Unified Mechanics Theory

Cemal Basaran
Introduction
to Unified
Mechanics Theory
with Applications
Second Edition
Introduction to Unified Mechanics Theory
with Applications
Y
Thermodynamic State Index

(TSI) axis
0.999
=
5 years old 100 years old
with stage4
cancer
0
X
5 years old
100 years old
Z
Time
Cemal Basaran
Introduction to Unified
Mechanics Theory
with Applications
Second Edition
Cemal Basaran
Department of Civil, Structural, and Environmental Engineering
University at Buffalo, SUNY
Buffalo, NY, USA
ISBN 978-3-031-18620-2 ISBN 978-3-031-18621-9 (eBook)

https://doi.org/10.1007/978-3-031-18621-9
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland
AG 2021, 2022
This work is subject to copyright. All rights are solely and exclusively licensed 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.
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.
The publisher, the authors, and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Foreword
A new scientific truth does not triumph by convincing opponents and making them see the
light, but rather because its opponents eventually die, and a new generation grows up that is
familiar with it.
1918 Nobel Prize in Physics for his work on quantum theory.
Friedrich-Wilhelms-Universität Max Planck

Berlin, Germany
v
Preface
This book is the culmination of my research over the last 28 years at the University at
Buffalo, The State University of New York. Most of the book relies heavily on the
work of my PhD students and our joint publications. This second edition is based on
the first edition. However, there are two new chapters in this second edition, Chap. 9
on fatigue and Chap. 10 on corrosion-fatigue interaction. Also, there are more
example problems added. Chapter 4 on unified mechanics theory has also been
augmented with more and better explanations.
The traditional [Newtonian] continuum mechanics portion of this book is pri-
marily based on the seminal textbook by Malvern (1969). Since I consider Prof.
Lawrence Malvern’s textbook one of the best written on continuum mechanics, I did
not find any reason to change the formulation or his presentation of the basics. Yet,
when I was a student, I found his book hard to follow because it was written at a high
level. Therefore, I have tried to simplify many of the formulae derivation steps to
make them easier for students to understand. While I did change the order of the
topics which I copied from Malvern (1969), I have tried to stay true to the message
Prof. Malvern had in his formulation.
I have tried to explain the unified mechanics theory in the simplest terms possible.
To accomplish this objective, I found it necessary to include the entire paper by
Sharp and Matschinsky (2015), which is the English translation of Boltzmann’s
(1877) original paper. While I acknowledge that including papers from a journal in
its entirety is a very unorthodox practice, it was necessary, because Boltzmann’s
original paper is misunderstood. Therefore, I had to present his entire derivation
annotated with my explanations and comments. To this day, many physics books
still refer to Boltzmann’s second law equation to be valid for gasses only, with no
reference to his derivation. Boltzmann did not make any such assumption in his
formulation that would have invalidated the applicability of his formulation to
solids. For over a quarter century, my PhD students and I have provided in the
literature the experimental proof of Boltzmann’s second law equation for solids.
Sharp and Matschinsky (2015) have also addressed this major misunderstanding in
their paper. I quote Sharp and Matschinsky (2015) directly. The fact that Boltzmann
vii
viii Preface
does not consider the interaction between the particles only increases the possible
number of complexions, as Boltzmann defines them. Interparticle interactions, like
friction or molecular bonding, would only reduce the number of possible complex-
ions. Therefore, Boltzmann’s formulation can be considered an upper bound limit of
possible permutations. However, Boltzmann’s mathematical derivation is valid for
solids, gasses, and liquids. It was not possible to include Boltzmann’s paper without
Max Planck’s (1900) paper because they complement each other.
In introducing the basics of thermodynamics, I have relied heavily on Callen
(1985). While there are many books on thermodynamics within the literature, I
found the work of Callen (1985) to be the most comprehensive and understandable
from a graduate student’s perspective. I also kept in mind that unfortunately most
mechanics students never take a basic course on thermodynamics at the undergrad-
uate level. Therefore, the chapter on thermodynamics is based primarily on the work
of Callen (1985).
The rest of the chapters are all taken from my work with my former PhD students
to whom I am eternally grateful. My former PhD students who contributed to the
development of this theory over the years are Drs. ChengYong Yan, Rumpa
Chandaroy, Hong Tang, Ying Zhao, Yujun Wen, Shihua Nie, Juan Gomez, Eray
Mustafa Gunel, Shidong Li, Wei Yao, Hua Ye, Ming Hui Lin, Michael Sellers,
Mohammed Abdul Hamid, Yong Chang Lee, Noushad Bin Jamal, and Hsiao Wei
Lee. I am also indebted to Ms. Shifan Cheng for the graphics in the book.
I am especially grateful to the US Navy, Office of Naval Research, for believing
in me and sponsoring most of the work reported in this book. Their continuous
support, [starting in 1997 with the ONR Young Investigator Award Program under
program director Dr. Roshdy Barsoum], has been an immense help, and without it,
this idea would have never come to a realization.
Buffalo, NY, USA Cemal Basaran
References
Callen, H. B. (1985). Thermodynamics and an introduction to thermostatistics (2nd ed.). Wiley.

Malvern L. E. (1969). Introduction to the mechanics of continuous medium. Prentice-Hall.
Planck, M. (1900). Zur Theorie des Gesetzes der Energieverteilung im Normalspectrum.
Verhandlungen der Deutschen Physikalischen Gesellschaft, 2, 237; Translated in ter Haar,
D. (1967). On the theory of the energy distribution law of the normal spectrum (PDF). The
old quantum theory (pp. 82). Pergamon Press. LCCN 66029628.
Sharp and Matschinsky, “Boltzmann, L. (1877). Sitzungberichte der Kaiserlichen Akademie der
Wissenschaften. Mathematisch-Naturwissen Classe. Abt. II, LXXVI, 373–435 (Wien. Ber. 1877,
76:373–435)”. Reprinted in Wiss. Abhandlungen, Vol. II, reprint 42, pp. 164–223, Barth, Leipzig,
1909 [Kim Sharp* and Franz Matschinsky, Translation of Ludwig Boltzmann’s Paper “On the
relationship between the second fundamental theorem of the mechanical theory of heat and
probability calculations regarding the conditions for thermal equilibrium” Sitzungberichte der
Kaiserlichen Akademie der Wissenschaften. Mathematisch-Naturwissen Classe. Abt. II, LXXVI,
1877, 373–435 (Wien. Ber. 1877, 76:373–435)]. Reprinted in Wiss. Abhandlungen, Vol. II,
reprint 42, pp. 164–223, Barth, Leipzig, 1909. Entropy, 2015, 17, 1971–2009.
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 What Is the Mechanics of Continuous Medium? . . . . . . . . . . . . 1
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Stress and Strain in Continuum . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1 Newton’s Universal Laws of Motion . . . . . . . . . . . . . . . . . . . . 5
2.1.1 First Universal Law of Motion . . . . . . . . . . . . . . . . . . 5
2.1.2 Second Universal Law of Motion . . . . . . . . . . . . . . . . 8
2.1.3 Third Universal Law of Motion . . . . . . . . . . . . . . . . . . 9
2.1.4 Range of Validity of Newton’s Universal
Laws of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.5 Relation to the Thermodynamics
and Conservation Laws . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2 Stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.1 Definitions of Stress and Traction . . . . . . . . . . . . . . . . 12
2.2.2 Stress Vector on an Arbitrary Plane . . . . . . . . . . . . . . . 15
2.2.3 Symmetry of Stress Tensor . . . . . . . . . . . . . . . . . . . . . 19
2.2.4 Couple Stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.5 Principal Stresses and Principal Axes . . . . . . . . . . . . . . 24
2.2.6 Stress Tensor Invariants . . . . . . . . . . . . . . . . . . . . . . . 27
2.2.7 Octahedral Plane and Octahedral Stresses . . . . . . . . . . 30
2.3 Deformation and Strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.3.1 Small Strain Definition . . . . . . . . . . . . . . . . . . . . . . . . 34
2.3.2 Small Strain and Small Rotation Formulation . . . . . . . . 38
2.3.3 Small Strain and Rotation in 3-D . . . . . . . . . . . . . . . . . 47
2.4 Kinematics of Continuous Medium . . . . . . . . . . . . . . . . . . . . . 48
2.4.1 Material (Local) Description . . . . . . . . . . . . . . . . . . . . 48
2.4.2 Referential Description (Lagrangian Description) . . . . . 48
2.4.3 Spatial Description (Eulerian Description) . . . . . . . . . . 49
2.4.4 Material Time Derivative in Spatial Coordinates
(Substantial Derivative) . . . . . . . . . . . . . . . . . . . . . . . 49
ix
x Contents
2.5 Rate of Deformation Tensor and Rate of Spin Tensor . . . . . . . . 53

2.5.1 Comparison of Rate of Deformation Tensor, D,
and Time Derivative of the Strain Tensor, ε_ . . . . . . . . . 54
2.5.2 True Strain (Natural Strain) (Logarithmic Strain) . . . . . 55
2.6 Finite Strain and Deformation . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.6.1 Green Deformation Tensor, C, Cauchy
Deformation Tensor, B-1 . . . . . . . . . . . . . . . . . . . . . . 63
2.6.2 Relation Between Deformation, Strain,
and Deformation–Gradient Tensors . . . . . . . . . . . . . . . 64
2.6.3 Comparing Small Strain and Large (Finite)
Strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.6.4 Strain Rate and Rate of Deformation Tensor
Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
2.6.5 Relation Between, the Spatial Gradient of Velocity
Tensor, L and the Deformation–Gradient
Tensor, F . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2.7 Rotation and Stretch Tensors in Finite Strain . . . . . . . . . . . . . . . 72
2.8 Compatibility Conditions in Continuum Mechanics . . . . . . . . . . 73
2.9 Piola–Kirchhoff Stress Tensors . . . . . . . . . . . . . . . . . . . . . . . . 76
2.9.1 First Piola–Kirchhoff Stress Tensor σ 0 . . . . . . . . . . . . . 76
2.9.2 Second Piola–Kirchhoff Stress Tensor σ~ . . . . . . . . . . . 77
2.10 Direct Relation Between Cauchy Stress Tensor
and Piola–Kirchhoff Stress Tensors . . . . . . . . . . . . . . . . . . . . . 78
2.11 Conservation of Mass Principle . . . . . . . . . . . . . . . . . . . . . . . . 79
2.12 The Incompressible Materials . . . . . . . . . . . . . . . . . . . . . . . . . 80
2.13 Conservation of Momentum Principle . . . . . . . . . . . . . . . . . . . 81
2.14 Conservation of Moment of Momentum Principle . . . . . . . . . . . 84
2.15 Lagrangian Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
2.16 Hamilton’s Principle: The Principle of Stationary Action . . . . . . 88
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
3 Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.1 Thermodynamic Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.2 First Law of Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.2.1 Work Done on the System (Power Input) . . . . . . . . . . . 94
3.2.2 Heat Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.3 Second Law of Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . 98
3.3.1 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
3.3.2 Quantification of Entropy in Thermodynamics . . . . . . . 103
3.3.3 Gibbs–Duhem Relation . . . . . . . . . . . . . . . . . . . . . . . . 104
3.3.4 Euler Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3.3.5 Entropy Production in Irreversible Process . . . . . . . . . . 108
3.3.6 Clausius–Duhem Inequality . . . . . . . . . . . . . . . . . . . . 109
3.3.7 Traditional Use of Entropy as a Functional
in Continuum Mechanics . . . . . . . . . . . . . . . . . . . . . . 112
3.3.8 Entropy as a Measure of Disorder . . . . . . . . . . . . . . . . 114
Contents xi
3.3.9 Thermodynamic Potential . . . . . . . . . . . . . . . . . . . . . . 118

3.3.10 Time-Independent Dissipation
(Instantaneous Dissipation) . . . . . . . . . . . . . . . . . . . . . 127
3.3.11 Dissipation Power and Onsager Reciprocal
Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
4 Unified Mechanics Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
4.1 Literature Review of Use of Thermodynamics
in Continuum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
4.2 Laws of Unified Mechanics Theory . . . . . . . . . . . . . . . . . . . . . 150
4.2.1 Second Law of Unified Mechanics Theory . . . . . . . . . . 151
4.2.2 Third Law of Unified Mechanics Theory . . . . . . . . . . . 153
4.3 Evolution of Thermodynamic State Index (Φ) . . . . . . . . . . . . . . 153
4.3.1 On the Relationship Between the Second
Fundamental Theorem of The Mechanical Theory
of Heat and Probability Calculations Regarding
the Conditions For Thermal Equilibrium,
by Ludwig Boltzmann (1877) . . . . . . . . . . . . . . . . . . . 156
4.3.2 Critic of Boltzmann’s Mathematical Derivation . . . . . . 199
4.3.3 On the Law of Distribution of Energy
in the Normal Spectrum, [Max Planck, (1901)
Annalen Der Physik, Vol. 4, p. 553 Ff, 1901] . . . . . . . . 201
4.3.4 Thermodynamic State Index (TSI) in Unified
Mechanics Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
4.4 Experimental Verification Example . . . . . . . . . . . . . . . . . . . . . 215
4.4.1 Tension-Compression Cyclic Loading . . . . . . . . . . . . . 215
4.4.2 Monotonic Loading Test . . . . . . . . . . . . . . . . . . . . . . . 217
4.5 Dynamic Equilibrium Equations in Unified Mechanics
Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
4.5.1 Derivation of the Dynamic Equilibrium
Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
4.5.2 Vibration of a 1-DOF Mass-Spring System . . . . . . . . . 221
4.5.3 Entropy Generation in Sliding Friction Contact:
Thermodynamic Fundamental Equation . . . . . . . . . . . . 224
4.5.4 Comparison with Newtonian Mechanics Results . . . . . . 226
4.5.5 Comparison with Experimental Data . . . . . . . . . . . . . . 229
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
5 Unified Mechanics of Thermomechanical Analysis . . . . . . . . . . . . . 237
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
5.2 Unified Mechanics Theory-Based Constitutive Modeling . . . . . . 237
5.2.1 Flow Theory and Yield Criteria . . . . . . . . . . . . . . . . . . 237
5.2.2 Effective Stress Concept and Strain Equivalence
Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
5.2.3 Unified Mechanics Theory Implementation . . . . . . . . . 243
xii Contents
5.3 Return Mapping Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . 243

5.3.1 Linearization (Consistent Jacobian) . . . . . . . . . . . . . . . 246
5.4 Thermodynamic Fundamental Equation
in Thermomechanical Problems . . . . . . . . . . . . . . . . . . . . . . . . 249
5.5 Conservation Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
5.5.1 Conservation of Mass . . . . . . . . . . . . . . . . . . . . . . . . . 251
5.5.2 Momentum Principle in Newtonian Mechanics . . . . . . . 253
5.5.3 Conservation of Energy . . . . . . . . . . . . . . . . . . . . . . . 255
5.6 Entropy Law: Second Law of Thermodynamics . . . . . . . . . . . . 257
5.7 Fully Coupled Thermomechanical Equations . . . . . . . . . . . . . . 263
5.8 Numerical Validation of the Thermomechanical
Constitutive Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
5.8.1 Thin Layer Solder Joint: Monotonic and Fatigue
Shear Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
5.9 Thermomechanical Analysis of Cosserat Continuum:
Length Scale Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
5.9.1 Cosserat Couple Stress Theory . . . . . . . . . . . . . . . . . . 274
5.9.2 Toupin-Mindlin Higher-Order Stress Theory . . . . . . . . 276
5.9.3 Equilibrium Equations . . . . . . . . . . . . . . . . . . . . . . . . 278
5.9.4 Finite Element Method Implementation . . . . . . . . . . . . 281
5.9.5 General Couple Stress Theory: Variational
Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
5.9.6 Reduced Couple Stress Theory: Variational
Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
5.9.7 Reduced Couple Stress Theory: Mixed
Variational Principle . . . . . . . . . . . . . . . . . . . . . . . . . . 284
5.9.8 General Couple Stress Theory Implementation . . . . . . . 285
5.10 Cosserat Continuum Implementation in Unified Mechanics
Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
5.10.1 Rate-Independent Plasticity Without Degradation . . . . . 289
5.10.2 Rate-Dependent Plasticity (Viscoplasticity)
Without Degradation . . . . . . . . . . . . . . . . . . . . . . . . . 295
5.10.3 Introducing the Thermodynamic State Index . . . . . . . . 296
5.10.4 Entropy Generation Rate in Cosserat Continuum . . . . . 298
5.10.5 Integration Algorithms . . . . . . . . . . . . . . . . . . . . . . . . 299
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
6 Thermomechanical Analysis of Particle-Filled Composites . . . . . . . 309
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
6.2 Ensemble-Volume Averaged Micromechanical
Field Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
6.3 Noninteracting Solution for Two-Phase Composites . . . . . . . . . 316
6.3.1 Average Stress Norm in Matrix . . . . . . . . . . . . . . . . . . 316
6.3.2 Average Stress in Filler Particles . . . . . . . . . . . . . . . . . 322
Contents xiii
6.4 Pairwise Interacting Solution for Two-Phase Composites . . . . . . 323

6.4.1 Approximate Solution of Two-Phase Interaction . . . . . . 323
6.4.2 Ensemble-Average Stress Norm in the Matrix . . . . . . . 327
6.4.3 Ensemble-Average Stress in the Filler Particles . . . . . . 329
6.5 Noninteracting Solution for Three-Phase Composites . . . . . . . . 331
6.5.1 Effective Elastic Modulus of Multiphase
Composites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332
6.5.2 Ensemble-Average Stress Norm in the Matrix . . . . . . . 332
6.5.3 Ensemble-Average Stress in Filler Particles . . . . . . . . . 336
6.6 Effective Thermomechanical Properties . . . . . . . . . . . . . . . . . . 338
6.6.1 Effective Bulk Modulus . . . . . . . . . . . . . . . . . . . . . . . 339
6.6.2 Effective Coefficient of Thermal Expansion
(ECTE) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
6.6.3 Effective Shear Modulus . . . . . . . . . . . . . . . . . . . . . . . 340
6.6.4 Effective Young’s Modulus and Effective
Poisson’s Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
6.6.5 Numerical Examples . . . . . . . . . . . . . . . . . . . . . . . . . 342
6.7 Micromechanical Constitutive Model of the Particulate
Composite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
6.7.1 Modeling Procedures for Particulate Composites . . . . . 346
6.7.2 Elastic Properties of Particulate Composites . . . . . . . . . 348
6.7.3 A Viscoplasticity Model . . . . . . . . . . . . . . . . . . . . . . . 350
6.7.4 Thermodynamic State Index . . . . . . . . . . . . . . . . . . . . 354
6.7.5 Solution Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
6.7.6 Consistent Elastic-Viscoplastic Tangent Modulus . . . . . 359
6.8 Verification Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
6.8.1 Material Properties of ATH . . . . . . . . . . . . . . . . . . . . . 361
6.8.2 Properties of PMMA . . . . . . . . . . . . . . . . . . . . . . . . . 362
6.8.3 Properties of Matrix-Filler Interphase . . . . . . . . . . . . . . 362
6.8.4 Properties of Particulate Composites . . . . . . . . . . . . . . 362
6.8.5 Finite Element Simulation Results . . . . . . . . . . . . . . . . 366
6.8.6 Cyclic Stress-Strain Response . . . . . . . . . . . . . . . . . . . 366
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
7 Unified Micromechanics of Finite Deformations . . . . . . . . . . . . . . . 375
7.1 Introduction to Finite Deformations . . . . . . . . . . . . . . . . . . . . . 375
7.2 Frame of Reference Indifference . . . . . . . . . . . . . . . . . . . . . . . 379
7.3 Unified Mechanics Theory Formulation for Finite Strain . . . . . . 383
7.3.1 Thermodynamic Restrictions . . . . . . . . . . . . . . . . . . . . 384
7.3.2 Constitutive Relations . . . . . . . . . . . . . . . . . . . . . . . . . 392
7.4 Thermodynamics State Index . . . . . . . . . . . . . . . . . . . . . . . . . . 403
7.5 Definition of Material Properties . . . . . . . . . . . . . . . . . . . . . . . 405
7.6 Applications of Finite Deformation Models . . . . . . . . . . . . . . . 409
7.6.1 Material Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . 409
xiv Contents
7.7 Numerical Implementation of Dual-Mechanism Model . . . . . . . 415

7.7.1 Simulating Isothermal Stretching of PMMA . . . . . . . . . 416
7.7.2 Simulating Non-isothermal Stretching of PMMA . . . . . 422
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
8 Unified Mechanics of Metals Under High Electrical Current
Density: Electromigration and Thermomigration . . . . . . . . . . . . . . 427
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
8.2 Physics of Electromigration Process . . . . . . . . . . . . . . . . . . . . . 427
8.2.1 Driving Forces of Electromigration Process . . . . . . . . . 428
8.2.2 Laws Governing Electromigration
and Thermomigration . . . . . . . . . . . . . . . . . . . . . . . . . 429
8.2.3 Electromigration Electron Wind Force . . . . . . . . . . . . . 432
8.2.4 Temperature Gradient Diffusion Driving Force . . . . . . . 435
8.2.5 Stress Gradient Diffusion Driving Force . . . . . . . . . . . 436
8.3 Laws of Conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 438
8.3.1 Vacancy Conservation . . . . . . . . . . . . . . . . . . . . . . . . 439
8.4 Newtonian Mechanics Force Equilibrium . . . . . . . . . . . . . . . . . 443
8.5 Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444
8.6 Electrical Conduction Equations . . . . . . . . . . . . . . . . . . . . . . . . 447
8.7 Thermodynamic Fundamental Equation for Electromigration
and Thermomigration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449
8.7.1 Entropy Balance Equations . . . . . . . . . . . . . . . . . . . . . 450
8.8 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 456
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 457
9 Fatigue of Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459
9.1 Predicting High Cycle Fatigue Life of Metals . . . . . . . . . . . . . . 460
9.1.1 Thermodynamic Fundamental Equation . . . . . . . . . . . . 460
9.1.2 Entropy Generation Mechanisms . . . . . . . . . . . . . . . . . 461
9.1.3 Comparison Between Different Entropy
Generation Mechanisms . . . . . . . . . . . . . . . . . . . . . . . 473
9.1.4 Temperature Evolution due to Mechanical Work . . . . . 475
9.1.5 Entropy and TSI Calculations . . . . . . . . . . . . . . . . . . . 477
9.2 Predicting Ultrasonic Vibration Fatigue Life . . . . . . . . . . . . . . . 480
9.2.1 Thermodynamic Fundamental Equation . . . . . . . . . . . . 481
9.2.2 Temperature Evolution due to Mechanical Work . . . . . 489
9.2.3 Comparison Between Different Entropy
Generation Mechanisms . . . . . . . . . . . . . . . . . . . . . . . 492
9.2.4 Computing Thermodynamic State Index . . . . . . . . . . . 494
9.3 Predicting Low Cycle Fatigue Life . . . . . . . . . . . . . . . . . . . . . . 496
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 501
Contents xv
10 Corrosion-Fatigue Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505

10.1 Corrosion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
of Corrosion-Fatigue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 506
10.2.1 Entropy Generation Mechanisms During
Corrosion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 506
10.2.2 Entropy Generation Mechanisms: Fatigue . . . . . . . . . . 511
10.3 Thermodynamic State Index (TSI) . . . . . . . . . . . . . . . . . . . . . . 513
10.4 Comparing Simulation Results and Test Data . . . . . . . . . . . . . . 513
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517
Abbreviations1
γ Shear strain
δij Kronecker delta
ε Total strain tensor
W Vorticity tensor
ε_ p Plastic strain rate tensor
ρ Mass density
σ Total stress tensor
S Deviatoric stress tensor
σO First Piola-Kirchhoff stress tensor
σ~ Second Piola-Kirchhoff tensor
φ Dissipation potential
Φ Thermodynamic state index
TSI Thermodynamic state index
Ψ Helmholtz-free energy
b Body force per unit mass
B-1 Cauchy deformation tensor
C Green deformation tensor
D Rate of deformation tensor
E Lagrange strain tensor
E Euler strain tensor
ETotal Total energy
F Deformation gradient tensor
F Plasticity yield surface
F~ Second Piola-Kirchhoff force
g Gibbs function
1
Most of the time I tried to use the same character to define a variable. However, due to the large
number of variables, some characters were used more than once to define different variables. To
clarify the variables, they are re-defined after each equation.
xvii
xviii Abbreviations
h Enthalpy
I Impulse
I 1, I 2, I 3 Total stress tensor invariants
J2D Second invariant of the deviatoric stress tensor
J3D Third invariant of the deviatoric stress tensor
k, or kB Boltzmann’s constant
K Kinetic energy
V Potential energy
L Spatial gradient of velocity
m mass
mij the distributed moment in couple stress theory
nij Direction cosines
p Momentum
q Heat flux vector
Q Plastic potential
Q Total heat input
s Entropy
t Time
u, v, w Displacements along spatial coordinates, x, y, z, respectively
i, j, k Unit vectors along spatial coordinates, x, y, z, respectively
U Strain energy
~v Velocity
A Area
V Volume
Chapter 1
Introduction
1.1 What Is the Mechanics of Continuous Medium?
Probably one of the best definitions of mechanics of a continuous medium is

provided by Malvern (1969) in his seminal work. “Continuum mechanics is a branch
of mechanics concerned with the deformation or flow of solids, liquids, and gasses.”
In continuum mechanics, molecular structure and electronic structure of the material
are ignored. When molecular structure must be taken into account molecular
dynamics and when the electronic structure must be considered quantum mechanics
are used. In continuum mechanics, it is assumed that the material is continuous
without empty spaces. It is also assumed that all differential equations governing the
continuous medium are continuous functions, except at boundaries between contin-
uous regions. It is also assumed that the derivatives of the differential equations are
continuous as well. A material satisfying these requirements is considered a contin-
uous medium. It is important to point out that “space” does not mean there cannot be
atomic vacancies. There will always be atomic vacancies. It just means that there
cannot be a space proportional to the size of the object being studied.
The concept of a continuous medium allows us to define stress and strain at an
“imaginary point,” a geometric point in space assumed to be occupying no volume.
When we say a point in continuum mechanics, we do not mean an atomic point. It is
an imaginary point with no volume. Continuum stresses and strains are defined at
this point. Atomic stresses are defined differently, which is outside the scope of this
book. This approach allows us to use differential calculus to study non-uniform
distributions of strain. This assumption of the continuous medium allows us to study
deformation in most engineering problems but not all. When the continuous medium
assumption is not satisfied or molecular dynamics [movement of atoms] or electron-
ics structure influences the mechanical behavior, then continuum mechanics cannot
be used. In these latter instances, a multi-scale mechanics analysis becomes
essential.
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 1

C. Basaran, Introduction to Unified Mechanics Theory with Applications,
https://doi.org/10.1007/978-3-031-18621-9_1
2 1 Introduction
The traditional definition of the continuous medium also makes three assump-
tions (Malvern, 1969):
Continuity A material is continuous if it fills the space that it occupies, leaving no
pores or empty spaces, and if furthermore, its properties are describable by contin-
uous functions.
Homogeneity A homogenous material has identical properties at all points. The
size is larger than the Representative Volume Element (RVE) of the material. RVE is
the smallest volume over which a property measurement can be made that will yield
a value representative of the whole (Hill, 1965).
Isotropy A material is isotropic with respect to certain properties if these properties
are the same in all directions in space.
Only the first assumption is needed to define concepts of stress and strain. The last
two assumptions are needed when we introduce stress-strain relations (constitutive
relations) into continuum mechanics equations.
Of course, anisotropic, inhomogeneous, and discontinuous systems can also be
analyzed by continuum mechanics by representing them as discrete pieces, as it is
done in computational mechanics, such as the finite element method.
Chapter 2 of this book covers only the general principles of continuum mechan-
ics, and associated definitions, such as stress and strain. The fundamentals of
continuum mechanics, presented in Chap. 2, have been well established by Isaac
Newton (1687), Leonhard Euler (1736), Cauchy, Green, Truesdell, and Toupin
(1960), Truesdell and Noll (1965), Truesdell (1965, 1966), and others. Maugin
(2016) published a comprehensive account of the foundations of continuum mechan-
ics as well as a complete list of references to earlier work. Therefore, there is no
attempt in this book to cover earlier fundamental developments.
Continuum mechanics cannot be discussed without including material constitu-
tive equations. While material behavior modeling is a very important topic of major
current research efforts, it is not the focus of this book. It is covered in the
applications of the unified mechanics theory sections, Chaps. 5, 6, 7, 8, and 9.
Chapter 3 of the book covers the basics of thermodynamics. This chapter starts
with a comprehensive literature survey of the use of thermodynamics in continuum
mechanics in the last 150 years. While we do not claim to have included every paper
published on the topic, the survey covers all significant developments. Energy
dissipation and degradation of materials and structures are discussed in the context
of thermodynamics. Since not all engineering students are required to take the course
in thermodynamics, the reader has assumed a beginner, as such a piece of elementary
information is included.
Chapter 4 presents the formulation of the unified mechanics theory. Great effort
has been made to include all the details of the formulation. Chapter 4 also discusses
the motivation behind the development of the unified mechanics theory. Essentially,
the present-day mechanics equations are all based on the three laws of motion of
Isaac Newton. However, these three laws are “incomplete,” because they do not
include the laws of thermodynamics. The following examples can help us explain
References 3
what we mean by “incomplete.” For example, if a soccer ball is given an initial

acceleration by a kick according to Newton’s second law, that acceleration is
constant; it does not degrade. The ball will travel with that acceleration forever
according to F = ma. However, that is not true. In the same fashion, Newton’s third
law (also Hooke’s law, which was published by Robert Hooke 10 years before
Newton’s Principia) assumes that reaction due to an applied load will be constant,
without degradation of the material. Of course, materials degrade, and response
displacement does not remain constant. Fortunately, energy loss in the ball and
material degradation can be modeled by the laws of thermodynamics. Basaran and
Yan published the first paper in (1998) using Boltzmann’s entropy generation rate as
a metric for degradation in microelectronics solder joints. They implemented it by
modifying Newton’s laws with Boltzmann’s second law of thermodynamics formu-
lation. Since then, the theory has gone through significant consolidation by exper-
imental verifications and mathematical derivations by many researchers around the
world. These developments are included in Chap. 4.
Chapter 5 presents a formulation for thermo-mechanical analysis using unified
mechanics theory. The chapter includes non-local and local mechanics formulations.
Cosserat continuum is explained in great detail and used for introducing length scale
into the continuum mechanics formulation. It is presented with simple steps from a
beginner graduate student’s perspective
Chapter 6 is about particle-filled composite materials’ formulation using the
unified mechanics theory. Particle-filled acrylic composite formulation and imple-
mentation are presented in the context of small strain formulation.
Chapter 7 covers the finite strain formulation of unified mechanics theory.
Constitutive modeling of polymer is presented in detail.
Chapter 8 presents the electro-thermo-mechanical loading application of the
unified mechanics theory. Electromigration and thermomigration are covered in
detail.
Chapter 9 covers the derivation of thermodynamic fundamental equations for
fatigue in metals with examples.
Chapter 10 includes a discussion on corrosion-fatigue interaction and derivation
of the fundamental equation for corrosion in metals.
I provided a list of references at the end of each chapter. While they are not
complete, they can help a researcher in the right direction.
References
Hill, R. (1965). Continuum micro-mechanics of elastoplastic polycrystals. Journal of the Mechan-

ics and Physics of Solids, 13(2), 89–101.
Malvern, L. E. (1969). Introduction to the mechanics of continuous medium. Prentice-Hall.
Basaran, C., & Yan, C. Y. (1998). A thermodynamic framework for damage mechanics of solder
joints. Transaction of ASME, Journal of Electronic Packaging, 120, 379–384.
Maugin, G. A. (2016). Continuum mechanics through the ages - from the renaissance to the
twentieth century. Springer.
4 1 Introduction
Newton, I. (1687). Philosophiæ Naturalis Principia Mathematica, .

Euler, L. (1736). Mechanica sive motus scientia analytice exposita
Trusdell, C., & Toupin, R. (1960). The classical field theories. In S. Flügge (Ed.), Principles of
classical mechanics and field theory / Prinzipien der Klassischen Mechanik und Feldtheorie.
Springer.
Truesdell, C., & Walter Noll, W. (1965). The non-linear field theories of mechanics. Springer.
Truesdell, C. (1965). The elements of continuum mechanics. Springer-Verlag.
Truesdell, C. (1966). Rational mechanics of materials. In Six lectures on modern natural philoso-
phy. Springer.
Chapter 2
Stress and Strain in Continuum
2.1 Newton’s Universal Laws of Motion
It is essential to start the discussion about continuum mechanics, with the universal
laws of motion as given by Isaac Newton. Understanding their implications in
historical context is essential to understanding these simple and probably the most
famous physics equations. Of course, Newton’s universal laws of motion are very
well known to most readers of this book; however, their historical context and
evolution are not. Therefore, we find it necessary to quote these basics from Sir
Isaac Newton’s own Wikipedia website with their sources. The information below is
quoted directly with little or no editing.
Philosophiæ Naturalis Principia Mathematica (Latin for Mathematical Principles
of Natural Philosophy), often referred to as simply the Principia, is a work in three
books by Isaac Newton, in Latin, first published 5 July 1687. After annotating and
correcting his copy of the first edition, Newton published two further editions, in
1713 and 1726 (Harvard UP, 1972). The Principia states Newton’s universal laws of
motion, forming the foundation of classical mechanics; Newton’s law of universal
gravitation; and a derivation of Kepler’s laws of planetary motion (which Kepler first
obtained empirically) (Cohen, 2002; Galili & Tseitlin, 2003; Kleppner & Kolenkow,
1973; Plastino & Muzzio, 1992; Resnick et al., 1992) (Fig. 2.1).
2.1.1 First Universal Law of Motion
The original text in Latin reads:

Newton’s original Latin text in Principia states:
Lex I: Corpus omne perseverare in statu suo quiescendi vel movendi uniformiter in directum,
nisi quatenus a viribus impressis cogitur statum illum mutare.

https://doi.org/10.1007/978-3-031-18621-9_2
6 2 Stress and Strain in Continuum
Fig. 2.1 Original 1687

Principia Mathematica,
where Sir Isaac Newton
published the three laws, in
Latin
Translated to English, this reads:

Law I: Everybody persists in its state of being at rest or of moving uniformly straight
forward, except insofar as it is compelled to change its state by force impressed.
(Modern Translation) Law I: In an inertial frame of reference, an object either

remains at rest or continues to move at a constant velocity, unless acted upon by an
external force.
The Greek philosopher Aristotle (384–322 BC) articulated that all objects have a
natural place in the universe. He believed that heavy objects like boulders wanted to
be at rest on the Earth and that light objects like smoke wanted to be at rest in the sky
and the stars wanted to remain in the heavens. He also believed that a body was in its
natural state when it was at rest, and for the body to move in a straight line at a
constant speed an external force was needed continually to propel it; otherwise, it
would stop moving. Galileo Galilei (1564–1642) however, realized that a force is
necessary to change the velocity of a body, i.e., acceleration, but no force is needed
to maintain its velocity. In other words, Galileo stated that, in the absence of a force,
a moving object will continue moving. The tendency of objects to resist changes in
motion was what Johannes Kepler (1571–1630) had called inertia. Kepler’s insight
was refined by Newton, who made it into his first law, also known as the “law of
inertia”—no force means no acceleration, and hence the body will maintain its
velocity. Newton’s first law is a restatement of the law of inertia which Galileo
had already described; hence, Newton appropriately gave credit to Galileo.
2.1 Newton’s Universal Laws of Motion 7
The law of inertia was discovered by different philosophers and scientists inde-
pendently, including Thomas Hobbes (1588–1679) in his Leviathan (Hobbes, 1651).
The seventeenth-century philosopher and mathematician René Descartes
(1596–1650) also formulated the law, although he did not perform any experiments
to confirm it (Hellingman, 1992; Resnick & Halliday, 1977).
2.1.1.1 Formulation of the First Law
The first law, also known as the law of inertia, states that if the net force (the vector
sum of all forces acting on an object) is zero, then the velocity of the object is
constant. Velocity is a vector quantity, which expresses both the object’s speed and
the direction of its motion; therefore, the statement that the object’s velocity is
constant is a statement that both its speed and the direction of its motion are constant.
The first law can be stated mathematically, when the mass is a non-zero constant,
as,
d~v
F = 0, =0
dt
where F is the total net force acting on the object, d~v=dt is the time derivative of
velocity. Consequently, an object that is at rest will stay at rest unless an external
force acts upon it. An object that is in motion will not change its velocity unless an
external force acts upon it. This is known as the uniform motion. An object continues
to do whatever it happens to be doing unless a force is exerted upon it. If it is at rest, it
continues in a state of rest (demonstrated when a tablecloth is skillfully whipped
from under dishes on a tabletop and the dishes remain in their initial state of rest),
where the inertia of dishes balances the force exerted by the table cloth. If an object is
moving, it continues to move without turning or changing its speed. This is evident
in space probes that continuously move in outer space. Change in motion must be
imposed against the tendency of an object to retain its state of motion. In the absence
of a total net force, a moving object tends to move along a straight-line path
indefinitely.
Newton stated the first law of motion to establish a frame of reference for which
the other laws are applicable. The first law of motion postulates the existence of at
least one frame of reference called a Newtonian or inertial frame of reference,
relative to which the motion of a particle, not subjected to a net force, is a straight
line at a constant speed (Newton, 1999). Newton’s first law is often referred to as the
law of inertia. Thus, a condition necessary for the uniform motion of a particle
relative to an inertial frame of reference is that the total net force acting on it is zero.
In this sense, the first law can be restated as (Motte’s 1729 translation):
In every material universe, the motion of a particle in a preferential reference frame X is
determined by the action of forces whose total vanished for all times when and only when the
velocity of the particle is constant in X. That is, a particle initially at rest or in uniform
motion in the preferential frame X continues in that state unless compelled by forces to
change it.
“Newton’s first and second laws are valid only in an inertial reference frame. Any
reference frame that is in uniform motion with respect to an inertial frame is also
inertial, i.e., Galilean invariance or the principle of Newtonian relativity” (Resnick &
Halliday, 1977). If someone kicks a ball on a moving train car, the acceleration the
ball will gain will be with respect to the train not with respect to the ground.
2.1.2 Second Universal Law of Motion
Isaac Newton’s original Latin text reads:

Lex II: Mutationem motus proportionalem esse vi motrici impressae, et fieri secundum
lineam rectam qua vis illa imprimitur.
This was translated by Motte in 1729 as:

Law II: The alteration of motion is ever proportional to the motive force impress’d; and is
made in the direction of the right line in which that force is impress’d.
(Modern translation) Law II: In an inertial reference frame, the vector sum of the
forces “F” on an object is equal to the mass “m” of that object multiplied by the
acceleration “a” of the object:
According to modern ideas of how Newton was using his terminology, this is
understood, in modern terms, as an equivalent of:
The change of momentum, ½d ðm~vÞ, of a body, is proportional to the impulse (Fdt) impressed
on the body and happens along the straight line on which that impulse (Fdt) is impressed.
This may be expressed by the formula F = p′ where p′ is the time derivative of the
momentum ½p = d ðm~vÞ. This equation can be seen in the Wren Library of Trinity
College, Cambridge, UK, in a glass case in which Newton’s manuscript is open to
the relevant page.
Motte’s 1729 translation of Newton’s Latin manuscript continued with Newton’s
commentary on the second law of motion, reading:
If a force generates a motion, a double force will generate double the motion, a triple force
triple the motion, whether that force be impressed altogether and at once, or gradually and
successively. And this motion (being always directed the same way with the generating
force), if the body moved before, is added to or subtracted from the former motion, according
to as they directly conspire with or are directly contrary to each other; or obliquely joined,
when they are oblique, so as to produce a new motion compounded from the determination
of both.
The scientific intention in which Newton used his terminology, and how he
defined the second law and intended it to be understood, has been extensively
studied by historians of science.
2.1.2.1 Formulation of Newton’s Second Law
The second law states that the rate of change of momentum of a body is directly
proportional to the force applied, and this change in dp
dt momentum takes place in the
direction of the applied force, F,
dp dðm~vÞ
F= =
dt dt
The second law can also be stated in terms of an object’s acceleration. It is

important to note that a general expression for Newton’s second law for variable
mass systems by treating the mass as a variable cannot be derived. However, F = dp dt
can be used to analyze variable mass systems only if we apply it to an entire system
of constant mass having parts among which there is an interchange of mass. Since
Newton’s second law is valid only for constant-mass systems (Beatty, 2006; Thorn-
ton, 2004; Hellingman, 1992), mass, m, can be taken outside the differentiation
operator by the constant factor rule in differentiation. Thus,
d~v
F=m = ma
dt
An impulse is defined by a force F acting on an object over an interval of time Δt

and can be expressed by
Z
I= Fdt
Δt
Since force is the time derivative of momentum, it follows that,
I = Δp = mΔ~v
this relation between impulse and momentum is closer to Newton’s definition of the
second law.
2.1.3 Third Universal Law of Motion
Newton’s original Latin text reads:

Lex III: Actioni contrariam semper et æqualem esse reactionem: sive corporum duorum
actiones in se mutuo semper esse æquales et in partes contrarias dirigi.
Direct translation to English, this reads:

Law III: To every action, there is always opposed an equal reaction: or the mutual actions of
two bodies upon each other are always equal and directed to contrary parts.
(Modern translation) Law III: When one body exerts a force on a second body
F12 = F21, the second body simultaneously exerts a force F21 equal in magnitude and
opposite in direction to the first body.
Newton’s explanatory comment to this law:
Whatever draws or presses another is as much drawn or pressed by that other. If you press a
stone with your finger, the finger is also pressed by the stone. If a horse draws a stone tied to
a rope, the horse (if I may so say) will be equally drawn back towards the stone: for the
distended rope, by the same endeavor to relax or unbend itself, will draw the horse as much
towards the stone, as it does the stone towards the horse, and will obstruct the progress of the
one as much as it advances that of the other. If a body impinges upon another, and by its
force changes the motion of the other, that body also (because of the equality of the mutual
pressure) will undergo an equal change, in its own motion, toward the contrary part. The
changes made by these actions are equal, not in the velocities but in the motions of the
bodies; that is to say if the bodies are not hindered by any other impediments. For, as the
motions are equally changed, the changes of the velocities made toward contrary parts are
reciprocally proportional to the bodies. This law takes place also in attractions, as will be
proved in the next scholium (Fairlie & Cayley, 1965).
F 12 = F 21
In the above, as usual, motion is Newton’s term for momentum, hence his careful
distinction between motion and velocity.
Newton used the third law to derive the law of conservation of momentum
(Cohen, 1995), from a deeper perspective; however, conservation of momentum is
the more fundamental idea (derived via Noether’s theorem from Galilean invari-
ance), and holds in cases where Newton’s third law appears to fail, for instance,
when force fields, as well as particles, carry momentum, and in quantum mechanics.
An example illustration of Newton’s third law is in which two people push
against each other. The first person on the left exerts a normal force F12 on the
second person directed towards the right, and the second person exerts a normal
force F21 on the first person directed towards the left. The magnitudes of both forces
are equal, but they have opposite directions, as dictated by Newton’s third law.
The third law states that all forces between two objects exist in equal magnitude
and opposite direction: if one object A exerts a force FA on a second object B, then B
simultaneously exerts a force FB on A, and the two forces are equal in magnitude and
opposite in direction: FA = - FB. The third law means that all forces are interactions
between different bodies, or different regions within one body, and thus that there is
no such thing as a force that is not accompanied by an equal and opposite force. In
some situations, the magnitude and direction of the forces are determined entirely by
one of the two bodies, say Body A; the force exerted by Body A on Body B is called
the “action,” and the force exerted by Body B on Body A is called the “reaction.”
This law is sometimes referred to as the action–reaction law, with FA called the
“action” and FB the “reaction.” In other situations, the magnitude and directions of
the forces are determined jointly by both bodies, and it is not necessary to identify
one force as the “action” and the other as the “reaction.” The action and the reaction
are simultaneous, and it does not matter which is called the action and which is called
the reaction; both forces are part of a single interaction, and neither force exists
without the other (Cohen, 1967).
The two forces in Newton’s third law are of the same type (e.g., if the road exerts
a forward frictional force on an accelerating car’s tires, then it is also a frictional
force that Newton’s third law predicts for the tires pushing backward on the road).
From a conceptual standpoint, Newton’s third law is seen when a person walks:
they push against the floor, and the floor pushes against the person. Similarly, the
tires of a car apply a shear force on the road while the road pushes back on the tires
with a re-action shear force—the tires and road simultaneously push against each
other. The reaction force is responsible for the motion of the person or the car. These
forces depend on friction; however, if a person or car is on ice, there will be no
motion because of failure to produce the needed reaction force.
It is important to point out that Robert Hooke published his law in 1676. Hooke’s
law which is more sophisticated than Newton’s third law, for deformable bodies as
F = ku where k is the stiffness of a one-dimensional spring and u is the displacement
response in the direction of the force. It is easy to see the action–reaction concept in
Hooke’s law. “Hooke stated in his 1678 work that he was aware of the law in 1660,
which is 17 years before Newton published his Principia in 1687” (Robert Hooke’s
Wikipedia page).
In their original form, Newton’s laws of motion are not adequate to characterize the motion
of rigid bodies and deformable bodies. Leonhard Euler in 1750 introduced a generalization
of Newton’s laws of motion for rigid bodies called Euler’s laws of motion, later applied as
well for deformable bodies assumed as a continuum. If a body is represented as an
assemblage of discrete particles, each governed by Newton’s laws of motion, then Euler’s
laws can be derived from Newton’s laws. Euler’s laws can, however, be taken as axioms
describing the laws of motion for extended bodies, independently of any particle structure
(Lubliner, 2008; Leonard Euler’s Wikipedia page).
2.1.4 Range of Validity of Newton’s Universal Laws

of Motion
Newton’s laws were verified by experiments and observations for over 200 years, and they
are excellent approximations at the scales and speeds of most of the motions we observe in
daily life. Newton’s laws of motion, together with his law of universal gravitation and the
mathematical techniques of calculus, provided for the first time a unified quantitative
explanation for a wide range of physical phenomena.
These three laws hold for macroscopic objects under everyday conditions. However,
Newton’s laws are inappropriate for use in certain circumstances, most notably, at very high
speeds (in special relativity, the Lorentz factor must be included in the expression for
momentum along with the rest mass and velocity) or in very strong gravitational fields.
Therefore, Newton’s laws cannot be used to explain phenomena such as conduction of
electricity in a semiconductor, optical properties of materials, errors in non-relativistically
corrected GPS systems, and superconductivity. Explanation of these phenomena requires
more sophisticated physical theories, including general relativity and quantum field theory.
However, in quantum mechanics, concepts such as force, momentum, and position are
defined by linear operators that operate on the quantum state; at speeds that are much lower
than the speed of light, Newton’s laws are just as exact for these operators as they are for
classical objects.
2.1.5 Relation to the Thermodynamics

and Conservation Laws
In modern physics, the laws of conservation of momentum, energy, and angular momentum
are of more general validity than Newton’s laws, since they apply to both light and matter,
and both classical and non-classical physics. Because force is the time derivative of
momentum, the concept of force is redundant and subordinate to the conservation of
momentum, and is not used in fundamental theories (e.g., quantum mechanics, quantum
electrodynamics, general relativity, etc.). Other forces, such as gravity, also arise from
momentum conservation. Newton stated the third law within a world-view that assumed
instantaneous action at a distance between material particles. However, he was prepared for
philosophical criticism of this action at a distance, and it was in this context that he stated the
famous phrase ‘I feign no hypotheses.’ The discovery of the second law of thermodynamics
in the 19th century showed that not every physical quantity is conserved over time, thus
disproving the validity of inducing the opposite metaphysical view from Newton’s laws.
Hence, a ‘steady-state’ worldview based solely on Newton’s laws and the conservation laws
does not take entropy into account.
This brings us to the main topic of this book. In unified mechanics, theory entropy
generation is incorporated into Newton’s universal laws of motion.
2.2 Stress
2.2.1 Definitions of Stress and Traction
The external surface force per unit area applied to the body is called the traction, t. It
is a distributed external load. The vector ΔF acting at an imaginary point in the
interior surface is called the internal force vector (Fig. 2.2).
ΔA is the area of the internal imaginary point Q; ΔF is the force vector acting on
this area as a result of external loads applied to the body. We choose to use the term
imaginary point intentionally because later we will assume that the area of this point
goes to zero to be able to define stress at a point.
Fig. 2.2 Definition of stress at an imaginary point Q in the internal surface at cross-section cut a-a
2.2 Stress 13
The average stress σ acting on ΔA is the vector ΔFΔA , ΔF is the vector sum of the
forces acting at point Q, and ΔA is the area of the chosen imaginary point. The
average stress is also a vector having the same direction as ΔF.
Assuming that ΔA is selected so that it represents a point with zero volume as the
value of ΔA approaches zero.
lim
ΔA → 0
of course, ΔA can never be equal to zero. If it were zero, stress value would be
infinite. We let ΔA approach zero to be able to define stress at an imaginary point.

