances is assumed independent

1 Fundamental Aspects of Viscoelastic Response

Fig. 1.5 Distinction

Now the response of a viscoelastic specimen is expected to lie between that of

Mechanical Response of Viscoelastic Materials

Fig. 1.6 Qualitative

Fig. 1.7 Distinction

1 Fundamental Aspects of Viscoelastic Response

Fig. 1.8 Conceptual time

Specifically, to establish the creep compliance at a constant temperature:

Material Response to Cyclic Strain Loading

As before, we use expressions (1.1) and (1.2) to establish, respectively, the

By definition, the creep compliance of linear viscoelastic materials is assumed independent of

Mechanical Response of Viscoelastic Materials

Fig. 1.9 Qualitative

= A(t) o sin( t+)

Material Response to Constant Strain Rate Loading

1 Fundamental Aspects of Viscoelastic Response

Fig. 1.10 Qualitative

Energy Storage and Dissipation