ΔF
σ = lim
ΔA → 0 ΔA
Because the area corresponding to internal force is approaching zero, the numer-
ator, ΔF also approaches zero. However, the fraction, in general, approaches a finite
limit

ΔF
σ = lim
k → 1 ΔA k
where k represents the number of points in the body. This is a basic postulate of
continuum mechanics that such a limit exists and is independent of the area used
(Malvern, 1969).
It is important to explain what we mean by an “imaginary point” with zero
volume, as the area ΔA, it is natural to assume that this point represents an atom.
It cannot be smaller than an atom. However, this point does not represent an atom.
Because at the atomic level fundamental definitions of continuum do not exist, and
this definition of stress at the atomic level is not true. This “imaginary point”
therefore represents a point in the continuum, but not an atom or any other quantum
mechanics’ matter. Since quantum mechanics is outside the scope of this book, we
will refer the reader to a textbook on quantum mechanics.
In x, y, z Cartesian coordinate system stresses at point Q on the cross-section that
has x axis as its normal can be defined by
ΔF x dF x
σ xx = lim =
ΔAx → 0 ΔAx dAx
ΔF y dF y
σ xy = lim =
ΔAx → 0 ΔAx dAx
ΔF z dF z
σ xz = lim =
ΔAx → 0 ΔAx dAx
Fig. 2.3 Definition of stress

at point Q
Subscript x in dAx indicates that x axis is normal to area dAx (Fig. 2.3).
In the same fashion, we can define stress vectors acting on other surfaces, y, and
z. As a result, the state of stress at a point is defined by a second-order tensor with
nine vector components
2 3
σ xx σ xy σ xz
6 7
σ = 4 σ yx σ yy σ yz 5
σ zx σ zy σ zz
Normal stresses, σ ii, are positive when the vector is in the positive axis direction.
Shear stresses σ ij are positive when the vector component is in the positive direction
on the positive face of the block and the negative direction on the negative face of the
block. Shear stresses σ ij are negative when the vector component is in the negative
direction on the positive face of the block and the positive direction on the negative
face of the block.
In this book, tensile stress is defined to be positive, and compression is defined as
negative. However, we realize that this definition is not universal. For instance, in
geo-mechanics tension is negative and compression is positive.
While tension and compression lead to completely different results when applied
to any object, negative shear stress and positive shear stress are considered to be the
same kind of loading in opposite directions from a physical point of view. There is
no compression shear or tension shear.
2.2 Stress 15
2.2.2 Stress Vector on an Arbitrary Plane
Let’s define a plane ABC at an arbitrary slope passing through point Q (Figs. 2.4 and
2.5).
To obtain the equations governing the state of stress at an arbitrarily oriented
plane ABC, we will use the conservation of momentum principle. Forces acting on
the tetrahedron are five vectors representing the resultant force on each of the faces
(ΔAx, ΔAy, ΔAz, ΔA) and the resultant body force ρbΔV, where ρ is mass density, b is
body force per unit mass, and ΔV is the volume of the tetrahedron.
σ (n) is the average value of stress on the oblique face which has n as its normal.
σ is the average stress on the area ΔAx that has x axis as its normal, similarly σ ( y)
(x)
and σ (z) are defined.

Equilibrium of the tetrahedron and stress vector components on the inclined plane
can be derived from the conservation of momentum principle of a collection of
particles. While conservation of momentum is discussed later in the book, assuming
students have a rudimentary knowledge of the principle we will use it at this stage to
derive the equilibrium equations.
The momentum principle of a collection of particles states that the vector sum of
all external forces acting on the free body is equal to the rate of change of the total
momentum. The total momentum of the collection of particles in a given volume is
given by
Fig. 2.4 Stresses acting at a point in Cartesian coordinates

Fig. 2.5 (a) An arbitrary tetrahedron QABC at point Q (b) free-body diagram of
tetrahedron QABC
Z Z
~vρdV = ~vdm
ΔV Δm
where dm is the mass of the tetrahedron, dV is the volume of ρ is mass density, and ~v
is the velocity of the particle.
The time rate of change of the total momentum is given by
d~v d~v
Δm = ðρΔV Þ
dt dt
It is assumed that Δm does not change over time. The conservation of momentum
principle yields the following equilibrium equation for the free body shown in
Fig. 2.6.
d~v
σ ðnÞ ΔA þ ρbΔV - σ ðxÞ ΔAx - σ ðyÞ ΔAy - σ ðzÞ ΔAz = ðρΔV Þ
dt
We need to calculate the height and volume of the arbitrary tetrahedron at point
Q. Then we let the height h go to zero, to be able to define the imaginary point Q,
h h h
cos α = , cos β = , cos γ =
OA OB OC
nx = cosα, ny = cosβ, nz = cosγ
Three direction cosines, nx, ny, nz, are also bound by the following relation:
n2x þ n2y þ n2z = 1

2.2 Stress 17
Fig. 2.6 Geometry of the tetrahedron at a point
As a result
h = OA nx = OB ny = OC nz
The volume of the pyramid is given by
1
ΔV = hΔA
3
substituting h in the previous equation yields,
1 1 1
ΔV = ðOA cos αÞΔA = ðOB cos βÞΔA = ðOC cos γ ÞΔA
3 3 3
Note that we can define the direction cosines also using the relationship between
the areas of the faces of the tetrahedron
ΔAx
cosα =
ΔA
because ΔAx is the projection of the inclined area of interest ΔA on y - z plane with
x axis its normal. Hence
1
ΔV = OA ΔAx
3
Similarly, we can write
1 1
ΔV = OB ΔAy and ΔV = OC ΔAz
3 3
substituting ΔV, ΔAx, ΔAy, and ΔAz in the conservation of momentum equation,
we obtain

1 1 d~v
σ ðnÞ ðΔAÞ þ ρb hΔA = σ ðxÞ ΔA nx þ σ ðyÞ ΔA ny þ σ ðzÞ ΔA nz þ ρ hΔA
3 3 dt
Eliminating ΔA from each term leads to the general equilibrium equation.
1 1 d~v
σ ðnÞ þ ρb h = σ ðxÞ nx þ σ ðyÞ ny þ σ ðzÞ nz þ ρ h
3 3 dt
dt = 0, then, we
Let’s assume, there is no body force b = 0, and no acceleration d~
v
can write
σ ðnÞ = σ ðxÞ nx þ σ ðyÞ ny þ σ ðzÞ nz ð2:1Þ
In some textbooks, Eq. (2.1) is obtained by allowing the tetrahedron h to go to

zero in the equilibrium equation, to define an imaginary point. Equation (2.1) defines
the stress at a point on an arbitrary oblique plane passing through point Q. This
equation was derived from the conservation of momentum principle of a collection
of particles. Hence it applies to solid mechanics as well as fluid mechanics. How-
ever, it is important to point out that this is not an equilibrium of stresses equation.
There is an equilibrium of forces but not stresses. If direction cosines are replaced by
their respective equations in terms of areas, this equation becomes equilibrium of
forces at point Q. We should again emphasize that this equation is only in terms of
stresses and direction cosines. It is not possible to write equilibrium of stresses.
The resultant stress on any inclined plane can be determined from the stress tensor
of the point and the direction cosines of the inclined plane.
Eq. (2.1) is a vector equation because stresses are vectors. The corresponding
algebraic equations yield,
ðnÞ
σx = σxx nx þ σyx ny þ σzx nz
ðnÞ ð2:2Þ
σy = σxy nx þ σyy ny þ σzy nz
σðnÞ
z = σxz nx þ σyz ny þ σzz nz
where σ ðxnÞ is resultant vector component on n plane in the (x) direction, and we have
the same definition for y and z directions. We can write a second-order tensor, σ ij,
Cauchy stress tensor from these algebraic equations. In indicial notation algebraic
equations can be written as
2.2 Stress 19
ðnÞ
σ j = σ ij ni i = x,y,z j = x,y,z
in matrix notation {σ (n)} = {n}[σ]. Remember that { } defines a vector and [ ] defines
a matrix.
Cauchy stress tensor in any other Cartesian coordinate system can easily be
obtained. Assume that the new coordinate system is rotated with respect to the
original coordinate system by α, β, γ angles. The components of the new rotated
stress tensor σ ij can be defined in terms of the three stress vectors acting across the
three planes normal to the new coordinates.
Components of the new stress tensor σ ij will be
ðnÞ
σ i = σ ij nj ∙
σ ij is a second-order tensor (linear vector function); therefore the components of the

rotated stress tensor σ ij can be obtained by the tensor transformation equations.
σ ij = nki nlj σ kl
or in matrix notation
½σ = ½N T ½σ ½N ½σ = ½N ½σ ½N T
where [N] is the matrix of direction cosines nki = cos ðxi , xk Þ of the angles between
the new and old axes. [N]T is its transpose. The matrix of direction cosine is given by
2 3
cos ðx, xÞ cos ðx, yÞ cos ðx, zÞ
6 7
½N = 4 cos ðy, xÞ cos ðy, yÞ cos ðy, zÞ 5
cos ðz, xÞ cos ðz, yÞ cos ðz, zÞ
2.2.3 Symmetry of Stress Tensor
Normal stresses act orthogonal to the surface they are acting on. However, shear
stresses act parallel to the plane on each surface of the cubic control volume, which
represents a point in space with zero volume.
Shear stresses are assumed to be positive when they are on the positive face of the
cube and acting in the positive coordinate axis direction. On the negative face, shear
stress is positive if the direction of the stress is in the negative direction of the axis.
Fig. 2.7 (a) Linear shear stresses acting at point Q. (b) x-y plane view of point Q
When there are no distributed couple-stresses (moments) (body or surface) or the

material has no length scale effect, all off-diagonal terms of the stress tensor are
equal, due to moment equilibrium at point Q
σ xy = σ yx , σ xz = σ zx , σ yz = σ yz
Writing equilibrium of moment about z axis leads to ∑Mz = 0

σ xy dydz dx - σ yx dxdz dy = 0
σ xy = σ yx
Similarly writing ∑My = 0 and ∑Mx = 0 equilibrium equations lead to
σ xz = σ zx and σ zy = σ yz
However, we should make it clear that we obtained this result because we

assumed that at point Q only linear stress vectors are acting and there are no
couple-stress vectors (distributed moments) acting at point Q or material does not
exhibit a length-scale effect that leads to differential shear stresses on opposite sides
of the point Q. Therefore, in the absence of distributed moment acting at a point, the
Cauchy stress tensor is symmetric (Fig. 2.7).
2.2 Stress 21
2.2.4 Couple Stresses
Couple stress at a point can exist due to external loads, like a magnetic field, due to
materials microstructure, or due to the small size of the structure which can cause
very large deformation gradients. If there are the couple-stresses mxz on x plane and
myz on y plane and we write the moment equilibrium with respect to z axis,

σ xy dydz dx þ mxz - σ yx dxdz dy þ myz = 0
Therefore σ xy ≠ σ yx, similarly σ xz ≠ σ zx, σ yz ≠ σ zy.

The skew-symmetric second-order couple stress tensor can be given by
2 3
1 1
6 0 m - mxz 7
6 2 xy 2 7
6 1 1 7
½m = 6 - myx 0 myz 7
6 2 2 7
4 5
1 1
m - m 0
2 zx 2 zy
It is important to point out that the ½ multiplier in front of the couple-stress

(moment) is due to the relation between shear stresses and normal stresses [2εxy = γ xy]
(Fig. 2.8).
Fig. 2.8 Couple stresses y

acting in conjunction with
linear shear stresses in the
x-y plane
myz
σyx
σxy
mxz dy
mxz
σxy x
σyx
myz
dz
dx
z
Example 2.1 Figure 2.1E
Fig. 2.1E State of stress at a point
Equilibrium of forces acting on the wedge on the left can be given by
Aσ ðnÞ = σ ðxÞ nx A þ σ ðyÞ ny A = σ ðxÞ Ax þ σ ðyÞ Ay
For the contribution of forces in x direction to the resultant force on the inclined
plane (n), we can write the following relation, where A represents the area along the
inclined surface:
σðnÞ
x = σxx nx þ σyx ny
Aσ ðxnÞ = - 5 MPa ðA cos 45Þ - 2 MPa ðA cos 45Þ
σðnÞ
x = - 5 MPa cos 45 - 2 MPa cos 45 = - 4:95 MPa

Find the stresses acting on the ABC [111] plane [i.e., plane intercepts x = 1,
y = 1, z = 1].
2 3 2 3
σ xx σ xy σ xz 16 - 40 25
6 7 6 7
σ = 4 σ yx σ yy σ yz 5 = 4 - 40 42 20 5 MPa
σ zx σ zy σ zz 25 20 25
Projection of ABC plane on x, y, z planes is the same shown below, OBC, OAB,
OAC
2.2 Stress 23
y
σ yy
σ yz σy
x
B
σ xy
σ zy
σ xx
σ zz σ xz x O A
σ zx x
C
z
z
Fig. 2.2E Stresses acting on the point
B
B
1 1
√2 √2 √2 Area of OBC = 2
1 ∙1 = 2
1
OBC = OAB = OAC
C A O A
√2 1
Area of ABC = 1
2
√ 2 ∙ √ 2 Cos 30 = 0.866
OBC OAC OAB

nx = = 0.5774 , ny = = 0.5774 , nz = = 0.5774
ABC ABC ABC
3 2 3 2
nx 0:5774
6 7 6 7
Direction Cosines are n = 4 ny 5 = 4 0:5774 5 which is normal to the [111]
nz 0:5774
plane. Therefore, we can define the stresses on the [111] plane utilizing its normal
n. The dot product between the tensor and the unit vector gives us the components of
σ on the surface defined by the normal n
2 3 2 3 2 3 2 3
σnx 16 - 40 25 0:5774 0:5774
6 σn 7 6 7 6 7 6 7
4 y 5 = σ ∙ n = 4 - 40 42 20 5 ∙ 4 0:5774 5 = 4 12:703 5 MPa
σnz 25 20 25 0:5774 40:418
y
( )
B
( )
ABC Plane x
A
2 ðnÞ
3 2 3
σx 0:5774
6 7 6 7
σðnÞ = nT ∙ 4 σðnÞ
y 5 = ½0:5774 0:5774 0:5774 ∙ 4 12:703 5 = 31:00 MPa
σðnÞ
z
40:418
2 2 2
σðnÞ = σðnÞ
n þ σðnÞ
s
2 2 2 2
σðnÞ = σðnÞ
x þ σðnÞ y þ σðnÞ z = 1795:5685
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
σs = σðnÞ2 - σ2n = 28:879 MPa
2.2.5 Principal Stresses and Principal Axes
The stress tensor given in the original Cartesian coordinate system changes values if
we change the orientation of the coordinate system, i.e., simply rotating it around the
origin O. The inclined plane (Fig. 2.9) represents the orientation of a new coordinate
system. Therefore, the stress tensor becomes a function of the rotation angle with
respect to the original coordinate system. One of these orientations will have the
maximum normal stress values, also called principal stresses.
For a three-dimensional stress tensor, these principal values and principal orien-
tation can be identified by eigenvalue analysis,
Normal and shear stress components acting on the inclined face of the tetrahedron
are a function of the orientation of the inclined plane. If the stress tensor is
symmetric, we can choose an orientation of the new coordinate axes such that the
shear stress components vanish in this new coordinate system. These special axes are
called principal axes or principal directions. The three planes that are orthogonal to
three principal axes are called the principal planes. On principal planes, all shear
stresses are zero. There are only normal stresses on the principal planes. Maximum
2.2 Stress 25
Fig. 2.9 Arbitrary

tetrahedron in a stress cube
normal stress is defined as the algebraically largest of the three principal stresses.
The minimum normal stress is defined as the algebraically smallest of the three
principal stresses. There is always such a set of three mutually orthogonal directions
at any point if the state of stress at a point is a symmetric second-order tensor. This is
a property common to all symmetric second-order tensors.
The principal planes and principal stresses can be found by traditional eigenvalue
analysis.
Let [I] be a matrix of unit vectors in one of the arbitrary orientations of the
coordinate system. Let λ be the principal stress components in the new orientation,
whose normals are given by ½n.
Since there is no shear stress on the principal planes, the stress vector on the
principal plane will be parallel to the normal
ðnÞ
σ i = λni
where i = x, y, z.
In matrix notation
n o
σ ðnÞ = λfng
Since
n o
σ ðnÞ = fng½σ
Therefore, we can write,
f ng ½ σ = λ f n g
This can be written as
fng½½σ - λ½I = 0 ð2:3Þ
where [I] is an identity matrix

2 3
1 0 0
6 7
I =40 1 05
0 0 1
fng is a row matrix {n1, n2, n3} where ni defines the direction cosine. λ is the
principal stress value sought. Thus
2 3
λ 0 0
6 7
λ½I = 4 0 λ 05
0 0 λ
Equation (3.3) has a solution only if the determinant of the matrix is equal to zero
since the direction cosines vector cannot be zero, and because n21 þ n22 þ n23 = 1 must
be satisfied.
Hence,
½½σ - λ½I = 0

ðσ xx - λÞ σ xy
σ xz

σ yx σ yy - λ σ yz
=0

σ zx σ zy ðσ zz - λÞ
This is a cubic equation for the unknown λ. If the stress matrix is symmetric and
has real stress values (as opposed to imaginary values), the three roots of the cubic
equation
λ3 þ aλ2 þ bλ þ c = 0
are all real numbers.

The three roots of the cubic equation λ1, λ2, λ3 are the three principal stresses
λ1 = σ 1 , λ2 = σ 2 , and λ3 = σ 3 :
2.2 Stress 27
To find the principal directions, we can use Eq. (2.3).
fng½½σ - λ½I = 0
If we substitute σ 1 for λ in the above equation, these three equations reduce to

only two linearly independent equations. These two equations can be solved with the
help of the n21 þ nn22 þ n23 = 1 requirement,
o to determine the direction cosines of the

normal nð1Þ = nxð1Þ nðy1Þ nzð1Þ to the plane which σ 1 acts on.
h When we arei solving for direction cosines, since one of the equations
n2x þ n2y þ n2z = 1 is quadratic, two solutions will be found representing two
oppositely directed normals to the same plane. The choice of which one is positive
is
n arbitrary. Ino the same manner, n the processo is repeated for σ 2 to find
nðx2Þ , nðy2Þ , nðz2Þ and for σ 3 to find nðx3Þ , nðy3Þ , nðz3Þ . The third principal direction
can also be found by taking a direction orthogonal to the first two directions.
When all the roots of the characteristic equation are different, then these are three
unique orthogonal principal directions. However, when we find the roots of the
characteristic equation, if two of them are equal and one is different, then the
direction of the different root is unique; however, the other two must be arbitrarily
chosen as any two perpendicular axes to each other and perpendicular to the unique
axis. Three axes must define a right-handed system. If all three principal stresses are
equal, the state of stress is said to be the hydrostatic state of stress. Because this is the
only state of stress that can exist in fluids.
2.2.6 Stress Tensor Invariants
Stress invariant is a scalar quantity, and it is an intrinsic property of a stress tensor

that does not depend on the coordinate system. When a stress tensor is rotated, each
vector entry in the tensor matrix changes its value; however stress invariants stay the
same. This is because in Eigen Eq. (2.4), three roots are independent of the
coordinate system.
The determinant of the characteristic equation can be found by

ðσ xx - λÞ σ xy σ xz

σ yx σ yy - λ σ yz =0

σ zx σ zy ðσ zz - λÞ
λ 3 - I 1 λ2 þ I 2 λ - I 3 = 0 ð2:4Þ
where I1, I2, and I3 are the first, second, and third invariants of the stress tensor.
Invariants are given by the following equations:
I 1 = σ xx þ σ yy þ σ zz = trσ ð2:5Þ
Or in indicial notation
I 1 = σ ii

I 2 = σ xx σ yy þ σ yy σ zz þ σ zz σ xx - σ 2xy - σ 2xz - σ 2yz
Or in indicial notation
1
I2 = σ ii σ jj - σ ij σ ij ð2:6Þ
2
where i = x, y, z j = x, y, z
h i
I 3 = σ xx σ yy σ zz þ 2σ xy σ yz σ xz - σ 2xz σ yy - σ 2yz σ xx - σ 2xy σ zz
or
2 3
σxx σxy σxz
6 7
I 3 = 4 σyx σyy σyz 5 = det j σ j ð2:7Þ
σzx σzy σzz
2.2.6.1 Stress Invariants in Principal Axes
In the principal planes, these are no shear stresses; therefore the invariant terms
become simpler:
I 1 = σ1 þ σ2 þ σ3 ð2:8Þ
I 2 = σ1σ2 þ σ2σ3 þ σ3σ1 ð2:9Þ
I 3 = σ1σ2σ3 ð2:10Þ
Stress invariants are mostly used in the constitutive modeling of materials. Any
function constructed with invariants is also invariant with respect to coordinate
systems.
2.2 Stress 29
2.2.6.2 Representation of Stress Tensor in Spherical and Deviatoric

Components
We can represent any stress tensor as a summation of the hydrostatic stress tensor
and the remainder, which is called the deviatoric stress tensor.
The hydrostatic stress part for any stress tensor is given by
1
p= σ þ σ yy þ σ zz ð2:11Þ
3 xx
Using this definition, we can split the stress tensor into two parts.
2 3 2 3
p 0 0 ðσ xx - pÞ σ xy σ xz
6 7 6 7
½σ = 4 0 p 0 5 þ 4 σ yx σ yy - p σ yz 5 ð2:12Þ
0 0 p σ zx σ zy ðσ zz - pÞ
In Eq. (2.12) the first tensor is called the hydrostatic stress tensor, and the second
tensor is called the deviatoric stress tensor. Hydrostatic tensor is also called spherical
stress tensor. We will use Sij to represent deviatoric stress tensor,
2 3
Sxx Sxy Sxz
6 7
Sij = 4 Syx Syy Syz 5
Szx Szy Szz
where
1
Sxx = σ xx - σ þ σ yy þ σ zz Sxy = σ xy , Sxz = σ xz , Syz = σ yz
3 xx
In the same fashion Syy and Szz can be written. In Cartesian form, the deviatoric
stress tensor is given by
1
Sij = σ ij - σ δ
3 kk ij
where δij is Kronecker delta, when
if i = j δij = 1, if i ≠ j δik = 0
Summation over repeated indices rule is utilized, σ kk = σ 11 + σ 22 + σ 33

Hydrostatic stress is the same in all three directions; therefore, in isotropic
materials, hydrostatic stress causes elastic volumetric change only. However,
deviatoric stress is different in all directions; as a result it can cause shape change
and inelastic (irreversible) deformation. In anisotropic materials both hydrostatic and
deviatoric stresses can lead to inelastic shape change.
2.2.6.3 Invariants of the Deviatoric Stress Tensor
Principal deviatoric stresses are in the same planes as the total stress tensor principal
planes. Following the same procedure we used for the total stress tensor, we can
obtain the characteristic [eigen] equation of the deviatoric stress tensor as follows:
2 3
ðSxx - λÞ Sxy Sxz
6 7
4 Syx Syy - λ Syz 5=0
Szx Szy ðSzz - λÞ
This equation can be expanded to the following form:
λ3 - J 2D λ - J 3D = 0
where J2D and J3D are the second and third invariants of the deviatoric stress tensor,
respectively. The first invariant of the deviatoric stress tensor is zero J1D = 0,
1
J 2D = Sij Sij
2
h i
1 2 2
= σ xx - σ yy þ σ yy - σ zz þ ðσ xx - σ zz Þ2 þ σ 2xy þ σ 2xz þ σ 2yz
6
h i
1
J 2D = S2xx þ S2yy þ S2zz þ S2xy þ S2xz þ S2yz
2
1
J 3D = Sij Sjk Ski
3

Sxx Sxy Sxz

J 3D = Syx Syy Syz

Szx Szy Szz
It is important to point out that the second invariant of the deviatoric stress tensor
is a representation of shear stress; however physical interpretation of the third
invariant is more complex.
2.2.7 Octahedral Plane and Octahedral Stresses
The plane that makes equal angles with the three principal directions is called the
octahedral plane. The projection of the stress vector on the octahedral plane is called
the octahedral shear stress. The magnitude of the normal stress on the octahedral
plane is constant.
2.2 Stress 31
Octahedral shear stress is given by the following relation:

h i h 2 i
1 2 2 2 2
τ2oct = σ xx - σ yy þ σ yy - σ zz þ ðσ xx - σ zz Þ2 þ σ xy þ ðσ xz Þ2 þ σ yz
9 3
Normal stress on the octahedral plane is given by
1
σ oct = σ þ σ yy þ σ zz
3 xx
The second invariant of the deviatoric stress tensor is related to octahedral shear
stress by the following relation:
3 2
J 2D = τ
2 oct
In most metals plasticity happens mainly due to shear stresses; hence deviatoric
stress tensor invariants and octahedral stresses are utilized often in defining yield
surfaces.
Stresses acting on the point shown above are given below:
σ xx = 4 MPa, σ xy = 3 MPa, σ xz = 2 MPa

σ yy = 5 MPa, σ yz = 1 MPa, σ zz = 6 MPa,
Fig. 2.3E (a) Stresses acting on the point. (b) Principal directions
(i) Calculate the normal and shear stresses acting on the octahedral plane [i.e., 1,1,1
plane with respect to the principal axes] using the stress transformation equa-
tions, and then compare those values with the formula given below:
1
σ oct = σ þ σ yy þ σ zz
3 xx
h i h i
1 2 2 2 2 2
τ2oct = σ xx - σ yy þ σ yy - σ zz þ ðσ xx - σ zz Þ2 þ σ xy þ σ yz þ ðσ zx Þ2
9 3
(ii) Verify that J 2D = 32 τ2:

oct
First, we need to calculate principal stresses and principal directions using the
Eigen equations,

ð4 - λÞ 3 2

3 ð5 - λÞ 1 =0

2 1 ð6 - λÞ
This yields a third-order polynomial in λ. The principal stresses are these roots.
Principal stresses are λ1 = σ 1 = 9.0000 MPa, λ2 = σ 2 = 4.7321 MPa, λ3 = σ 3 =
1.2679 MPa
Direction cosines for octahedral plane [111] normal with respect to the principal
directions were calculated earlier in Example 2.2.
2 3
0:5774
6 7
n = 4 0:5774 5
0:5774
Using transformation equations σ oct = nT nσ = n21 σ 1 þ n22 σ 2 þ n23 σ 3
σ oct = 0:57742 9:0000 þ 0:57742 4:7321 þ 0:57742 1:2679 = 5:0000 MPa
The norm of the stress vector acting on the octahedral plane is
2
σðnÞ = ðσ1 n1 Þ2 þ ðσ2 n2 Þ2 þ ðσ3 n3 Þ2
= 0:57742 9:00002 þ 0:57742 4:73212 þ 0:57742 1:26792
= 35:0061
2.3 Deformation and Strain 33
Also
2
σðnÞ = σ2oct þ τ2oct
Hence
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
τoct = σ 2 - σ 2oct = 3:1623 MPa
And now using the formula given above we can recalculate the stresses on the
octahedral plane
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
h i h i
1 2
τoct = ð 4 - 5 Þ 2 þ ð 5 - 6Þ 2 þ ð 4 - 6Þ 2 þ ð 3 Þ 2 þ ð 1 Þ 2 þ ð 2 Þ 2
9 3
= 3:1623 MPa
1
σ oct = ½4 þ 5 þ 6 = 5:0000 MPa
3
h i
1 2 2
J 2D = σ xx - σ yy þ σ yy - σ zz þ ðσ xx - σ zz Þ2 þ σ 2xy þ σ 2xz þ σ yz
2
6
(iii) h i
1
= ð4 - 5Þ2 þ ð5 - 6Þ2 þ ð4 - 6Þ2 þ 32 þ 22 þ 12 = 15
6
Verify
3 2
J 2D =τ
2 oct
3
15 = 3:16232 = 15
2
2.3 Deformation and Strain
Let’s assume, we have a deformable solid object defined by corners A to H. Under

external loading, the corner points displace to new locations given by points A′ to H′
(Fig. 2.10).
Using Cartesian coordinates x, y, z, we can easily measure the relative displace-
ment of corner points (or any point in the object) with respect to their original
location. However, just knowing the relative displacement dr of each point in the
deformable body does not help us to understand the strain imposed on the body due
Fig. 2.10 Definition of strain deformation and rigid body motion
to the new configuration. Therefore, we must establish a strain definition. There are
many definitions of strain. We can choose any definition of strain and calculate the
strain in the deformable body. The point we are trying to emphasize is that strain is a
definition (a human construct) calculated using measured displacement quantities.
Therefore, strain cannot be measured directly.
Strain formulation without any constraints on the magnitude of the strain gives a
more accurate representation of the deformation. However, the mathematical
description of finite strain (large strain) is quite complex. Therefore, for many
engineering problems, the small-strain theory is used, which is satisfactory with a
reasonable degree of error.
First, we will present the small strain formulation and then a more complex large
strain (finite strain) formulation. Rigid body motion does not lead to any strain.
2.3.1 Small Strain Definition
There is no magic number that defines the boundary between small-strain theory and
large strain theory. However, any strain value less than 2% is usually, but not always,
considered a small strain. The decision to use the small or large strain formulation
depends on the problem at hand and the material properties. Now, we will define
strains with respect to the initial (undeformed) configuration.
2.3.1.1 Elementary Definition of Pure Uniaxial Strain
At the unit defined by ABCD, corners B and C are stretched to new locations B′ and
C′ (Fig. 2.11).
As a result, small-strain Exx is the change in length per unit of initial length in the
x-direction. In the same fashion, we can define the unit extension per unit length for
the y and z directions Eyy and Ezz
Fig. 2.11 Uniaxial

extension in one dimension
Δu Δv Δw
εxx = ε = ε =
Δx yy Δy zz Δz
where Δv and Δw are extensions in the y and z directions, respectively. We

let Δx, Δy, and Δz approach zero to be able to define the strain at an imaginary
point. Strain and stress are always defined at a point of zero volume. The term
imaginary is coined to reflect the zero-volume feature of the point.
2.3.1.2 Pure Shear Strain
The shear strain is defined as the change in the initial right angle at point A
(Fig. 2.12).
π
γ xy = - Ψ = θ1 þ θ2
2
for small angles, [θ < 0.5 radian]
tan θ = θ
Δv Δu
tan θ1 = θ1 = and tan θ2 = θ2 =
Δx Δy
As a result, we can write,
Δv Δu
γ xy = θ1 þ θ2 = þ radian
Δx Δy
Fig. 2.12 Shear strain

definition in two dimensions
Since we define our strain at a point in the limit dimensions, zero volume of the
differential element Δx and Δy approach zero. Hence, we can define strain in
differential form as
∂u ∂v ∂u ∂v
Exx = Eyy = γ = þ
∂x ∂y xy ∂y ∂x
where u is the displacement in x axis and v is the displacement in y axis. It is

important to point out that the initial (undeformed) position is the independent
variable and the reference state in these small strain calculations. In other words,
the change in per unit length is with respect to the initial length. If we utilize the
deformed position as the reference state and as the independent variable, we obtain a
different formulation of strain. Of, course these two reference states will lead to two
numerically different strain values.
Example 2.4
Calculate the shear strain at point B in the given triangle-shaped plate after defor-
mation. Before deformation, the lengths of edges AB and BC are 300 mm and
400 mm, respectively. Point B moves u = 1 mm in x axis direction and v = 3 mm in
y axis direction to point B′. Calculate the shear strain at point B using the change in
angle and using the formula.
3 mm 1 mm
tanα = α = 0:01 rad tanβ =
ð300 mm þ 1 mmÞ ð400 mm þ 3 mmÞ
β = 0:0025 rad
h i h i
π π 400 π 300
γ xy = - π - - atan - 0:01 - π - - atan - 0:0025
2 2 300 2 400
= α þ β = 0:0125 rad
∂u ∂v 1 mm 3 mm
γ xy = þ = þ = 0:0125 rad
∂y ∂x 400 mm 300 mm
2.3.1.3 Pure Rigid Body Motion
Let us assume the cantilever beam shown in Fig. 2.13 is supported at point A and
subjected to vertical and axial loads at point B. Point C in the beam will be subjected
to rigid-body displacement and rigid-body rotation, but it will not experience any
strain, because point C undergoes rigid body motion, only.
The point C undergoes displacement in x and y axes directions as well as rotation
around z axis. However, these motions will not lead to any strain or stress at point C
(Fig. 2.14).
Fig. 2.13 Definition of pure

rigid body motion
Fig. 2.14 Pure rigid body

rotation
Fig. 2.15 Spatial

description of the new
location of point A
Fig. 2.16 Material (local)

B’
coordinate system
description
B A’
A
2.3.2 Small Strain and Small Rotation Formulation
We will start with defining spatial coordinate and material coordinate systems. Let’s
say f(u, t) is a function that defines the displacement of point A located at a particular
point A′ in space at time t.
This spatial description is called Eulerian description. Another description is
material field description, also called Lagrangian description in many textbooks. In
this case function g(r, t) defines the displacement of point A at time t regardless of
where the point A is located in space. This is called material or local description
(Figs. 2.15 and 2.16).
The coordinates of a point in a local reference coordinate system (configuration) r

are referred to as material coordinates. In three dimensions we will use r, s, t for
material coordinates.
The coordinates of a point in the spatial coordinate system x, y, z are referred to as
spatial coordinates. Of course, we can link the material coordinate system and spatial
coordinate system. In computational solid mechanics, such as in the finite element
method, we prefer to use material coordinates because it is more convenient.
2.3.2.1 Definition of Material (Local) Coordinates
In Fig. 2.17 we define a coordinate system that is shaped in the original shape of the
line AB. We place the origin of our local coordinate axis r at point A, where r = 0,
and at point B the value of the local coordinate is r = 1. By normalizing the
coordinate between zero and one, we can define the displacement of any point
between A and B with respect to initial coordinates easily. Of course, we can define
displacement of point A using the global Cartesian coordinate system x, y and time,
or we can use the material (local) coordinate system (which is usually called the local
coordinate system in finite element method terminology). Local coordinate r is in the
axis direction of the initial one-dimensional element shown in Fig. 2.18. The origin
of the local coordinate system is at point A, or it can be located at any point between
A and B. As a result, if point A moves with respect to the initial location due to
loading origin A also moves. Therefore coordinate r will define the relative dis-
placement of point A with respect to B at any time regardless of where the point A is.
We will use the terminology local coordinates rather than material coordinates.
For a two-dimensional case let x - y axes define our global (spatial) Cartesian
coordinate system and let r define the local coordinate system. Let dr define an
arbitrary infinitesimal line element vector (Fig. 2.18).
Where r is the material (local) coordinate axis. It is always in the natural direction
of the member, u is the displacement vector along the spatial axis x, and v is the
displacement vector along the spatial y axis.
Fig. 2.17 Material (local)

coordinate system
Fig. 2.18 Local (material)

coordinate system, r
The local coordinate relative displacement dr in spatial Cartesian coordinate

system x, y can be given by
du du dx du dy
= þ
dr dx dr dy dr
ð2:13Þ
dv dv dx dv dy
= þ
dr dx dr dy dr
It is important to point out that the left-hand side of this equation represents strain
components with respect to the initial length. Equation (2.13) can be written in a
matrix form:
2 3 2 3
du du du 2 dx 3
6 dr 7 6 dx dy 776 dr 7
6 7 6 76 7
6 7=6 6 7
6 7 6 76 7
4 5 6
4 dv
74 5
dv dv 5 dy
dr dx dy dr
Or
du
~
=J n
dr
where dudr is the column vector on the left-hand side containing relative displacement
vector components, ñ is the column vector on the right-hand side containing the
direction cosines, Cosα = dx dr , Cosβ = dr . J is called the displacement–gradient
dy
matrix or the Jacobian matrix, or the unit relative displacement matrix. It may be
considered as an operator which operates on a direction vector ~ n to yield the unit
relative displacement components for an infinitesimal line AB. This process may be
considered akin to the transformation from spatial coordinate system x, y to local
(material) coordinate system r. Assume that length of AB → 0; as a result, all

derivatives are evaluated at point B, because we need to define strains at a point.
We can split the Jacobian matrix into two components. One component will
include deformations around point B, and the other will include rigid-body relative
displacements. It is assumed that rigid body relative displacements do not cause any
strain on the body. This splitting process can be accomplished in the following
manner.
2 3 2 3
∂u ∂u ∂u 1 ∂u ∂v
6 þ
6 ∂x ∂y 7 ∂x 2 ∂y ∂x 7
6 7=6 6
7
7
4 ∂v ∂v 5 4 1 ∂v ∂u ∂v 5
þ
∂x ∂y 2 ∂x ∂y ∂y
2 3
1 ∂u ∂v
6 0 -
6 2 ∂y ∂x 7 7
þ6 7
4 1 ∂v ∂u 5
- 0
2 ∂x ∂y
Earlier we defined small strain components as follows:

∂u ∂v ∂u ∂v
Exx = , Eyy = , γ = þ
∂x ∂y xy ∂y ∂x
For convenience, we will define shear strain Exy as half the decrease in the angle
γ xy to be able to define strain vectors as components of a second-order strain tensor ε.
1
Exy = γ
2 xy
As a result, the first matrix is the strain matrix, and the second matrix is the
rotation matrix:

1 ∂u ∂v 1 ∂v ∂u
Ωxy = - Ωyx = - = - Ωxy
2 ∂y ∂x 2 ∂x ∂y
As a result, the Jacobian matrix can be written as
Exx Exy 0 Ωxy

J= þ
Eyx Eyy Ωyx 0
In our stress and strain definitions, the first subscript defines the plane they belong
to and the second subscript defines the direction. Now we can show that the strain
matrix leads to deformations (shape change) and the rotation matrix leads to rigid
body motion only with no shape change (strain) (Fig. 2.19).
We can write earlier transformation equations between local (material) coordi-
nates and spatial coordinates to find the unit relative displacement of B with respect
to A.
Fig. 2.19 Deformed

differential element without
rigid body motion
2 3
2 3 du 1 du dv 2 3
du þ dx
6 7
6 dr 7 6 dx 2 dy dx 76 dr 7
4 5=6 74 5
dv 4 1 dv du dv 5 dy
dr þ dr
2 dx dy dy
dx
Assume γ 1 is very small cos γ 1 = = cos 0 = 1,
dr
and
dy π
= cos = 0
dr 2
As a result, we can write,

2 3
du
6 dr 7 Exx Exy 1 Exx
4 5= =
dv Eyx Eyy 0 Eyx
dr
Thus for AB
du dv
Exx = , E =
dr yx dr
It is important to point out that these equations are only valid for small deforma-
tions. Similarly, for AD we can write
Assuming γ 2 is very small.
dx π dy
= cos = 0, = cos 0 = 1
dr 2 dr
2 3
du
6 dr 7 Exx Exy 0 Exy
4 5= =
dv Eyx Eyy 1 Eyy
dr
du dv
= Exy = Eyy
dr dr
2.3.2.2 Small Strain in Local Coordinates
Displacement of corner point C to point C′ is plotted in detail Figs. 2.20, 2.21, and
2.22.
Small strain along local axis r can be written as
Fig. 2.20 Deformation of

ABCD into A′B′C′D′ in
local and global coordinates
Fig. 2.21 Displacement of

point C to C′
Fig. 2.22 Displacement of a point in material (local) and spatial coordinates

Exx dx þ γ xy dy cos θ þ Eyy dy sin θ
Err =
dr
where dr is the initial length of AC in local coordinates

dx dy dy
Err = Exx þ γ xy cos θ þ Eyy sin θ
dr dr dr
Knowing that cos θ = dxdr sin θ = dr

dy
Inserting these equations in εrr yields
Err = Exx cos 2 θ þ γ xy sin θ cos θ þ Eyy sin 2 θ
Using trigonometric relations,
1
cos 2 θ = ð1 þ cos 2θÞ
2
1
sin 2 θ = ð1 - cos 2θÞ
2
2 sin θ cos θ = sin 2θ
Strain along r axis can also be expressed as
Exx þ Eyy Exx - Eyy γ xy

Err = þ cos 2θ þ sin 2θ
2 2 2
We could also obtain components of displacement CC′ in r and s local coordi-

nates by using the dot product of vector CC′ and unit vectors along r and s.
Fig. 2.23 Unit vectors in local (material) r-s coordinates. And spatial coordinates x, y
Material coordinate system unit vectors can be represented in terms of spatial

coordinates’ unit vectors as follows (Fig. 2.23):
l = i cos θ þ j sin θ
m = - i sin θ þ j cos θ
The displacement in r and s directions can be found by the dot product.

In r direction,

CC0 l = Exx dx þ γ xy dy i þ Eyy dy j ½icosθ þ jsinθ
= Exx dx þ γ xy dy cosθ þ Eyy dysinθ
In s direction,

CC0 m = Exx dx þ γ xy dy i þ Eyy dy j ∙ ½ - i sinθ þ jcosθ

= - Exx dx þ γ xy dy sinθ þ Eyy dycosθ
From here we can obtain,
CC0 l
Err =
dr
and
CC 0 m
Ess =
ds
Fig. 2.24 Shear strain in

local (material) coordinates
2.3.2.3 Shear Strain in Local (Material) Coordinates
Displacement along s axis can be formulated by the following summation

(Fig. 2.24):

Eyy dy cos θ - Exx dx þ γxy sin θ
We can also obtain the displacement of point C along s axis from the arc length.
Assuming small deformations and small strains

αdr = Eyy dy cos θ - Exx dx þ γ xy dy sin θ

dy dx dy
α = Eyy cos θ - Exx þ γ xy sin θ
dr dr dr

α = Eyy sin θ cos θ - Exx cosθ þ γ xy sin θ sin θ

Rotation of s axis (β) can be found by substituting θ þ π2 in the equation for α.
β = - Eyy sin θ cos θ þ Exx sin θ cos θ - γ xy cos 2 θ
In Fig. 2.25, it is assumed that the shear strain is positive. Therefore α is positive
when it is counterclockwise, and β is positive when it is clockwise.
Fig. 2.25 Relations

between local (material) and
spatial coordinates
γ rs = ðα - βÞ
= Eyy sin θ cos θ - Exx sin θ cos θ - γ xy sin 2 θ
- - Eyy sin θ cos θ þ Exx sin θ cos θ - γ xy cos 2 θ

γ rs = - 2 Exx - Eyy sin θ cos θ þ γ xy cos 2 θ - sin 2 θ
2.3.3 Small Strain and Rotation in 3-D
Using the derivation, we used for a two-dimensional case we can derive the
formulation for a three-dimensional case. In three dimensions, we will use u for
displacement along x, v for displacement along y, and w for displacement along
z axis. r axis is again our local (material) coordinate axis
2 3
8 9 ∂u ∂u ∂u 8 dx 9
>
> du >
> 6 ∂x > >
> >
>
> >
> 6 ∂y ∂z 7
7>> dr >
>
< dr >
> = 6 7>< > =
dv 6 ∂v ∂v ∂v 7 dy
=6 7
>
> dr >
> 6 ∂x ∂y ∂z 7> dr >
>
>
> >
> 6 7>> dz >
> >
: dw >
> ; 4 ∂w ∂w ∂w 5>: > ;
dr ∂x ∂y ∂z dr
Again Jacobian matrix J can be written as
J=E þ Ω
where
2 3
∂u 1 ∂u ∂v
1 ∂u ∂w
6 þ þ
6 ∂x 2 ∂z ∂x 7
2 ∂y ∂x
6 77
6 1 ∂u ∂v ∂v 1 ∂v ∂w 7
E=6 þ þ 7
6 2 ∂y ∂x ∂y 2 ∂z ∂y 7
6 7
4 1 ∂u ∂w 1 ∂w ∂v ∂w 5
þ þ
2 ∂z ∂x 2 ∂y ∂z ∂z
2 3
1 ∂u ∂v 1 ∂u ∂w
6 0 - -
6 2 ∂y ∂x 2 ∂z ∂x 7
6 7
7
6 1 ∂u ∂v 1 ∂v ∂w 7
Ω=6 - - 0 - 7
6 2 ∂y ∂x 2 ∂z ∂y 7
6 7
4 1 ∂u ∂w
1 ∂v ∂w
5
- - - - 0
2 ∂z ∂x 2 ∂z ∂y
When components of Ω are small, compared to 1 radian, it represents rigid body

rotation.
2.4 Kinematics of Continuous Medium
There are three distinct descriptions of the displacements in a continuum.
2.4.1 Material (Local) Description
The variables of this coordinate system are the particle at time t, regardless of where
the point is in spatial coordinates. Note that not the position of the particle but the
particle itself is the variable.
2.4.2 Referential Description (Lagrangian Description)
This description usually utilizes a reference configuration for description. The

independent variables are the r, s, t coordinates and time with respect to a reference
configuration and reference state. The chosen reference configuration is usually the
initial (unstressed) state at t = 0. When the reference configuration is chosen to be the
initial undeformed shape at t = 0, the referential description is also called the
Lagrangian description. Many mechanics researchers consider Lagrangian and
2.4 Kinematics of Continuous Medium 49
material description to be the same. It is assumed that the origin of the coordinate
system is attached to the particle.
2.4.3 Spatial Description (Eulerian Description)
The spatial description is defined by the current location x, y, z of the object at time t,
which are considered the independent variables. This is the most common coordinate
description used in fluid mechanics. Because spatial description uses a fixed region
of space rather than a body of matter [a particle]. For most students, it is usually
difficult to visualize the different descriptions. Here we will quote an explanation
from Malvern (1969) “The spatial description is especially useful in fluid mechanics,
where we may observe a flow in a wind tunnel or a channel, a fixed region [window]
in space. We can imagine perfect instrumentation with which an observer records the
fluid velocity at a field point in space [a window] as a function of time. If this is done
for every point in the region, we have the spatial description of the velocity ~v as a
function of position x,y,z and time.”
A Comparison of Lagrangian and Eulerian Descriptions
“In computational solid mechanics, all particles in a body must be defined by their
location in space from the initial unstressed (undeformed) state to the final config-
uration. Therefore, Lagrangian (material) description is preferred. This approach
stands in contrast to Eulerian formulation, which is usually used in the analysis of
fluid mechanics problems, in which attention is focused on the motion of the fluid
through a stationary control volume (a fixed window). Considering the analysis of
solids and structures, a Lagrangian formulation usually represents a more natural and
effective analysis approach than an Eulerian formulation. For example, using an
Eulerian formulation of a structural problem with large displacements, new control
volumes have to be created (because the boundaries of the solid change continuously
[due to large displacements])” (Bathe, 1996). Of course, the transformation of
coordinates between these descriptions can easily be accomplished.
Of course, the transformation between these descriptions can be done with a
simple transformation formulation (Fig. 2.28).
2.4.4 Material Time Derivative in Spatial Coordinates

(Substantial Derivative)
The material time derivative is the time derivative with the material coordinates held
constant. If we know the complete motion of all particles, x = f(r, t), we can calculate
any particle’s velocity simply by taking the partial time derivative with material
coordinates r, s, t held constant.
2
∂ ∂x ∂~v ∂
~v = Gðr, t Þ = f ðr, t Þ and a = f ðr, t Þ
∂t ∂t r ∂t r ∂t
The partial time derivative with the material coordinates r, s, t held constant is
completely different than the partial time derivative with spatial coordinates x, y,
z held constant. In spatial description velocity ~v = gðx, t Þ is given by, hence

∂~v ∂
gðx, t Þ
∂t x ∂t
And this is different from the acceleration given by

∂~v ∂
Gðr, t Þ
∂t r ∂t
∂~v
∂t x
is called the local rate of change of velocity, because spatial coordinates are
fixed. It is the rate of change of the reading of a velocity meter located at the fixed
x, y, z coordinates, which is different than the acceleration of the particle just now
passing through the fixed window with x, y, x coordinates.
In a steady-state flow, the local rate of change is everywhere zero, but this does
not mean that the acceleration of the particle is zero. Even in a steady-state flow, the
velocity varies in general from point to point and a particle changes its velocity as it
moves from one point of constant velocity to another point of different constant
velocity, [like due to changing pipe diameter]. Since the universal laws of motion are
written for particle accelerations and not for local rates of change, we need the
particle accelerations in the spatial description. This can be accomplished by utiliz-
ing the material derivatives.
The material time derivative is the time derivative with the material coordinates
held constant. However, if the spatial description ~v= gðx ,t) is known but not the
referential description ~v = Gðr ,t), we cannot use ∂x ∂t r
to calculate the material
derivative. We assume that there exists a differentiable and unique function defining
the relations between spatial and material coordinates, x(r, t). Then by substituting
this function, x(r, t) into the spatial description of the velocity ~v = gðx,t) we obtain,
~v = g½xðr, t Þ, t
Or in rectangular Cartesian coordinates
~vi = gi ½xðr, s, t, timeÞ, yðr, s, t, timeÞ, zðr, s, t, timeÞ, time
where
iis particle
number. And by the chain rule
∂~vi ∂~vi ∂~vi ∂xj
∂t
= ∂t þ ∂xj ∂t (summation on repeated indice j only)
r x r
2.4 Kinematics of Continuous Medium 51
∂x
∂xj
Knowing that ~v = ∂t r
in Cartesian coordinates ~vj = ∂t r
, substituting it above
in the chain rule yields
ai = ð∂~
vi
∂t r
Þ = ð∂~
vi
∂t x
Þ þ ~vj ∂~
vi
∂xj
(summation on repeated indice j only)
This equation can be written in matrix notation as
d~v ∂~v
a= þ v~ grad~v
dt ∂t
where grad~v is the gradient with respect to the spatial coordinates, (x, y, z) which is
sometimes denoted ∇x~v to distinguish it from the gradient ∇r ~v which is with respect
to the material coordinates (r, s, t)
This material derivative can also be used for any other properties. For example, if
the spatial description of density is given by the scalar function ρ(x, t), then the rate
of change of the density in the neighborhood of the particle instantaneously at x is
given by

dρ ∂ρ
= þ v~ gradρ
dt r ∂t x

∂ρ ∂ρ
= þ ~vj
∂t x ∂xj

∂ρ
The first term ∂t x
gives the local rate of change of the density in the neigh-
∂ρ
borhood of the point (x, y, z), while the second term ~vj ∂x j
gives the convective rate of
change of the density in the neighborhood of a particle as it moves to a place with a
different density. Hence we can write the following equation as the material deriv-
ative operator:

d ∂
= þ ~v grad
dt ∂t
which can be applied to a scalar, a vector, or a tensor function of the spatial position
(x, y, z and t).
The material derivative has also been described very succinctly, [in MIT 2.20
Marine Hydrodynamics lecture notes], by quantifying the time rate of change of a
fluid property as it travels through a given flow field [flow window]. The material
derivative is the time derivative of a property x of a particle; hence it is Lagrangian.
However, we can write the material derivative in the Eulerian frame of reference.
The time rate of change of property, x defined in the Eulerian frame of reference
can be given by
Fig. 2.26 Material (local), spatial, and referential coordinate system description
df ðxðtÞ,tÞ f ðx þ ~v δt,t þ δtÞ - f ðx,tÞ

= lim
dt δt → 0 δt
Writing Taylor’s series expansion of this variation and considering that ṽδt = δx
yields
∂f ðx,tÞ
f ðx þ ~vδt,t þ δtÞ = f ðx,tÞ þ δt þ δx ∇ fðx,tÞ þ higher order terms
∂t
From these two equations, we conclude that the material derivative of a property
f as experienced by a particle traveling with a velocity of ṽ is
df ∂f
= þ ~v ∙ ∇f
dt ∂t
Generalized notation is given in the following form:
d ∂
þ ~v ∇
dt ∂t
½Lagrangian ½Eulerian
2.5 Rate of Deformation Tensor and Rate of Spin Tensor 53
Fig. 2.27 Relative velocity,

d~v, of point A relative to
point B
Where r and s are local (material) coordinate system axes that move with the
particle. O′ is the origin of the referential (Lagrangian) coordinate system. O is the
origin of the spatial description (Eulerian) coordinate system (Fig. 2.26).
2.5 Rate of Deformation Tensor and Rate of Spin Tensor
The rate of deformation tensor, D, is also called the stretching tensor or velocity of
strain (strain rate) tensor. The spin tensor, W, is also called the vorticity tensor.
We realize that the rate of deformation tensor is defined differently by different
authors. We will subscribe to Malvern’s (1969) definition, and use his formulation
and derivation with few modifications (Fig. 2.27).
The relative velocity components of d~vi of point, A relative to point B can be
given by
∂~vi
d v~i = dxm
∂xm
or in matrix form
½dv~i = ½~vi,m ½dxm ð2:14Þ
or in tensorial notation
d~v = L dx = dx LT

where L is Lim = ~vi,m LT im = ~vm,i m and i can have values of 1, 2, and 3.
The components of tensor L are spatial gradients of the velocity. Lim = ~vi,m can be
written as the sum of a symmetric tensor D, which we will call the rate-of-
deformation tensor (also called stretching tensor), and a skew-symmetric tensor

W called the spin tensor (also called vorticity) tensor.
Hence we can write
L=D þ W
1 1
D = L þ LT and W = L - LT
2 2
or in Cartesian coordinates
1 1
Dim = ð~vi,m þ ~vm,i Þ and W im = ð~vi,m - ~vm,i Þ
2 2
W spin matrix is different from the Ω rotation matrix. They should not be
confused with each other. The same thing is true, for the rate of deformation tensor
D which is not a strain-rate matrix. The rate of deformation tensor, D, allows us to
define the relative velocity of point A with respect to point B. As a result, Eq. (2.14)
can be given by
d~vi = Dim dxm þ W im dxm
2.5.1 Comparison of Rate of Deformation Tensor, D,

and Time Derivative of the Strain Tensor, ε_
The small strain tensor ε is defined in terms of local coordinates (material or

Lagrangian description) (initial coordinates). Therefore, it is given by

1 ∂ui ∂uj
Eij = þ
2 ∂r j ∂r i
where ri represents local (material) coordinates [Lagrangian coordinates], ui, repre-

sents the amount of deformation in each axis. Therefore, time derivative of the small
strain tensor would be given by

dEij 1 ∂~vi ∂~vj
= þ
dt 2 ∂r j ∂r i
Since we define the spatial gradient of velocity as follows
∂~vi d ∂u
= ð iÞ
∂xj dt ∂xj
2.5 Rate of Deformation Tensor and Rate of Spin Tensor 55
On the other hand, the rate of deformation tensor, D, is given by

1 ∂~vi ∂~vj
Dij = þ
2 ∂xj ∂xi
The time derivative of the strain tensor is with respect to local coordinates
(material coordinates-Lagrangian description), while the rate of deformation tensor
is defined with respect to spatial coordinates x, y, z (Eulerian description). For small
displacement and small strain problems, both tensors are the same. However, D, the
rate of deformation tensor is necessary for large displacement and large strain
problems.
Where i = 1, 2, 3, j = 1, 2, 3. It is important to point out that for indicial notation
axes are represented by x, or by xi. However, when convenient x, y, z spatial
coordinates and r, s, t are also used throughout the book.
2.5.2 True Strain (Natural Strain) (Logarithmic Strain)
Engineering strain εE is defined by elongation divided by the initial length
L - L0
dEE =
L0
in incremental form as
Z L
dL E 1
dEE = E = dL
L0 L0 L0
However, if we use the instantaneous length (new length after deformation) in the
denominator, we obtain the true strain (natural or logarithmic strain):
dL
dεT =
L
Here the increment of true strain is defined by the change in length per unit of
instantaneous (new) length. To find the total true strain, we can integrate the
increment
ZL ZL
dL L L0 þ ΔL
dE = E =
T T
= ln = ln = ln 1 þ EE
L L0 L0
L0 L0
where ln is the natural logarithm. The relation between true strain and engineering
strain can also be written as

eE = 1 þ εE or εE = eε - 1
T T
From the Taylor series, we can write the following relation:
X1
xk x2 x3
ex = =1 þ x þ þ þ ⋯...
k=0
k! 2! 3!
As a result, we can write the following relation between true strain and engineer-
ing strain:
2 3
ET εT
εT þ þ þ ⋯:: = εE
2! 3!
In three dimensions for the small strain, this can be generalized as follows:
dεTij = Dij dt
where Dij is the rate-of-deformation tensor, and dt is the time increment.

It is important to point out that increments of natural strain dεTij , are components
of a Cartesian tensor; as a result the transformation formulas and principal axis
theory all apply. The quantities εTij defined by integration are not components of a
Cartesian tensor, because during the deformation principal axes continuously rotate
at each increment (Malvern, 1969).
Fig. 2.28 Description of Lagrangian and Eulerian descriptions of a displacement vector

2.6 Finite Strain and Deformation 57
2.6 Finite Strain and Deformation
There are many definitions of finite (large) strain. They can be categorized into two
classes:
1. Defining strain with respect to undeformed original configuration and geometry.
This approach is called Lagrangian formulation.
2. Defining strain with respect to deformed configuration and geometry. This
approach is called the Eulerian formulation.
Large (finite) strain formulation is the easiest to define in terms of a deformation–
gradient tensor, F. However, the deformation gradient tensor includes the strain
tensor and the rotation tensor. As a result, this can make it tricky to be used in
material modeling. On the other hand, strain tensor only includes strains where x, y,
z are global Cartesian, coordinates (referential description) and r, s, t is the local
coordinate (material description) system (Fig. 2.28).
In Fig. 2.28 the vector joining point A in the undeformed location and deformed
location is the displacement vector. The displacement vector can be defined in the
Lagrangian description as follows:
uðr, t Þ = cðt Þ þ xðr, t Þ - r
If we want to define the displacement vector in the Eulerian description, then we

can write
uðx, t Þ = cðt Þ þ x - rðx, t Þ
where c(t) is the rigid body motion. In the finite element method, the origin of the
material (local) coordinate system is usually in the center or at a corner of an
element. However, for some special-purpose elements, it is placed at a special
location, but always inside the element.
Fig. 2.29 Normalized local

(material) coordinates
system and spatial
coordinates
The partial derivative of the displacement vector with respect to material coordi-
nates yields the material displacement gradient tensor,
∂ui ∂x
= i - δik = F ik - δik
∂r ∂r k
where Fik is the deformation–gradient tensor and δik is Kronecker’s delta.

If we take the partial derivative of the displacement vector with respect to spatial
coordinates, it yields the spatial displacement–gradient tensor
∂uj ∂rj
= δjk - = δjk - F jk- 1
∂x ∂xk
The deformation gradient tensor, F, describes the stretches and rotations that the
material point has undergone from time t0 to t.
The deformation equation for a point can be defined in the local (material)
coordinate system and can be mapped onto the spatial coordinate system. For
example, we can assume the following relationship (Figs. 2.28 and
2.29), between the material and spatial coordinates,
x = xA þ ðxB - xA Þr
y = yA þ ðyB - yA Þr
we will derive the finite strain formulation using Malvern’s (1969) formulation and
notation with few changes if any. We will define deformation–gradient tensor F as a
tensor whose components are the partial derivatives with respect to material coordi-
nates in the Lagrangian formulation. F, deformation–gradient tensor is defined in
terms of the undeformed configuration. x, y, z spatial coordinates are represented by
bold x. Local (material) coordinates r, s, t are represented by bold r.
The relation between the spatial and local (material) description deformations is
given by employing the deformation–gradient tensor F, which is defined as the
tensor whose rectangular Cartesian components are the partial derivatives ∂xi=∂ri and
which operates on an arbitrary infinitesimal material vector dr located at r, to
associate it with a vector dx located at point x as follows:
dx = F dr or dx = dr FT
In indicial matrix notation
∂xi
fdxi g = dr j
∂r j
Therefore
2 3
∂x ∂x ∂x
6 ∂r ∂s ∂t 7
6 7
∂xi 6 ∂y ∂y ∂y 7
F ij = =6
6 ∂r
7
∂r j 6 ∂s ∂t 7
7
4 ∂z ∂z ∂z 5
∂r ∂s ∂t
In Fij the first index identifies the row, and the second index identifies the column.
In Fig. 2.28 we see that the deformation is the difference between the position
vector before and after deformation, where c is the rigid body motion.
x=r þ u
2 3
∂u ∂u ∂u
6 ∂r ∂s ∂t 7
6 7
6 ∂v ∂v ∂v 7
F=6
6 ∂r
7þI
6 ∂s ∂t 7
7
4 ∂w ∂w ∂w 5
∂r ∂s ∂t
where I is the third-order unity matrix.

When Eulerian description is used, and deformation gradient is defined with
respect to deformed configuration (the spatial deformation gradient F-1) relations
between spatial coordinate x and local (material) coordinates r are given by
dr = F - 1 dx
or in indicial matrix notation
∂r i
fdr i g = dxj
∂xj
From the last equation, we can infer that the spatial deformation gradient, F-1 at
point x is the inverse of the material deformation–gradient tensor F,
Therefore,
F F-1 = I
can be written between the material deformation–gradient tensor and the spatial
deformation–gradient tensor, where I is the unit matrix.
Fig. 2.30 Axial strain

definition in
one-dimensional line
element
Strain is a human construct. It is not a real physical quantity that can be measured
directly, like displacement or time. It is defined to be able to formulate the response
of a continuum using mathematics. Because strain is not a direct physical quantity, it
can be defined in many ways. For large strain formulation the strain will be defined
as one-half the change in squared length of the local (material) vector AB as follows
(Fig. 2.30):
" 2 # " # " #
1 AB j - jAB 1 2drAB þ dr
0 2 2 2 2
1 jAB þ dr j - jABj dr 1 dr 2
2
= 2
= 2
= þ
2 jABj 2 jABj 2 jABj AB 2 AB
1 2
Axial strain = Er þ E
2 r
For Lagrangian formulation (referential description) in three dimensions, the

strain will be defined as
ðdsÞ2 - ðdSÞ2 = 2dr E dr
where dS = AB is the initial undeformed length and ds = AB′ is the deformed length
(Fig. 2.30). In indicial notation the last equation can be given by
ðdsÞ2 - ðdSÞ2 = 2dr i Eij dr j
For Eulerian formulation (spatial description)
ðdsÞ2 - ðdSÞ2 = 2dx E dx
in indicial notation
ðdsÞ2 - ðdSÞ2 = 2dxi Eij dxj

Example 2-5: Deformation–Gradient Tensor

In the following examples, it is assumed that the relation between material (local)
coordinates and spatial coordinates are given by linear interpolation functions.
For the quadrilateral element shown above material (local) coordinates of the
nodes [1, 2, 3, 4] are given as follows:
r 1 = 1,s1 = 1,r 2 = - 1,s2 = 1,r 3 = - 1,s3 = - 1,r 4 = 1,s4 = - 1
Nodal linear interpolation functions are given by
1
Ni = ð1 þ r i r Þð1 þ si sÞ
4
where i is the node number and ri, and si are the local coordinates of node i. Hence,
we can define the relation between spatial and material coordinates by
Xi = 4 Xi = 4
xðr, sÞ = N x , yðr, sÞ =
i=1 i i i=1
N i yi
Pure Stretching Deformation–Gradient Tensor

Pure stretching in the x and y directions. Assume 100% elongation in the x-direction
and a 50% elongation in the y-direction.
1 1
xðr, sÞ = ð1 þ r Þð1 þ sÞ 4 þ 0 þ 0 þ ð1 þ r Þð1 - sÞ4 = 2 þ 2r
4 4
1 1 3 3
yðr, sÞ = ð1 þ r Þð1 þ sÞ 3 þ ð1 - r Þð1 þ sÞ 3 þ 0 þ 0 = þ s
4 4 2 2
The deformation–gradient tensor is

2 3
∂x ∂x
6 ∂r ∂s 7 = 2 0
F=4 5
∂y ∂y 0 1:5
∂r ∂s
Simple Shear Deformation–Gradient Tensor
1 1
xðr, sÞ = ð 1 þ r Þ ð 1 þ sÞ 2 þ 0 þ 0 þ ð 1 þ r Þ ð 1 - sÞ 2 = 1 þ r
4 4
1 1 1 3 r
yðr, sÞ = ð1 þ r Þð1 þ sÞ 3 þ ð1 - rÞð1 þ sÞ 2 þ 0 þ ð1 þ r Þð1 - sÞ 1 = þ þ s
4 4 4 2 2

2 3
∂x ∂x
6 ∂r ∂s 7 = 1:0 0:0
F=4 5
∂y ∂y 0:5 1:0
∂r ∂s
The non-zero off-diagonal terms reflect shear, and 1.0 diagonal terms reflect zero
normal strain.
Pure Shear Deformation–Gradient Tensorr
1 1 1 3 1
xðr, sÞ = ð1 þ r Þð1 þ sÞ 3 þ ð1 - r Þð1 þ sÞ 1 þ 0 þ ð1 þ r Þð1 - sÞ2 = þ r þ s
4 4 4 2 2
1 1 1 3 r
yðr, sÞ = ð1 þ r Þð1 þ sÞ 3 þ ð1 - r Þð1 þ sÞ 2 þ 0 þ ð1 þ r Þð1 - sÞ 1 = þ þ s
4 4 4 2 2

2 3
∂x ∂x
6 ∂r ∂s 7 = 1:0 0:5
F=4 5
∂y ∂y 0:5 1:0
∂r ∂s
The non-zero off-diagonal terms reflect shear, and 1.0 diagonal terms reflect zero
normal strain.
2.6.1 Green Deformation Tensor, C, Cauchy Deformation

Tensor, B-1
Green deformation tensor C and Cauchy deformation tensor B-1 are related to the
strain tensors. Instead of giving one-half the change in squared length per unit
squared initial length, Green deformation tensor, C, refers to the undeformed
configuration, and its tensor entries provide the new squared length (ds)2 of the
element into which the given initial vector dr is deformed, while the Cauchy
deformation tensor, B-1, gives the initial squared length (dS)2 of a vector dx defined
in the deformed configuration.
As a result, we can establish the Green deformation, C tensor, C, as follows:
ðdsÞ2 = dr C dr
or in indicial notation as
ðdsÞ2 = dr i Cij dr j
In the same manner Cauchy deformation tensor B-1 can be given by
ðdSÞ2 = dx B - 1 dx
or in indicial notation as
-1
ðdSÞ2 = dxi Bij dxj
comparing Lagrangian strain tensor, E, and Green deformation tensor, C, we

observe that
2E = C - I or 2Eij = C ij - δij
In the same manner, comparing Eulerian strain tensor E and Cauchy deformation
tensor B-1 we observe that
-1
2E = I - B - 1 or 2E ij = δij - Bij
There is no special reason for defining the Cauchy deformation tensor as B-1. It
could be named just about any character. However, we are following the notation
used by Cauchy in 1827. In the formulation given above, both the Green deforma-
tion tensor, C, and Cauchy deformation tensor, B-1, reduce to unit tensor when
strain is zero (Malvern, 1969).
2.6.2 Relation Between Deformation, Strain,

and Deformation–Gradient Tensors
The square of the new length (ds)2 can be written as
ðdsÞ2 = dx dx
where dx is the vector in the deformed configuration. On the other hand relation
between the dx and the undeformed vector dr in the local (material) coordinates is
given by
dx = F dr
Hence, we can write

ðdsÞ2 = dr FT ðF drÞ = dr FT F dr
Similarly, the square of the original length (dS)2 can be written as
ðdSÞ2 = dr dr
Because dr = F-1 dx, we can write

h T i h T i
ðdSÞ2 = dx F - 1 F - 1 dx = dx F - 1 F - 1 dx
Let’s compare these later equations with our strain definition equations where
Green deformation tensor C is given by
ðdsÞ2 = dr C dr
Therefore, the Green deformation tensor is also defined by
C = FT F
and Cauchy deformation tensor is defined by
ðdSÞ2 = dx B - 1 dx
Therefore, we can write

h T - 1 i
B-1 = F-1 F
in indicial notation, we can write the following relation:
∂xk ∂xk
C ij =
∂r i r j
and
∂r k ∂r k
Bij- 1 =
∂xi ∂xj
Summation on repeated indices applies in both of these equations.

The Lagrangian strain tensor can then be obtained as follows:
2E = C - I
Since C = FT F
1 T
E= F F-I
2
or indicial notation
1 ∂xk ∂xk
Eij = - δij
2 ∂r i ∂r j
and Eulerian strain tensor can be given by
2E = I - B - 1
h T i
Since B-1 = F-1 F-1
h T i
1
E = I - F-1 F-1
2
In indicial notation
1 ∂r ∂r k
E⋆
ij = δ - k
2 ij ∂xi ∂xj
Both Green deformation tensor C and Lagrangian strain tensor E are symmetric
tensors. Therefore, they both have three eigenvalues (principal values) in Eigen
directions (principal directions). Also, principal directions of C and E coincide,
because in principal directions off-diagonal terms are zero in Green deformation
tensor C. Then they have to be zero in strain tensor E. Of course, the same arguments
can be made between Cauchy deformation tensor B-1 and Euler strain tensor E.
However, principal (Eigen) directions of [B-1 and E] and [C and E] will not
coincide.
2.6.3 Comparing Small Strain and Large (Finite) Strain
For the sake of simplicity, we will assume that material (local) coordinate axes and
spatial coordinate axes are parallel (Fig. 2.31).
xi = spatial coordinate, ri = local (material) coordinate.
Displacement ui is expressed in local (material) coordinates, ui = ui(r, s, t, time).
Fig. 2.31 Spatial and local

(material) coordinates
We can define the Lagrangian strain as follows:
1 ∂xk ∂xk
Eij = - δij
2 ∂r i ∂r j
Substituting xi = ri + ui where ui = ui(r, s, t, time)
1 ∂ui ∂uj ∂uk ∂uk

Eij = þ þ
2 ∂r j ∂r i ∂r i ∂r j

1 ∂u ∂u ∂v ∂v ∂w ∂w
E 11 = 1þ 1þ þ þ -1
2 ∂r ∂r ∂r ∂r ∂r ∂r
or
" 2 2 #
2
∂u 1 ∂u ∂v ∂w
E11 = þ þ þ
∂r 2 ∂r ∂r ∂r
Similarly
" 2 2 #
2
∂v 1 ∂u∂v ∂w
E 22 = þ þ þ
∂s 2 ∂s∂s ∂s
" 2 2 #
2
∂w 1 ∂u ∂v ∂w
E 33 = þ þ þ
∂t 2 ∂t ∂t ∂t
off-diagonal (shear-strain terms)

1 ∂u ∂u ∂v ∂v ∂w ∂w
E 12 = 1þ þ 1þ þ -0
2 ∂r ∂s ∂r ∂s ∂r ∂s
or

1 ∂u ∂v 1 ∂u ∂u ∂v ∂v ∂w ∂w
E12 = þ þ þ þ
2 ∂s ∂r 2 ∂r ∂s ∂r ∂s ∂r ∂s

1 ∂u ∂w 1 ∂u ∂u ∂v ∂v ∂w ∂w
E13 = þ þ þ þ
2 ∂t ∂r 2 ∂r ∂t ∂r ∂t ∂r ∂t

1 ∂v ∂w 1 ∂u ∂u ∂v ∂v ∂w ∂w
E 23 = þ þ þ þ
2 ∂t ∂s 2 ∂s ∂t ∂s ∂t ∂s ∂t
In indicial notation rj refers to r, s, t local coordinates; ui refers to

u, v, w displacement vectors in x, y, z spatial coordinates. The first term in large
(finite) strain formulation gives the small strain formulation.
In the same way, the Euler strain formulation can be given by
1 ∂ui ∂uj ∂uk ∂uk

Eij = þ -
2 ∂xj ∂xi ∂xi ∂xj
Again ui represents u, v, w for subscripts 1,2,3. xi represents x, y, z spatial

coordinates for 1, 2, and 3.
The only difference between Lagrangian strain tensor and Eulerian strain tensor is
the fact that in Lagrangian strain tensor all derivatives are with respect to local
(material) (undeformed) coordinates; on the other hand Eulerian strain tensor com-
ponents are with respect to spatial (deformed) coordinates.
For small displacement and small strain cases, the difference between the
Lagrangian strain and Eulerian strain is small.
Example 2.6
One corner, node 1, of the thin, t = 1 cm, steel plate shown below is deformed. All
other three corners are fixed. Calculate the Lagrangian strain tensor, E, and the
Eulerian strain tensor, E for the element, as a function of material (local) coordi-
nates using the deformation–gradient matrix, F. Assume node 3 coordinates are
y = 1 cm, x = 1 cm (Fig. 2.6E).
We assume that deformation has a linear distribution in the element. As a result,
the relations, between the material (local) coordinates and the spatial coordinates,
can be given by the following relations:
Fig. 2.6E Rectangle-

shaped thin plate elementt
1 1 1
x= ð1 þ rÞð1 þ sÞ 4 þ ð1 - r Þð1 þ sÞð1Þ þ ð1 - rÞð1 - sÞð1Þ
4 4 4
1
þ ð1 þ r Þð1 - sÞ ð3Þ
4
1 1 1
y = ð1 þ rÞð1 þ sÞ ð3:5Þ þ ð1 - r Þð1 þ sÞ ð3Þ þ ð1 - r Þð1 - sÞ ð1Þ
4 4 4
1
þ ð 1 þ r Þ ð 1 - s Þ ð 1Þ
4
Lagrangian strain tensor E is given by
1 T
E= F ∙F-I
2
Eulerian strain tensor E is given by

h T i
1
E = I - F-1 ∙ F-1
2
Deformation–gradient matrix F is given by

2 3
∂x ∂x
6 ∂r ∂s 7
F=4 5
∂y ∂y
∂r ∂s
x=u þ r
2 3
∂u ∂u
6 ∂r ∂s 7
6 7
F=6 7þI
4 ∂v ∂v 5
∂r ∂s
X4 X4 X4 X4
x= xN
i=1 i i
y= yN
i=1 i i
u= uN
i=1 i i
v= vN
i=1 i i
∂x X4 ∂N i ∂y X4 ∂N i
= x
i = 1 i ∂r
= y
∂r ∂r i = 1 i ∂r
∂x X4 ∂N i ∂y X4 ∂N i
= x
i = 1 i ∂s
= y
∂s ∂s i = 1 i ∂s
where xi and yi are coordinates of a point after deformation.

As a result, deformation–gradient matrix F is
1 ð10 þ 2sÞ ð2 þ 2r Þ
F=
8 ð 1 þ sÞ ð9 þ r Þ
Note that deformation–gradient matrix F is a function of material coordinates,

r and s. Because strain is a property of a point, we use material coordinates of the
point to define the strain at that point. Then we can calculate our strain tensors using
the following relations.
Lagrangian strain tensor is given by
1 T
E= F F-I
2
Eulerian strain tensor is given by

h T i
1
E = I - F-1 F-1
2
We can find strain at any point by substituting the local coordinates of the point
r = ri, s = si, for example, if we want to know the strain at node 1 where r = 1 and
s = 1, we substitute these coordinates for r and s.
2.6.4 Strain Rate and Rate of Deformation Tensor Relations
In Lagrangian formulation strain rate tensor E_ is given by
d d d
ðdsÞ2 - ðdSÞ2 = ð2dr E drÞ
dt dt dt
dr and dS are constant with respect to time because they represent the initial
dimensions. Therefore, we can write
d dE
ðdsÞ2 = 2dr dr
dt dt
On the other hand, the rate of deformation tensor D is given by
d
ðdsÞ2 = 2dx D dx
dt
Also, we have already shown that dx = F dr. Hence, we write the following
relation:
d
ðdsÞ2 = 2ðdrFT Þ D ðF drÞ
dt
= 2dr ðFT D FÞ dr
subtracting the strain rate equation from the rate of deformation
dE
2dr dr - 2dr FT D F dr = 0
dt
dE T
2dr - F D F dr = 0
dt
of course, dr = 0 is the trivial solution. Hence the following relation must be

satisfied:
dE
= FT D F
dt
must be satisfied or in indicial notation
dE ij ∂xm ∂x
= Dmn n
dt ∂r i ∂r j
The relationship between Green deformation rate tensor C and strain rate tensor
E_ can be given by
dC dE
=2
dt dt
When displacement–gradient components are small compared to unity, the strain

rate is equal to the rate of deformation.
dE
=D
dt
in indicial notation
dE ij
= Dij
dt
2.6.5 Relation Between, the Spatial Gradient of Velocity

Tensor, L and the Deformation–Gradient Tensor, F
The definition of deformation–gradient tensor F is given by
or dx = F dr
∂xi
dxi = dr j
∂r j
The rate of change of the deformation gradient is F_ . The spatial gradient of

velocity tensor L is given by
dv = L dx
Hence
∂~vi
Lij = ~vi,j =
∂xj
∂xi
If we write a time derivative of F = ∂r j
we obtain

_F = d ∂xi = ∂_xi = ∂_xi ∂xm
dt ∂r j ∂r j ∂xm ∂r j

_F = ∂vi ∂xm and since ∂vi = d ∂xi ∂r j
∂xm ∂r j ∂xm dt ∂r j ∂xm
F_ = L F and L = F_ F - 1
Relations between Euler strain rate tensor E, the rate of deformation tensor D,
and the spatial gradients of velocity tensor L can be given by

E_ = D - E L þ LT E
Proof of this equation is provided by Malvern (1969) on page 163.
2.7 Rotation and Stretch Tensors in Finite Strain
When strain is finite (large), the symmetric and skew-symmetric parts of the
displacement–gradient matrix, J, cannot be represented by additive decomposition
of a pure strain matrix and a pure rotation matrix. However, other types of
2.8 Compatibility Conditions in Continuum Mechanics 73
multiplication decompositions are possible, where one of the two tensors will
represent a rigid body rotation and the second tensor will be a symmetric positive-
definite.
2.8 Compatibility Conditions in Continuum Mechanics
When the displacements are known, the strain field can easily be calculated. For the
case of small strain, relations are given by

1 ∂ui ∂uj
Eij = þ
2 ∂r j ∂r i
where ui represent displacement vectors u, v, w, and ri represents local (material)

coordinate axes r, s, t. There are nine strain equations, but because of symmetry only
six of these are linearly independent.
For any given strain field to be admissible, certain compatibility conditions that
guarantee the continuum character of the medium must be satisfied. This require-
ment is also due to mathematics. If there are six known strain equations, it is not
possible to obtain three unknown displacement components as unique values. The
number of unknowns and the number of linearly independent equations must be the
same. St. Venant’s compatibility equations must be satisfied by the six strain
equations to find unique displacement values.
In a three-dimensional solid mechanics boundary value problem, St. Venant’s
compatibility equation provides us with six equations. There are also three force
equilibrium equations.
In addition, there are six stress-strain constitutive relations. As a result, in total
there are 15 equations and 12 unknowns (6 stresses and 6 strains). However, only
three of the compatibility equations are linearly independent. Hence, we have
12 equations and 12 unknowns.
However, if the displacements are unknown then compatibility equations are not
needed, because there are 15 equations (6 strain-displacement relations, 6 stress-
strain constitutive relations, and 3 force equilibrium equations). On the unknown
side, there are 3 unknown displacements, 6 unknown stresses, and 6 unknown
strains. Therefore, the number of equations is equal to the number of unknowns.
In most computational solid mechanics boundary value problems, usually, the
nodal displacements are the primary unknowns, as in the displacement-based finite
element method. As a result, we do not need the compatibility equations. However,
in force-based analysis methods, compatibility functions become necessary.
In summary, when displacements are not explicitly retained as the primary
unknowns, the compatibility conditions (this should be called the compatibility of
the displacement field) are needed to make sure that the strain field yields a
continuous displacement field that has a single displacement value at any point in
the continuum (Fig. 2.32).
Fig. 2.32 Strain-

displacement relation
When displacements and strains are small, the distinction between referential and
spatial descriptions is negligible for small strain calculations. Assuming small strain
and small deformations, we can define small strain in spatial x, y, z coordinate system
as follows:
∂u
Exx =
∂x
∂v
Eyy =
∂y
∂w
Ezz =
∂x

1 ∂u ∂v
Exy = þ
2 ∂y ∂x

1 ∂u ∂w
Exz = þ
2 ∂z ∂x

1 ∂v ∂w
Eyz = þ
2 ∂z ∂y
2
∂ E
If we assume that ∂x∂yxy exists
" #
2
∂ Exy 3
1 ∂ u
3
∂ v
= þ
∂x∂y 2 ∂2 y∂x ∂x2 ∂y
we can also write second derivatives of Exx and Eyy and sum them up
2
∂ Exx ∂ Eyy
2 3 3
∂ u ∂ v
þ = 2 þ 2
∂y 2 ∂x 2 ∂y ∂x ∂x ∂y
2.8 Compatibility Conditions in Continuum Mechanics 75
Hence
2 2
∂ Exx ∂ Eyy ∂ Exy
2
þ =2
∂y 2 ∂x 2 ∂x∂y
We can repeat this process for Exz and Eyz. Then we have
" #
2 2 2
∂ Exy ∂2 Ezx 1 ∂ ∂u ∂v ∂ ∂w ∂u
þ = þ þ þ
∂x∂z ∂y∂x 2 ∂x∂z ∂y ∂x ∂y∂x ∂x ∂z
3 3 3 3
1 ∂ u ∂ v ∂ w ∂ u
= þ þ þ
2 ∂x∂y∂z ∂x2 ∂y ∂x2 ∂y ∂x∂y∂z
2 2
1 ∂ ∂u ∂ ∂v ∂w
= 2 þ 2 þ
2 ∂y∂z ∂x ∂x ∂y ∂y
" #
2
1 ∂ Exx ∂ Eyz
2
= 2 þ :
2 ∂y∂z ∂x2

∂ ∂Exy ∂ 1 ∂Eyz
2
∂ Exx
= þ -
∂y∂z ∂x ∂z ∂y 2 ∂x
A similar process can be repeated for other shear strain pairs Eyz and Exy, Eyz and
Exz.
These equations prove the integrability of the strain field. Based on this derivation
the following six St. Venant’s compatibility equations can be given:
2 2
∂ Exx ∂ Eyy ∂ Exy
2
þ -2 =0
∂y 2 ∂x 2 ∂x∂y
2 2
∂ Eyy ∂2 Ezz ∂ Eyz
þ -2 =0
∂z2 ∂y 2 ∂y∂z
2 2 2
∂ Ezz ∂ Exx ∂ Ezx
þ -2 =0
∂x2 ∂z2 ∂z∂x

2
∂ Exx ∂ ∂Eyz ∂Ezx ∂Exy
- þ - þ þ =0
∂y∂z ∂x ∂x ∂y ∂z
2
∂ Eyy ∂ ∂Eyz ∂Ezx ∂Exy
- þ - þ =0
∂z∂x ∂y ∂x ∂y ∂z

∂ ∂Eyz ∂Ezx ∂Exy
2
∂ Ezz
- þ þ - =0
∂x∂y ∂z ∂x ∂y ∂z
These six compatibility equations must be satisfied to ensure a compatible strain

field and to ensure that only a single-valued displacement field exists at any point.
Therefore, the compatibility conditions are necessary and sufficient conditions for
unique displacement values.
While there are six compatibility equations, however only three of them can be
linearly independent because they are obtained from three independent displace-
ments u, v, w.
2.9 Piola–Kirchhoff Stress Tensors
Cauchy stress tensor is defined in the spatial coordinates x, y, z (in the deformed
configuration). Euler strain is also defined in spatial position in the deformed
configuration. Therefore, using the Euler strain tensor definition with the Cauchy
stress tensor is appropriate.
On the other hand, the Lagrangian formulation is defined in material coordinates
r, s, t (in the undeformed configuration) or any other reference state configuration
that does not change. In this case, it is also more appropriate to define the Piola–
Kirchhoff stress tensor in local (material) reference coordinates. Two Piola–Kirch-
hoff stress definitions are using the undeformed (or any other stationary reference)
state as the basis for formulation.
2.9.1 First Piola–Kirchhoff Stress Tensor σ0
The First Piola–Kirchhoff stress tensor is also called the Lagrangian stress tensor.
The derivatives are defined with respect to material (local) coordinates. However,
this stress tensor is not symmetric, even when there is no distributed body moment
(non-polar case) in the system.
Stress is defined as a force per unit undeformed area. Keep in mind that stress is a
human construct not a directly measurable physical quantity. Thus, it is possible to
define it as we please. The first Piola–Kirchhoff stress tensor is defined by the
following equation (Fig. 2.33):
dF
n0 σ 0 =
dA0
where σ 0 is the first Piola–Kirchhoff stress tensor and dF is the actual force acting
on the deformed configuration dA; however, the force is divided by the initial
undeformed area dA0.
2.9 Piola–Kirchhoff Stress Tensors 77
Fig. 2.33 Piola–Kirchhoff stress definition
dF can also be written using Cauchy stress tensor σ as dF = (n σ)dA. Hence we

can write the following equation between the first Piola–Kirchhoff stress tensor and
Cauchy stress tensor,
ðn0 σ 0 ÞdA0 = dF = ðn σ ÞdA
2.9.2 Second Piola–Kirchhoff Stress Tensor σ~
~ that is related to the

Second Piola–Kirchhoff stress tensor, σ~ uses a pseudo-force d F
actual force dF by the inverse of the deformation–gradient tensor F-1.
~ = F - 1 dF
dF
Remember that material (local) coordinates and spatial coordinates are also
related by
dr = F - 1 dx
Then we can write
~
dF
ðn0 σ~Þ =
dA0
Using the relation, we defined above, we can write
F - 1 dF
ðn0 σ~Þ =
dA0
At the same time, dF can be also defined in terms of Cauchy stress tensor as
T
F - 1 ðn σ ÞdA ðn σ Þ F - 1 dA
ðn0 σ~Þ = =
dA0 dA0
The Piola–Kirchhoff stress tensors are sometimes called pseudo-stress tensors.
2.10 Direct Relation Between Cauchy Stress Tensor

and Piola–Kirchhoff Stress Tensors
The following relation can be given:
ρ0 - 1 T T
σ~ = F σ F - 1 and σ~ = σ 0 F - 1
ρ
Details of the long derivations for these relations are provided by Malvern (1969).
The second Piola–Kirchhoff stress tensor is usually preferred in finite-strain
elasticity problems.
Newtonian equation of motion in undeformed (reference) state using the first
Piola–Kirchhoff stress tensor is given by
Z Z Z
d2 x
n0 σ dA þ
0
ρ0 b0 dV = ρ0 dV
dt 2
A0 V0 V0
Transforming the surface integral to volume integral by the Gauss theorem

(divergence theorem) in indicial notation, we can write
∂σ 0ij d2 x
þ ρ0 b0j = ρ0 2i
∂r i dt
where ri is the material (reference) coordinate system axis. Since the relation
between the first and second Piola–Kirchhoff stress tensors is given by
σ 0 = σ~ FT
or in indicial notation
σ0ij = ˜σik xi,k

2.11 Conservation of Mass Principle 79
by substituting these equations in the equilibrium equations given above, we obtain

the equation of motion in terms of the second Piola–Kirchhoff stress tensor:
d 2 xk
½˜σij xk,j i þ ρ0 b0k = ρ0
dt 2
2.11 Conservation of Mass Principle
The rate of change in mass is given by

Z
∂m ∂ρ
= dV
∂t ∂t
V
where ρ is the material mass density, which can be a function of x, y, z coordinates

and time. The flux is the rate of mass flowing through an area dA with a velocity of
~vn = ~v n can be given by
Z Z
ρ~vn dA = ρ~v n dA
A A
Using the divergence theorem (Gauss’s theorem) the integral over a closed
surface can be converted to an integral over a volume bounded by the closed surface
by the following equation:
Z Z
n × ~v dA = — × ~vdV
A V
where n is the vector normal to the surface, ~v is velocity vector — × ~v is given by

i j k

∂ ∂ ∂
~
—×v=

∂x ∂y ∂z
~v ~vy ~vz
x
where i, j, k are the unit vectors in Cartesian coordinates x, y, z. Hence,

Z Z
ρ~v n dA = — ðρ~vÞ dV
A V
The time rate of change in mass is defined by

Z
∂ρ
dV
∂t
V
Equating the time rate of change in mass to the rate of inflow through area A
yields
Z Z
∂ρ
dV = - — ðρ~vÞ dV
∂t
V V
The negative on the right-hand side is because normal of the surface is defined as
positive outward. Therefore, inflow is in the negative normal direction.
Hence, we can write
Z
∂ρ
þ — ðρ~vÞ dV = 0
∂t
V
Here the integral must be zero for an arbitrary volume. Therefore the integrand
must be equal to zero.
∂ρ
þ — ðρ~vÞ = 0
∂t
This equation is called the continuity equation, as a result of the conservation of

mass principle. Or in indicial notation, it can be given as
∂ρ ∂~vx ∂~vy ∂~vz

þρ þ þ =0
∂t ∂x ∂y ∂z
2.12 The Incompressible Materials
When the material is incompressible, we assume that density does not change over
time,
dρ
=0
dt
2.13 Conservation of Momentum Principle 81
Therefore
∂~vx ∂~vx ∂~vz

þ þ =0
∂x ∂y ∂z
must be satisfied. It is important to point out that in the theory of plasticity during
plastic deformation (flow), some materials, like metals, are considered to be
incompressible.
2.13 Conservation of Momentum Principle
The equilibrium of forces equation is obtained by using the conservation of momentum

principle. This principle applies to a collection of particles as well as a continuous medium.
This principle states that the vector sum of all the external forces acting on the free-body is
equal to the rate of change of the total momentum (Malvern, 1969).
Force is the time derivative of momentum; therefore the concept of force is

redundant and subordinate to the conservation of momentum. In this section, we
will derive the equilibrium of forces from the conservation of momentum. According
to the conservation of momentum principle for a collection of particles the time rate
of change of the total momentum of a given set of particles equals the vector sum of
all the external forces acting on the particles is considered to be the modern
interpretation of Newton’s second law. It is assumed that Newton’s third law of
action and reaction and Hooke’s law govern the internal forces.
The continuum form of conservation of momentum principle is the basis for the
Newtonian continuum mechanics. In this book, the term “Newtonian Mechanics”
refers to all mechanics formulations based on Newton’s universal laws of motion.
Consider the cube shown in Fig. 2.34, occupying volume V and bounded by
surface A. There is an external surface loading vector σ (n), per unit area on any
arbitrary plane, and b is the body force per unit mass.
Fig. 2.34 Conservation of

momentum
Momentum, p, is defined by
Z
p= ρ ~v dV
where ρ is mass density, ~v is velocity vector, and dV is the volume. The time rate of
change of the total momentum of a portion of the particles can be given by the
material derivative of the integral, which is defined as the time rate of change of any
quantity for a portion of material. Hence, the time rate of change of the total
momentum is
Z
dp d
= ρ~vdV
dt dt
External forces acting on the particles can be given by the summation of surface
tractions and body forces
Z Z
σ ðnÞ dA þ ρbdV
A V
According to the conservation of momentum principle, the time rate of change of

the total momentum equals the vector sum of all the external forces
Z Z Z
d
σ ðnÞ dA þ ρbdV = ρ~vdV
A V dt
Using a Cartesian coordinate system where x, y, and z are represented by indices i.

Z Z Z
ðnÞ d
σ i dA þ ρbi dV = ρ~
vi dV
A V dt
Components of external surface loading vector σ (n) can be given by
ðnÞ
σ i = σ ij nj
Or in Cartesian coordinates
σ ðxnÞ = σ xx nx þ σ xy ny þ σ xz nz
σ ðynÞ = σ yx nx þ σ yy ny þ σ yz nz
σ ðznÞ = σ zx nx þ σ zy nz þ σ zz nz
2.13 Conservation of Momentum Principle 83
where nx,nx,nx, are direction cosines between the normal to the arbitrary plane and
spatial coordinates. In matrix notation, this can also be written as
n o
σ ðnÞ = fng½σ
2 3
n oT n oT σxx σxy σxz
ðnÞ ðnÞ ðnÞ ðnÞ 6 7
σ = σx σy σz fng = fnx ny nz g ½σ = 4 σyx σyy σyz 5
σzx σzy σzz
[σ] is the second-order Cauchy stress tensor, which is a linear vector function.
Using the divergence theorem, we can write the following conversion:
Z Z
ðnÞ ∂σ ji
σ i dA = dV
A V ∂xj
Meanwhile, due to the conservation of mass material time derivative of volume

integral can be written as Malvern (1969):
Z Z
d d~vi
ρ~vi dV = ρ dV
dt V dt
Substituting the last two relations in the conservation of momentum equation

yields
Z Z
∂σ ji d~v
þ ρbi dV = ρ i dV
V ∂xj V dt
Hence the conservation of momentum principle in Newtonian mechanics takes

the final form in the following equation:
Z
∂σ ji d~vi
þ ρbi - ρ dV =
V ∂xj dt
For any arbitrary volume, we can write Cauchy’s equation of motion in Newto-
nian mechanics as follows:
∂σ ji d~v
þ ρbi = ρ i
∂xj dt
Of course, for static equilibrium right-hand side of this equation is equal to zero.
Hence
∂σ ji
þ ρbi = 0
∂xj
In any other Cartesian coordinate system rotated with respect to the original
coordinate system stress tensor can be defined using the tensor transformation
equations.
½σ = ½N T ½σ ½N
where [N] is the matrix of direction cosines nik = cos ðxr , xi Þ of the angles between
the axes of the original and rotated coordinate systems. Superscript T is used to
denote transpose. A bold character indicates a matrix or a vector. Sometimes
brackets are used to distinguish a matrix from a vector.
2.14 Conservation of Moment of Momentum Principle
The time rate of change of the total moment of momentum for a collection of masses
is equal to the vector sum of the moments of the external forces acting on these
masses. Assuming there are no distributed moment couples, we can write
Z Z Z
d ðnÞ
ðr × ρ~vÞdV = r×σ dA þ ðr × ρbÞdV
dt
v A V
using vector product definition
a × b = eijr aj br = ejri aj br
we can re-write the conservation of momentum of moment principle in indicial

notation as follows:
Z Z Z
d
eijr xj ρ~vr dV = eijr xj σ ðrnÞ dA þ eijr xj br ρdV
dt
V A V
We have defined σ ðrnÞ previously for an arbitrary direction as
ðnÞ
σ i = σ ij nj
We can also transform the surface integral to a volume integral using the relation
given by the divergence (Gauss) theorem
Z Z
∂
eijk nj~vk dA = eijk ð~vk ÞdV
∂xj
A V
2.14 Conservation of Moment of Momentum Principle 85
where ρ is the mass density of the material per unit volume. We can transform the
moment of momentum principle into the following form:
Z Z
d ∂ xj σ kr
eijr x ~v ρdV = eijr þ xj ρbr dV
dt j r ∂xk
V V
The time derivative of displacement is velocity; hence we can substitute ~vi = dxi
dt
Z Z
d~v ∂σ kr
eijr ~vj~vr þ xj r ρdV = eijr xj þ ρbr þ σ kr δjk dV
dt ∂xk
V V
eijr ~vj~vr = 0 because ~vj~vr is symmetric for indices jr, while eijr is anti-symmetric.
Knowing that Cauchy’s equation of motion is given by
∂σ kr d~v
þ ρbr = ρ r
∂xk dt
Substituting this equilibrium equation of motion above yields

Z
eijr σ kr δjk dV = 0
V
which is equal to
Z
eijr σ jr dV = 0
V
However, the last equation is equal to zero. Because for an arbitrary volume, V, at
each point
eijr σ jr = 0
for
• i = 1: σ 23 = σ 32
• i = 2: σ 31 = σ 13
• i = 3: σ 12 = σ 21
This proves that the stress tensor is symmetric when there are no couple (moment)
stresses. This proof also establishes the symmetry of the stress tensor in general
without any assumption of equilibrium or uniformity of the stress distribution.
2.15 Lagrangian Mechanics
In this section, we intend to introduce Lagrangian mechanics briefly. Lagrangian

mechanics introduced in 1788 by Joseph-Louis Lagrange is a different formulation
of classical Newtonian mechanics. Lagrangian mechanics allows the inclusion of
non-conservative forces. However, it does not establish any additional laws of
mechanics in addition to Newton’s.
Lagrangian mechanics uses energy as the variable rather than forces. The main
equation in non-relativistic Lagrangian mechanics is the Lagrangian, defined by the
following equation:
L=T -V
where T is the kinetic energy of the system, and V is the potential energy of the
system. The minus sign in the equation is counterintuitive to the total energy of the
system. So Lagrangian is not the total energy. There is no unique Lagrangian for all
physical systems. Any function which yields the equation of motion can be defined
as a Lagrangian. Also, for dissipative forces, another function must be introduced
alongside L, if Newtonian mechanics is used that dissipative force is usually an
empirical function. In unified mechanics theory, we will introduce the thermody-
namic state index into the Lagrangian; hence, there is no need for an empirical
dissipation/degradation constraint functional alongside the Lagrangian.
Lagrangian mechanics can only be applied to systems whose constraints, if any, are all
holonomic. Three examples of nonholonomic constraints are, when the constraint equations
are nonintegrable when the constraints have inequalities, or with complicated
non-conservative forces like friction. In contrast, the unified mechanics theory does not
have this restriction. Unified mechanics theory applies to all systems, holonomic and
nonholonomic. Because there are no constraint functionals in the unified mechanics theory.
Nonholonomic constraints require special treatment, and one may have to revert to Newto-
nian mechanics, or use other methods (Hand & Finch, 2008).
The Lagrange equations of the first kind for a single particle are given by
X n
∂L d ∂L ∂f i
- þ λi =0
∂x dt ∂x_ i
∂x
where coordinates x, y, z are represented by x, dot over x_ indicates time derivative, λi

are the Lagrange multiplier for the constraint (dissipation) function fi, and n is the
number of constraints needed to satisfy the equilibrium. The kinetic energy is
defined by the velocity using the material time derivative of its position with respect
to time.
The Lagrange equation of the second type is referred to as the Euler-Lagrange
equation, given by

∂L d ∂L
- =0
∂r dt ∂_r
2.15 Lagrangian Mechanics 87
where r is the vector of a point in the material configuration space. The spatial
position coordinates x, y, z, can be represented as functions of the material (gener-
alized) coordinates r, s, t. . . . and time. Generalized coordinates can be more than
3 that is the reason for additional dots. However, for a simple force deformation
mechanics problem, they can be the minimum r, s, t.
Example 2.7
For a mass m attached to the tip of a linear-elastic 1-dimensional spring, Fig. 2.7E,
obtain the equilibrium equation using the Lagrangian mechanics.
For this problem, we can use the Euler-Lagrange equation. Assume that gener-
alized coordinates and position coordinates have the same origin and we have small
deformations. Then we can write

∂L d ∂L
- =0
∂x dt ∂x_
1 1
L=T -V = m_x2 - kx2
2 2
Substituting Lagrangian in the Euler-Lagrange equation above yields
- kx - m€x = 0
which is the un-damped dynamic equilibrium equation of motion. We can of course

obtain the same equation using the second and third laws of Newton and
Hooke’s law.
If we use the Lagrange equation of the first kind and add empirical damping
[dissipation] constraint function as follows:
f = λc_xx
Fig. 2.7E Single degree of

freedom system
X
m
Taking Lagrange multiplier λ = - 1, using the Lagrange equations of the first

kind we arrive at
- kx - m€x - c_x = 0
It is important to point out that the constraint function is not unique but empirical
based on test data.
2.16 Hamilton’s Principle: The Principle of Stationary

Action
We do not intend to make this section comprehensive, but just introductory. The
concept has been discussed in detail by Morin (2007). We intend to quote a summary
from Morin (2007).
Let’s define a quantity called the action, S, with dimensions of (energy) × (time)
and given by the following equation:
Z t2
S= Ldt
t1
where L is the Lagrangian and dt is the time interval. Let’s assume we are dealing
with a simple one-dimensional problem, and x is the only coordinate.
Let’s pose the following question: What function x(t) yields a stationary value of
S. A stationary value is a local minimum, the minimum, or a saddle point. The
answer is given by the following theorem.
Theorem If the function x0(t) yields a stationary value of S, then the following
relation must be satisfied.

d ∂L ∂L
=
dt ∂_x0 ∂x0
It is assumed that we are considering only functions that have their endpoints
fixed. Meaning
xðt 1 Þ = x1 and xðt 2 Þ = x2
Finally, Hamilton’s principle (the principle of stationary action) can be given by

the following definition:
The path of a particle is the one that yields a stationary value of the action, S.
This principle is equivalent to F = ma because the above theorem shows that if and only if
we have a stationary value of S, then the Euler-Lagrange equation holds. And Euler-
Lagrange equations are equivalent to F = ma. Therefore. “Stationary action” is equivalent
to F = ma (Morin, 2007).
References 89
Fig. 2.8E Ball

dropped from a platform
Example 2.8
Consider a ball being thrown from a balcony to the ground. There are many possible
paths the ball can take as shown in Fig. 2.8E. The vertical distance of the ball from
the balcony at any time is defined by x(t). Each path has a different function x(t). We
can substitute each one of the x(t) functions in Lagrangian to yield an S value.
Z t2
S= Ldt
t1
However, only one of the paths xðt Þ = 12 gt 2 will yield a stationary value of the
action, S.
Here we conclude the discussion on basic concepts of continuum mechanics.
References
[In Latin] Isaac Newton’s Philosophiae Naturalis Principia Mathematica: the Third edition (1726)
with variant readings, assembled and ed. by Alexandre Koyré and I Bernard Cohen with the
assistance of Anne Whitman (Cambridge, MA, 1972, Harvard UP). Among versions of the
Principia online: Volume 1 of the 1729 English translation is available as an online scan; limited
parts of the 1729 translation (misidentified as based on the 1687 edition) have also been
transcribed online.
Bathe, K. J. (1996). Finite element procedures. Prentice-Hall.
Beatty, M. F. (2006). Principles of engineering mechanics volume 2 of principles of engineering
mechanics: Dynamics-the analysis of motion (p. 24). Springer. ISBN 0-387-23704-6.
Cohen, I. B. (1967). Newton’s Second Law and the concept of force in the principia. In The annus
Mirabilis of Sir Isaac Newton 1666–1966. The MIT Press.
Cohen, I. B. (1995). Science and the Founding fathers: Science in the political thought of Jefferson,
Franklin, Adams, and Madison (p. 117). W.W. Norton. ISBN 978-0-393-24715-2.
Cohen, I. B. (P. M. Harman, & A. E. Shapiro, Eds.). (2002). The investigation of difficult things:
Essays on Newton and the history of the exact sciences in honor of D.T. Whiteside (p. 353).
Cambridge University Press. ISBN 0-521-89266-X.
Fairlie, G., & Cayley, E. (1965). The life of a genius (p. 163). Hodder and Stoughton.
Galili, I., & Tseitlin, M. (2003). Newton’s first law: Text, translations, interpretations and physics
education. Science & Education., 12(1), 45–73. Bibcode:2003Sc&Ed..12...45G. https://doi.org/
10.1023/A:1022632600805
Hand, L. N., & Finch, J. D. (2008). Analytical mechanics. Cambridge University Press.
Hellingman, C. (1992). Newton’s third law revisited. Physics Education, 27(2), 112–115.
Bibcode:1992PhyEd..27..112H. https://doi.org/10.1088/0031-9120/27/2/011
Hobbes, T. (1651). Leviathan. Oxford University Press.
Kleppner, D., & Kolenkow, R. (1973). An introduction to mechanics (pp. 133–134). McGraw-Hill.
ISBN 0-07-035048-5.
Lubliner, J. (2008). Plasticity theory (Rev. ed.) (PDF). Dover Publications. ISBN 0-86-46290-0.
Archived from the original (PDF) on 31 March 2010.
Malvern, L. (1969). E introduction to the mechanics of continuous medium. Prentice-Hall.
Morin, D. (2007). Introduction to classical mechanics: With problems and solutions. Cambridge
University Press.
Newton, I. (1999). The principia. A new translation by I. B. Cohen and A. Whitman, University of
California Press, Berkeley.
Newton, I. (1687). Philosophiæ Naturalis Principia Mathematica (Newton’s personally annotated
1st edition)
Newton, I. Principia, Corollary III to the laws of motion
Plastino, A. R., & Muzzio, J. C. (1992). On the use and abuse of Newton’s second law for variable
mass problems. Celestial Mechanics and Dynamical Astronomy. Netherlands: Kluwer Aca-
demic Publishers, 53(3), 227–232. Bibcode:1992CeMDA..53..227P.
Resnick, & Halliday. (1977). Physics (3rd ed., pp. 78–79). Wiley. Any single force is only one
aspect of mutual interaction between two bodies.
Resnick, R., Halliday, D., & Krane, K. S. (1992). Physics (Vol. 1, 4th ed., p. 83). Wiley.
The Mathematical Principles of Natural Philosophy. (1687). Encyclopædia Britannica
Thornton, M. (2004). Classical dynamics of particles and systems (5th ed., p. 53). Brooks/Cole.
ISBN 0-534-40896-6.
Wikipedia Sir Isaac Newton webpage. (2001). https://en.wikipedia.org/wiki/Isaac_Newton
Chapter 3
Thermodynamics
Newtonian mechanics is a study of the determination of the location of points in

space-time (x, y, z axes and time) coordinate system after being subjected to external
forces. Since Newton’s 1687 formulations, many other mechanics theories were
proposed. However, they are all based on Newton’s universal laws of motion and are
referred to as Newtonian mechanics. In this book term “Newtonian mechanics” is
used to refer to all mechanics theories that use Newton’s universal laws of motion.
In the universal laws of motion of Newtonian mechanics, an object being studied
is assumed to be ageless, and energy loss (entropy generation) is not considered.
Lagrangian mechanics considers dissipation by adding an empirical constraint
function to the Lagrangian.
Thermodynamics is about the past, present, and future of matter, under a given set
of initial conditions, boundary conditions, and loading paths. Directly quoting
Callen’s (1985) excellent description “Thermodynamics is concerned with the
macroscopic consequences of the myriads of atomic coordinates [events] that, by
the coarseness of macroscopic observations, do not appear explicitly in a macro-
scopic description of the system.” Therefore, thermodynamics studies of nature at
the macroscopic scale reflect what happens at the atomic scale. However, “Among
the many consequences of the ‘hidden’ atomic modes of motion, the most evident is
the ability of these modes to act as a repository for energy” Callen (1985).
Thermodynamics is a universal subject. As such, it is assumed that laws of
thermodynamics are universally valid for all systems at the macro level under
equilibrium conditions.
Therefore, we can treat any structure or system or any piece of material as a
thermodynamic system. The size of the system and time scale is very important for
the laws of thermodynamics. In this book, we are only concerned with continuum
mechanics where the universal laws of motion of Newton are valid. The boundary
between quantum mechanics and continuum mechanics is not an easily quantifiable
one. There is a gray area. However, continuum mechanics is the primary topic in
this book.

https://doi.org/10.1007/978-3-031-18621-9_3
92 3 Thermodynamics
When we treat a continuum system (solid or fluid), we will assume that is a closed
system. This means that for a given period (time increment) there is no exchange of
matter or energy with its surroundings. This can be considered as a fixed control box
in space and time. The control box is insulated, to be able to satisfy thermodynamic
conservation laws.
3.1 Thermodynamic Equilibrium
All macroscopic systems have “memory.” This memory is always with respect to
their stress-free state, which is the minimum energy state. All macroscopic systems
tend to evolve toward a minimum energy point, which is defined by their intrinsic
properties. Evolution towards these minimum energy points can be slow or fast,
depending on internal and external factors.
These minimum energy points of the system are called thermodynamic equilib-
rium states. These states usually are asymptotic states. Hence, they are static. Callen
(1985) provides a good description of thermodynamic equilibrium from the atomic
point of view, “the macroscopic thermodynamic equilibrium state is associated with
incessant and rapid transitions among all the atomic states consistent with the given
boundary conditions. If the transition mechanism among the atomic states is suffi-
ciently effective, the system passes rapidly through all representative atomic states in
the course of a macroscopic observation; such a system is in thermodynamic
equilibrium. In actuality, few systems are in absolute and true thermodynamic
equilibrium. In absolute thermodynamic equilibrium, from the atomic point of
view, all radioactive materials would have decayed completely and nuclear reactions
would have transmuted all nuclei to the most stable of isotopes. Such processes
would take cosmic times to complete, and generally can be ignored. A system that
has completed the relevant [significant] processes of spontaneous evolution and that
can be described by a reasonably small number of parameters can be considered to
be in metastable thermodynamic equilibrium. Such a limited thermodynamic equi-
librium is sufficient for the application of [laws of] thermodynamics.” This is the
definition of thermodynamics equilibrium we use in the rest of the book.
From a practical point of view, a system is in equilibrium if it satisfies the laws of
thermodynamics. However, this justification is circular. That is to say if a system
satisfies the laws of thermodynamics it is said to be in equilibrium.
A succinct definition of thermodynamics with an example is given by Callen
(1985) “Thermodynamics is the determination of the equilibrium state that eventu-
ally results in a closed system” (Fig. 3.1).
Let us assume we have a container separated into two sections with a moveable rigid wall.
The container wall is assumed to be impermeable to matter, and adiabatic (heat does not
enter or leave the container). This container is the definition of a closed system in thermo-
dynamics. Initially, the separation wall is fixed. If we release the separation wall, it will move
to a new location due to the gradient of pressure on each side of the wall.
3.2 First Law of Thermodynamics 93
Fig. 3.1 Thermodynamic chambers. P pressure, V volume, N number of particles, T temperature,

C concentration
Then if we remove the adiabatic coating from the separation wall, the heat can freely flow
between two sections of the container. Now we drill holes in the separation wall, the matter
can flow freely between two sections depending on the concentration gradient. Every time a
constraint is removed, a spontaneous process will take place that will result in a new
equilibrium state. Initial values of pressure, volume, temperature, chemical concentration,
and all other parameters in each chamber will take on new thermodynamic equilibrium
values. The primary problem in thermodynamics is the computation of these parameters at
the thermodynamic equilibrium.
It is important to point out again, what we mean by a closed system. Because our
thermodynamic equilibrium is defined for a closed system. According to thermodynamics,
a system is considered to be closed if it cannot exchange any energy, matter, or anything
whatsoever with its surrounding. This assumption does not prevent us from solving any
engineering problem with fluctuating loads or masses. Because the problem is always solved
in incremental time steps. At each step, equilibrium must be satisfied (Callen, 1985).
3.2 First Law of Thermodynamics
This law of thermodynamics states that total energy is conserved in every process.
Therefore, it is sometimes referred to as the law of conservation of energy. The law is
independent of the path taken between the initial and final states of the system.
This law is generalized from experimental observations. Therefore, it is an
empirical conclusion. While mathematically it may be difficult to prove this law
universally, we assume its validity because there is no way to disprove it mathemat-
ically or experimentally [at least until now]. Now, we can formulate the first law of
thermodynamics for application in continuum mechanics. The conservation of
energy can be written as
d
ðu þ kÞ = W input þ Qinput ð3:1Þ
dt
94 3 Thermodynamics
where u is the internal potential energy, k is macro-level kinetic energy, Winput is the
work done by the system, due to external loads, and Qinput is energy added to the
system by heat transfer. Both Winput and Qinput do not have exact differentials.
Therefore,
I
W input dt ≠ 0 ð3:2Þ
and
I
Qinput dt ≠ 0 ð3:3Þ
However
I

W input þ Qinput dt = 0: ð3:4Þ
Here, exact differential means that there exists a continuously differentiable

function, called the potential function. In physical terms, if a system is loaded heated,
H returned to its initial state, Winput and Qinput, will not be recovered completely.
and
dt denotes the integral through the loading cycle. However, the total energy of the
system is an exact differential
I
ðdu þ dk Þdt = 0 ð3:5Þ
3.2.1 Work Done on the System (Power Input)
The work done on the system by external (forces) surface tractions and body forces
can be given by
Z Z
W input = σ ðnÞ ~v dA þ ρb ~v dV ð3:6Þ
A V
where σ (n) is the external surface traction per unit area, ~v is the velocity field in the
body, and b is the per unit mass body forces, and ρ is the mass density. Concentrated
point loads are also represented by surface tractions. We can write this equation in
indicial notation as
Z Z
ðnÞ
W input = σ j ~vj dA þ ρ bj~vj dV ð3:7Þ
A V
ðnÞ
The relation between external surface traction σ j and internal stresses σ ij are
given by
ðnÞ
σ j = σ ij ni ð3:8Þ
The surface integral can be transformed to volume integral by the divergence

theorem:
Z Z
ðnÞ ∂σ ij ∂~vj
σ j ~vj dA = ~vj þ σ ij dV ð3:9Þ
∂xi ∂xi
A V
Hence Winput can be written as

Z
∂σ ij ∂~vj
W input = ~vj þ ρbj þ σ ij dV ð3:10Þ
∂xi ∂xi
V
where xj denotes Cartesian coordinate system axes. Cauchy’s equation of motion is

given by

∂σ ij d~vj
þ ρbj = ρ ð3:11Þ
∂xi dt
We can substitute this equilibrium equation in Winput, which yields

Z
d~vj ∂~vj
W input = ~vj ρ þ σ ij dV ð3:12Þ
dt ∂xi
V

d~v
vj ρ dtj can be written as

d~vj d 1
~vj ρ = ρ~v ~v ð3:13Þ
dt dt 2 j j
Hence,
Z h i Z
d 1 ∂~vj
W input = ρ~v ~v dV þ σ ij dV ð3:14Þ
dt 2 j j ∂xi
V V
∂v
where ∂xji is the spatial gradient of velocity.
96 3 Thermodynamics
If we assume that there are no distributed moments (couple stresses) and/or

polarized fields acting on the system, then, the stress tensor σ ij is symmetric. Earlier
spatial gradient of velocity tensor L was given by
∂~vi
=L=D þ W ð3:15Þ
∂xj
where D is the rate of deformation tensor and W is the rate of spin tensor.
We have shown that for non-polar cases the stress matrix σ ij is symmetric; on the
other hand, the rate of spin tensor Wij is skew-symmetric. Therefore, we can write the
following relations:
σ ij W ij = 0 and σ ij Dij ≠ 0 ð3:16Þ
Substituting these equations into Winput leads to

Z Z
d 1
W input = ρ~v ~v dV þ σ ij Dij dV ð3:17Þ
dt 2 j j
V V
The first term represents the kinetic energy, and the second term represents the
internal (strain) energy of the system. Here we should point out that this equation
is based on Cauchy’s equilibrium equation, which is based on Newtonian
mechanics laws.
3.2.2 Heat Input
The heat input, Q, has two parts. One is heat conduction through the surface coming
from outside, and the second part is the distributed internal heat generation with a
source strength of r per unit mass.
Internal heat generation is due to internal scattering, friction, or chemical reac-
tions. These two components of the heat travel in opposite directions. One is coming
from outside through the surface, and the second part is directly generated by the
material
Z Z
Qinput = q ndA - ρrdV ð3:18Þ
A V
where q is the heat flux vector, and n is the surface normal vector. The negative sign
due to internal heat generation is in the opposite direction of the heat coming from
outside, and it is outward. More importantly, r is generated internally by the system
due to atomic level scattering and internal friction.
We can substitute all these terms in the first law of thermodynamics equation. For
the general case, the total energy of the system will be the summation of the kinetic
energy and internal potential energy. Kinetic energy here is associated with the
macroscopic level velocity of the matter. It does not refer to the kinetic energy of the
atoms. That is included in the internal potential energy, which refers to many forms
of energy stored in the lattice.
For arbitrary values of Winput and Qinput we can write the following relations:
dE = dU þ dK = Qinput þ W input dt ð3:19Þ
Total energy rate dE_ total can be given by

Z h i
dE dU dK d 1
= þ = ρ u þ ρ ~v ~v dV ð3:20Þ
dt dt dt dt V 2
where u is the internal strain energy per unit mass. We can write the first law of
thermodynamics by remembering that Qinput is negative when heat enters the system
from outside and Winput is positive when the work is done by the system.
dK
Since d U_ = Qinput þ W input - ð3:21Þ
dt
2 3
Z Z Z
d 6 7
ρudV = 4 - q ndA þ ρrdV 5
dt
V A V
2 3
Z Z Z
6d 1 7 d 1
þ4 ρ~v ~vdV þ σ : DdV 5 - ρ ~v ~vdV ð3:22Þ
dt 2 dt 2
V V V
Kinetic energy terms cancel each other. Using the divergence theorem, we can
transform surface integral to volume integral:
Z Z
q ndA = ∇ qdV ð3:23Þ
A V
We finally collect all terms on one side and set it to equal to zero:
Z
du
ρ þ ∇ q - ρr - σ : D dV = 0 ð3:24Þ
dt
V
98 3 Thermodynamics
Finally, the energy conservation law in indicial notation can be written as

du ∂qj
ρ = σ ij Dij þ ρr - ð3:25Þ
dt xj
where u is the internal potential energy per unit mass. The left side of the equation is
the rate of increase in the internal potential energy. The right side of the equation
represents the input power that is converted into internal work as mechanical work
and as heat generation, however, excluding the kinetic energy at the macro level. The
last term is negative because it represents heat inflow per unit volume through the
boundaries.
We should point out that this equation is a very simple form of the first law of
thermodynamics because it assumes that only Winput is due to surface traction
(distributed loads). If we have distributed fields such as electro-magnetic and
chemical loads, they must be included in this equation. These load cases will be
covered in Chap. 8. Moreover, the first law of thermodynamics given above does not
include energy loss terms, because mechanical work terms are defined according to
Newtonian mechanics. In the next chapter, the same equation will be derived for
unified mechanics theory where energy loss will be included.
3.3 Second Law of Thermodynamics
There are several ways to define the second law; however, we believe Kelvin-
Planck’s statement is the most concise and classical description. According to this
statement, “it is impossible to construct an engine that will produce no other effect
than the extraction of heat from a single heat reservoir and the performance of an
equivalent amount of work.” In other words, we can state that it is not possible to
construct an engine that has an efficiency of 100%, meaning that the input energy
and output energy for the intended work cannot be equal. There will always be
energy loss for unintended work. This energy loss will only occur in the positive
direction, meaning that we can only dissipate energy, but cannot gain energy in a
closed system. This law of nature is expressed mathematically as an inequality
stating that the internal entropy production is always nonnegative and is positive
for an irreversible process. This inequality is called Clausius–Duhem inequality.
Over simplified definition of entropy is that entropy quantifies how much energy
is unavailable for work. Another way of stating the second law of thermodynamics is
that the entropy of all-natural processes increases. The concept of entropy will allow
us to tie probability and statistics to continuum mechanics.
In the first law of thermodynamics, we saw interconvertibility of energy between
different forms such as heat and mechanical work. In Newtonian mechanics, kinetic
energy and potential energy may be converted from one to other with no energy loss
[Fdt = d(mv)]. The transformation can proceed in either direction, like a pendulum
3.3 Second Law of Thermodynamics 99
Fig. 3.2 Twin chambers
with no friction that swings until eternity. Of course, in addition to dissipation of

kinetic energy due to friction, there is also degradation of the pendulum material due
to stresses induced by self-weight. On August 5, 1976, the Great Westminster clock
in London, UK, also known as Big Ben, experienced fatigue failure after 117 years
in use.
The second law of thermodynamics states that a fatigue-free pendulum is not
possible. Each time a pendulum swings, it loses some of its energy; as a result there
is less total energy to continue swinging. Complete reversal is not possible. Fric-
tional dissipation and degradation in materials are irreversible processes.
The second law of thermodynamics also defines a preferred direction in any
natural process. This is clearly stated by the definition of the second law given by
Clausius (1850). He stated that the heat never flows on its own from the colder
system to the warmer. This defines a preferred direction in the flow of heat between
two bodies at different temperatures (Fig. 3.2). At this point, it is important to define
what temperature means. While at the macroscale in continuum mechanics, we
consider objects stationary, at the atomic scale this is not true. Even if an object is
not subjected to any dynamic loading, if the temperature of the object is above zero
Kelvin, all its atoms are vibrating continuously. This phenomenon is called atomic
vibrations. At any given point in time, not all atoms vibrate at the same frequency
and amplitude, nor with the same energy. Vibrational modes of atoms are called
phonons. All modes can be determined by solving the eigenvalue equation, just like
in continuum mechanics. Atoms will be vibrating at different levels of energy;
however, the average vibrational energy of all atoms in the object will be fluctuating
around a constant energy level. This constant average energy level will increase
when the temperature of the object increases. Therefore, the temperature is just a
measure of the average vibrational energy of all atoms in a body. Going back to the
second law of thermodynamics, consider two containers filled with the same gas.
Assume the temperature in A is lower and in B is higher. Low energy atoms
(colder atoms) can’t move towards B. Because atoms in B have higher energy levels,
their movement towards A is possible. This directional preference is another state-
ment of the second law of thermodynamics.
100 3 Thermodynamics
3.3.1 Entropy
The term “entropy” was first used by Clausius (1850). The second law of thermo-
dynamics is related to a variable called entropy, s, which is a mathematical quantity.
Therefore, entropy allows us to quantify the second law of thermodynamics. While
the concept of entropy is not easily understood from a continuum mechanics point of
view, it can be explained by utilizing molecular behavior.
First, we should clarify that the second law of thermodynamics cannot be proven
mathematically, but it is a generalization from observations (inductive reasoning).
There are several simplified descriptions of entropy while they are not mathematical;
however, they help in explaining an abstract quantity (Fig. 3.3).
Assume we have a ball at point A. It has a potential energy of mgh. Once the ball
is slightly pushed, it rolls down and travels along the flat surface. At the point, B the
ball will have no potential energy but just kinetic energy. Finally, at some point C the
ball will come to a stop. It will have zero energy.
As the ball is traveling from point A to point C, the amount of energy available for
work continually decreases. Hence, the amount of energy unavailable for work
(entropy) increases. This latter one is called entropy. In other words, “production
of entropy is associated with dissipation (or consumption) of the capacity for
spontaneous change when any process occurs” (DeHoff, 1993). The entropy of the
closed system increases or remains constant in all processes. Therefore, entropy
always increases. Energy and entropy are governed by different laws. Energy can
enter or exit a body but it cannot be created inside a body. However, entropy can be
created inside the body.
Callen (1985) formulates the interpretation of entropy in a set of postulates
depending upon a posteriori justification. These postulates are based on the basic
principle that the minimization of energy function happens at equilibrium. In the
following section, we will quote these postulates as defined by Callen (1985).
Postulate I
There exist particular states, called equilibrium states, of simple systems that mac-
roscopically are characterized completely by the internal energy and mole numbers
of the chemical components.
Fig. 3.3 Different states of

energy
Postulate II
There exists a function, called the entropy (s), of the extensive parameters of any
composite system, defined for all equilibrium states and having the following
property. The values assumed by the extensive parameters in the absence of an
internal constraint are those that maximize the entropy over the manifold of
constrained equilibrium states.
An extensive parameter in a homogeneous system is proportional to the total mass
of the system.
An intensive variable is defined as a variable that has the same value everywhere
in a homogeneous system. The densities of extensive variables (e.g. u,s,~v ) are
intensive variables.
This postulate assumes the existence of a function called entropy only for the
thermodynamic equilibrium states. In the absence of constraints, the system is free to
select anyone number of states; however, the state of maximum entropy is always
selected by the system.
As stated earlier, the basic problem of thermodynamics is “The single,
all-encompassing problem of thermodynamic is the determination of the equilibrium
state that eventually results after the removal of internal constraints in a closed
composite system” Callen (1985). This basic problem can be solved if the entropy
of the system is known as a function of the intrinsic parameters of the system. The
relation that gives the entropy as a function of the system parameters is known as a
thermodynamic fundamental relation. It, therefore, follows that if the fundamental
relation of a particular system is known, all conceivable thermodynamic information
about the system can be obtained from it. Essentially if the fundamental relation of a
system is known, every thermodynamic attribute is completely and precisely deter-
mined.” Callen (1985) did an outstanding job in establishing the postulator basis for
thermodynamics.
Postulate III
The entropy of a composite system is additive over the constituent subsystems. The
entropy is continuous and differentiable and is a monotonically increasing function
of the energy.
The additive property of entropy is just a summation of the entropies of the
constituent subsystems.
Xn Xn
s= s =
i=1 i
s ðU i ,
i=1 i
V i, Ni, . . . Nr Þ ð3:26Þ
where n is the number of subsystems.

The entropy of each subsystem is a function of extensive parameters of that
subsystem alone when the additive property is used for spatially separate subsys-
tems. The following property must be satisfied. “The entropy of a simple system is a
homogeneous first-order function of the extensive parameters.” In other words, if all
the extensive parameters are multiplied by a constant, say λ, the entropy is multiplied
by the same constant.
sðλU, λV, λN . . . λN r Þ = λsðU, V, N 1 . . . N r Þ ð3:27Þ
Postulate II implies that the partial derivative of entropy with respect to internal
energy is a positive quantity.

∂s
>0 ð3:28Þ
∂U V,N i
The possibility of negative values of this derivative is a topic of research. It was

first mentioned by Ramsey (1956). However, such instances are non-equilibrium
states. They can only be produced in unique systems, and most importantly, they
decay away very quickly, [usually at the atomic level time scale]. Therefore, these
non-equilibrium transient points do not invalidate the positive value of the partial
derivative for steady-state equilibrium states. Since entropy is a continuous, differ-
entiable, and monotonic function of energy, the opposite is also true. We can express
energy as a single-valued, continuous, and differentiable function of entropy and
other system parameters, as follows.
U = U ðs, V, N 1 , . . . N r Þ ð3:29Þ
where V is the volume and N1, . . .Nr are the entire system parameters.
Postulate IV
At zero Kelvin temperature, the entropy of any system vanishes in the state for which
∂U
=0 ð3:30Þ
∂s
Essentially this postulate states that when atoms have no vibrational energy, it is
not possible to generate entropy. Therefore, at zero Kelvin temperature systems have
zero entropy generation.
Postulate IV is an extension of the Nernst postulate or also known as the third law
of thermodynamics.
These four postulates are the rational basis for the development of thermodynam-
ics. Based on these postulates, we can solve any thermodynamics problem. Since
everything organic or inorganic is a thermodynamics system, these postulates are
applicable. Using these postulates, we can solve any problem that we can derive the
thermodynamic fundamental relation, in the following procedure.
Let us assume the thermodynamic fundamental equations governing each of the
constituent systems (mechanisms) are known. These fundamental governing equa-
tions allow us to calculate the entropy generation in each sub-system (mechanisms)

when these mechanisms are in equilibrium. If the entire system is in a constrained
equilibrium state, the total entropy is obtained by the addition of the individual
entropies for each subsystem (a mechanism). Entropy generation in each subsystem
(a micro mechanism) is calculated. Therefore, the total entropy is a function of the
various parameters of the subsystems (micro-mechanisms). Taking straightforward
differentiation of the entropy function, we can compute the extrema of the function.
Of course, extrema can be minimum, maximum, or a horizontal inflection point.
Then by taking the second derivative, we classify these equilibrium states as stable
equilibrium, unstable equilibrium, or meta-stable equilibrium.
When a system is in a stable equilibrium point for energy, it is also in a stable
equilibrium state for entropy function (fundamental relation). Minimization of
energy function corresponds to maximization of entropy function because energy
is a function of the system parameters and entropy (Fig. 3.3).
3.3.2 Quantification of Entropy in Thermodynamics
Carnot in 1824 defined entropy for a hypothetical reversible engine, where he shows
the mechanical equivalence of heat energy. Entropy is a state variable defined by

dQ
ds = - ðfor a reversible processÞ ð3:31Þ
T
where Q is heat energy in Joule or Calorie and T is the temperature in Kelvin. In

1851, William Thomson (Baron Kelvin) and Clausius published Carnot’s work
posthumously, where it is shown that entropy is a state variable. “If the integral of
a quantity around any closed path is zero, that quantity is called a state variable, that
it has a value that is characteristic only of the state of the system regardless of how
that state was arrived at” (Halliday & Resnick, 1966).
I
ds = 0 ðfor a reversible processÞ ð3:32Þ
We should point out that neither heat energy Q nor temperature T is a perfect
differential of any function. However, ds = dQ T is a perfect differential of a function.
The fact that entropy is a state function can easily be proven for an ideal gas. Entropy
will return to its initial value whenever the temperature returns to its initial value, for
any reversible process. Of course, a reversible process does not exist in real life. It is
just an imaginary idealization created for expedience in the formulation.
3.3.3 Gibbs–Duhem Relation
The fundamental equation for total internal energy U is a function of entropy,

volume, and extensive parameters given by
U = U ðS, V, N 1 , . . . . . . . . . N r Þ ð3:33Þ
Taking the first differential of the total energy, we can write

∂U ∂U ∂U ∂U
dU = dS þ dV þ dN 1 þ ⋯ þ dN r ð3:34Þ
∂S ∂V ∂N 1 ∂N r
Partial derivatives for individual terms are defined as follows:

∂U
= T, ðtemperatureÞ ð3:35Þ
∂S

∂U
- = P, ðpressureÞ ð3:36Þ
∂V

∂U
= μi , electro - chemical - mechanical potential of the ith component ð3:37Þ
∂N i
Temperature, T, pressure, P, and electro-chemical potential of any

subcomponent, μi, are called intensive parameters. Because they have the same
value at all points in a homogeneous system, of course, this does not mean that we
cannot use them for heterogeneous systems. In computational mechanics, we
discretize a heterogeneous system into small pieces of homogeneous elements.
Based on these definitions, for a homogeneous system in equilibrium, the differential
of total internal energy can be given by
dU = TdS - PdV þ μ1 dN 1 þ ⋯ þ μr dN r ð3:38Þ
The formal thermodynamic definition of temperature (at the macro level) agrees
with the intuitive definition given by (∂U
∂S
= T). We should point out that this is the
macro definition of temperature since at the atomic level temperature is defined by
the amplitude of intensity of vibrational energy of atoms. In addition, the definition
of pressure given above agrees with the mechanics’ definition of pressure. [-PdV] is
called quasi-static work.
dW m = - PdV ð3:39Þ
The term quasi-static here used is in contrast to a dynamic load where a load is
applied very fast; as a result, inertia forces are not negligible.
It is necessary to discuss the negative sign in quasi-static work. The quasi-static
work is assumed positive if it increases the total energy of the system.
If the change in volume dV is negative, work done on the system is positive

increasing its energy, because of the negative sign in the equation.
dW M = - Pð - dV Þ = þ PdV ð3:40Þ
Heat flux can be defined quantitatively with the help of quasi-static work. For a
quasi-static process at a constant number of moles, the heat energy dQ can be defined
by
dQ = dU - dW m ð3:41aÞ
or
dQ = dU þ PdV ð3:41bÞ
In thermodynamics terms, heat and heat flux are used interchangeably. In ther-
modynamics terminology “heat” like “work” is only a form of energy. It is assumed
that once the energy is transferred to a system either in the form of heat or as
mechanical work, it is indistinguishable from any energy that may have been
transferred directly to the system. It is important to remind that Joule is the unit
for energy and work. Therefore, there cannot be a distinction between variables with
the same units.
The total energy U of a state is not just a sum of work (dW) and heat (dQ).
Because dWM and dQ are imperfect differentials, which means that the differential is
path-dependent. [Exact differential is path independent.]
The integrals of dWM and dQ is the work and heat for that particular path of the
process. Their sum is the total energy difference ΔU. However, ΔU is independent
of the path taken.
We should point out that when the fundamental equation is written as U = U(S,
V, N1, . . .Nr) the total energy U is a dependent variable, and entropy S is an
independent variable. This is the fundamental equation unified mechanics theory is
based on. Since entropy is an independent variable, it can only be defined in space on
its axis. The relation U = U(S, V, N1, . . .Nr) is also named the energetic fundamental
relation. However, S = S(U, V, N1, . . .Nr) is said to be entropic fundamental relation,
which is referred to as thermodynamic fundamental relation in this book.
The remaining terms in dU equation represent an increase of internal energy
associated with the addition of matter to a system. These terms are called quasi-static
electro-chemical work. The term matter here does not exclude electron flow as a
matter due to an electrical bias. The quasi-static electro-chemical work is given by
Xr
dW c = μN
i=1 i i
ð3:42Þ
Therefore, energy equilibrium energy becomes a summation of heat, mechanical

work, and electro-chemical work.
dU = dQ þ dW M þ dW C ð3:43Þ
3.3.4 Euler Equation
Quoting Callen’s (1985) formulation, the homogeneous first-order property of the

energetic fundamental relation permits that equation be written in a convenient form
called the Euler form.
U ðλS, λX 1 , . . . , λX r Þ = λU ðS, X 1 , . . . X r Þ ð3:44Þ
Following the same differentiation process and using the same definitions we
used earlier in this section for the energy equilibrium equation and differentiating
with respect to λ yields

∂U ∂ðλSÞ ∂U ∂λX i
þ þ ⋯ = U ðS, X 1 , . . . X r Þ ð3:45Þ
∂ðλSÞ ∂λ ∂ðλX i Þ ∂λ
Assuming λ = 1. Then
∂U Xt ∂U
Sþ X =U
i = 1 ∂X i i
ð3:46Þ
∂S
Xr
U = TS þ PX
i=1 i i
ð3:47Þ
For a simple system, this equation becomes
U = TS - PV þ μ1 N 1 þ ⋯ þ μr N r ð3:48Þ
The last two equations are called Euler relations. A differential form of the
relation among the intensive parameters can be obtained directly from the Euler
relation and is known as the Gibbs–Duhem relation.
Taking differentiation of Euler relation given by Eq. (3.47) leads to
Xr Xr
dU = TdS þ SdT þ P dX i þ
i=1 i i=1
X i dPi ð3:49Þ
Earlier we derived that,

Xr
dU = TdS þ P dX i
i=1 i
ð3:50Þ
We should note that in the derivation of the last equation, we separated quasi-
static mechanical work and quasi-static electro-chemical work. Here they are both
represented in one term:
dU = dQ þ dW M þ dW C ð3:51Þ
Xr
P dX i
i=1 i
ð3:52Þ
Substituting this last equation in the differential above, we end up with

Xr
SdT þ i=1
X i dPi = 0 ð3:53Þ
This equation is referred to as the Gibbs–Duhem relation. For a single component

system, this equation can be written as
SdT - VdP þ Ndμ = 0 ð3:54Þ
The Gibbs–Duhem relation presents the relationship among the intensive param-
eters in differential form. The number of linearly independent intensive parameters
that have independent variations is called the number of thermodynamic degrees of
freedom of any system, which are mapped onto the thermodynamic state index axis
in the unified mechanics theory, which is discussed in the next chapter.
At this point, it is necessary to restate the thermodynamic formalism in energy
representation. This fundamental equation is
U = U ðS, V, N 1 , . . . N r Þ ð3:55Þ
which contains all thermodynamic information of any system. Relying on our

definitions the equations of thermodynamic state can be given by
T = T ðS, V, N 1 , . . . N r Þ
P = PðS, V, N 1 , . . . N r Þ
ð3:56Þ
μ1 = μ1 ðS, V, N 1 , . . . N r Þ
μr = μr ðS, V, N 1 , . . . N r Þ
When all equations of thermodynamic state are combined, it is equivalent to the

thermodynamic fundamental equation.
Assuming we have a simple single-component system in which two equations of
the state are known, the Gibbs–Duhem relation can be integrated to obtain the third
equation of state.
As pointed out by Callen (1985), it is possible to express internal energy
U without entropy S, term. However, such a total energy equation is not a funda-
mental energetic relation since it does not contain all thermodynamic information
about the system.
Many textbooks refer to Gibb’s relation in the following form, which Gibbs
defined only for the case of a fluid,
du = Tds - Pdv ð3:57Þ
where u is the internal energy density (specific internal energy), s is entropy for unit
volume, P is the thermodynamic pressure, and v is the specific volume of the fluid.
3.3.5 Entropy Production in Irreversible Process
The cantilever beam shown in Fig. 3.4 is subjected to linear elastic static cyclic
loading. Let us assume that the force is small enough to produce a stress level well
below the macro yield stress. A student observing this beam will see that after every
load removal, the beam returns to the original shape. To the observer, this gives the
impression that deformation under elastic loading is a perpetual reversible process.
However, after a certain number of cycles, the cantilever beam will fatigue and fail
regardless of how small the stresses are. It will probably break away from the support
where stresses are maximum. This simple experiment shows that while we observe
macro behavior to be reversible, however, our observation at the macro level is not
accurate. If the linear elastic loading-unloading cycle were reversible, the beam
would never fatigue, and deflections from one cycle to another cycle would not
change in magnitude. In earlier pages, we cited the fatigue failure of Big Ben [clock
in London] after 117 years in service.
According to the second law of thermodynamics, reversible processes are imag-
inary. They cannot happen in the real domain. What happens in the lattice is when
the beam is mechanically loaded, there is strain in the lattice. However, when the
load is removed, atoms do not go back to their original lattice site. They go to a
lattice site with a lower energy level, to minimize their energy level. In the process,
they leave behind a vacancy. As a result, the process is not reversible. Keep in mind
that the thermodynamically reversible process is where the initial and the final states
Fig. 3.4 A cantilever beam

subjected to cycling loading
are the same. If an atom moves to a new lattice site with lower energy, state and is left
behind a vacancy that is a different thermodynamic state than the initial one. There is
a positive entropy generation, which is a quantitative measure of dissipation in the
system. DeHoff (1993) states that “changes in the real world are always accompa-
nied by friction and something [energy] that is dissipated. When the pebble is
dropped into a pond its kinetic energy at impact is dispersed by the wave motion
throughout the pond and the [initial kinetic energy] is dissipated in the absorbing
water of the pond.” In this irreversible process, the energy cannot be recovered. “The
entropy production in a system is a quantitative measure of this energy dissipation
[mechanisms].”
Eddington (1958) referred to entropy as “entropy is time’s arrow,” The total
entropy of any system plus its surroundings always increases with increasing time.
However, the entropy-time relationship is not linear, and each one is linearly
independent. Time can change without a change in entropy, as in at zero Kelvin or
in an undisturbed system. However, time cannot change unless there is a change in
entropy. If we had a process where entropy is destroyed, then the direction of the
time arrow must be reversed for that process to happen. However, this is not possible
in a real system. A reader may question the accuracy of the theory of elasticity where
all mechanisms are assumed to be reversible. Because of the second law of thermo-
dynamics, the theory of elasticity provides an approximate solution to an imaginary
system that cannot exist in reality. In the words of DeHoff (1993) “All real processes
have a finite rate in response to finite influences; such real processes are called
irreversible to emphasize the contrast with imaginary reversible processes. Real
processes are irreversible and suffer dissipations that result in the production of
entropy and thus a permanent (irreversible) change in the system.” Therefore, for an
irreversible process, we can write
Z2
dQ
ΔS = S2 - S1 = >0 ð3:58Þ
T irreversible
1
Of course, here it is assumed that it is a closed system and it is an adiabatic

process.
While we acknowledge that all real processes are irreversible, sometimes for the
sake of simplicity a system is modeled as an imaginary reversible process, to be able
to get an understanding of the system near time = 0.
3.3.6 Clausius–Duhem Inequality
There are different approaches to deriving this inequality. We will quote the deri-
vation given by Malvern (1969). This inequality is considered a simple quantitative
expression of the second law of thermodynamics. The thermodynamic theory
explains the relationship between energy flowing into a system and the entropy
generation in the system and its surroundings (Fig. 3.5).
Heat input in a system is given by the difference between the heat conduction
coming from outside through the surface A and distributed internal heat generation
r per unit mass:
Z Z
Qinput = - q ndA þ ρrdV ð3:59Þ
A V
Entropy was defined by dS = dQ

T. Hence entropy input rate can be given by
Z Z
q ρr
- ndA þ dV ð3:60Þ
T T
A V
We are assuming that the system is closed. If the system is open in a control
surface, [fixed in space], additional entropy input due to mass flux would be
accounted for by
Z
- ρsV ndA ð3:61Þ
A
where s is entropy per unit mass. According to the second law of thermodynamics,
entropy increase rate in the system ≥ entropy input rate
Z Z Z
d q r
ρsdV ≥ - ndA þ ρdV ð3:62Þ
dt T T
V A V
This is the integral form of the Clausius–Duhem inequality. This equation means
that in an irreversible process internal entropy production is always greater than
input. In this equation, outward normal is a positive sign. An equal sign ensures that
the equation holds for the reversible [imaginary] process.
Fig. 3.5 Heat exchange for

a system
We can transform the surface integral into a volume integral. Since volume can be
an arbitrary value, we can write the local version of the Clausius–Duhem inequality
that must be satisfied at each point in any given volume as follows:

ds r 1 q
≥ - div ð3:63Þ
dt T ρ T
or
ds r 1 q
γ - þ divq - 2 gradT ≥ 0 ð3:64Þ
dt T ρT ρT
where γ is the internal entropy production rate per unit mass.

A stronger assumption of inequality was proposed by Truesdell and Noll (1985):
ds r 1 1
- þ divq ≥ 0 - q gradT ≥ 0 ð3:65Þ
dt T ρT ρT 2
where the first inequality represents the local entropy production by the system and
the second inequality represents entropy production by heat conduction.
The Truesdell and Noll (1985) interpretation of the second law of thermodynam-
ics is important for the unified mechanics theory because it is assumed that disorder
in the system increases only due to internal entropy production.
Entropy production by heat conduction in a free (unconstrained) system, Fig. 3.6,
cannot lead to irreversible disorder in the system because the system travels between
stress-free equilibrium states. Of course, the entropy of the entire surrounding
including the system increases. The interatomic equilibrium distance in a free
(unconstrained) system is a function of temperature; therefore due to the heat
conduction temperature of the system increase (atoms have higher vibrational
energy) the equilibrium inter-atomic distance increases. However, there will be no
strain (or stress) exerted on the atoms. Therefore, internal work will be zero; as a
result, internal entropy production will be zero. If the atoms move from one thermal
equilibrium to another thermal equilibrium, it is assumed that there will be no
internal entropy generation. Of course, this assumption is not 100% true. When
Fig. 3.6 Expansion of a cantilever beam due to increasing ambient temperature

Fig. 3.7 Repulsive and

attractive bonding energy Repulsive energy
between two isolated atoms
as a function of interatomic r0
distance. (After Callister Jr Interatomic disctance r
& Rethwisch, 2010)
Net Energy EN
Attractive energy
there is heat transport in a lattice, there is always scattering with phonons. We are
assuming this scattering-induced internal entropy generation is so small that we can
ignore it.
When there are no external constraints blocking atoms moving freely at each
temperature, they have a stress-free equilibrium interatomic separation distance r0.
When the interatomic distance is r0, attraction/repulsion atomic bond energy is at an
extremum (maximum), and the force acting on the atoms is zero (Fig. 3.7). There-
fore, internal work is not possible, because internal work and internal entropy
production are proportional to the dislocation of an atom from the equilibrium lattice
site and force acting on the atom. Of course, when a system is at equilibrium and the
force acting on the atom is zero, the work done must be zero. However, this
argument assumes that the system is free to move and boundary conditions do not
impose any constraint on the system that prevents it from moving freely. If the
cantilever shown in Fig. 3.6 is fixed on both sides, of course, there will be internal
work and internal entropy generation.
Clausius–Duhem inequality is just a quantitative statement of the second law of
thermodynamics; hence, it must be satisfied by every process in organic or inorganic
systems. The coordinate system assumed in the derivation of the formulation in
previous pages is arbitrary; however the positive definite property of the inequality is
not arbitrary. It must be satisfying for all and any coordinate system.
3.3.7 Traditional Use of Entropy as a Functional

in Continuum Mechanics
The unified mechanics theory uses entropy as a linearly independent state variable.
Furthermore, entropy (thermodynamics state index) defines a new additional 5th
linearly independent axis in addition to Newtonian space-time axes. This is needed
to be able to define the location of a system in a space-time-thermodynamic state
index coordinate system. This is needed to be able to define the coordinates of a
system in a space-time-thermodynamic state index coordinate system. While there
are similarities between the traditional interpretation of entropy in Newtonian
(continuum) mechanics and unified mechanics, there are also major differences. In
Newtonian (continuum) mechanics, the derivative of displacement with respect to
entropy is taken as zero. In the unified mechanics theory, derivatives with respect to
entropy are not zero. When derivatives with respect to entropy are taken as zero,
unified mechanics theory collapses to Newtonian mechanics. We assume that the
derivative of time with respect to entropy is small enough to be considered zero for
most engineering problems.
The traditional use of entropy in Newtonian (continuum) Mechanics is considered
extensively by Malvern (1969), Coleman (1964), and Coleman and Mizel (1964,
1967). Here we summarize their work. “In Newtonian continuum mechanics, the use
of the Clausius-Duhem inequality differs from the usual second law of thermody-
namics, where the entropy is a state function determined by the instantaneous values
of the other state variables. In Newtonian continuum mechanics, entropy is not
explicitly postulated to be a state function. As a result, in Newtonian continuum
mechanics entropy maximization (and minimization of entropy generation rate) does
not affect system state variables or system properties. In simple terms, in Newtonian
continuum mechanics, dissipation and degradation of the system are not directly
included in the laws of Newton, but as empirical test data curve-fitting constraints.
Traditionally many researchers and authors of thermodynamics have postulated that
entropy is a state function of all the state variables including possibly some variables
that are not observable by macro-scale specimen testing.”
We must clarify what we mean by “state variables.” For a simple ideal gas,
pressure, temperature, and volume are thermodynamic variables. However, the
internal energy and the entropy are state variables (also called state functions).
State variables are expressed employing thermodynamics variables. For a variable
to be classified as a state variable (also called state function), it must have an exact
differential, based on a vanishing integral in an arbitrary cycle. This definition
assumes that the material is capable of going through an arbitrary cycle and being
restored to its initial state. As an example, we can plastically deform a steel beam,
and then melt it or anneal it and return it to its original state. Therefore, the total
energy of the system is a state variable because it has an exact differential:
I

Pinput þ Qinput dt = 0 ð3:66Þ
Malvern (1969) states that if the “state” of a continuum is taken to be defined by a

limited number of explicitly enumerated macroscopic state variables, observable at
least in principle, and then the entropy must in general depend on the history of this
limited number of state variables and not merely on their current values. This is of
course the case for any solid that experiences inelastic deformations. Because plastic
strain is path-dependent, therefore, entropy is also of course path-dependent.
Coleman (1964) and Malvern (1969) define entropy as a function of the history of
deformation gradient tensor F, temperature, and the instantaneous value of temper-
ature gradient. However, this definition is for thermo-mechanical loading only.
Malvern (1969) makes the distinction between function and functional by defin-
ing that a function depends only on the instantaneous values of its variables.
Functional is a function of the history of the variables. Malvern (1969) citing a
derivation by Coleman and Mizel (1964) defines the caloric equation of state [called
thermodynamic fundamental equation in this book], in the following form:
s = f ðu, F Þ ð3:67Þ
where entropy is a function of internal energy and deformation gradient. However,

this definition ignores other electro-thermo-chemical-radiation mechanisms that
could contribute to entropy generation. Earlier Truesdell and Toupin (1960) postu-
lated an energetic equation of state as follows:
u = uðs, V i , . . . , V n Þ ð3:68Þ
Here s is entropy per unit mass and Vi is a thermodynamic substate variable

accounting for all thermos-electro-mechanical-chemical-radiation processes. Of
course, the inverse of the last Eq. (3.68) also holds true:
s = sðu, V 1 , . . . , V n Þ ð3:69Þ
which is a more general thermodynamic fundamental [caloric] equation of the state.

The terminology naming these equations is not uniform in the literature. However,
they have the same meaning.
3.3.8 Entropy as a Measure of Disorder
Order and disorder are relative terms, meaning that they are defined with respect to
an initial reference state. For example, in Fig. 3.8a, the initial location of each piece,
such as A7, can be considered in the ordered state. In general, the initial or strain-free
state of a system is considered as ordered. Of course, this does not mean that the
strain-free state has no disorder at the atomic scale. It is just a benchmark [reference]
state. Any change from the initial reference state is an increase in disorder. For a
quantifiable disorder to happen in a system, there must be some inhomogeneity. For
example, if we have a box (Fig 3.8b) with four identical baseballs, regardless of how
much we shake the box, the final “disordered” state will be identical to the initially
ordered state. As a result, there is no way to quantify a disorder.
There are 24 distinct configurations possible for these balls in the container.
However, the positions the balls could take are not observable at the macroscale,
because the balls are identical. Fortunately, most systems and materials at the micro
aggregate scale are not made up of identical crystals. Now assume that we mark the
balls as A, B, C, D (Fig. 3.8c), and initially we place them in the box in alphabetical
order. We assume that the initial alphabetical order configuration is a reference
“ordered” state. In this case, after we shake the box among 4 ! = 24 possible
configurations, the initially ordered state is only possible 1/24 of the time. That
means 23/24 times it will be a disordered state. If we have 26 baseballs marked A to
Z, the possibility of getting the ordered state is one in millions. All organic and
inorganic systems are made up of a large number of atoms (or molecules). When
subjected to any external load (disturbance), they move from their initial ordered
Fig. 3.8 Description of order and disorder
state to a new configuration that we call “disordered state,” of course, there is an

energy cost associated with changing the ordered state. It does not happen on its own
with no external energy input. “In statistical mechanics, the entropy of a state is
related to the probability of the occurrence of that state among all the possible states
that could occur” (Malvern, 1969). This is only rational because entropy generation
requires work. As more entropy is generated, other configurations that are possible
are achieved. Malvern (1969) states “Thus, increasing entropy is associated with
increasing disorder. The second law of thermodynamics seems to imply an almost
metaphysical principle of preference for disorder.” In nature, it is observed that
changes of state are more likely to occur in the direction of greater disorder.
However, the second law of thermodynamics also implies that when the disorder
is maximum, entropy is also at maximum, and the entropy generation rate is zero.
The balls in a box example are an illustration of this argument. An engineering-
related example would be having a standard 15 cm in diameter and 39 cm in height
concrete cylinder, where each gravel, a sand particle is numbered. Under compres-
sion loading, this concrete cylinder will go through many disordered states before
failure.
Callen (1985) explains eloquently why entropy is a measure of disorder, quanti-
tatively. We will summarize his approach. Callen (1985) states that in statistical
mechanics the number of microstates [arrangement of A, B, C, D balls in our
example, Fig. 3.8] among which the system undergoes transitions and which thereby
share a uniform probability of occupation, increases to the maximum permitted by
the constraints. This statement is strikingly reminiscent of the entropy postulate of

thermodynamics, according to which the entropy increases to the maximum permit-
ted by the imposed constraints, for a closed system. It is possible to conclude that
entropy can be identified with the number of microstates in a closed system.
Callen (1985) points out that entropy is additive, but the number of microstates is
multiplicative. The number of microstates available to two systems is the product of
the number of microstates available to each system. If we have two dices, each has
six microstates. But two dices rolled together have 6 × 6 = 36 microstates. Callen
(1985) points out that to interpret the entropy, then we require an additive quantity
that measures the number of microstates available to a closed [isolated] system.
Boltzmann (1877) [English translation by Sharp and Matschinsky (2015)] and
Planck (1900a, b, c, d) suggested that this problem could be solved by identifying
the entropy with the logarithm of the number of available microstates because the
logarithm of a product being the sum of the logarithms:
ln ða bÞ = ln a þ ln b ð3:70Þ
Thus, the following equation resulted:
s = k ln w ð3:71Þ
where w is the number of microstates consistent with the macroscopic constraints of

the closed [isolated] system. The confusion about w being the number of microstates
or the probability of a microstate is discussed in the next chapter. Boltzmann
constant k is defined by
k = R=N A = 1:3807 × 10 - 23 J=K ð3:72Þ
is used to define temperature on the Kelvin scale and to ensure consistency of units
on both sides of the equation. R = 8:31 molK
Joule
is the gas constant; NA = 6.022 × 1023 mol-
ecules/mol is the Avogadros number. Equation (3.71) is considered the basis for
statistical mechanics. Callen (1985) refers to Eq. (3.71) as a postulate that is dramatic
in its brevity, simplicity, and completeness.
According to Boltzmann, the equation calculating the natural logarithm of the
number of microstates available to the system and multiplying with a constant
k yields the entropy. However, the opposite is also true to calculate the number of
microstates. In Unified Mechanics Theory, this second approach is used, since
entropy is a function of internal state variables and all active mechanisms; it can
directly be calculated because entropy is the most general caloric equation [funda-
mental equation] of state. Callen (1985) refers to the Boltzmann equation as the
statistical mechanics in the microcanonical ensemble.
Entropy calculations in continuum mechanics are not readily available for all
mechanisms. Of course, very often we do not know what these mechanisms are. This
is expected to be an active research field soon. Entropy in thermodynamics and
statistical physics are the same thing. Statistical mechanics interpretation of entropy
where natural progress towards more disorder establishes a concrete and understand-
able concept of entropy. However, in statistical mechanics to be able to calculate the
entropy all active mechanisms and processes taking place in the material must be
accounted for. Then entropy generation for each process must be calculated. Unfor-
tunately, this is a very primitive field in the mechanics of materials, because all
material constitutive models at continuum scale are based on empirical functions
obtained by testing of materials. There are no explicit formulas for entropy gener-
ation due to thermo-mechanical-electrical-chemical-radiation loads. This is a wide-
open research topic and a very challenging one, to say the least, because the material
modeling field has always been based on empirical observations. For example, we
crash a concrete cylinder, and model it with macro measurement variables, such as
stress, strain, and curve fitting. However, macroscale testing does not give us much
information about the actual micro-mechanisms that are responsible for degradation
and final failure. These are the micro-mechanisms that we need to know in detail to
compute entropy generation. It is always much easier to curve fit to a test data and
use it than to understand the actual mechanisms and processes leading to failure.
Theoretical physicist and mathematician Freeman Dyson in Scientific American
September (1954) suggested that heat is disordered energy. The author gave an
example of a flying rifle bullet. The bullet has kinetic energy but no disorder. After
the bullet hits a steel plate target, its kinetic energy is transferred to random motions
of the atoms in the plate and bullet. This disordered energy makes itself felt in the
form of heat. However, the author ignores other mechanisms, because heat is not the
only product of the initial kinetic energy. There are plastic deformations, melting,
phase change, thermomigration, and other entropy generation mechanisms that
could result from an impact.
Freeman Dyson (1954) also adds that the quantity of disorder is measured in
terms of the mathematical concept called entropy; of course, entropy is a function of
energy. Therefore, we can compute disorder in terms of energy using Boltzmann’s
equation. It is important to point out that disorder is the perfect way to express the
degradation of all organic and inorganic systems. Because as the entropy increases
the amount of disorder from the initial “ordered,” reference state increases. The
degradation of the material is nothing but another disordered state. Realignment of
atoms in any system under loading just takes one of the possible disordered states. Of
course, because of a very large number of possible disordered states, the system
cannot get into every possible disordered configuration before entropy becomes
maximum, because of the closed-isolated system requirement. This is easier to
explain with an example. Say we have a box with 26 baseballs marked A to Z.
Initially they are all in alphabetical order. Say the box is 10 kg. We are allowed to
spend 100 Joule energy to shake the box. Assume we have probably 10 chances to
shake the box. Of course, for our closed system when 100 joules are consumed, the
maximum entropy is reached. However, we could not discover all possible disorder
states that 26 baseballs could take. In the same fashion when a material fails in a
fatigue loading or monotonic loading, the failure is always through a different path.
Yet the maximum entropy, the value to reach the failure is always the same.
Experimental verifications have been provided by Naderi et al. (2009), Yun and
Modarres (2019), and Tu and Gusak (2019). Failure or increase in disorder in a
system can be considered travel over an energy terrain. A soccer ball sitting in a
Fig. 3.9 Energy terrain along different paths
valley after a kick will travel. However, there are many paths it can take. Each one of
them is possible. Each one represents a disorder path. However, the ball will finally
come to stop at a valley. We can write the fundamental relation. Of course, we have
to know what path the ball will take to use the properties relevant to that path. Then
we can exactly predict in which valley the ball will stop (Fig. 3.9).
3.3.9 Thermodynamic Potential
In Newtonian continuum mechanics, the choice of thermodynamic potential is

subjective. One can use different thermodynamic potentials depending on the
problem. However, there are some mathematical conditions that all thermodynamic
potentials must satisfy. To be able to define thermodynamic potentials, first we must
define thermodynamic variables that are also called state variables or independent
variables.
Thermodynamic Variables
The thermodynamic local state of a system at any given time and space can be
defined by the thermodynamic variables using the thermodynamic fundamental
relations. The evolution of a thermodynamic state under external loading can be
defined by a succession of equilibrium states.
Each valley [microstate] in the energy landscape can be an equilibrium point.
However, that does not mean that it is at a maximum entropy point. We should point
out that, as Lemaitre and Chaboche (1990) suggest, the ultra-rapid phenomenon for
which the time scale of the evolution is in the same order as the atomic relaxation
time for a return to thermodynamic equilibrium (atomic vibrations) is excluded from
this theory’s field of applications because in continuum mechanics time scale to
reach thermodynamic equilibrium is much longer than atomic vibration periods.
Any physical phenomenon can be defined with the appropriate choice of ther-
modynamic variables. Any physical process defined by thermodynamics variables is
admissible if it satisfies all laws of thermodynamics.
It is essential to point out that when degradation is introduced into Newton’s laws
of motion, Lemaitre and Chaboche (1990) opined that there is no objective way of
choosing the internal variables best suited to study a phenomenon. This is an opinion
in which the choice is dictated by empirical experience (laboratory observations),
“physical feeling,” and very often by the type of application. This approach is the
only option when Newtonian mechanics is used for Kachanov-type empirical dam-
age mechanics models. However, in the unified mechanics theory, all thermody-
namics variables must be chosen objectively to be able to assemble the fundamental
equation based on physics and mathematics. They must represent all active micro-
mechanisms responsible for internal entropy production. In Lemaitre and
Chaboche’s (1990) approach, which is based on Newtonian Mechanics, thermody-
namics variables are up to the scientist/engineer to decide so that experimental data
can be fit into the same curve. In unified mechanics theory, all micro-mechanisms
that are responsible for entropy generation must be included, and they must be
physically and mathematically formulated in the fundamental equation without a
curve fitting a function to a test data.
Observable Thermodynamics Variables
1. Temperature
2. Space coordinates
While some authors define total strain or stress as observable variables, we do not
subscribe to this school of thought, because strain and stress are human constructs,
and definitions, not physical quantities. We can only measure displacement and then
calculate strain and stress. If something is calculated, it is not a directly observable
variable.
Independent Internal Thermodynamic Variables
1. Entropy
2. Internal energy
3. Crystal structure (phase)
Formulation of Thermodynamic Potential
Postulate: There exists a thermodynamic potential function defined by thermody-
namic variables. The thermodynamic state laws are derived from this potential.
Thermodynamic potential must be concave [f ′′(X) < 0] with respect to temperature
and convex [f ′′(X) > 0] with respect to other variables. These convex and concave
requirements ensure the thermodynamic stability requirement imposed by the
Clausius–Duhem inequality. Malvern (1969) proposes the following thermody-
namic potentials given in Table 3.1.
The Helmholtz free specific energy potential Ψ is the portion of the internal
energy that is available for doing work at a constant temperature. In continuum
Table 3.1 Thermodynamic potentials (Malvern 1969)

Thermodynamic independent
Potential Relation to u variables
Internal energy u u s, vj
Helmholtz free energy Ψ Ψ = u - sT T, vj
Enthalpy h h = u - τjvj s, τj
Free enthalpy, or Gibbs g g = u - sT - τjvj T, τj
function =h - sT
mechanics, free energy Ψ is a function of thermodynamic state variables defined

earlier
Ψ = u - sT ð3:73Þ
Enthalpy, h, is the portion of the internal energy, u, that can be released as heat
when the thermodynamic tensions are held constant. Enthalpy can also be given by
h = u - τ i vi ð3:74Þ
where and vi is a thermodynamic tension conjugate and τi is a thermodynamic

tension defined by

∂u
τi = ð3:75Þ
∂vi S
In differential form, thermodynamic potentials can be given by
du = Tds þ τi dvi ð3:76Þ

dΨ = - sdT þ τi dvi ð3:77Þ
dh = Tds - vi dτi ð3:78Þ
dg = - sdT - vi dτi ð3:79Þ
Assuming state variables defined by the subscripts are held constant, the follow-
ing relationships can be written:

∂u
T= ð3:80Þ
∂s vi

∂u
τi = ð3:81Þ
∂vi s

∂Ψ
s= - ð3:82Þ
∂T vi

∂Ψ
τi = ð3:83Þ
∂vi T

∂h
T= ð3:84Þ
∂s τi

∂h
vi = - ð3:85Þ
∂τi s

∂g
s= - ð3:86Þ
∂T τi

∂g
vi = - ð3:87Þ
∂τi T
It is assumed that thermodynamic tensions can be identified with the Piola–

Kirchhoff stress tensors, and with recoverable work assumption. Malvern (1969)
states that “The concept of recoverable work is bound to the existence of a caloric
equation [fundamental equation] of state and only thermodynamic tensions derived
from a potential defined by such an equation do work not contributing to entropy
production. The assumed existence of such a caloric equation [fundamental equa-
tion] of state does not imply our knowledge of such a formula for it.” We subscribe
to the school of thought that if there is such a recoverable work, it would violate the
second law of thermodynamics. Hence, fully recoverable work [reversible thermo-
dynamic process] and thermodynamic tensions are just abstract concepts used for
expediency in our mathematical framework.
Here, it is assumed that internal energy u is a potential for the thermodynamic
tensions when entropy is constant. Of course, this assumption eliminates using
internal energy as a thermodynamic potential in any real process, because the
derivative of internal energy with respect to entropy is not zero. However,
Helmholtz’s free energy density is a thermodynamic potential in an isothermal
process. Since most engineering problems are solved in incremental format, Helm-
holtz free energy is more appropriate as a thermodynamic potential. It should be
clarified that the term thermodynamic tension is not the same as universal thermo-
dynamic force or just any stress tensor.
Helmholtz free energy can be implemented as a thermodynamic potential, for an
elastic-plastic solid under thermo-mechanical loads. Helmholtz free energy for an
elastic-plastic solid is given by
Ψ = Ψð½D - Dp , T Þ = ΨðDe , T Þ ð3:88Þ
Then we can write the following relation based on the deformation gradient tensor
which is a summation of elastic and plastic parts:
∂Ψ ∂Ψ ∂Ψ
= - ð3:89Þ
∂De ∂D ∂Dp
We can also write
_ = ∂Ψ : De þ ∂Ψ T_
Ψ ð3:90Þ
∂De ∂T
Thermodynamic potential must satisfy Clausius–Duhem inequality, which is

given by
ds r 1 q gradT
- þ divq - : 2 ≥ 0 ð3:91Þ
dt T ρT ρ T
If we substitute r from the conservation of energy equation, we obtain the

fundamental inequality that contains both the first and second laws of
thermodynamics

ds 1 q 1 du
þ div - - σ : D þ divq ≥ 0 ð3:92Þ
dt ρ T T dt
Since

q divq q gradT
div = - ð3:93Þ
T T T2
Multiplying each term with T,

ds du gradT
ρ T - þ σ : D-q ≥0 ð3:94Þ
dt dt T
Helmholtz-specific free energy is given by
Ψ = u - Ts ð3:95Þ
Differentiating Eq. (3.95) we obtain
dΨ du ds dT
= -T -s ð3:96Þ
dt dt dt dt
We can write

dΨ dT ds du
- þs =T - ð3:97Þ
dt dt dt dt
If we substitute this new relationship with fundamental inequality, we get


∂Ψ dT gradT
σ : D-ρ þs -q ≥0 ð3:98Þ
dt dt T
We should point out that in the conservation of energy time derivative of internal
energy is multiplied by density (ρ), which is dropped in this derivative for consis-
tency of units. Now we can substitute dΨ
dt in the fundamental inequality. Meanwhile,
we can split the deformation–gradient tensor, D, into elastic and plastic components:
D = De þ D p ð3:99Þ
Hence, we can write

∂Ψ ∂Ψ _ _ gradT
σ : ðDp þ De Þ - ρ : D e
þ T þ sT -q ≥0 ð3:100Þ
∂De ∂T T

∂Ψ e p ∂Ψ _ gradT
σ -ρ : D þ σ : D - ρ s þ T -q ≥0 ð3:101Þ
∂De ∂T T
If we assume a small strain formulation under pseudo-static loading, a

deformation–gradient tensor is equal to a small strain rate tensor:
De þ DP = ε_ e þ ε_ p ð3:102Þ
In many continuum mechanics textbooks for thermo-elastic materials, the fol-

lowing assumption is made for T_ = 0:
∂Ψ
sþ =0 ð3:103Þ
∂T
However, Eq. (3.103) assumes that τidvi= 0, which is not true. Because entropy
generation happens due to many mechanisms, not just temperature change. More-
over, the reversible process is imaginary; it cannot happen in real life, without
violating the second law of thermodynamics. There is always entropy generation.
Therefore, thermo-elastic laws are defined as an imaginary process where τidvi=
0 ε_ p = 0,grad T = 0 and T is arbitrary. Then the following relations define thermo-
elastic laws in Newtonian mechanics.
∂Ψ ∂Ψ
σ =ρ and s = - ð3:104Þ
∂εe ∂T
The stress-Helmholtz free energy relation in Eq. (3.104) assumes that stress is
independent of change in entropy, of course, which is not true.
Thermodynamic Forces
Thermodynamic potential Ψ(D, T, Vi, . . .Vr) is a function of thermodynamic state
variables. In constitutive modeling, the concept of thermodynamic force is very
convenient for material modeling. It is defined by

∂Ψ
Ai = ρ ð3:105Þ
∂V i
where Ψ is the thermodynamic specific free energy potential.
Ψ = Ψðε, s, T, V k Þ ð3:106Þ
Ψ is a function of observable state variables and internal variables.

Of course, the vector of this force is normal to the thermodynamic potential Ψ
surface.
Dissipation Potential
In Newtonian mechanics-based constitutive modeling, a phenomenological dissipa-
tion potential is defined. This is also the basis for yield surface or plastic potential
used in the theory of plasticity. Lemaitre and Chaboche (1990) indicate correctly that
the whole problem of modeling a phenomenon (material behavior) lies in the
empirical determination of the analytical expressions for the thermodynamic poten-
tial Ψ and dissipation potential φ or its dual φ, and their identification in character-
istic experiments. Therefore, constitutive modeling of Newtonian mechanics of
materials is an empirical process.
In Newtonian mechanics, it is postulated that there exists a dissipation potential
(also called pseudopotential in some books). This dissipation potential is expressed
as a continuous and convex scalar-valued function of the state variables E_ P , -
→
V_ k , - q : Of course, this definition limits the potential to be valid for thermo-
T
mechanical loading only. Hence, it can be given by
_ V_k , q=T
φ e, ð3:107Þ
The state variable Vk is used to define any internal state variable. Dissipation
potential is assumed to have a zero value at the origin of the state variables.
However, this is not a mathematical requirement. It is possible to define a dissipation
function with a non-zero value at the origin. The complementary relationships
between internal state variables and associated variables can be defined by
employing dissipation potential. The normality rule requires that the first derivative
of the dissipation potential with respect to state variables is normal to the potential
surface and directed outwards:
∂φ ∂φ → ∂φ
σ= Ak = - g=- ð3:108Þ
∂_E ∂Vk ∂ðq=T Þ
Table 3.2 Dissipation flux The flux of state variables Dual variables
variables and dual variables
e_ σ
(Lemaitre and Chaboche
1990) -Vk Ak
-
→
q g = grad T
T
We should point out that in most continuum mechanics textbooks dissipation

potential is a function of plastic strain, not elastic strain. However, this approach
assumes that elastic strain does not cause dissipation and is completely recoverable.
However, this approach is not correct, because there is dissipation during elastic
response. If there were no dissipation during elastic response, there would not be
fatigue under elastic loading.
According to Lemaitre and Chaboche (1990), the thermodynamics forces are the
!
components of the vector gradφ , which are normal to the φ surface in the state
variable space.
The term “thermodynamic force” is an abstract force concept that is supposed to
satisfy the laws of thermodynamics. However, when elastic deformations, all
entropy generating mechanisms are ignored they do not satisfy the laws of thermo-
dynamics, in the strict sense. Therefore, it is more appropriate to call these forces
pseudo-thermodynamics forces.
It is easier to formulate material models that constitute laws in state variables that
can easily be calculated or measured. Therefore, the Legendre–Fenchel transforma-
tion can be used to define the complementary potential corresponding to dissipation
potential. Dissipation variables for the thermomechanical system are given in
Table 3.2.
→
The corresponding potential is φ σ, Ak , g , which is the dual of the dissipation

→
potential φ E_ , V_k , q =T . For details of the Legendre–Fenchel transformation of the
dissipation, potential readers are referred to Lemaitre and Chaboche (1990).
The corresponding potential must be differentiable. The normality rule also
applies to the corresponding potential. Therefore, constitute relation can be given,
∂φ
e_ p = ð3:109Þ
∂σ
∂φ
- V_k = ð3:110Þ
∂Ak
q ∂φ
- = ð3:111Þ
T ∂g
The potentials φ and φ must be non-negative, convex, and zero at the origin to
satisfy Clausius–Duhem inequality. According to Lemaitre and Chaboche (1990),
the normality rule is sufficient to ensure the satisfaction of the second principle of
thermodynamics, (assuming all entropy generating mechanisms are represented in

the potential) but it is not a necessary condition. However, this assumption cannot be
substantiated mathematically because it ignores entropy generation due to other
mechanisms. This rule applies to “generalized standard materials” under thermo-
mechanical loads only. Lemaitre and Chaboche (1985) define standard material as

that for which only the first of the above three rules E_ P = ∂φ ∂σ
applies. This first
relation yields the plasticity or viscoplasticity. The standard material here would
only include metals under thermo-mechanical loading.
Determination of dissipation potential in Newtonian mechanics is an empirical
process. The dissipation potential represents the energy dissipated in the system by
heat, mechanical work at the lattice level, and all other mechanisms. While potential
cannot be directly measured, state variables that define the potential can be measured
or indirectly calculated. Therefore, these potentials are defined in terms of state
_ V_k , -T q or dual variables σ, Ak, g.
variables e,
The dissipation potentials can be written as a function of the rate of state variables
or as a function of state variables themselves: [Note that in Newtonian mechanics
dissipation potential is not a function of entropy directly. It is usually a function of
strain, or stress.]

q
_ V_k ,
φ e, or φðe, T, V k Þ ð3:112aÞ
T
φ ðσ, Ak , gÞ or φ ðE, T, V k Þ ð3:112bÞ
Decoupling of Intrinsic and Thermal Dissipation

The dissipation process involves reversal mechanisms, because we assume that
thermal loading in an unconstrained system does not lead to any irreversible
dissipation. Thermal dissipation is usually treated separately. Therefore, the dissi-
pation potential can be written as the sum of two terms one representing intrinsic
dissipation and the other thermal dissipation.
φ = φ1 ðσ, Ak Þ þ φ2 ðqÞ ð3:113Þ
The second law of thermodynamics must be satisfied by each dissipation term.

The following inequalities now must be satisfied:
∂φ1 ∂φ1
σ : eP - Ak V_ k = σ : þ Ak ≥0 ð3:114aÞ
∂σ ∂Ak
q ∂φ
-g - -g 2 ≥0 ð3:114bÞ
T ∂g
3.3.10 Time-Independent Dissipation (Instantaneous

Dissipation)
When we were discussing the laws of thermodynamics, it was stipulated that energy
exchange could not happen instantaneously. However, in constitutive modeling of
materials assuming that the material response is independent of the rate simplifies the
modeling and analysis significantly. Therefore, such a simplification yields the
theory of plasticity. As stated by Lemaitre and Chaboche (1990) when the dissipa-
tion potential φ e_ p , V_k is a positive, homogeneous function of degree one, its dual
function φ(σ, Ak) is non-differentiable.
Convexity of the dissipation potential φ(σ, Ak) is checked with an indicator
function:
When
φ = 0 if f < 0 E_ P = 0 ð3:115aÞ
φ = þ 1 if f = 0 E_ P ≠ 0 ð3:115bÞ
The indicator function f (also referred to as yield criteria) must be convex and
defined by the same dual variables,Ak f(σ, Ak). Based on the normality rule, it is
possible to show that
(
p ∂F _ f_ = 0
e_ = λ if ð3:116Þ
∂σ f =0
where F is a potential function (yield surface/function) that is equal to f in the case of

associative plasticity theories and λ_ is a scalar multiplier determined by the consis-
tency condition of
f_ = 0:
Equations describing the normality rule are then given by
∂F _ ∂f
e_ P = λ_ =λ ð3:117Þ
∂σ ∂σ
∂F ∂f
- V_k = λ_ = λ_ ð3:118Þ
∂Ak ∂Ak
when the plastic strain increment is not normal to F function (yield surface). Then
the plastic strain is normal to a new function Q, plastic potential. However, this is
called non-associative plasticity, because the increment of the plastic strain is not
normal to the yield function F but to a plastic potential function Q.
Non-associative plasticity is governed by
∂Q Q=0
e_ P = λ_ if ð3:119Þ
∂σ Q_ = 0, if < 0 → E_ P = 0
The theory of plasticity fundamentals and implementation details are discussed in

the following chapters.
3.3.11 Dissipation Power and Onsager Reciprocal Relations
Earlier in the chapter, we discussed separating internal entropy production into two
parts, namely, mechanical work dissipation and the dissipation due to heat
conduction.
Specific entropy, s, the equation was derived earlier as

ds 1 q 1 du
þ div - - σ ij Dij þ qi,i ≥ 0 ð3:120Þ
dt ρ T T dt
Remembering the first law of thermodynamics
du
ρ = σ ij Dij þ ρr - qi,i ð3:121Þ
dt
We can write specific entropy rates as
ds 1 r 1
= σ D þ - q ð3:122Þ
dt ρT ij ij T ρT i,i
Earlier we also derived specific entropy production as
ds r 1 q
γ - þ q - i gradT ð3:123Þ
dt T ρT i,i ρT 2
Hence, we can write
1 q
γ= σ D - gradT ð3:124Þ
ρT ij ij ρT 2
Thus, internal entropy generation is separated into two parts; the first part is due to
mechanical work dissipation, and the second part is due to irreversible heat conduc-
tion. The strong form of Clausius–Duhem inequality requires
1
σ D >0 ð3:125Þ
ρT ij ij
qi
- gradT > 0 ð3:126Þ
ρT 2
From here onward, we will use Malvern’s (1969) description of Onsager recip-
rocal relations, as follows:
“The terms in the internal entropy production are called generalized irreversible
forces X and fluxes J. Selection of force and flux term is arbitrary. However, it is
assumed that the dot product of the generalized irreversible force X and the gener-
alized flux vector J must give the dissipation power, per unit mass”. Applying this
concept to the formulation above, we can assign:
1
Generalized irreversible forces, ρT σ ij and - ρT1 2 gradT:
Generalized fluxes Dij and qi
Hence, we can write
γ = X i ∙ J = X im Jm ð3:127Þ
Constitutive equations give the fluxes as functions of the forces or vice versa.
Hence, we can write a phenomenological relation between generalized irreversible
fluxes and forces as follows:
J m = Lmk X ik ð3:128aÞ
X ik = akm J m ð3:128bÞ
where it is assumed that coefficients satisfy the Onsager reciprocal relations

(Onsager 1931)
Lmk = Lkm and akm = amk ð3:129Þ
“It is pointed out that in certain cases the Onsager reciprocal relations must be
replaced by amk(b) = akm(-b), or amk = - akm, for example in systems affected by a
magnetic field b” (de Groot 1952; Malvern 1969). The symmetry of the constitutive
matrix may also be affected by other factors such as material anisotropy or length
scale affects. Substitution of the phenomenological constitutive relations into the
entropy production yields two quadratic equations
1
γ= L X i X i ≥ 0 and γ = akm J k J m ≥ 0 ð3:130Þ
T mk m k
These quadratic forms must be positive-definite. A necessary and sufficient

condition for the positive definiteness of a quadratic equation having an asymmetric
coefficient matrix with real elements is that all the eigenvalues |Lmk - λδmk| = 0 or |
akm - λδkm| = 0 must be positive. The second quadratic equation is called a
dissipation function, Q
Table 3.3 Examples of different thermodynamic forces and corresponding fluxes in some dissi-
pative processes
Thermodynamic force, Thermodynamic
Primary mechanism X flux Examples
Heat conduction Temperature gradient Heat flux Fatigue, creep,
wear
Plastic deformation of Stress Plastic strain Fatigue, creep,
solids wear
Chemical reaction Reaction affinity Reaction rate Corrosion, wear
Mass diffusion Chemical potential Diffusion flux Wear, creep
Electrochemical reaction Electrochemical Corrosion rate Corrosion
potential density
After Imaninan and Modarres (2016)
Q = akm J k J m ð3:131Þ
Based on the above definition, the generalized irreversible force can be given by
the following relation:
1 ∂Q
X ik = ð3:132Þ
2 ∂J k
Ziegler (1963) has shown that these phenomenological relations, Eqs. (3.128a,
3.128b, 3.131) and Onsager relation if they have symmetric dissipation function,
follow from a principle of the maximum rate of entropy production or maximum
dissipation power, which he deduces for “quasi-static processes” by a statistical
mechanics approach modified to include irreversible processes” (Malvern 1969).
Table 3.3 lists some of the thermodynamic forces and fluxes for some dissipation
mechanisms.
Here we conclude our discussion of some fundamental concepts of thermody-
namics as it relates to continuum mechanics. For further understanding of thermo-
dynamics concepts, readers are encouraged to refer to thermodynamics books listed
in the References section of this chapter.
References
Anahita Imanian, Mohammad Modarres (2016) Thermodynamics as a fundamental science of

reliability, 2016/12, Proceedings of the Institution of Mechanical Engineers, Part O: Journal
of Risk and Reliability, Volume 230, Issue 6, Pages 598–608, Publisher SAGE Publications
Boltzmann, L. (1877). Sitzungberichte der Kaiserlichen Akademie der Wissenschaften.
Mathematisch-Naturwissen Classe. Abt. II, LXXVI, 373–435 (Wien. Ber. 1877, 76:373–435).
Reprinted in Wiss. Abhandlungen, Vol. II, reprint 42, pp. 164–223, Barth, Leipzig, 1909 [Kim
Sharp* and Franz Matschinsky, Translation of Ludwig Boltzmann’s Paper “On the Relationship
between the Second Fundamental Theorem of the Mechanical Theory of Heat and Probability
References 131
Calculations Regarding the Conditions for Thermal Equilibrium” Sitzungberichte der

Kaiserlichen Akademie der Wissenschaften. Mathematisch-Naturwissen Classe. Abt. II, 1877,
LXXVI, 373–435 (Wien. Ber. 1877, 76:373–435)]. Reprinted in Wiss. Abhandlungen, Vol. II,
reprint 42, pp. 164–223, Barth, Leipzig, 1909 Entropy, 2015, 17, 1971–2009.
Callen, H. B. (1985). Thermodynamics and an introduction to thermostatistics (2nd ed.). Wiley.
Callister, W. D., Jr., & Rethwisch, D. (2010). Materials science and engineering and introduction.
Wiley.
Clausius, R. (1850). Ueber die bewegende Kraft der Wärme und die Gesetze, welche sich daraus für
die Wärmelehre selbst ableiten lassen. Annalen der Physik. 79 (4): 368–397, 500–524.
Coleman, B. D. (1964). Thermodynamics of materials with memory. Archive for Rational Mechan-
ics and Analysis, 17, 1–46.
Coleman, B. D., & Mizel, V. J. (1964). Existence of caloric equations of state in thermodynamics.
The Journal of Chemical Physics, 40, 1116–1125.
Coleman, B. D., & Mizel, V. J. (1967). Existence of entropy as a consequence of asymptotic
stability. Archive for Rational Mechanics and Analysis, 25, 243.
De Groot, S. R. (1952), Thermodynamics of Irreversible Processes, Published by North Holland/
Interscience, Amsterdam/NY, 1952, First Reprint, 1952
DeHoff, R. T. (1993). Thermodynamics in materials science. McGraw Hill.
Eddington, A. S. (1958). The nature of the physical world. University of Michigan Press.
Freeman J. Dyson (1954), Scientific American, Vol. 191, No. 3 (September 1954), pp. 58–63
(6 pages)
Halliday, D., & Resnick, R. (1966). Physics. Wiley.
Jean Lemaitre, Jean-Louis Chaboche, Mechanics of Solid Materials, Cambridge University Press,
1985
Karl Ziegler (1963), The Nobel Prize in Chemistry 1963, Prize motivation: for their discoveries in
the field of the chemistry and technology of high polymers, https://www.nobelprize.org/prizes/
chemistry/1963/ziegler/facts/
Lars Onsager (1931), Reciprocal Relations in Irreversible Processes. I., Phys. Rev. 37, 405,
15 February 1931
Lemaitre, J., & Chaboche, J. L. (1990). Mechanics of solid materials. Cambridge University Press.
Naderi, M., Amiri, M., & Khonsari, M. M. (2009). On the thermodynamic entropy of fatigue
fracture. Proceedings of the Royal Society A: Mathematical, Physical and Engineering,
466(2114), 423–438.
Planck, M. (1900a). Über eine Verbesserung der Wienschen Spektralgleichung. Verhandlungen der
Deutschen Physikalischen Gesellschaft, 2, 202–204. Translated in ter Haar, D. (1967). On an
improvement of Wien’s equation for the spectrum (PDF). The old quantum theory (pp. 79–81).
Pergamon Press. LCCN 66029628.
Planck, M. (1900b). Zur Theorie des Gesetzes der Energieverteilung im Normalspectrum.
Verhandlungen der Deutschen Physikalischen Gesellschaft, 2, 237. Translated in ter Haar,
D. (1967). On the theory of the energy distribution law of the normal spectrum (PDF). The
old quantum theory (pp. 82). Pergamon Press. LCCN 66029628.
Planck, M. (1900c). Entropie und Temperatur strahlender Wärme (Entropy and temperature of
radiant heat). Annalen der Physik, 306(4), 719–737. Bibcode:1900AnP...306..719P. https://doi.
org/10.1002/andp.19003060410
Planck, M. (1900d). Über irreversible Strahlungsvorgänge (On irreversible radiation processes)
(PDF). Annalen der Physik, 306(1), 69–122. Bibcode:1900AnP...306...69P. https://doi.org/10.
1002/andp.19003060105
Ramsey, N. F. (1956, July 1). Thermodynamics and statistical mechanics at negative absolute
temperatures. Physics Review, 103, 20.
Truesdell, C., & Noll, W. (1985). The non-linear field theories of mechanics. Springer.
Truesdell, C. and Toupin, R.A. (1960) The Classical Field Theories. In: Flugge, S., Ed., Handbuck
der Physik, Vol. III/I, Springer-Verlag, Berlin.
Tu, K. N., & Gusak, A. M. (2019). A unified model of mean-time-to-failure for electromigration,
thermomigration, and stress-migration based on entropy production. Journal of Applied Phys-
ics, 126, 075109.
Yun, H., & Modarres, M. (2019). Measures of entropy to characterize fatigue damage in metallic
materials. Entropy, 21(8), 804.
Chapter 4
Unified Mechanics Theory
The term unified is used to describe the unification of universal laws of motion of
Newton and laws of thermodynamics ab initio level. As we discussed in earlier
chapters, Newton’s laws do not account for energy loss or degradation. They only
govern what happens to a system in the initial moment a load is applied to a brand-
new structure. However, the laws of thermodynamics control what happens after the
initial moment over time. Historically, continuum mechanics is based on the laws of
Newton only, and phenomenological test data fit empirical models that are supposed
to satisfy thermodynamics’ laws introduce energy loss, dissipation, and degradation.
In the next section, we will discuss the earlier work on using thermodynamics in
continuum mechanics.
4.1 Literature Review of Use of Thermodynamics

in Continuum Mechanics
Efforts to unify Newtonian mechanics and thermodynamics have been attempted

since the first formulation of the laws of thermodynamics, because Newton’s laws do
not account for dissipation of energy and degradation researchers used different
methods to incorporate energy dissipation into mechanics formulations to be able to
formulate degradation, fracture, fatigue, and life span predictions. Most of these
efforts have been based on phenomenological curve fitting methods, where an
empirical potential is used to curve fit the dissipation, failure, or degradation test
data. There is no unifying theme among these models other than fitting a function to
a test data and using it in conjunction with Newton’s laws. For the sake of com-
pleteness, we will summarize some of them. While we do not claim to include every
effort, we will summarize the most well-known work in the field.
Lagrangian mechanics (1788) [long before Thermodynamics laws were written]
can be considered the first attempt to include non-conservative forces, into Newton’s
https://doi.org/10.1007/978-3-031-18621-9_4
134 4 Unified Mechanics Theory
laws. Quoting from Haddad’s (2017) seminal work on the literature review of
thermodynamics, “In an attempt to generalize classical thermodynamics to
non-equilibrium thermodynamics, Onsager (1931, 1932) developed reciprocity the-
orems for irreversible processes based on the concept of a local equilibrium that can
be described in terms of state variables that are predicated on linear approximations
of thermodynamic equilibrium variables. Onsager’s theorem pertains to the thermo-
dynamics of linear systems, wherein a symmetric reciprocal relation applies between
forces and fluxes.
In particular, the force exerted by the thermal gradient causes a flow or flux of
matter in thermo-diffusion. Conversely, a concentration gradient causes a heat flow,
an effect that has been experimentally verified for linear transport processes involv-
ing thermo-diffusion, thermo-electric, and thermo-magnetic effects. Classical irre-
versible thermodynamics as originally developed by Onsager characterizes the rate
of entropy production of irreversible processes as a sum of the product of fluxes with
their associated forces, postulating a linear relationship between the fluxes and
forces. The thermodynamic fluxes in the Onsager formulation include the effects
of heat conduction, the flow of matter (i.e., diffusion), mechanical dissipation (i.e.,
viscosity), and chemical reactions. This thermodynamic theory, however, is only
correct for near thermodynamic equilibrium processes wherein a local and linear
instantaneous relation between the fluxes and forces holds.”
“Building on Onsager’s classical irreversible thermodynamic theory, Prigogine
(1955, 1961, 1968) developed a thermodynamic theory of dissipative
non-equilibrium structures. This theory involves kinetics describing the behavior
of systems that are far away from thermodynamic equilibrium states. Prigogine’s
thermodynamics lacks the thermodynamic fundamental equation] function of the
system, and hence, his concept of entropy for a system away from [thermodynamic]
equilibrium does not have a total differential. Furthermore, Prigogine’s characteri-
zation of dissipative structures is predicated on a linear expansion of the entropy
function about a particular [thermodynamic] equilibrium, and hence, is limited to the
neighborhood of the [thermodynamic] equilibrium. This is a severe restriction on the
applicability of this theory. In addition, his entropy cannot be calculated nor deter-
mined” (Prigogine & Herman, 1971; Prigogine & Nicolis, 1977; Haddad, 2017).
Prigogine’s work (Prigogine & Defay, 1954; Prigogine, 1955, 1957, 1961;
Prigogine & Herman, 1971; Prigogine & Nicolis, 1977) on dissipative structures
and their role in thermodynamic systems far from equilibrium won him the Nobel
Prize in Chemistry in 1977. In engineering mechanics, most of our states are near the
thermodynamic equilibrium point. Therefore, Prigogine’s work is important to
understand. Quoting from Prigogine’s own Wikipedia webpage, “Prigogine proved
that dissipation of energy in chemical systems results in the emergence of new
structures due to internal self-re-organization [in the previous sentence if we replace
“chemical” with “mechanical”, it still holds]. Prigogine (1955) drew connections
between dissipative structures and the Rayleigh-Bénard instability, and the Turing
mechanism. Turing’s (1952) mechanism describes how patterns in nature such as
stripes and spots can arise naturally out of a homogeneous uniform state. Rayleigh-
Bénard instability is a type of natural convection, occurring in a plane horizontal
4.1 Literature Review of Use of Thermodynamics in Continuum Mechanics 135
layer of fluid heated from below, in which the fluid develops a regular pattern of
convection cells known as Bénard cells (Getling, 1998; Koschmieder, 1993).”
“Prigogine’s dissipative structure theory led to pioneering research in self-
organizing systems, as well as philosophical inquiries into the formation of com-
plexity on biological entities and the quest for a creative and irreversible role of time
in the natural sciences. Prigogine’s formal concept of self-organization was used also
as a “complementary bridge” between General Systems Theory and thermodynam-
ics, conciliating the cloudiness of some important systems theory concepts with
scientific rigor.” “Definitions of self-organization and general systems theory are
described as, ‘self-organization’, also called (in the social sciences) spontaneous
order, is a process where some form of overall order arises from local interactions
between parts of an initially ‘disordered’ system. The term ‘disorder’ here defines
the initial state. The process can be spontaneous when sufficient energy is available,
not need control by any external agent. It is often triggered by random fluctuations,
amplified by positive feedback. The resulting organization is wholly decentralized,
distributed over all the components of the system, [i.e., cooling molten metal yields a
solid polycrystal metal is a good example]. As such, the organization is typically
robust and able to survive or self-repair substantial perturbations. Chaos theory
discusses self-organization in terms of islands of predictability in a sea of chaotic
unpredictability. General systems theory is the interdisciplinary study of systems. A
system is a cohesive conglomeration of interrelated and interdependent parts that is
either natural or fabricated. Every system is delineated by its spatial and temporal
boundaries, surrounded, and influenced by its environment, described by its struc-
ture and purpose or nature, and expressed in its functioning. In terms of its effects, a
system can be more than the sum of its parts if it expresses synergy or emergent
behavior. Changing one part of the system usually affects other parts and the whole
system, with predictable patterns of behavior. For systems that are self-learning and
self-adapting, the growth and adaptation depend upon how well the system is
adjusted to its environment. Some systems function mainly to support other systems
by aiding in the maintenance of the other system to prevent failure. The goal of
systems theory is systematically discovering a system’s dynamics, constraints, and
conditions and elucidating principles (purpose, measure, methods, tools, etc.) that
can be discerned and applied to systems at every level of nesting, and in every field
for achieving optimized equifinality” (Beven, 2006) quoting from Wikipedia to
define, “equifinality is the principle that in open systems a given end state can be
reached by many potential means. It also means that a goal can be reached in many
ways. The same final state may be achieved via many different loading paths.
However, in closed systems, a direct cause-and-effect relationship exists between
the initial condition and the final state of the system. Biological and social systems
are open systems, however, operate quite differently.” Simply, the idea of
equifinality implies that the same results may be achieved with different initial
conditions and in many ways.
Entropy has also been used beyond positive sciences. The seminal paper on the
topic was published by Jaynes (1957) who proved that statistical mechanics, which
is based on Boltzmann’s entropy formulation, could be generalized to information
theory independent of experimental verification. Jaynes (1957), using

von-Neumann–Shannon’s definition of entropy as a measure of uncertainty
represented by a probability distribution, showed that entropy becomes a primitive
concept more fundamental than energy. Jaynes (1957) also proved that thermody-
namic entropy is identical to the information-theory entropy of the probability
distribution except for the presence of Boltzmann’s constant.
Valanis (1971) established the irreversibility and existence of entropy as a state
function in Newtonian mechanics formulation. Valanis (1971) was able to prove this
for reversible and irreversible systems and processes, irrespective of the constitutive
equations of the system. He postulated that, in the case of a reversible system,
entropy is a function of the deformation gradients (strains) and temperature; in the
case of an irreversible system, it is also a function of n internal variables necessary to
describe the irreversibility of the system. Valanis (1971) based his proof of the
existence of entropy as a consequence of the integrability of the differential form of
the first law using an extended form of the Caratheodory conjecture, to the effect that
in the neighborhood of a thermodynamic state there exist other states, which are not
accessible by processes, which are reversible and adiabatic. Valanis (1971) proposed
the endochronic plasticity theory, which deals with the plastic response of materials
employing memory integrals, expressed in terms of memory kernels. Formulation of
this theory is based on thermodynamics concepts and provides a unified point of
view to describe the elastic-plastic behavior of materials since it places no require-
ment for a yield surface and “loading function” to distinguish between loading and
unloading. A key ingredient of the theory is that the deformation history is defined
with respect to a deformation memory scale called intrinsic time. In the original
version of the endochronic theory, proposed by Valanis (1971), the intrinsic time
was defined as the path length in the total strain space. The so-called endochronic
theory violates the second law of thermodynamics and leads to constitutive relations,
which characterize inherently unstable materials (Rivli, 1981). Aiming the correc-
tion of this deficiency, a new version of the endochronic theory was developed by
Valanis and Komkov (1980) in which the intrinsic time was defined as the path
length in the plastic strain space.
Rice (1971) used an internal-variable thermodynamic formalism for the descrip-
tion of the microstructural rearrangements to characterize metal plasticity. Rice
(1971) postulated that the theoretical foundations of constitutive relations at finite
strain for metals exhibiting inelasticity are a consequence of specific structural
rearrangements on the microscale of the material, such as metals deforming plasti-
cally through dislocation motion. Rice (1971) is a generalization of his earlier work
by Kestin and Rice (1970), where it is assumed that each microstructural arrange-
ment proceeds at a rate governed by its associated thermodynamic force. This work
became the framework for the time-dependent inelastic behavior in terms of a flow
potential and reduces to statements on the normality of plastic strain increment to
yield surface in the time-independent case. Rice’s (1971) formulation assumes
inelastic deformation under macroscopically homogeneous strain and temperature
as a sequence of constrained equilibrium states. Using equilibrium thermodynamic
formalism and thermodynamic potentials, Rice (1971) relates changes in the local
structural rearrangements to corresponding changes in the macroscopic stress or

strain states. However, Rice (1971) assumes that a discrete set of scalar internal
variables characterize the state of internal rearrangement. Each internal variable
characterizes a specific structural rearrangement. Unfortunately, this classification
does not satisfy the second law of thermodynamics. According to the second law of
thermodynamics, not the individual internal variables like stress or strain, but only
the fundamental relation [entropy generation rate] determines the new structural
rearrangement. Rice (1971) also assumes that if various equilibrium states are
considered each corresponding to the same set of values for the thermodynamic
internal variables, then neighboring states are related by the usual laws of thermo-
elasticity. This later assumption also violates the second law of thermodynamics,
because according to the second law of thermodynamics even thermo-elastic defor-
mation leads to irreversible entropy generation. The easiest way to explain this is
fatigue under elastic loading or molecular dynamics simulations under elastic load-
ing. According to the traditional thermo-elasticity assumption, if there is no inelastic
deformation, thermodynamics internal state variables do not change. As a result,
materials could never fatigue under elastic loading. Of course, this violates the
second law of thermodynamics, because there is no reversible thermodynamic
process in metals or nature in general.
Rice (1971) postulates that “if neighboring constrained equilibrium states
corresponding to different sets of internal variables are considered, we must write,
[using his variables without adopting to our notation]

V 0 S : δE - f α δξα þ θδ V 0 η = δ V 0 u ð4:1Þ
In the formulation, V0 denotes volume at some reference state at a temperature θ0,

S is Kirchoff stress (symmetric), δE is the increment of macroscopic Lagrange
(or material) strain tensor, fα defines thermodynamic forces ( f1, f2, . . ., fn acting on
internal variables, δξα set of (total number is unspecified as n) thermodynamic
internal state variables that characterize the state of internal rearrangement, θ is
temperature, δ(V0η) is the increment of entropy, and δ(V0u) is the increment of
internal energy. This is the most important equation in Rice’s (1971) theoretical
framework. The rest of the theoretical framework is based on this equation. This
equation finally leads to defining a yield surface as a potential, which is the most
important ingredient for the theory of incremental plasticity. Unfortunately, in this
equation given by Rice (1971), there is no relation between entropy and thermody-
namic forces ( f1, f2, . . ., fn ) acting on internal variables in the formulation. Deriv-
atives of these forces with respect to entropy are considered zero. As a result, in
practice defining these thermodynamic forces and internal variables becomes a trial-
error process (or even art). This point was also emphasized in Chap. 3 when we
discussed thermodynamic potential. Because of this problem, yield surface is an
empirical function with different coefficients for different materials for different
loading paths for different temperatures, different strain rates, and different length
scales or geometries of the materials. Of course, it is possible to come up with a
different yield surface for the same material, using different constants. Therefore, it
would be more accurate to call these thermodynamic forces pseudo-thermodynamics
forces, since they are empirical, and many are not directly related to the laws of
thermodynamics. While Rice (1971) interprets, fαδξα as the work of increments of
internal variables, in practice when we obtain a yield surface from experiments many
of our exponents, or “material constants” have no physical meaning to justify as an
internal thermodynamic variable, because two different scientists can have signifi-
cantly different number of constants to define a yield surface for the same material.
Our argument becomes clearer in the following formulation by Rice (1971).”
Free energy ϕ and its Legendre transform ψ (complementary energy) are given by
the following relations:
ϕ = ϕðE, θ, ξÞ = u - θη
∂ϕ ∂ϕ
ψ = ψ ðS, θ, ξÞ = E : -ϕ=E : - u þ θη ð4:2Þ
∂E ∂E
Based on definitions given in Eq. (4.2) and assuming that entropy is constant, the
variation of complementary energy can be written as
1
δψ = E : δS þ f α δξα þ ηδθ ð4:3Þ
V0
Of course, in real life when a variation is applied, in addition, to stress δS internal

variables δξα and temperature δθ, and entropy will of course change. Nevertheless,
here entropy change is assumed zero, θδη = 0. If the internal variables δξα = 0 are
also constant, Rice (1971) derives the thermo-elastic constitutive equation based on
Newtonian Mechanics, where it is assumed that no elastic deformation can generate
entropy and all derivatives with respect to entropy are zero:
∂ψ ðS, θ, ξÞ ∂ϕðE, θ, ξÞ
E= and S = ð4:4Þ
∂S ∂E
Thermodynamics forces associated with internal variables are defined by
∂ψ ðS, θ, ξÞ ∂ϕðE, θ, ξÞ
f α = V0 = - V0 ð4:5Þ
∂ξα ∂ξα
Using Maxwell’s relations, which are a set of equations derivable from the
symmetry of second derivatives of potentials. Rice (1971) arrives at
∂E ðS, θ, ξÞ 1 ∂f α ðS, θ, ξÞ
= 0 ð4:6Þ
∂ξα V ∂S
Equation (4.6) serves as the central equation in the rest of Rice (1971) in the
development of the theoretical foundations of inelastic constitutive theory. A flow
potential for the inelastic strain rate is proposed by postulating that at any given
temperature and pattern of internal rearrangement within the material, the rate at
which any specific structural rearrangement occurs is fully determined by the
thermodynamic forces. The concept, used by Rice (1971), is identical to the basis
for the unified mechanics theory where pseudo-dynamic forces are replaced with the
thermodynamic fundamental equation of the material. It is assumed that any specific
microstructural configuration can be fully determined by the thermodynamic force
associated with that microstructural rearrangement, as follows:
ξ_ β is a function of f β ,θ,ξ ðfor β = 1, 2, . . . , nÞ ð4:7Þ
Rice (1971) further postulates that the current temperature and pattern of internal
microstructural rearrangement may enter the kinetic equations as parameters, but the
influence of the macroscopic stress state on a given microstructural rearrangement
appears only through the fact that the associated force is dependent on stress. The
author acknowledges that this is not the most general class of kinetic equations;
however, it does represent conventional metal plasticity behavior where the associ-
ated shear stress governs slip on a crystallographic system or at the discrete dislo-
cation. Based on Newtonian mechanics, it is assumed that the force on a given
segment of the dislocation line governs its motion. Rice (1971) recast the kinematic
equations in an integral form as follows:
Z f
∂
ξ_ β = ξ_ α ðf , θ, ξÞ df α ð4:8Þ
∂f β 0
where the integral is carried out at fixed values of θ and ξ and defines a point function
since each term in the integrand is an exact differential. Rice (1971) further postu-
lates that thermodynamic forces may be viewed as functions of the macroscopic
stress, S, temperature θ,and internal variables ξ and then defines a function
Z f
1
ΩðS, θ, ξÞ = ξ_ α ðf , θ, ξÞ df α ð4:9Þ
V0 0
The stress derivative of this function is given by
∂ΩðS, θ, ξÞ 1 ∂f ðS, θ, ξÞ
= 0 ξ_ α ðf , θ, ξÞ α ð4:10Þ
∂S V ∂S
Rice (1971) elegantly shows that the right-hand side of this equation is equal to
the inelastic strain rate in the following way.
Let δE be the difference (variation) in strain between the neighboring constrained
equilibrium states, differing by δS, δθ, δξ. It is assumed that the variations in the
internal variables correspond to variations in the macroscopic strain. An inelastic or

plastic portion (δE)p of the strain difference is defined as that part which would result
from the change in internal variables if stress and temperature were held fixed:
∂EðS, θ, ξÞ 1 ∂f α ðS, θ, ξÞ
ðδE Þp = δξα = 0 δξα ð4:11Þ
∂ξα V ∂S
Similarly, an elastic (or thermo-elastic) portion (δE)e is defined as that which

would result from the change in stress and temperature, if the other internal variables
were held fixed. As a result, the following relations can be written:
δE = ðδE Þe þ ðδE Þp ð4:12Þ
From the complementary energy, we can write
2 2
∂ ψ ∂ ψ
ðδE Þe = : δS þ δθ ð4:13Þ
∂S∂S ∂S∂θ
Based on the viewpoint of the classical theory of irreversible processes, e.g., De

Groot and Mazur (1962a), Rice (1971) assumes that macroscopically homogeneous
deformation processes may be suitably approximated as the sequence of constrained
thermodynamic equilibrium states, each fully characterized by values for E, θ, ξ at
the corresponding instant. As a result, Rice (1971) assumes that all the preceding
relations are valid during this process. Thus, the following relations can be written:
e p
E_ = E_ þ E_ ð4:14Þ
p 1 ∂f α ðS, θ, ξÞ _
E_ = 0 δ ξα ð4:15Þ
V ∂S
An analogous expression for the elastic portion can be written from Eq. (4.13).
Therefore, Ω is called a flow potential, (or more commonly known as a yield
surface). According to the normality rule, it can be shown that the inelastic portion of
the strain-rate vector is normal to the surface of constant flow potential in stress
space.
p ∂ΩðS, θ, ξÞ
E_ = ð4:16Þ
∂S
Rice (1971) defines thermodynamics restriction on the formulation in the follow-

ing way. Following the earlier stated assumption that material is taken from one
constrained thermodynamic equilibrium state to another by an irreversible process
extending from time t1 to t2. Then the first and second laws of thermodynamics in
classical form dealing exclusively with the comparison of constrained equilibrium
states, are given by
Z t 2 Z Z

T z_ dA þ qdA dt = V 0 u2 - u1 ð4:17Þ
t1 A A
Z t 2 Z

ðq=θÞdA dt ≤ V 0 η2 - η1 ð4:18Þ
t1 A
where T is the surface stress vector (traction), q is the heat supplied to the surface per
unit area (positive inward), z_ is the deformation rate, u is the internal energy, and η is
the entropy. Time integral from t1 to t2 follows the irreversible process. It is
important to point out that elastic deformation here is considered a reversible
process; hence, fatigue under elastic stresses is not possible. Alternatively, it is
assumed that during a molecular dynamics simulation under elastic loading atoms
return to their original lattice site. Based on earlier work by Kestin and Rice (1970),
following the classical approach to irreversible processes for a macroscopically
homogeneous system with internal variables, Rice (1971) views all actual processes
as a sequence of constrained equilibrium states. The work of surface stresses can be
written in terms of internal stresses and strains as V 0 S : E_ for the constrained
equilibrium states and the temperature is assumed uniform in the material. Then
the first and second laws of equilibrium thermodynamics can be generalized for any
actual homogenous process as follows:
S : E_ þ Q_ = u_
Z
1
Q = 0 q dA ≤ θη_ ð4:19Þ
V A
where Q is the total heat supply rate per unit reference volume. Rice (1971) further
defines entropy generation rate due to internal variables and conjugate thermody-
namic forces as
1
σ= f α ξ_ α ≥ 0 ð4:20Þ
θV 0
Then, Rice (1971) defines the total change in entropy as, the summation of Q, the
total heat supply rate per unit reference volume plus the entropy generation due to
internal variables and thermodynamic forces as
Q þ θσ = θη_ ð4:21Þ
Q is defined as positive inward, in Rice’s (1971) formulation, which we are quoting

without modification. It is assumed that heat entering from outside does not increase
the irreversible entropy generation. On the other hand, heat generated in the system
and dissipating outward leads to irreversible entropy generation.
The second law requires that the non-negative work rate of associated forces on
internal variables is given by
∂ϕðE, θ, ξÞ
ξ_ α ≥0 ð4:22Þ
∂ξα
Remember in Eq. (4.5) fα was defined as
∂ϕðE, θ, ξÞ
f α = - V0 ð4:23Þ
∂ξα
Rice (1971) concludes “That is, the rates of internal arrangement occurring during
a process must be such that the free energy would decrease if strain and temperature
were held fixed at current values.”
In Newtonian mechanics, Eqs. (4.22) and (4.23) are primarily used to require that
flow potential must be convex and smooth surface but are not included in Newton’s
universal laws to degrade the energy or material. Degradation is imposed by
empirical formulation utilizing a damage potential obtained from experimental
data. Dissipation is imposed by an empirical constant, like damping, or by an
empirical function.
Later Rice (1977) derived thermodynamic restrictions on the quasi-static growth
or healing of Griffith cracks. However, as presented they were nothing beyond
global restrictions on the detailed molecular kinetics of crack growth. Therefore,
there was no attempt to integrate these restrictions into Newton’s universal laws of
motion. Other work by Rice (1977) on the use of thermodynamics in the theory of
plasticity and fracture mechanics was based on or a derivative of his work
discussed here.
Bazant (1972) used thermodynamics for modeling the mechanics of interfaces
in concrete structures, where a thermodynamically consistent formulation of
interacting continua with surfaces based on surface thermodynamics was proposed.
Creep, shrinkage, and delayed thermal dilation and their interaction with adsorbed
water layers confined between two solid adsorbent surfaces were accounted for in
the formulation.
Kijalbaev and Chudnovsky (1970) and Chudnovsky (1973, 1984) proposed a
probabilistic model of the fracture process unifying the phenomenological study of
long-term strength degradation of materials, fracture mechanics, and statistical
approach to fracture mechanics. Chudnovsky (1984) used irreversible thermody-
namics to model the deterministic side of the failure phenomenon and stochastic
calculus to account for the failure mechanisms controlled by “chance,” particularly
the random roughness of fracture surfaces. Kijalbaev and Chudnovsky (1970) and
Chudnovsky (1973, 1984) derived the entropy production for an elastic medium
with damage. Citing Swalin (1972) on the invariance of the entropy jump during a
phase transition with respect to stresses and temperature, Chudnovsky (1973)
hypothesized that the entropy jump invariance can be used to predict the local
failure. The author proposed pairing of a phenomenological critical damage param-
eter and entropy jump, ΔS, requirement as a local failure criterion. It is important to
point out that Chudnovsky proposed calculating entropy generation in a fracture
mechanics problem for all mechanisms, damage, translation, rotation isotropic

expansion, and distortion of the active crack zone.
Klamecki (1980a, b, 1984) proposed an entropy-based model of plastic deforma-
tion energy dissipation in sliding. In the entropy production rate author included both
microstructural changes and internal heat generation. This is the first paper in the
literature we could find where a partial derivative of entropy was taken with respect
to internal energy, microstructure generated entropy, surface area, and mass. As a
result, Klamecki (1980a, 1984) postulated that the temperature component of inter-
nal energy U, microstructural energy, G, surface energy, and chemical potential are
represented in entropy calculations, and defined entropy S by
S = SðU, G, A, M Þ ð4:24Þ
∂S ∂S ∂S ∂S
dS = dU þ dG þ dA þ dM ð4:25Þ
∂U ∂G ∂A ∂M
And
TdS = dU þ ϕdG þ γdA þ μdM ð4:26Þ
[Note: In the derivation of Eq. (4.26) Klamecki assumes U = ST.] where the
temperature, internal energy, microstructural energy, surface energy, and chemical
potential are represented by T, U, ϕ, γ, and μ, respectively. Klamecki (1980a, b,
1984) defined the conservation of energy by
dU = δQ þ δW ð4:27Þ
where Q represents heat and W represents mechanical work. δ is used to imply the
equation is for a process [variation]. Work, W, is assumed to be due to plastic
deformation only, and then taken as stress multiplied by plastic strain. The heat,
Q, has two components: one is assumed to be due to internal heat generation, R, in
the system, and the other one heat flows into the system due to temperature gradient.
As a result, internal energy can be written as
dU = K i ð∇T Þi þ R þ Mρ - 1 τij dεij ð4:28Þ
where K is the thermal conductivity and ρ is the mass density. Substituting internal
energy given by Eq. (4.28) in entropy Eq. (4.26) leads to
TdS = K i ð∇T Þi þ R þ Mρ - 1 τij dεij þ ϕdG þ γdA þ μdM ð4:29Þ
As Klamecki (1984) points out this definition of entropy is very valuable since it
contains, or may be further generalized to contain, all the energy dissipation mech-
anisms in sliding bodies. Of course, there is nothing in the equation that limits it to
siding bodies only. Klamecki (1984) speculated that there are three regimes of
system behavior, which are of interest when the intent is to describe changes in the
system by the use of entropy generation. At thermodynamic equilibrium, the system
entropy is at maximum, and the change in entropy with time is zero. When the
system is not in equilibrium because of a continual supply of energy to it, it can be
near equilibrium or far from the equilibrium state. Klamecki (1984) asserted that the
process occurring in the nonequilibrium states could be analyzed by studying the
entropy production in the system. The entropy created in the system is given by
TdS = R þ Mρ - 1 τij dεij þ ϕdG þ γdA þ μdM ð4:30Þ
These terms have the general form of a thermodynamic force F, multiplied by a

corresponding flux, J. Then entropy production can be expressed using the Onsager
relation
X
n
S= Fi J i ð4:31Þ
i=1
where n is the number of internal entropy generation mechanisms. Entropy produc-

tion rate is defined by
n
dS X dJ dF
S_ = = Fi i þ J i i ð4:32Þ
dt i=1
dt dt
Here, Klamecki (1984) postulates that the first term in the summation represents
the entropy generation rate far from the equilibrium state where thermodynamic
force does not change over time. The second term represents the entropy generation
near equilibrium. As a result, the summation can be summarized as follows:
S_ = S_ J þ S_ F ð4:33Þ
In near equilibrium, entropy production rate S_ F will continuously get smaller,

according to the second law of thermodynamics. Thermodynamic forces change in
such a way that a stationary state of the minimum rate of entropy generation results
in equilibrium, as stipulated by the second law of thermodynamics.
S_ F = 0 ð4:34Þ
Klamecki (1984) attributes this principle, Eq. (4.33), to Glansdorff and Prigogine
(1971a, b). As stated by Klamecki (1984) far from the thermodynamic equilibrium,
no widely accepted entropy production evolution criterion has been formulated so
far. Laws of thermodynamics are valid around equilibrium points only. Since in
engineering mechanics most of our problems are around equilibrium points and
solved as a sequence of constrained equilibrium points, not having laws governing

far from equilibrium is not a concern.
Evans et al. (1993), proposed the fluctuation theorem (FT). A laboratory exper-
iment that verified the validity of the FT was carried out in 2002, where a plastic bead
was pulled through a solution by a laser. Fluctuations in the velocity were recorded
that were opposite to what the second law of thermodynamics would dictate for
macroscopic systems (Wang et al., 2002; Chalmers, 2016; Gerstner, 2002; Searles &
Evans, 2004).
The fluctuation theorem does not state or prove that the second law of thermo-
dynamics is wrong or invalid. The second law of thermodynamics is valid for
macroscopic systems at equilibrium or near equilibrium. Rivas and Martin-Delgado
(2017) have also encountered a partial violation of the second law of thermodynam-
ics in a quantum system known as the Hofstadter lattice. Ostoja-Starzewski (2016),
Ostoja-Starzewski and Raghavan (2016) proved violations of the second law are
relevant as the length and/or time scales become very small. The second law then
needs to be replaced by the fluctuation theorem, and mathematically, the irreversible
entropy is a sub-martingale. As indicated above, these are far from thermodynamic
equilibrium states. This partial violation has no place within the framework of
classical thermodynamics because it is a spontaneous event that does not affect the
laws of thermodynamics at the macroscale at equilibrium or near equilibrium states.
Klamecki (1984) summarized these concepts in a very simple figure. The author
postulates that entropy production is a function of l thermodynamic variables and so
can be represented by a surface in (l + 1) dimensional space. For ease of plotting only
two variables are used in Fig. 4.1. The equilibrium exists at the state of maximum
entropy and zero entropy generation rate and is represented by point E in Fig. 4.1. If
the system is maintained in a non-equilibrium state by supplying energy to it, two
processes are usually considered. Near equilibrium, the system will evolve from its
initial state A to state B at the minimum rate of entropy generation. If the system is
far from equilibrium, it may move from its initial state of C to state D but at the
present, no description of this state or the process in terms of entropy production is
available in the literature. The only accepted governing principle is energy conser-
vation for systems far from equilibrium.
Whaley (1983) proposed a model for fatigue crack nucleation using irreversible
thermodynamics to quantify the damage caused by plastic strain. Whaley’s (1983)
model is based on the hypothesis that entropy gain which results from dynamic
irreversible plastic strain is a material constant and the plastic strain can be used as a
parameter for an empirical function curve-fit to fatigue test data to simulate fatigue
behavior of metals. Whaley (1983) also postulated that structural damage by fatigue
could be quantified by a single material parameter, the critical entropy threshold of
fracture. Whaley (1983) postulated that the critical entropy threshold of fracture is
therefore related to the irreversible part of the fracture energy. The critical entropy
threshold of static fracture is a random variable and the variability can be quantified
by a confidence interval. According to Whaley (1983), the confidence interval for the
entropy threshold just comes from the variance of the plastic strain. It is noteworthy
that Whaley (1983) assumed that only plastic strain generates irreversible entropy; of
Fig. 4.1 Graphical

representation of entropy
generation associated with
system states at equilibrium
(state E) near equilibrium
(state B) and far from
equilibrium (state D) as a
function of thermodynamic
state variables 1 and
2. (After Klamecki (1984))
course, that is not accurate, because using plastic strain alone as a metric violates the
second law since if there is no increase in plastic strain but loading under the same
stress level, there will be no entropy generation according to this model.
Naderi et al. (2010) were able to prove experimentally that total cumulative
entropy generation is constant at the time of failure and is independent of sample
geometry, load, and loading frequency. They named this critical entropy value
Fatigue Fracture Entropy (FFE). Figure 4.2 shows experimental fatigue fracture
entropy invariance with respect to geometry and loading frequency. Total accumu-
lated entropy of metals, undergoing repeated cyclic load as it reaches the point of
fracture is a constant value, independent of load amplitude, geometry, size of the
specimen, frequency, and stress state; Naderi et al. (2010) experimentally validated
the invariance of fatigue fracture entropy with respect to loading path and displace-
ment loading amplitude for Al 6061-T6 samples.
In Fig. 4.2 fracture fatigue entropy remains at about 4 MJm-3 K-1 for both
tension-compression and bending fatigue. Displacement amplitude is varied from
25 to 50 mm. A filled square represents tension-compression; a filled circle repre-
sents bending; a filled star represents torsion data.
Amiri and Khonsari (2012) and Jang and Khonsari (2018) also did extensive
experiments to show that fatigue fracture entropy remains constant independent of
loading condition, loading frequency, and sample geometry. Their samples were
AISI 1018 carbon steel and Al 7075-T6. Liakat and Khonsari (2015) measured
fatigue fracture entropy of un-notched and V-notched specimens. They observed that
fatigue fracture entropy remains constant.
Imanian and Modarres (2015, 2018) and Yun and Modarres (2019) experimen-
tally proved the concept of using entropy as a degradation metric; authors discussed
the entropic characterization of the corrosion-fatigue degradation mechanism. They
proposed an entropy-based damage prognostics and health management technique
for integrity assessment and remaining useful life prediction of aluminum 7075-
T651 specimens. Their experimental validations proved that using entropy as a
Fig. 4.2 Fatigue fracture entropy versus the number of cycles to failure for different bending
fatigue tests of Al 6061-T6 with different specimen thickness, frequencies, and displacement
amplitudes. Fatigue fracture entropy remains at roughly 4 MJm-3K-1, regardless of thickness,
load, and frequency. Displacement amplitude varied from 25 to 50 mm. Filled circle, thickness =
6.35 mm, f = 10 Hz; filled diamond, thickness = 3.00 mm, f = 10 Hz; filled star, thickness =
4.82 mm, f = 10 Hz; unfilled circle, thickness = 6.35 mm, f = 6.5 Hz; unfilled triangle, thickness =
4.82 mm, f = 12.5 Hz; unfilled star, thickness = 6.35 mm, f = 6.5 Hz; unfilled diamond, thickness =
6.35 mm, f = 12.5 Hz. (After Naderi et al. (2010))
thermodynamic state function for damage characterization is an effective way of

handling the endurance threshold uncertainties for life prediction purposes. The
authors also derived the thermodynamic fundamental equation for corrosion fatigue.
Imanian and Modarres (2015, 2018) and Yun and Modarres (2019) also
performed experiments where multiple entropy generation mechanisms were con-
tributing to the damage evolution of corrosion-fatigue total entropy for different
loading conditions. They measured the cumulative final value of fracture corrosion-
fatigue entropy. The final entropy value is between 0.7 and 1.5 MJm-3 K-1. The
authors observed that there is a narrow band of distribution of entropy to failure data
points [fracture entropies] irrespective of loading condition. They concluded that
entropy can quantify the uncertainties associated with microstate variabilities. Fur-
thermore, they stated that it reveals the independence of entropy to the loading
condition (i.e., failure path). Fracture entropy’s slim distribution band can be
interpreted due to uncertainties, such as instrumental measurement errors, the legit-
imacy of the assumptions considered in entropy evaluation, weak control of the
experimental, operational, and environmental conditions, and human error.
Sosnovskiy and Sherbakov (2015, 2016) postulated the existence of a generalized
theory of evolution based on the concept of tribo-fatigue entropy. The essence of the
proposed approach is that tribo-fatigue entropy is determined by the process of
degradation of any system due to thermodynamic mechanical effects causing the
change in the state. Sosnovskiy and Sherbakov (2016) derived a framework for the
law of entropy increase in the general form. They also provided extensive experi-
mental validation for the theory of what they called “mechanothermodynamics.”
They stated, “It is shown that mechanothermodynamics—a generalized physical
discipline—is possible by constructing a bridge between Newtonian mechanics and
thermodynamics. The entropy is the bridge between thermodynamics to mechanics.”
The first and the second principles of mechanothermodynamics were presented.
They formulated the foundation of the general theory of degradation evolution of
mechanothermodynamic systems for:
• Energy theory of limiting states
• Energy theory of damage
• Foundations of the theory of electrochemical damage
They provided mathematical fundamentals of the theory of interaction between
damage caused by loads of different nature (mechanical, thermal, etc.). The authors
proposed a single function for the critical damage (limiting) state of metals and
polymeric materials operating in different conditions. The analysis of 136 laboratory
experiment results showed that logarithmic function is fundamental: it is valid for
low-, average-, and high-strength pure metals, alloys, and polymers over a wide
range of temperatures of medium (from helium temperature to 0.8 TS, where TS is
the material melting temperature) and mechanical stresses (up to the strength limit
for single static loading) while the fatigue life was of the order of 106–108 cycles.
Mechanothermodynamics uses the same idea as the unified mechanics theory to
unify Newtonian mechanics and thermodynamics. That is using entropy as a bridge
to connect Newtonian mechanics and thermodynamics, which was first published in
Basaran and Yan (1998). Sosnovskiy and Sherbakov (2015, 2016) used an empirical
logarithmic function to fit the experimental evolution of degradation. It is not clear
why they do not use the Boltzmann equation, since their logarithmic function can be
obtained directly from the Boltzmann equation. The term tribo-fatigue-entropy in
their work refers to entropy generation in tribology [which is the focus of their
research] and the fatigue process. They surmise that if an analogy between light and
strain energy is justified, then the strain energy absorption law may be like
Bouguer’s light absorption law. This law, which is also exponential, becomes the
basis for their empirical degradation function formulation.
Haddad (2017) published probably one of the most comprehensive reviews of the
history of thermodynamics from its classical to its postmodern forms. Haddad et al.
(2005) and Haddad (2019) also provided a general systems theory framework for
thermodynamics which attempts to harmonize thermodynamics with classical New-
tonian mechanics. The main idea used by Haddad (2019) to unify mechanics and
thermodynamics is attributed to Basaran and Yan (1998). Haddad stated, “The
dynamical system’s notion of entropy proposed by Haddad et al. (2005, Haddad
(2019), Basaran and Yan (1998), Basaran and Nie (2004)), Sosnovskiy and
Sherbakov (2016) involving an analytical description of an objective property of
matter can potentially offer a conceptual advantage over the subjective quantum
expressions for entropy proposed in the literature (e.g., Daróczy entropy, Hartley
entropy, Rényi entropy, von Neumann entropy, infinite-norm entropy) involving a
measure of information. An even more important benefit of the dynamical systems

representation of thermodynamics is the potential for developing a unified classical
and quantum theory that encompasses both mechanics and thermodynamics without
the need for statistical (subjective or informational) probabilities.”
Cuadras et al. (2015) proposed a method to characterize electrical resistor damage
based on entropy measurements. They postulated that irreversible entropy and the
rate at which it is generated are more convenient parameters than resistance for
describing damage because they are essentially positive in virtue of the second law
of thermodynamics, whereas resistance may increase or decrease depending on the
degradation mechanism. They tested commercial resistors to characterize the dam-
age induced by power surges. Resistors were biased with constant and pulsed
voltage signals, leading to power dissipation in the range of 4–8 W, which is well
above the 0.25 W nominal power to initiate failure. Entropy was inferred from the
added power and temperature evolution. They studied the relationship between
resistance, entropy, and damage. They stated that the power surge dissipates into
heat (Joule effect) and damages the resistor. They observed that there is a correlation
between entropy generation rate and resistor failure. Cuadras et al. (2015) concluded
that damage could be conveniently assessed from irreversible entropy generation.
Cuadras et al. (2016) proposed a method to monitor the aging and damage of
capacitors based on their irreversible entropy generation rate (Fig. 4.3). The authors
overstressed several electrolytic capacitors in the range of 33 me – 100 mF and
monitored their entropy generation rate. A strong relationship between capacitor
degradation and entropy generation rate was observed. Therefore, they proposed a
threshold entropy generation rate as an indicator of capacitor time-to-failure. This
magnitude is related to both capacitor parameters and a damage indicator such as
entropy. They validated the model as a function of capacitance, geometry, and rated
voltage. Moreover, they identified different failure modes, such as heating, electro-
lyte dry-up, and gasification from the dependence of entropy generation rate with
temperature.
Fig. 4.3 Capacitor entropy 0.100 4

generation rate versus time 3 S
rate
(seconds) for different
P(W)
2
voltage levels. (After 0.075
1
Srate(WK−1)
Cuadras et al. (2016)) 0

280 300 320 340 360 380
0.050 T(K)
20 V
25 V
30 V
0.025
35 V
40 V
0.000
0 10 20 30 40 50
t(s)
Cuadras et al. (2017) proposed a method to assess the degradation and aging of
light-emitting diodes (LEDs) based on an irreversible entropy generation rate.
Researchers degraded several LEDs and monitored their entropy generation rate in
accelerated tests. They compared the thermo-electrical results with the optical light
emission evolution during degradation. They found a good relationship between
aging and entropy generation rate because they both are related to device parameters
and optical performance. They proposed a threshold of entropy generation rate as a
reliable damage indicator of LED end-of-life that can avoid the need to perform
optical measurements to assess optical aging. The method is far more physics-based
and beyond the typical empirical statistical models based on curve fitting to test data
for lifetime prediction provided in the literature. Cuadras et al. (2017) tested different
LED colors and electrical stresses to validate the electrical LED model and analyzed
the degradation mechanisms of the devices to validate the model.
There has been significant interest in using entropy generation rate as a damage
metric to predict degradation, fracture, and fatigue life prediction, such as Basaran
and Chandaroy (2002); Basaran and Tang (2002); Basaran et al. (2003, 2005,
2008a, b); Tang and Basaran (2003); Gomez and Basaran (2005, 2006); Lin and
Basaran (2005); Gomez et al. (2006); Li et al. (2008); Li and Basaran (2009); Gunel
and Basaran (2010, 2011a, b); Basaran and Lin (2007a, b, 2008); Basaran and Nie
(2007); Bin Jamal et al. (2020); Sherbakov and Sosnovskiy (2010); Sosnovskiy
(1987, 1999, 2004, 2005, 2007, 2009); Sosnovskiy and Sherbakov (2012, 2017,
2019); Temfack and Basaran (2015); Wang and Yao (2019); Yao and Basaran
(2012, 2013a, b, c); Yun and Modarres (2019); Wang and Yao (2017), Guo et al.
(2018), Zhang et al. (2018), Wang et al. (2019), and Osara and Bryant (2019a, b).
Young and Subbarayan (2019a, b) and Suhir (2019) used the Boltzmann-Arrhenius-
Zhurkov equation to predict the evolution of time to failure, while this approach
essentially is based on Basaran and Yan’s (1998) concept. Suhir advocates using the
Boltzmann equation independent of entropy and Newtonian mechanics. Unfortu-
nately, this approach reduces to using Boltzmann distribution just as an evolution
function for an empirical model curve-fit to test data. Hsiao and Liang (2018)
developed a sensor that monitors entropy generation in real-time and that can give
real-time information on system aging and prediction for further estimating the
failure of electrical or any mechanical system.
We do not claim to have included a comprehensive survey of the literature on the
topic; however, we tried to give a brief historical perspective of the efforts.
4.2 Laws of Unified Mechanics Theory
Laws of unified mechanics theory are not new laws of nature. They are the unifica-
tion of the existing laws of Newton and thermodynamics at the ab initio level.
The first law of thermodynamics is also known as the law of conservation of
energy. Newton’s first law is about a state where the summation of all externally
applied forces is zero. Newton’s universal first law of motion can be summarized as
4.2 Laws of Unified Mechanics Theory 151
“An object at rest stays at rest and an object in motion stays in motion with the same
speed and the same direction unless acted upon by an unbalanced force.” Newton’s
first law of motion is also known as the law of inertia, exemplified by the simple
example of pulling the tablecloth swiftly from the table while leaving all cups and
plates in place due to their inertia forces.
The first law is intended for a system with no dissipative, or unbalanced forces
acting on the system. Therefore, we will start numerating laws of unified mechanics
with the second law.
4.2.1 Second Law of Unified Mechanics Theory
The second law of unified mechanics theory is the ab initio level unification of the
second law of Newton and the second law of thermodynamics in Boltzmann’s
formulation.
“The change of momentum of a body is proportional to the impulse impressed on
the body and happens along the straight line on which that impulse is impressed.
Degradation of the input impulse takes place according to the second law of
thermodynamics. The rate of degradation of impulse is directly proportional to the
entropy generation rate in the system along the path chosen. The entropy generation
rate of the system can be mapped onto a linearly independent non-dimensional axis
called the Thermodynamic State Index (TSI) which can have coordinate values only
between 0 and 1 (Fig. 4.4). Thermodynamic state index (1 - Φ) is the normalized
non-dimensional form of the second law of thermodynamics in Boltzmann entropy
formulation.” As a result, unification of the second laws in the Newtonian sense can
be given by
F dt ð1 - ΦÞ = dðmvÞ ð4:35Þ
Generalizing Eq. (4.35) for multiple sources of impulse, [i.e., mechanical, mag-
netic, electric, chemical, etc.] and assuming that the mass is not a function of time,
and there are n number of different impulses acting on the body, including both
conservative and non-conservative, then the second law of the unified mechanics
theory becomes
n Z
X
½F ð1 - ΦÞdt i = mdv ð4:36Þ
i
Combining the second laws of Newton and thermodynamics requires the modi-
fication of the Newtonian space-time coordinate system. A new axis must be added
to be able to define the thermodynamic state. As a result, the motion of any particle in
the universe can be defined only in a five-dimensional space that has five linearly
independent axes. None of these axes can represent the information of other axes.
Fig. 4.4 The coordinate system in unified mechanics theory
The thermodynamic state index is the normalized form of the second law of
thermodynamics as given by the Boltzmann entropy formulation. Derivation of the
(1 - Φ) term is provided in the following section. The additional axes Thermody-
namics State Index (TSI) is necessary to locate the thermodynamic state of a particle.
Space-time coordinates of a point can be defined by Newton’s universal laws of
motion. However, thermodynamic state index axis coordinate cannot be defined by a
space-time coordinate system. We find it necessary to give the following example.
Let us assume there is a 5-year-old boy in Istanbul and a 100-year-old man in
New York. Using the space-time Cartesian coordinate system, we can define their
location by x, y, z coordinates and their age on the time axis. However, this does not
give any information about their thermodynamic state. Let us assume that a 5-year-
old boy has a terminal illness and is expected to die in a few days and 100-year-old is
expected to die in a few days. This information cannot be represented in a space-time
coordinate system. However, on the TSI axis, both persons will be at Φ = 0.999
coordinate.
Another example is, if a soccer ball is given an initial acceleration with a kick, F,
it will have an initial acceleration of a = F/m but eventually, it will come to a stop.
Let’s assume the ball was stationary initially in the terrain profile shown in Fig. 4.6.
Depending on the path it follows, it will come to a stop at one of the valleys. The
amplitude and direction of the initial acceleration of the ball are governed by the
second law of Newton and the slowing down process of the ball is governed by the
laws of thermodynamics, which is represented by (1 - Φ) term.
4.3 Evolution of Thermodynamic State Index (Φ) 153
4.2.2 Third Law of Unified Mechanics Theory
“To every action, there is always opposed an equal reaction. The initial reaction of a
body in response to action will change over time as the stiffness of the reacting
system degrades over time according to the second law of thermodynamics and its
thermodynamic fundamental equation.” The reaction will be governed by the third
law of the unified mechanics theory, given by
h i
dU 21 d 1
F 12 = F 21 = = k 21 ½1 - Φu221 ð4:37Þ
du21 du21 2
dΦ
For problems where du 21
is assumed to be negligible in each increment, then we
can write the following simplified form of the third law:
F 12 = F 21 = ½k21 ð1 - ΦÞ u21 ð4:38Þ
where U21 is the strain energy of the reacting member, Φ is the Thermodynamic
State Index, k21 is the stiffness of the reacting member, and u21 is the displacement in
the reacting member.
It is noteworthy that the third law of the unified mechanics theory includes both
Hooke’s law (1678) and Newton’s third law (1687).
4.3 Evolution of Thermodynamic State Index (Φ)
Materials, under externally applied loading, [i.e., mechanical, thermal, electrical,

chemical, etc.], change their thermodynamic state. This process will follow the laws
of thermodynamics. Simply put the first law of thermodynamics will govern the
conservation of energy and the second law of thermodynamics will govern the
entropy generation rate according to the thermodynamic fundamental equation of
the system. The evolution of entropy [disorder] in the system is given by
Boltzmann’s entropy equation. When the entropy is maximum and the entropy
generation rate is zero for a closed and isolated system, the change in thermody-
namic state index will come to a stop.
While the author of this book and many others have proven the validity of
Boltzmann’s equation in solids in the last 30 + years, extensively, Boltzmann’s
(1877) original work and its interpretation of disorder were misinterpreted until
recently as being only applicable to gasses. Recently, there has been a great deal
of interest in translating Boltzmann’s original papers into English. One recent such
paper is by Sharp and Matschinsky (2015). We find it essential to include the entirety
of their translation in this section. It is important to point out that, Sharp and
Matschinsky (2015) also state that “what Boltzmann wrote on these subjects is
rarely quoted directly, his methods are not fully appreciated, and key concepts
have been misinterpreted.” The following section about Boltzmann’s work is a direct
quotation from Sharp and Matschinsky (2015):
The translation provided here is for Boltzmann (1877). Previous work of Maxwell and
Boltzmann’s derivations was based on mechanical laws of motion and particle interaction of
gasses. However, this work by Boltzmann is much more general. His formulation requires
only that particles can exchange kinetic energy, but they do not specify the mechanism. As
Boltzmann predicted in the final sentences of this paper, his approach applied to not just
gasses, but liquids and solids of any composition. Indeed, the Boltzmann distribution has
also passed almost unchanged into the quantum world.
Boltzmann established the theoretical basis for statistical mechanics. Boltzmann

used three levels of hierarchy to describe the processes. At the highest level, there is
the macro-state, where the thermodynamic state variables such as temperature and
pressure can be observed directly. The second level in the hierarchy is where the
energy or velocity components of each molecule can be specified. He calls this
[energy or momentum of particles] Komplexion, which Sharp and Matschinky
(2015) translate as complexion. Finally, there is the third level at which the number
of molecules with each its energy level, velocity, and their [spatial] position is
specified. Here Boltzmann does not make any assumptions about the type [or the
state] of the molecules (Boltzmann’s w0, w1, etc.). Boltzmann calls a particular set of
w-values a Zustandeverteilung translated as a “state distribution” or “distribution of
states.” It is important to point out, as discussed by Sharp and Matschinsky (2015),
that a microstate is divided into two separate hierarchies in Boltzmann’s formulation,
both complexion and state distribution. “Boltzmann then uses permutation mathe-
matics to determine the distribution of states, which he denotes by p.” “Boltzmann
then shows how to find the distribution of states (w0max ,w1max , . . .) with the largest
number of complexions pmax subject to constraints on the temperature and number of
molecules. Boltzmann’s postulate is that (w0max ,w1max , . . .) is the most probable state
distribution and that this corresponds to thermal equilibrium.” Although Boltzmann
introduces discrete energy levels in Section I of his paper as a convenience and
considers them [discrete energy levels] unphysical, in Section II he shows that there
is no material difference compared to using continuous energy levels.
The statistical mechanics’ formulation of entropy is presented in Section V of
Boltzmann (1877). “Boltzmann shows that the statistical mechanics quantity he
denotes by Ω (multiplied by 2/3) is equal to the thermodynamic quantity entropy
(S) as defined by Clausius, with an additive constant. Boltzmann called, Ω, the
permutabilities translated here literally as ‘permutability measure.’ Boltzmann
defines permutability measure Ω as follows: A state distribution is specified by the
number of molecules having velocity components within some small interval u and
u + du, v and v + dv, w and w + dw for every u, v and w velocity values in the -
1 , + 1 range, and having [spatial] position coordinates in each small volume
(x, x + dx), (y+, y + dy), (z, z + dz) ranging over the total volume V. For each state
distribution, there are several possible complexions. [Combination of molecules
with a certain level of energy or velocity]. One state distribution has the most
complexions and so, therefore, is the most probable. Ω Is given by the [natural]
logarithm of the number of complexions for that state with the most complexions.”
“Boltzmann (1877) also clearly demonstrates that these are two distinct contri-
butions to entropy generation, one arising from the distribution of heat (kinetic
energy of atoms) and the distribution in space of atoms or molecules.” In the initial
effort to understand the nature of entropy, Carnot, Clausius, Maxwell Kelvin, and
others focused almost entirely on the contribution from heat only. “Boltzmann
unified the entropy due to thermal aspects with spatial distribution entropy into
one statistical mechanic’s formulation.” Boltzmann also discovered the third funda-
mental contribution to entropy, namely, radiation by deriving the Stefan-Boltzmann
Law (1884).
“Careful reading of Boltzmann (1877) is enlightening about some apparent
paradoxes subsequently encountered in the development of statistical mechanics.”
Unfortunately, many terms in Boltzmann’s equations and derivation have been
misinterpreted in many physics textbooks. “First, the variable, Ω, is not the loga-
rithm of a probability.” That would be obtained by dividing the number of com-
plexions p for a given state distribution, by the total number of complexions, J.
Boltzmann gives this a different symbol, w, from the first letter of the German word
for probability [wahrscheinlichkeit], but he does not use it. Confusingly, Planck later
chose to write Boltzmann’s equation for entropy as Planck (1901).
S = k ln w þ constant ð4:39Þ
k is Boltzmann’s constant, w is the probability that the system will exist in the state it
is in relative to all the possible states it could be in (Halliday & Resnick, 1966). On
the other hand, in most thermodynamics’ textbooks, Boltzmann’s equation is given
by
S = k ln Ω ð4:40Þ
where k is the Boltzmann’s constant and Ω is the number of microstates

corresponding to a given state with the macroscopic constraints (Callen, 1985; De
Hoff, 1993).
Sharp and Matschinky (2015) point out that Boltzmann’s Permutabilitätmass
(measure of permutability) method for counting possible micro states, there is no
need for a posteriori division by N! “to correct” the derivation using the “somewhat
mystical arguments of Gibbs and Planck.” Ehrenfest and Trakal (1921), van Kampen
(1984), Jaynes (1992), and Swendsen (2006) have pointed out that the correct
counting of micro states ~a la Boltzmann precludes the need for the spurious
indistinguishability term (/N!). “This fact has been ignored in most textbooks.”
Translation by Sharp and Matchinsky (2015) also clarifies one very important
point about Boltzmann’s derivation regarding non-equilibrium states. “lnΩ is not the
logarithm of a volume in momentum-coordinate-phase space occupied by the
system [Boltzmann here refers to a 5-dimensional space, the 5th axis being momen-
tum]. Boltzmann notes, using Liouville’s theorem, that dV [differential volume in
momentum-coordinate phases space] remains constant in time, and so it cannot
describe the entropy increase upon approach to the equilibrium that Boltzmann
was so concerned with. He thus avoids at the outset the considerable difficulty Gibbs
had accounting for changes in entropy with time. Boltzmann gave us, for the first
time, a definition of entropy applicable to every state (distribution) at equilibrium or
not. “Then the entropy of the initial and final states is not defined, but one can still
calculate the quantity which we have called the permutability measure. By exten-
sion, every complexion can then be assigned an entropy, using the permutability
measure of the state distribution to which that complexion belongs, Lebowitz
(1993), opening the door to the statistical mechanics of non-equilibrium states and
irreversible processes” (Sharp & Matchinsky, 2015).
Boltzmann’s work has historically been misinterpreted, assumed to apply to
gasses only, and ignored in the continuum mechanics field. Therefore, we feel
compelled to include his original paper in this book. The following section is the
English translation of Boltzmann (1877) by Sharp and Matschinky (2015) in its
entirety.
4.3.1 On the Relationship Between the Second Fundamental

Theorem of The Mechanical Theory of Heat
and Probability Calculations Regarding the Conditions
For Thermal Equilibrium, by Ludwig Boltzmann (1877)
The relationship between the second fundamental theorem [i.e., the second law of
thermodynamics] and calculations of probability became clear for the first time when
I demonstrated that the theorem’s analytical proof is only possible based on prob-
ability calculations. (I refer to my publication “Analytical proof of the second
fundamental theorem of the mechanical theory of heat derived from the laws of
equilibrium for kinetic energy” Wien. Ber. 63, p 8 reprinted ass Wiss. Abhand. Vol I,
reprint 20, pp 295 and my “Remarks about several problems in the mechanical
theory of heat” 3rd paragraph, Wiss. Abhand. Vol II, reprint 39.) [Boltzmann refers
to thermodynamics by the term “The mechanical theory of heat.”] [Wiss. Abhand
refers to Boltzmann’s collected works.] This relationship is also confirmed by
demonstrating that an exact proof of the fundamental theorem of the equilibrium
of heat is most easily obtained if one demonstrates that a certain quantity—which I
wish to define again as E—has to decrease as a result of the exchange of the kinetic
energy among the gaseous molecules and therefore reaching its minimum value for
the state of the equilibrium of heat. (Compare my “Additional studies about the
equilibrium of heat among gaseous molecules” Wiss. Abhand. Vol I, reprint 22, p
316.) The relationship between the second fundamental theorem and the laws of the
equilibrium of heat is made even more compelling in light of the developments in the
second paragraph of my “Remarks about several problems of the mechanical theory
of heat.” There, I mentioned for the first time the possibility of a unique way of
calculating the equilibrium of heat using the following formulation. It is clear that
every single uniform state distribution which establishes itself after a certain time
given a defined initial state is equally as probable as every single non-uniform state
distribution. Comparable to the situation in-game of Lotto [a board game] where
every single quintet is as improbable as the quintet 12345. The higher probability
that the state distribution becomes uniform with time arises only because there are far
more uniform than non-uniform state distributions. Furthermore: “It is even possible
to calculate the probabilities from the relationships of the number of different state
distributions. This approach would perhaps lead to an interesting method for the
calculation of the equilibrium of heat.” It is thereby indicated that it is possible to
calculate the state of the equilibrium of heat by finding the probability of the different
possible states of the system. The initial state in most cases is bound to be highly
improbable and from it, the system will always rapidly approach a more probable
state until it finally reaches the most probable state, i.e., that of the heat equilibrium.
If we apply this to the second basic theorem [i.e., the second law of thermodynam-
ics], we will be able to identify that quantity which is usually called entropy with the
probability of the particular state. Let’s assume a system of bodies that are in a state
of isolation with no interaction with other bodies, e.g., one body with higher and one
body with lower temperature and one so-called intermediate body which accom-
plishes the heat transfer between the two bodies; or choosing another example by
assuming a vessel with absolutely even and rigid walls one half of which is filled
with the air of low temperature and pressure whereas the other half is filled with the
air of high temperature and pressure. The hypothetical system of particles is assumed
to have a certain state at time zero. Through the interaction between the particles, the
state is changed. According to the second fundamental theorem, this change must
take place in such a way that the total entropy of the particles increases. This means
according to our present interpretation that nothing changes except that the proba-
bility of the overall state for all particles will get larger and larger. The system of
particles always changes from an improbable state to a probable state. It will become
clear later what this means. After the publication of my last treatise regarding this
topic, the same idea was taken up and developed further by Mr. Oskar Emil Meyer
totally independent of me [Die Kinetische Theorie der Gase, Breslau 1877, Seite
262]. He attempts to interpret, in the described manner, the equations of my
continued studies concerning the equilibrium of heat particles. However, the line
of reasoning of Mr. Meyer remained entirely unclear to me, and I will return to my
concerns with his approach on page 172 (Wiss. Abhand. Vol. II).
We have to take a different approach because it is our main purpose not to limit
our discussion to thermal equilibrium but to explore the relationship of this proba-
bilistic formulation to the second theorem of the mechanical theory of heat. We want
first to solve the problem which I referred to above and already defined in my
“Remarks on some problems of the mechanical theory of heat” [Wiss. Abhand.
reprint 39], namely, to calculate the probability of state distributions from the
number of different distributions. We want first to treat as simple a case as possible,
namely, a gas of rigid elastic spherical molecules trapped in a container with
absolutely elastic walls. (Which interact with central forces only within a certain
small distance, but not otherwise, the latter assumption, which includes the former as
a special case, does not change the calculations in the least.) Even in this case, the
application of probability theory is not easy. The number of molecules is not infinite,
in a mathematical sense, yet the number of velocities each molecule is capable of is
effectively infinite. Given this last condition, the calculations are very difficult; to
facilitate understanding, I will, as in earlier work, consider a limiting case.
Kinetic Energy Has Discrete Values
We assume initially that each molecule is only capable of assuming a finite number
of velocities, such as
1 2 3 p
0, , , ,⋯, ð4:41Þ
q q q q
where p and q are arbitrary finite numbers. Upon colliding, two molecules may
exchange velocities, but after the collision, both molecules still have one of the
above velocities, namely, [The fact that Boltzmann allows the exchange of energy
between molecules automatically makes his formulation general, i.e., irrespective of
the state of matter gas, liquid, or solid.]
1 2 p
0, or , or , etc till ð4:42Þ
q q q
This assumption does not correspond to any realistic mechanical model, but it is
easier to handle mathematically, and the actual problem to be solved is re-established
by letting p and q go to infinity.
Even if at first sight, this seems a very abstract way of treating the problem, it
rapidly leads to the desired objective, and when you consider that in nature all
infinities are [nothing] but limiting cases one assumes each molecule can behave in
this fashion only in the limiting case where each molecule can assume more and
more values of the velocity.
To continue, however, we will consider the kinetic energy, rather than the
velocity of the molecules. Each molecule can have only a finite number of values
for its kinetic energy. As a further simplification, we assume that the kinetic energies
of each molecule form an arithmetic progression, such as the following:
0, E, 2E, 3E, . . . , pE ð4:43Þ
We call P the largest possible value of the kinetic energy, pE. Before impact, each
of the two colliding molecules shall have a kinetic energy of
0, or E, or 2E, etc: pE ð4:44Þ
This means that after the collision, each molecule still has one of the above values
of kinetic energy. The number of molecules in the vessel is n. If we know how many
of these n molecules have a kinetic energy of zero, how many have a kinetic energy
of E, and so on, then we know the kinetic energy distribution. If at the beginning there
is some state distribution among the gas molecules, this will in general be changed
by the collisions. The laws governing this change have already been the subject of
my previous investigations. But right way, I note that this is not my intention here,
instead, I want to establish the probability of state distribution, regardless of how it is
created, or more specifically, I want to find all possible combinations of the ( p + 1)
kinetic energy values allowed to each of the n molecules and then establish how
many of these combinations correspond to each state distribution. [The term “state
distribution” is better translated/interpreted as the distribution of states into English.
However, I am keeping the original translation. Because the term refers to what
Boltzmann refers to as complexion, which is the distribution of [discrete] kinetic
energies of molecules]. The latter number [( p + 1)] then determines the likelihood of
the relevant state distribution, as I have already stated in my published “Remarks
about several problems in the mechanical theory of heat” (Wiss. Abhand. Vol II,
reprint 39, p 121).
As a preliminary, we will use a simpler schematic approach to the problem,
instead of the exact case. Suppose we have n molecules. Each of them can have
kinetic energy
0, E, 2E, 3E, . . . , pE ð4:45Þ
and suppose these energies are distributed in all possible ways among the
n molecules, such that the total energy is a constant, e.g., λE = L. Any such
distribution, in which the first molecule may have a kinetic energy of, e.g., 2E, the
second may have 6E, and so on, up to the last molecule, we call a complexion and so
that each individual complexion can be easily enumerated. We write them in
sequence (for convenience we divide through by E), specifying the kinetic energy
of each molecule. We seek the number P of complexions where w0 [number of]
molecules have kinetic energy 0, w1 molecules have kinetic energy E, w2 have kinetic
energy 2E, up to the wp which have kinetic energy pε. We said, earlier, that given
how many molecules have kinetic energy 0, how many have kinetic energy E, etc.,
this distribution among the molecules specifies the number of P of complexions for
that distribution; in other words, it determines the likelihood of that state distribution.
Dividing the number P by the number of all possible complexions, we get the
probability of the state distribution. [Boltzmann never performs this division. It is
done later by Planck (1901).]
Since the distribution of states does not determine kinetic energies exactly, the
goal is to describe the state distribution by writing as many zeros as molecules with
zero kinetic energy (w0), w1 ones for those with kinetic energy E, etc. All these zeros,
ones, etc. are the elements defining the state distribution. It is now immediately clear
that the number P for each state distribution is the same as the number of permuta-
tions of which the elements of the state distribution are capable and that is why the
number P is the desired measure of the permutability of the corresponding distribu-
tion of states. Once we have specified every possible complexion, we have also all
possible state distributions, the latter differing from the former only by immaterial
permutations of molecular labels. All those complexions which contain the same
Table 4.1 Possible permutations for different states

P P P
1. 0000007 7 6. 0000124 210 11. 0001222 140
2. 0000016 42 7. 0000133 105 12. 0011113 105
3. 0000025 42 8. 0000223 105 13. 0011122 210
4. 0000034 42 9. 0001114 140 14. 0111112 42
5. 0000115 105 10. 0001123 420 15. 1111111a 1
a
The state distributions are so arranged that read as a number, the rows are arranged in increasing
order
number of zeros, the same numbers of ones, etc., differing from each other merely by
different arrangements of elements, will result in the same state distribution, the
number of complexions forming the same state distribution, and which we have
denoted by P, must be equal to the number of permutations which the elements of the
state distribution are capable of. To give a simple numerical example, take =7, λ = 7
p = 7 so L = 7E, P = 7E. With seven molecules, there are eight possible values for the
kinetic energy 0, E, 2E, 3E, 4E, 5E, 6E, 7E to distribute in any possible way such that the
total kinetic energy = 7E. There are then 15 possible state distributions. We enu-
merate each of them in the above manner, producing the numbers listed in the second
column of the following table of state distributions (Table 4.1). The numbers in the
first column label the different state distributions.
In the last column, under the heading P is the number of possible permutations of
members for each state. The first state distribution, for example, has six molecules
with zero kinetic energy, and the seventh has kinetic energy 7E. So w0 = 6, w7 = 1,
w2 = w3 = w4 = w5 = w6 = 0 [w1 = 0 is also zero. However, Boltzmann does not
include it]. It is immaterial which molecule has kinetic energy 7E.
So, there are seven possible complexions, which represent this state distribution.
Denoting the sum of all possible complexions, 1716 by J then the probability of the
first state distribution is 7/J, similarly, the probability of the second state distribution
is 42/J, and the most probable state distribution is the tenth, as its elements permit the
greatest number of permutations. Hereon, we call the number of permutations the
relative likelihood of the state distribution; this can be defined in a different way,
which we next illustrate with a specific numerical example since generalization is
straightforward. Suppose we have an urn containing an infinite number of paper
slips. On each slip is one of the numbers 0, 1, 2, 3, 4, 5, 6, 7; each number is on the
same number of slips and has the same probability of being picked. We now draw
the first septet of slips and note the numbers on them. This septet provides a sample
state distribution with a kinetic energy of E times the number written on the first slip
for molecule 1 and so forth. We return the slips to the urn and draw a second septet,
which gives us a second state distribution, etc. After we draw a very large number of
septets, we reject all those for which the total does not equal seven. This still leaves a
large number of septets. Since each number has the same probability of occurrence
and the same elements in a different order from different complexions, each possible
complexion will occur equally often. By ordering the numbers within each septet by
size, we can classify each into one of the 15 cases tabulated above. So, the number of
septets that fall into the class 0000007 relative to the 0000016 class will be 7:42,
similarly for all the other septets. The most likely state distribution is the one, which
produces the most septets, namely, the 10th.
(Boltzmann’s footnote: If we divide the number of septets corresponding to a
particular state by the total number of septets, we obtain the probability of distribu-
tion. Instead of discarding all septets whose total is not 7 we could, after drawing a
slip, remove from the urn all those other slips for which a total of 7 is now
impossible, e.g., on drawing a slip with 6 on it, all other slips except those with
0 or l would be removed. If the first 6 slips all had 0 on them, only slips with 7 on
them would be left in the urn. One more thing should be noted at this point. We
construct all possible complexions. If we denote by w0 the arithmetic mean of all
values of w0 which belong to different complexions and form analogous expressions
w1 , w2 , . . . in the limit these quantities would also form the same state distribution.)
[Translators Sharp and Matschinsky’s note: Boltzmann’s comments on the results
of Mr. Oskar Meyer beginning “Ich will hier einige Worte € uber die von. Hrn. Oskar
Meyer. . .” on p 172 (Wiss. Ab.) and ending with “. . .Bearbeitung des allgemeinen
Problems zur€uckkehren,” on p 175 (Wiss. Ab.) are of historical interest only and are
omitted.]
The first task is to determine the permutation number, previously designated by P,
for any state distribution. Denoting by J the sum of the permutations P for all
possible state distributions, the quotient P/J is the state distribution’s probability,
henceforth denoted by W. [Boltzmann uses W for the initial of the German word,
Wahrscheinlichkeit.] We would first like to calculate the permutations P for the state
distribution characterized by w0 molecules with kinetic energy 0, w1 molecules with
the kinetic energy E, etc. It must be understood that
w0 þ w1 þ w2 þ ⋯ þ wp = n ð4:46Þ
w1 þ 2w2 þ 3w3 þ ⋯ þ pwp = λ ð4:47Þ
Because the total number of molecules is n, and the total kinetic energy is λE = L.
Describing the state distribution as before, a complexion has w0 molecules with zero
energy, w1 [is the number of molecules] with one unit [of energy], and so on. The
permutations, P, arise since of the n elements w0 are mutually identical, similarly,
with w1, w2 etc., elements. The total number of permutations is well known.
n!
P= ð4:48Þ
w0 !w1 ! . . .
The most likely state distribution will be for those w0, w1, . . . values for which
P is a maximum or since the numerator is a constant, for which the denominator is a
minimum. The values w0, w1 must simultaneously satisfy the two constraints (4.46)
and (4.47). Since the denominator of P is a product, it is easiest to determine the
minimum of its logarithm, that is the minimum of
M = ln ½w0 ! þ ln ½w1 ! þ ⋯ ð4:49aÞ
Here ln is the natural logarithm.1

It is natural in our problem that only integer values of w0, w1, . . . are meaningful.
However, to apply differential calculus, we will allow non-integer values and so find
the minimum of the expression
M 1 = ln Γðw0 þ 1Þ þ ln Γðw1 þ 1Þ þ ⋯ ð4:49bÞ
which is identical to (4.49a) for integer values of w0, w1, . . . . We then get the
non-integer values which for constraints (4.46) and (4.47) maximize M1.2 The
solution to the problem will in any case be obtained if for w0, w1, etc., we select
the closest set of integer values. If here and there a deviation of a few integers is
required, the nearest complexion is easily found.
The minimum of M1 is found by adding to both sides of the equation for M1
equation (4.46) multiplied by the constant h and Eq. (4.47) multiplied by the
constant k and setting the partial derivatives with respect to each of the variables
w0, w1, w2, . . . to zero. We thus obtain the following equations:
d ln Γðw0 þ 1Þ
þ h = 0, ð4:50aÞ
dw0
d ln Γðw1 þ 1Þ
þ h þ k = 0, ð4:50bÞ
dw1
d ln Γðw2 þ 1Þ
þ h þ 2k = 0 ð4:50cÞ
dw2
⋮ ⋮ ⋮

d ln Γ wp þ 1
þ h þ pk = 0, ð4:50dÞ
dwp
which leads to
d ln Γðw1 þ 1Þ d ln Γðw0 þ 1Þ
- ð4:50eÞ
dw1 dw0
- ð4:50fÞ
dw2 dw1
1
Translators’ footnote: The ambiguous symbol “l” [used by Boltzmann] for [natural] logarithm in
the original text has been replaced throughout by “ln .”
2
Translators’ footnote: The original text reads as “maximized but should mean minimized”
[because Boltzmann’s objective is to maximize P].
= - ⋯ ð4:50gÞ
dw3 dw2
The exact solution of the problem through evaluation of the gamma function
integral is very difficult; fortunately, the general solution for arbitrary finite values of
p and n does not interest us here, but only the solution for the limiting case of a larger
and larger number of molecules. Then the numbers w0, w1, w2 etc. become larger and
larger, so we introduce the function3
1
ϕðxÞ = ln Γðx þ 1Þ - xð ln x - 1Þ - ln 2π: ð4:51Þ
2
Then we can write the first equation of (4.50) as follows:
dϕðw1 Þ dϕðw0 Þ dϕðw2 Þ dϕðw1 Þ

ln w1 þ - ln w0 - = ln w2 þ - ln w1 - ð4:52Þ
dw1 dw0 dw2 dw1
Similarly, for the other equations of (4.50). It is also well known that
1 1
ϕð xÞ = - ln x þ þ ... ð4:53aÞ
2 12x
pffiffiffiffiffi
This series is not valid for x = 0, but here x! and 2π ðx=eÞx should have the same
value, and ϕ(x) = 0. Therefore, the problem of finding the minimum of w0 ! w1 ! w2 ! . . .
is replaced by the easier problem of finding the minimum of
pffiffiffiffiffi w0 w0 pffiffiffiffiffi w1 w1 pffiffiffiffiffi w2 w2
2π 2π 2π
e e e
Providing w is not zero, even at moderately large values of p and n both problems
have matching solutions. From Eq. (4.53a) it follows
dϕðw0 Þ 1 1
=- - - ... ð4:53bÞ
dw0 2w0 12w20
which for larger and larger values of w0 or lnw0 vanishes, the same also applies to the
other w′s, so Eq. (4.52) can be written as follows:
ln w1 - ln w0 = ln w2 - ln w1 ð4:54aÞ
Or

3
Boltzmann approximates ln x! by x ln x - x þ 12 ln ð2π Þ rather than x þ 12 ln x - x þ 12 ln ð2π Þ
as is now usual. For x ≪ 30,the relative difference is small.
w1 w2
= ð4:54bÞ
w0 w1
Likewise, the equations for the remaining w′s are
w2 w3 w4
= = = ... ð4:54cÞ
w1 w2 w3
One sees immediately that by neglecting the expression (4.53b) the minimum of
the denominator of
pffiffiffiffiffi nn
2π e
pffiffiffiffiffi w w0 pffiffiffiffiffi w
2π e 0
2π we1 1 . . .
is found instead of the minimum of the denominator of (4.48). So, for problems
involving w!, the use of a well-known papproximation
ffiffiffiffiffi (see Schlömilch’s
Comp. S. 438) amounts to the substitution of 2π ðw=eÞw for w!.
If we denote the common value of the quotient (4.54c) by, x we obtain
w1 = w0 x,w2 = w0 x2 , w3 = w0 x3 ,etc: ð4:55Þ
The two Eqs. (4.46) and (4.47) become

w0 1 þ x þ x2 þ . . . þ xp = n ð4:56Þ

w0 x þ 2x2 þ 3x3 þ . . . þ pxp = λ ð4:57Þ
One sees immediately that these equations differ negligibly from Equation
(42) and the preceding ones from my earlier work “Study of the thermal equilibrium
of gas molecules.” [Boltzmann refers to Eq. (42) in his earlier study, not Eq. (42) in
this derivation]
We can use the last equation to write
xpþ1 - 1
w0 =n ð4:58aÞ
x-1

d xpþ1 - 1
w0 x =λ ð4:58bÞ
dx x - 1
Carrying out the differentiation in the last equation
pxpþ1 - ðp þ 1Þ xp þ 1
w0 x =λ ð4:59Þ
ð x - 1Þ 2
Dividing this equation by Eqs. (4.58a, 4.58b) gives

pxpþ2 - ðp þ 1Þ xpþ1 þ x λ
= ð4:60aÞ
ðxpþ1 - 1Þ ðx - 1Þ n
Or
ðpn - λÞxpþ2 - ðpn þ n - λÞxpþ1 þ ðn þ λÞx - λ = 0 ð4:60bÞ
One can see immediately from Descartes’ theorem4 that this equation cannot have
more than three real positive roots, of which two are = 1. Again, it is easy to see that
both roots are not solutions to Eqs. (4.56) and (4.57) and do not solve the problem,
but that they showed up in the final equation merely because of multiplying by the
factor (x - 1)2. [This appears to be a typographic error. The ( p + 2)2 power makes
mathematical sense, Sharp and Matschinsky (2015)]. To be convinced of this, one
needs only to derive the final equation directly by dividing Eqs. (4.56) and (4.57).
Following this division and having removed the variable x from the denominator and
collecting powers of x throughout, we get the equation
ðnp - 1Þxp þ ðnp - n - λÞxp - 1 þ ðnp - 2n - λÞxp - 2 þ . . . ðn - λÞx - λ

= 0, ð4:61Þ
which is an equation of pth degree, and whose roots supply the solution to the
problem. Thus, Eqs. (4.60a, 4.60b) cannot have more positive roots than the solution
requires. Negative or complex roots have no meaning for the solution to the problem.
We note again that the largest allowed kinetic energy P = pE is very large compared
to the mean kinetic energy of a molecule
L λE
= ð4:62Þ
n n
From this, it follows that p is very large compared to λ/n. The polynomial
Eq. (4.61), which shares the same real roots with Eqs. (4.60a, 4.60b), is negative
for x = 0, x = 1; however, it has the value

p λ
nð p þ 1 Þ - , ð4:63Þ
2 n
which is positive and very large, since p is very large compared to n. The only
positive root occurs for x between 0 and 1, and we obtain it from the more convenient
Eq. (4.60b). Since x is a proper fraction, then the pth and ( p + 1)th powers are
smaller and can be neglected, in which case we obtain
4
Cardano’s formula.
λ
x= ð4:64Þ
nþλ
This is the value to which x tends for large p, and one can see the important fact
that for reasonably large values of p the value of x depends almost exclusively on the
ratio λ/n, and varies little with either λ or n providing their ratio is constant. Once one
has found x, it follows from Eq. (4.58b) that
1-x
w0 = n ð4:65Þ
1 - xpþ1
And Eq. (4.55) gives the values of the remaining w′s. It is seen from the quotients
w0 w1 w2
, , ,etc:
n n n
that the probabilities of the various kinetic energy values for larger p are again
dependent almost exclusively on the mean energy of the molecule. For infinitely
large p we obtain the following limiting values:
n2 n2 λ n2 λ 2
w0 = w1 = w2 = etc: ð4:66Þ
nþλ ðn þ λÞ2 ðn þ λÞ3
To establish whether we have a maximum or a minimum, we need to examine the

second variation of Eq. (4.49b). We note that w0, w1, w2 etc. are very large, so we can
use the approximation formula for lnΓ(w + 1)
1 1 1
w ð ln w - 1Þ - ln w - ln ð2π Þ þ þ etc:
2 2 12w
and neglecting terms, which have second or higher powers of w in the denominator,
obtain
ðδw0 Þ2 ðδw1 Þ2
δ2 M = þ þ ... ð4:67Þ
w0 w1
Therefore, we do have a minimum. I also want to remark on the size of the term
previously designated J. One easily finds that J is given by the following binomial
coefficient:
λþn - 1
J= λ
ð4:68Þ
When you neglect terms that diminish with increasing λ or n

λþn - 2 1
1 ð λ þ n - 1Þ
J = pffiffiffiffiffi ð4:69Þ
2π ðn - 1Þn - 2 λλþ2
1 1
Now λE/n is equal to the average kinetic energy μ of a molecule, therefore
λ μ
= ð4:70Þ
n E
So, for large numbers one has
2n - 1
λ 1þ 2λ
n-1
ðλ þ n - 1Þλþn - 2 = = λλþn - 2 = λλþn - 2 en - 1
1 1 1
1þ ð4:71Þ
λ
[Note: double equal is a relational operator used to compare two variable values
whether they are equal or not.] Therefore
1 λ n - 1 en - 1
J = pffiffiffiffiffi 1 , ð4:72Þ
2π ðn - 1Þn - 2
Therefore, neglecting diminishing terms
λ 1 1
ln J = n ln þ n - ln λ þ ln n - 1 - ln ð2π Þ ð4:73Þ
n 2 2
It goes without saying that these formulas are not derived here solely for finite
p and n values, because these are unlikely to be of any practical importance, but
rather to obtain formulas that provide the correct limiting values when p and
n become infinite.
Nevertheless, it may help to demonstrate, with specific examples of only moder-
ately large values of p and n that these formulas are quite accurate, and though
approximate, are of some value even here.
We first consider the earlier example, where n = λ = 7, i.e., the number of
molecules is seven, and the total kinetic energy is 7E, and so the mean kinetic energy
is E. Suppose first, that p = 7, so each molecule can only have 0, E, 2E, 3E, . . ., 7E of
kinetic energy. Then Eq. (4.60b) becomes
6x9 - 7x8 þ 2x - 1 = 0, ð4:74Þ
From which it follows
1 7 8
x= þ x - 3x9 : ð4:75Þ
2 2
Since x is close to 12, we can set x = 12 in the last two very small terms on the right-
hand side, and obtain
1 1 1 1
x= þ ð7 - 3Þ = þ 7 = 0:5078125 ð4:76Þ
2 29 2 2
You could easily substitute this value for x back into the right side of Eq. (4.75)
and obtain a better approximation for x; since we already have an approximate value
for x, a more rapid approach is to apply the ordinary Newton iteration method to
Eq. (4.74) which results in
x = 0:5088742 . . .
From this, one finds per Eq. (4.55)
w0 = 3:4535 w4 = 0:2316
w1 = 1:7574 w5 = 0:1178
w2 = 0:8943 w6 = 0:0599
w3 = 0:4551 w7 = 0:0304
These numbers satisfy the condition that

pffiffiffiffiffi w0 w0 pffiffiffiffiffi w1 w1
2π 2π . . . etc:
e e
is minimized, while the minimized variables w obey the two constraints
w0 þ w1 þ w2 þ w3 þ w4 þ w5 þ w6 þ w7 = 7, ð4:77aÞ
w1 þ 2w2 þ 3w3 þ 4w4 þ 5w5 þ 6w6 þ 7w7 = 7, ð4:77bÞ
which are minimum, incidentally because the first of Eq. (4.77a) coincides with the
minimum of ðw0 Þw0 ðw1 Þw1 . . .. This provides only an approximate solution to our
problem, which asks for so many (w0) zeros, so many (w1) ones, etc. with as many
permutations as the resulting complexion permits, while the w′s simultaneously
satisfies the constraints Eqs. (4.77a, 4.77b). Since p and n here are very small, one
hardly expects any great accuracy, yet you already get the solution to the permuta-
tion problem by taking the nearest integer for each w, except for w3, for which you
have to assign the value of 1 instead of 0.4551. In this manner, it is apparent
w0 = 3,w1 = 2,w2 = w3 = 1,w4 = w5 = w6 = w7 = 0
and we saw in the previous table that the complexion of 0001123 has the most
permutations. We now consider the same special case with n = λ = 7, but set p = 1;
that is, the molecules may have kinetic energies of 0, 1, 2, 3, . . ., 1. We know then
that the values of the variables w will vary little from those of the former case. In fact,
we obtain
1 7 w w1
x= ; w0 = = 3:5, w1 = 0 = 1:75, w2 = = 0:875,etc:
2 2 2 2
We consider a little more complicated example. Take n = 13, λ = 19, but we only
treat the simpler case where p = 1. Then we have
x = 19=32 w4 = 0:6560
w0 = 5:28125 w5 = 0:38950
w1 = 3:13574 w6 = 0:23133
w2 = 1:86815 ⋮
w3 = 1:10493 ⋮
Substituting here for the w′s the nearest integers, we obtain
w0 = 5,w1 = 3,w2 = 2,w3 = 1,w4 = 1,w5 = w6 . . . 0,
Already from the fact that w0 + w1 + . . . should = 13, it is seen that again one of
the w′s must be increased by one unit. From those w′s that is set to = 0, w5 differs
least from the next highest integer. We want therefore w5 = 1, and obtain the
complexion
0000011122345
whose digit sum is in fact = 19. The number of permutations this complexion is
capable of is
13! 13! 1
=
5!3!2! 4!3!2! 5
A complexion whose sum of digits is also = 19, and which one might suppose is
capable of very many permutations, would be the following:
0000111222334
The number of permutations is
13! 13! 1
=
4!3!3!2! 4!3!2! 6
This is less than the number of permutations of the first complexion we found
from the approximate formula. Likewise, we expect that the number of permutations
of the two complexions
0000111122335
And
0000111122344
is smaller still. This is, for both complexions
13! 13! 1
=
4!4!2!2! 4!3!2! 8
Other possible complexions are capable of still fewer permutations, and it would
be quite superfluous to follow these up here. It is seen from the examples given here
that the above formula, even for very small values of p and n gives values of w within
one or two units of the true values. In the mechanical theory of heat, we are always
dealing with extremely large numbers of molecules, so such small differences
disappear, and our approximate formula provides an exact solution to the problem.
We see also that the most likely state distribution is consistent with that known from
gases in thermal equilibrium. According to Eq. (4.66), the probability of having a
kinetic energy sE is given by
s
n2 λ
ws = ð4:78Þ
nþλ nþλ
Since λE/n is equal to the average kinetic [energy] of a molecule μ, which is finite,
so n is very small compared to λ. So, the following approximations
n2 n2 nE λ n E
= 1 - = e-λ = e-μ
n
= = , ð4:79Þ
nþλ λ μ nþλ λ
Hold. From which it follows that
nE - Esμ
ws = e , ð4:80Þ
μ
To achieve a mechanical theory of heat, these formulas must be developed

further, particularly through the introduction of differentials and some additional
considerations.
Kinetic Energies Exchange in a Continuous Manner

To introduce differentials into our formula, we wish to illustrate the problem in the
same manner as indicated on p 171 (Wiss. Abhand. Vol II) because this seems to be
the best way to clarify the matter. Here each molecule was only able to have one of
0, E, 2E, . . ., pE values for kinetic energy. We generated all possible complexions, i.e.,
all the ways of distributing 1 + p values of the kinetic energy among the molecules,
yet subject to the constraints of the problem, using a hypothetical urn containing
infinitely many paper slips. Equal numbers of paper slips have kinetic energy values
0, E, etc. written on them. To generate the first complexion, we draw a slip of paper
for each molecule and note the value of the kinetic energy assigned in this way to
each molecule. Very many complexions are generated in the same way, they are
assigned to this or that state distribution, and then we determine the most probable
one. That state distribution which has the most complexions, we consider as the most
likely or corresponding to thermal equilibrium. Proceeding to the continuous kinetic
energy case the most natural approach is as follows:
Taking E to be some very small value, we assume that in the urn are very many
slips of paper labeled with kinetic energy values between 0 and E. In the urn are also
equal numbers of paper slips labeled with kinetic energy values between E and 2E, 2E
and 3E up to infinity. Since E is very small, we can regard all molecules with kinetic
energy between x and x + E as having the same kinetic energy. The rest of the
calculation proceeds as in Section I above. We assume some complexion has been
drawn; w0 molecules have kinetic energy between 0 and E, w1 molecules have values
between E and 2E, w2 have values between 2E and 3E, etc.
Here, because the variables w0, w1, w2 etc. will be infinitely small, of the order of
magnitude of E, we prefer to write them as
w0 = E f ð0Þ; w1 = E f ðEÞ; w2 = E f ð2EÞ etc: ð4:81Þ
The probability of the state distribution in question is given, exactly as in

Section I, by the number of permutations that the elements of the state distribution
are capable of, e.g., by the number
n!
w0 !w1 !w2 ! . . .
Again, the most likely state distribution, which corresponds to thermal equilib-
rium, is defined by the maximum of this expression, that is, when the denominator is
minimized. We use again the reasonable approximation of Section I, replacing w!
with the expression
pffiffiffiffiffi w w
2π
e
pffiffiffiffiffi
We can omit the term 2π since it is a constant factor in the minimization; the
key again is to replace minimization of the denominator with minimization of its
logarithm; then we obtain the condition for thermal equilibrium that
M = w0 ln w0 þ w1 ln w1 þ w2 ln w2 þ . . . - n ð4:82Þ
is a minimum, while again satisfying the two constraints
n = w0 þ w1 þ w2 þ . . . ð4:83Þ
L = Ew1 þ 2Ew2 þ 3Ew3 þ . . . ð4:84Þ
which are identical to Eqs. (4.46) and (4.47) of Section I. Using Eq. (4.81) here, we
replace the variables w by the function f and obtain thereby
M = E½f ð0Þ ln f ð0Þ þ f ðEÞ ln f ðEÞ þ f ð2EÞ ln f ð2EÞ þ . . .

þ E ln E½f ð0Þ þ f ðEÞ þ f ð2EÞ þ . . . - n ð4:85Þ
and Eqs. (4.83) and (4.84) become
n = E½f ð0Þ þ f ðEÞ þ f ð2EÞ þ . . ., ð4:86Þ

L = E½E f ðEÞ þ 2E f ð2EÞ þ 3E f ð3EÞ . . .: ð4:87Þ
Using Eq. (4.86) the expression for M can also be written as
M = E½f ð0Þ ln f ð0Þ þ f ðEÞ ln f ðEÞ þ f ð2EÞ ln f ð2EÞ þ . . . - n þ n ln E ð4:88Þ
Since n and E are constant, (because E has the same value for all possible
complexions, and is constant between different state distributions), one can
minimize
M 0 = E½f ð 0Þ ln f ð0Þ þ f ðEÞ ln f ðEÞ þ f ð2EÞ ln f ð2EÞ þ . . . ð4:89Þ
instead. As E is made still smaller, the allowed values of kinetic energy approach a
continuum. For vanishingly small E, various sums in Eqs. (4.86, 4.87, 4.89) can be
written in the form of integrals, leading to the following equations:
Z 1
0
M = f ðxÞ ln f ðxÞdx ð4:90Þ
0
Z 1
n= f ðxÞdx, ð4:91Þ
0
Z 1
L= xf ðxÞdx, ð4:92Þ
0
The functional form of f(x) is sought which minimizes expression (4.90) subject
to the constraints (4.91) and (4.92), so one proceeds as follows: To the right side of
Eq. (4.90) one adds Eq. (4.91) multiplied by a constant k, and Eq. (4.92) multiplied
by a constant h. The resulting integral is
Z 1
½f ðxÞ ln f ðxÞ þ kf ðxÞ þ h xf ðxÞdx
0
where x is the independent variable, and f is the function to be varied. This result in
Z 1
½ ln f ðxÞ þ k þ hx δf ðxÞdx
0
Setting the quantity, which has been multiplied by δf(x) in square brackets =
0, and solving for the function f(x), we obtain
f ðxÞ = Ce - hx ð4:93Þ
Here the constant e-k - 1 is denoted by C for brevity. The second variation of M′
Z 1
0 ½δf ðxÞ2
δ M =
2
dx, ð4:94Þ
0 f ðxÞ
is necessarily positive, since f(x) is positive for all values of x lying between 0 and
1. By the calculus of variations M′ is a minimum. From Eq. (4.93), the probability
that the kinetic energy of a molecule lies between x and x + dx at thermal equilibrium
is
f ðxÞdx = Ce - hx dx ð4:95Þ
The probability that the velocity of a molecule lies between ω and ω + dω would
be
hmω2
Ce - 2 mωdω ð4:96Þ
where m is the mass of a molecule. Equation (4.96) gives the correct state distribu-
tion for elastic disks [bold emphasized by this author] moving in two dimensions,
for elastic cylinders with parallel axis moving in space, but not for elastic spheres,
which move in space. For the latter, the exponential function must be multiplied by
ω2dω not ωdω. To get the right state distribution for the latter case, we must set up
the initial distribution of paper slips in our urn in a different way. To this point, we
assumed that the number of paper slips labeled with kinetic energy values between
0 and E is the same as those between E and 2E. As also for slips with kinetic energies
between 2E and 3E, 3E and 4E, etc.
Now, however, let us assume that the three velocity components along the three
coordinate axes, rather than the kinetic energies, are written on the paper slips in the
urn. The idea is the same: There are the same number of slips with u between 0 and E,
v between 0 and ξ, and w between 0 and η. The number of slips with u between E and
2E, v between 0 and ξ, and w between 0 and η is the same. Similarly, the number for
which u is between E and 2E, v is between ξ and 2ξ, w is between 0 and η. Generally,
the number of slips for which u, v, w are between the limits u and u + E, v and v + ξ,
w and w + η are the same. Here u, v, w have any magnitude, while E, ξ, η are
infinitesimal constants. With this one modification of the problem, we end up with
the actual state distribution established in gas molecules.
(LB footnote: We can of course, instead of using finite quantities E, ξ, η and then
taking the limit as they go to zero, write du, dv, dw from the outset, then the
distribution of paper slips in the urn must be such that the number for which u, v,
w are between u and u + du, v and v + dv, w and w + dw are proportional to the
product dudvdw and independent of u, v, and w. The earlier distribution of slips in
the urn is characterized by the fact that although E could be replaced by dx, kinetic
energies between 0 and dx, dx and 2dx, 2dx and 3dx, etc. occurred on the same
number of slips.)
If we now define
wabc = E ζ η f ðaE, bζ, cηÞ ð4:97Þ
as the number of molecules of any complexion for which the velocity components lie
between the limits aE and (a + 1)E, bζ and (b + 1)ζ, cη and (c + 1)η, the number of
permutations, or complexions of these elements for any state distribution, becomes
n!
P = Qa = þp Qb = þq Qc = þr ð4:98Þ
a= -p b= -q c= -r wabc !
where we first assume u adopts only values between -pE and +pE, v between -qζ
and +qζ, w between -rη and +rη. Where again, the most likely state distribution
occurs when this expression, or if you will, its logarithm, is maximum. We again
substitute
pffiffiffiffiffi n n pffiffiffiffiffi w w
n! by 2π and w! by2π
e e
pffiffiffiffiffi
where you can again immediately omit the factors of 2π as they simply contribute
additive constants - 12 ln 2π to ln P ; omitting also the constant n ln n the term, the
requirement for the most probable state distribution is that the sum
= þp bX
aX = þq cX
= þr
- wabc ln wabc
a= -p b= -q c= -r
is a maximum, which only differs from ln P by an additive constant. The constraints

that the number of molecules=n and that the total kinetic energy=L, take the
following form:
aX
=p bX
=q X
c=r
n= wabc ð4:99Þ
a= -p b= -q c= -r
m X X X 2 2
a=p b=q c=r
L= a E þ b2 ζ 2 þ c2 η2 wabc ð4:100Þ
2 a= -p b= -q c= -r
Substituting for wabc using Eq. (4.97), one immediately sees that the triple sums can
in the limit be expressed as definite integrals; omitting an additive constant, the
quantity to be maximized becomes
Zþ1 Zþ1 Zþ1

Ω= - f ðu, v, wÞ ln f ðu, v, wÞ dudvdw, ð4:101Þ
-1 -1 -1
The two constraint equations become
Zþ1 Zþ1 Zþ1

n= f ðu, v, wÞ dudvdw, ð4:102Þ
-1 -1 -1
Zþ1 Zþ1 Zþ1

m
L= u2 þ v2 þ w2 f ðu, v, wÞ dudvdw, ð4:103Þ
2
-1 -1 -1
The variable Ω, which differs from the logarithm of the number of permutations only
by an additive constant, is of special importance for this work, and we call it the
permutability measure [bold emphasized by L. B.] I note, incidentally, that
suppression of the additive constants has the advantage that the total permutability
measure of two bodies is equal to the sum of the permutability measures of
each body.
Thus, it is the maximum of the quantity (4.101), subject to the constraints (4.102)
and (4.103), that is sought. No further explanation of this problem is needed here; it
is a special case of the problem I have already discussed in my treatise “On the
thermal equilibrium of gases on which external forces act”5 in the section which
5
Wien. Ber. (1875) 72:427–457 (Wiss. Abhand. Vol. II, reprint 32).
immediately precedes the appendix. There I provided evidence that this state distri-
bution corresponds to the condition of thermal equilibrium. [Boltzmann’s use of the
term thermal equilibrium is important. In mechanics, this corresponds to the equi-
librium of energy, irrespective of the state the matter is in.] Thus, one is justified in
saying that the most likely state distribution corresponds with the condition of
thermal equilibrium. If an urn is filled with slips of paper labeled in the manner
described earlier, the most likely sampling will correspond to the state distribution
for thermal equilibrium. We should not take this for granted, however, without first
defining what is meant by the most likely state distribution. For example, if the urn
were filled with slips labeled originally, then the statement would be incorrect.
The reasoning needed to arrive at the correct state distribution will not escape
those experienced in working with such problems. The same considerations apply to
the following circumstance: If we group all the molecules whose coordinates at a
particular time lie between the limits
ξ and ξ þ dξ,η and η þ dη,ζ and ζ þ dζ, ð4:104Þ
In addition, whose velocity components lie between the limits
u and u þ du,v and v þ dv,w and w þ dw, ð4:105Þ
and let these molecules collide with other molecules under specific conditions, after
a certain time their coordinates will lie between the limits
Ξ and Ξ þ dΞ,H and H þ dH,Z and Z þ dZ, ð4:106Þ
and their velocity components will lie between the limits
U and U þ dU,V and V þ dV,W and W þ dW, ð4:107Þ
Then at any time
dξ dη dζ du dv dw = dΞ dH dZ dU dV dW: ð4:108Þ
This is a general result. If at time zero the coordinates and velocity components of
arbitrary molecules (material points) lie between the limits (4.104) and (4.105) and
unspecified forces act between these molecules, [the fact that Boltzmann does not
make any assumption about the type of forces between the molecules, therefore,
shear forces between molecules are not excluded]. So that at time t the coordinates
and velocity components lie between the limits (4.106) and (4.107), then Eq. (4.108)
is still satisfied. [Note that here Boltzmann uses a 5-axes coordinate system to locate
the molecules, the 5th axis being the momentum.]
If instead of the velocity components, one uses the kinetic energy x and the
velocity direction defined by the two angles α and β, to describe the action of the
forces, these variables would initially lie between the limits
ξ and ξ þ dξ,η and η þ dη,ζ and ζ þ dζ,

x and x þ dx,α and α þ dα,β and β þ dβ,
And then after the action of the forces lie between the limits
Ξ and Ξ þ dΞ,H and H þ dH,Z and Z þ dZ,

X and X þ dX, A and A þ dA, B and B þ dB,
And so
pffiffiffi pffiffiffiffi
dξ dη dζ x dx φðα, βÞdα dβ = dΞ dH dZ X dX
φðA, BÞdA dB ð4:109Þ
So the product of the differentials du dv dw becomes dU dV dW. Therefore,

the list of slips in the urn must be labeled uniformly with velocity components lying
between u and u + du, v and v + dv, w and w + dw, whatever values u, v, w have.
Given a certain value of the coordinates, the p velocities
ffiffiffi must be pdescribed
ffiffiffiffi by the
corresponding “moments.” On the other hand, x dx goes over to X dX. With the
introduction of kinetic energy, slips must be labeled
pffiffiffi so that you have the same
number with kinetic energy between x and x þ x dx where dx is constant but x is
completely arbitrary. This last sentence is in agreement with my “Remarks on some
problems of the mechanical theory of heat” (Wiss. Abhand. Vol II, reprint 39, p
121), where I demonstrated that this is the only valid way to find the most likely state
distribution corresponding to the actual thermal equilibrium; here we have demon-
strated a posteriori that this leads to the correct state distribution for thermal
equilibrium, that which is the most likely in our sense.
Of course, it is easy to analyze those cases where other conditions exist besides
the principle of conservation of kinetic energy. Suppose, for example, a very large
number of molecules for whom (1) the total kinetic energy is constant; (2) the net
velocity of the center of gravity in the directions of the x-axis; (3) y-axis, and (4) z-
axis are given. The question arises, what is the most probable distribution of the
velocity components among the molecules, using the term in the previous sense? We
then have exactly the same problem, except with four constraints instead of one. The
solution gives us the most probable state distribution
f ðu, v, wÞ = Ce - h½ðu - αÞ þðv - βÞ þðw - γ Þ2

2 2
, ð4:110Þ
where C, h, β, γ are constants. This is in fact the state distribution for gas at thermal
equilibrium at a certain temperature, not at rest, but moving with a constant net
velocity. You can treat similar problems such as the rotation of gas in the same
manner, by adding in the appropriate constraint equations, which I have discussed in
my essay “On the definition and integration of the equations of molecular motion in
gases” (Wiss. Abhand. Vol II, reprint 36).
Some comment regarding the derivation of Eq. (4.101) from Eq. (4.98) is
required here. The formula for x! is
pffiffiffiffiffiffiffiffi x x 1 þ⋯
2πx e12x :
e
The substitution of this into Eq. (4.98) gives

pffiffiffiffiffi nþ1
e12nþ⋯
1
2π n 2
P= þ
ð4:111Þ
þp b = þq c = þr wabc þ2 1
pþqþrþ32 1
ð2π Þ Πaa =
= - p Πb = - q Πc = - r ðwabc Þ e12wabc
1 1
ln P = n þ ln n þ þ ⋯ - ðp þ q þ r þ 1Þ ln 2π -
2 12n
= þp bX
aX = þq cX
= þr ð4:112Þ
1 1
- wabc þ ln wabc þ þ⋯
a= -p b= -q c= -r
2 12wabc
First note, that in determining the magnitude of P , in the limit of a very large
number of molecules, n (and thus also of wabc), other small quantities such as E, ζ, η
can be treated as infinitesimals. So, all terms which have n or wabc in the denominator
can be neglected, and the 12 in the term wabc þ 12. The terms containing wabc scale with
the total mass of the gas, while the related 12 terms refer only to a single molecule. So,
the latter quantities can be neglected as the number of molecules increases. We then
get
= þp bX
aX = þq cX
= þr
ln P = n ln n - ðp þ q þ r þ 1Þ ln 2π - wabc ln wabc : ð4:113Þ
a= -p b= -q c= -r
Substituting [E ζ η f(aE, bζ, cη)], for wabc we obtain
aX
= þp bX
= þq
ln P = n ln n - ðp þ q þ r þ 1Þ ln 2π - n ln ðEζηÞ-
a= -p b= -q
cX
= þr
Eζηf ðaE, bζ, cηÞ ln f ðaE, bζ, cηÞ: ð4:114Þ
c= -r
One sees that aside from the triple sum, the terms on the right-hand side are
constant, and so can be omitted. We also let E, ζ, η decrease while p, q, r increase
infinitely, so the triple sum goes over into a triple integral over limits -1 to +1 and
from ln P we arrive immediately at the expression given by Eq. (4.101) for the
permutability measure Ω. The critical condition is that the number of molecules is
very large; this means that wabc is large compared to 12 ; also that the velocity
components between the limits aE and (a + 1)E, bζ and (b + 1)ζ, cη and (c + 1)η
are identical to those between the limits u and u + du, v and v + dv, w and w + dw.
This may appear strange at first sight, since the number of gas molecules is finite
albeit large, whereas du,dv, dw are mathematical differentials. But, on closer delib-
eration this assumption is self-evident. All applications of differential calculus to the
theory of gases are based on the same assumption, namely, diffusion, internal
friction, [bold emphasized by this author], heat conduction, etc. In each infinitesimal
volume element dxdydz there are still infinitely many gas molecules whose velocity
components lie between the limits u and u + du, v and v + dv, w and w + dw. The
above assumption is nothing more than that very many molecules have velocity
components lying within these limits for every u, v, w.
Consideration of Polyatomic Gas Molecules and External Forces
I will now generalize the formulas obtained so far, by first extending them to
so-called polyatomic gas molecules and then including external forces and thereby
finally beginning to extend the discussion to any solid and liquid. [Bold emphasized
by this author] In order not to consider too many examples, I will in each case deal
with the most important case, where, aside from the equation for kinetic energy,
there is no other constraint.
The first generalization can be applied to our formulas without difficulty. So far,
we assumed each molecule was an elastic sphere or a material point so that its
position in space was entirely defined by three variables (e.g., three orthogonal
coordinates). We know that this is not the case with real gas molecules. We shall
therefore assume that three coordinates are insufficient to completely specify the
position of all parts of a molecule in space; rather r variables will be necessary
p1 , p2 , p3 , . . . , pr ,
the so-called generalized coordinates. Three of them, p1, p2, p3 are the orthogonal
coordinates of the center of mass of the molecule; the others can be either the
coordinates of the individual atoms relative to the center of mass, the angular
direction, or whatever specifies the location of every part of the molecule. We will
also remove the restriction that only one type of gas molecule is present. We assume
instead, that there exists a second type whose every molecule has the generalized
coordinates
p´1 , ṕ2 , ṕ3 , . . . , ṕr0
For the third type, the generalized coordinates are
p001 , p}2 , p}3 , . . . , p}r}

If there are v + 1 types of molecules, the generalized coordinates of the final type
are
ð vÞ ðvÞ ðvÞ ðvÞ

p1 , p 2 , p 3 , . . . , p r ð v Þ :
The first three coordinates are always the orthogonal coordinates of the center of
mass. Of course, the necessary assumption is that many molecules of each type are
present. Let l be the total kinetic energy of the first type of gas; χ is its potential
energy6 (so that χ + l is constant if internal forces only are acting). Furthermore
q 1 , q2 , q3 , . . . , q r
are the momentum coordinates corresponding to p1, p2, . . ., pr. [Bold emphasized
by C.B. Note that, here Boltzmann defines momentum as a new additional axis’
coordinates.] We can think of l in terms of the coordinates p1, p2, . . ., pr and their
derivatives with respect to time
p_ 1 , p_ 2 , p_ 3 , . . . , p_ r ,
and denote the quantities c1 dl

dp_ 1 ,c2 dl
dp_ 2 by q1, q2, . . ., where c1, c2, . . . are arbitrary
constants. I would like to note here that in my essay “Remarks on some problems in
the mechanical theory of heat” Section III (Wiss. Abhand. Vol II, reprint 39) the
indexed variables designated pi referred to coordinate derivatives, while here they
have been designated by q; this mistake would probably not have caused any
misunderstanding.
We denote with the appropriate accents the analogous quantities for other types of
molecules. According to the calculations of Maxwell, Watson, and myself, in a state
of thermal equilibrium, the number of molecules for which the magnitudes of p4;
p5. . .pr, q1, q2, q3, . . ., qr lie between the limits
p4 and p4 þ dp4 ; p5 and p5 þ dp5 etc:qr and qr þ dqr ð4:115Þ
is given by
Ce - hðχþlÞ dp4 dp5 . . . dqr ð4:116Þ
where C and h are constants, independent of p and q. Analogous expressions hold of

course for the other molecular species with the same value of h, but different values
of C. Exactly the same equation as (4.116) is also obtained using the methods of
Sections I and II. Consider all those molecules of the first type, for which the
6
The quantity χ called “Kraftfunktion” or “Ergal” by Boltzmann is translated as potential energy.
variables p4, p5, . . ., qr at some time zero lies between the limits (4.115), after a lapse
of sometime t, the values of the same variables lie between the limits
P4 and P4 þ dP4 ,P5 and P5 þ dP5 ,etc: Qr and Qr þ dQr ð4:117Þ
The general principle already invoked gives the following equation:
dp4 dp5 . . . dqr = dP4 dP5 . . . dQr ð4:118aÞ
There is of course also
dp1 dp2 . . . dp3 = dP1 dP2 . . . dP3 ð4:118bÞ
So that, in fact, for the variables
p4 , p5 , . . . , q r
The product of their differentials does not change during a constant time interval.
Therefore, we must now imagine v + 1 urns. In the first are slips of paper, upon
which are written all possible values of the variables p4, p5, . . ., qr; and the number of
slips that have values within the limits of Eq. (4.115) is such that when divided by
the product dp4, dp5. . .qr it is a constant.
Similarly, for the labeling of the slips with the variables p04 , p05 , . . . , q0r in the
second urn, except that for the latter the constant can have a different value. The
same applies to the other urns. We draw from the first urn for each molecule of the
first type, from the second urn for molecules of the second type, etc. We now
suppose that the values of the variables for each molecule are determined by the
relevant drawings. It is of course entirely chance that determines the state distribu-
tions for the gas molecules, and we must first discard those state distributions, which
do not have the prescribed value for total kinetic energy. It will then be most likely
that the state distribution described by Eq. (4.116) will be drawn, i.e., that one
corresponding to thermal equilibrium. The proof of this is straightforward. So, the
results found in the first two sections can be readily generalized to this case.
We want to generalize the problem further, assuming that the gas is composed of
molecules specified exactly as before. But now so-called external forces are acting,
e.g., those like gravity, which originate outside the gas. [Of course, external forces
are not limited to gravity; they can be friction forces exerted by other molecules.] For
details on the nature of these external forces, and how to treat them, see my treatise
“On the thermal equilibrium of gases on which external forces act.”7 The essence of
the solution to the problem remains the same. Only now, the state distribution will no
longer be the same at all points of the vessel containing the gas; therefore
dp1 dp2 dp3 = dP1 dP2 dP3 will no longer hold. We will now understand the
7
Wien. Ber. (1875) 72:427–457.
generalized coordinates p1, p2, . . ., pr more generally to determine the absolute

position of the molecule in space and the relative position of its constituents.
The notion that p1, p2, p3 are just the orthogonal coordinates of the center of
gravity is dropped. The same is true for the molecules of all the other types of gas.
There is one further point to notice. Previously the only necessary condition was that
throughout the vessel very many molecules of each type were present; now it is
required that even in a small element of space, over which the external forces do not
vary significantly in either size or direction, very many molecules are present
(a condition, incidentally, which must hold for any theoretical treatment of problems
where external forces on gases come into play). This is because our method of
sampling presupposes that the states of many molecules can be considered equiva-
lent, in the sense that the state distribution is not changed when the states of these
molecules are exchanged. The probability of a state distribution is then determined
by the number of complexions of which this state distribution is capable.
This is why, for the case just considered, with v + 1 molecular species present,
v + 1 urns must be constructed.
We assume first that a complexion has been drawn where
w000... = f ð0, 0, 0, . . .Þαβγ . . . ð4:119Þ
molecules whose variables p1, p2, . . ., qr lie between limits 0 and α, 0 and β, 0 and γ,
etc. Furthermore, exactly
w10000... = f ðα, 0, 0, . . .Þαβγ . . . ð4:120Þ
Molecules have the same variables within the limits α and 2α, 0 and β, 0 and γ,
etc., and generally
wabc = f ðaα, bβ, cγ⋯kκÞαβγ⋯κ ð4:121aÞ
molecules with variables p1, p2, . . ., qr between limits
aα and ða þ 1Þα,bβ and ðb þ 1Þβ . . . kκ and ðk þ 1Þκ, ð4:121bÞ
These limits are so close that we can equate all the values in between, then it is as if
the variable p1 could only take the values 0, α, 2α, 3α, etc., variable p2 could take the
values 0, β, 2β, 3β, etc. Let n be the total number of molecules of the first type. We
again distinguish the variables for the other gases by the corresponding accents, so
that
n!n0 !n00 ! . . . nðvÞ !

P= ðvÞ
ð4:122Þ
Πwabc...k !Πw0a0 b0 ...k0 Πw00a00 b00 ...k00 ΠwaðvÞ bðvÞ ...kðvÞ !
is the possible number of permutations of the elements of this complexion, which we

call the permutability. The products are to be read so that the indices a, b, . . ., a′, b′,
. . . etc. run over all possible values, i.e., -1 to +1 for orthogonal coordinates, 0 to
2π for angular coordinates, and so on. Consider first, the case where p1 really can
take only the values 0, α, 2α, 3α, . . ., and similarly with the other variables; then
expression (4.122) is just the number of complexions this state distribution could
have; this number is, according to the assumptions made above, a measure of the
probability of the state distribution. Theffi variables w and n are all very large; we can
pffiffiffiffiffiffi
again, therefore, replace w! with 2π ðw=eÞw . We also denote the sum
n ln n þ n0 ln n0 þ . . . nðvÞ ln nðvÞ
pffiffiffiffiffi
By N, so we can also replace n! by 2π ðn=eÞn and then immediately take the
logarithm
hX X i
ln P = N - C ln 2π - wab... ln wab... þ w0a0 b0 ... ln w0a0 b0 ... þ . . . : ð4:123Þ
The sums are to be understood in the same sense as the products above. 2C is the
number of factorials in the denominator of Eq. (4.122) minus v + 1. Let us now
substitute the expression (4.121) for the variables w into Eq. (4.123) and then take
the limit of infinitesimal α, β, γ. . . .
Omitting unnecessary constants, the magnitude we obtain for the permutability
measure, denoted by Ω, is
Z Z
- . . . f ðp1 , p2 , . . ., qr Þ ln f ðp1 , p2 , . . ., qr Þdp1 dp2 , . . . ,dqr
Ω= Z Z

þ . . . f p01 , p02 , . . ., q0r0 ln f p01 , p02 , . . ., q0r0 dp01 dp02 , . . . ,dq ′ r0 þ . . .
ð4:124Þ
The integration is to extend over all possible values of the variables. I have in my
paper “On the thermal equilibrium of gases on which external forces act” demon-
strated that the expression in the square brackets is at a minimum for a gas in a state
of thermal equilibrium, including, of course, the kinetic energy constraint equation.
On the Conditions for the Maximum of the Power-Exponent Free Product
Determining the State Distribution Function
Before I go into the treatment of the second law, I want to concisely treat a problem
whose importance I believe I have shown in Section I, in the discussion of the work
of Mr. Oskar Emil Meyer on this subject, namely, the problem of finding the
maximum of the product of the probabilities of all possible states. However, I
want to deal with this problem only for mono-atomic gases, and with no other
constraint than the equation for the kinetic energy. We first consider the simplest
case where only a discrete number of kinetic energy values, 0, E, 2E, . . ., pE are
possible, and to start we use kinetic energies, not velocity components, as variables.
We again denote by w0, w1, w2, . . ., wv the number of molecules with kinetic energy
0, E, 2E, . . ., pE.
If we treat the subject in the usual way, the following relationship holds: The
quantity
B = w0 w1 w2 ⋯ wp ð4:125Þ
Or, if you prefer, the quantity
ln B = ln w0 þ ln w1 þ ln w2 þ ⋯ þ ln wp ð4:126Þ
Must be a maximum, with the constraints
n = w0 þ w1 þ w2 þ w2 þ ⋯ þ wp ð4:127Þ
And

L = w1 þ 2w2 þ 3w3 þ ⋯ þ pwp E: ð4:128Þ
If to Eq. (4.126) we add Eq. (4.127) multiplied by h and add Eq. (4.128) multiplied
by k, then set the partial derivatives of the sum with respect to w0, w1, w2. . . equal to
zero, we obtain the equations
1 1 1
þ h = 0, þ h þ k = 0, þ h þ 2k = 0 etc: ð4:129Þ
w0 w1 w2
From which, by elimination of the constants h and k
1 1 1 1 1 1
- = - = - = ... ð4:130Þ
w1 w0 w2 w1 w3 w2
Or
1 1 1 1
= a, = a þ b, = a þ 2b,⋯ = a þ pb: ð4:131Þ
w0 w1 w2 wp
Substituting these values into Eqs. (4.127) and (4.128), the two constants a and b can
be determined:
1 1 1 1
n= þ þ þ⋯þ ð4:132Þ
a a þ b a þ 2b a þ pb
E 2E 3E pE
L= þ þ þ⋯þ ð4:133Þ
a þ b a þ 2b a þ 3b a þ pb
The direct determination of the two unknowns a and b from these equations
would be extremely lengthy. The method of Regula falsi would provide a more rapid
solution for each special case; I have not troubled myself with such calculations but
will give here only a general discussion of how the expected solutions can be easily
obtained, keeping in mind that these methods can only provide an approximation
solution to the problem, since only positive integers are allowed, but fractional
values are not. The first point to note is that the problem ceases to have any meaning
as soon as the product p ( p + 1)/2 is greater than L/E. Because then it necessarily
follows that one of the w′s, and so also the product B, is zero.
Then there is no question of a maximum value for B. For the problem to make any
sense, an excessive value for the kinetic energy cannot be possible. If
pþ1 L
p = ð4:134Þ
2 E
Then all the w′s from w0 onwards must be equal to one for B to be non-zero. A
greater variation in values can occur only if smaller values of p are chosen. Then,
when n is large the above equations provide usable approximations. First, a will be
significantly smaller than b, so w0 is very large, and w1 will be much smaller; w2 will
be close to w1/2, w3 will be close to 2w2/3, etc. In general, the decrease in the variable
w with an increasing index will be fairly insignificant when the maximum of
w0 w1 w2. . . is sought, rather than the maximum of ww0 0 ww1 1 ww2 2 . . . . Given
much smaller p values, the value of a is not much less than b, so w0 is also not
that much larger than the other w′s; then w2 is greater than w1/2, w3 is greater than
(2/3)w2, etc. The decrease of w with increasing index is even less. Decreasing p still
further, a will dominate, and there will be hardly any decrease in w with increasing
index. Finally, b becomes negative, and the size of w will even increase with
increasing index. The following cases provide examples, for each of the integer
values of the w′s which maximize B are given.
n = 30,L = 30E,p = 5,w0 = 17,w1 = 5,w2 = 3,w3 = 2,w4 = 2,w5 = 1:

n = 31,L = 26E,p = 4,w0 = 18,w1 = 6,w2 = 3,w3 = 2,w4 = 2:
n = 40,L = 40E,p = 5,w0 = 23,w1 = 7,w2 = 3,w3 = 3,w4 = 2,w5 = 2:
n = 40,L = 40E,p = 6,w0 = 24,w1 = 6,w2 = 3,w3 = 3,w4 = 2,w5 = 1,w6 = 1:
n = 18,L = 45E,p = 5,w0 = 3,w1 = 3,w2 = 3,w3 = 3,w4 = 3,w5 = 3:
n = 23,L = 86E,p = 5,w0 = 1,w1 = 2,w2 = 2,w3 = 3,w4 = 4,w5 = 11:
ð4:135Þ
Let us now turn to the case where the value of the kinetic energy is continuous;
first, consider the kinetic energy x as the independent variable [bold emphasized by
this author], so the problem, in our view is the following: The expression
ZP
Q= ln f ðxÞdx ð4:136Þ
0
Becomes a maximum, while at the same time
ZP ZP
n= f ðxÞdx and L= xf ðxÞdx ð4:137Þ
0 0
are constant. P is also constant. I have purposefully set the upper integration limit to
P, not 1. It is then still straightforward to allow P to increase more and more.
Proceeding accordingly, we obtain:
ZP ZP
1
δ ½ ln x þ h f ðxÞ þ kx f ðxÞ dx = þ h þ kx dx δf = 0 ð4:138Þ
f
0 0
1 1
f=- = ð4:139Þ
h þ kx a þ bx
If we set h = - a; k = - b. To determine these two constants, we use the

equations
ZP
1 a þ bP
n= f ðxÞ dx = ln , ð4:140Þ
b a
0
ZP
P a a þ bP
L= xf ðxÞ dx = - ln , ð4:141Þ
b b2 a
0
which, writing a/b as α, leads to
P P
L þ αn = , bn = ln 1 þ ð4:142Þ
b α
And also
P
ðL þ αnÞ ln 1 þ = Pn: ð4:143Þ
α
From this transcendental equation, α must be determined, from which it is easy to

obtain a and b. Since Pn is the kinetic energy that the gas would have if every
molecule in it had the maximum possible kinetic energy P, we see immediately that
Pn is infinitely greater than L. L/n is the average kinetic energy of a molecule. It is
then easy to verify that P/α cannot be finite because then Pn/αn would be finite and in
the expression (L + αn), L could be neglected. But then in Eq. (4.143), only P/α terms
would remain, and only vanishingly small values of this term could satisfy the
equation, which is inconsistent with the original assumption. Nor can P/α be
vanishingly small because then L would again be vanishingly small compared to
αn. Furthermore
P
ln 1 þ ð4:144Þ
α
Could be expanded in powers of P/α, and the Eq. (4.143) would yield a finite
value for P/α. There remains only the possibility that P/α is very large. Since
αn Pn
ln
Pn αn
vanishes, Eq. (4.143) gives
a
= pe - L ,
nP
α= ð4:145Þ
b
from which follows
P p2 nP
b= , a = e- L ð4:146Þ
L L
By the approach used in this section, using the mean kinetic energy of a molecule,
these equations show that in the limit of increasing p, W, L the probability of
dispersion in kinetic energy remains indeterminate. We now want to consider a
second, more realistic problem. We take the three velocity components u, v,
w parallel to the three coordinate axes as the independent variables [bold empha-
sized by this author], and find the maximum of the expression
Zþ1 Zþ1 Zþ1

Q= ln f ðu, v, wÞdudvdw, ð4:147Þ
-1 -1 -1
While simultaneously the two expressions
Zþ1 Zþ1 Zþ1

n= f ðu, v, wÞdudvdw, ð4:148Þ
-1 -1 -1
Zþ1 Zþ1 Zþ1

L= u2 þ v2 þ w2 f ðu, v, wÞ dudvdw ð4:149Þ
-1 -1 -1
remain constant, integrating over u, v, w. If the velocity magnitude is

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ω= u2 þ v2 þ w2 ð4:150Þ
In addition, its direction is given by the two angles θ and φ (length and breadth8);
we have as is well known
dudvdw = ωdω sin θdθdφ, ð4:151Þ
Hence
ZP
Q = 4π ln f ðϱÞϱ2 dϱ ð4:152Þ
0
8
Translators note: Altitude and azimuth?
ZP
n = 4π ϱ2 f ðϱÞ dϱ ð4:153Þ
0
ZP
L = 4π ϱ4 f ðϱÞ dϱ ð4:154Þ
0
If there are no external forces [like friction or body forces], then clearly f(u, v, w) is
independent of the direction of the velocity. Instead of integrating to infinity, we
intentionally integrate to a finite value of P. Evaluating f(ϱ) just as we did for f(x)
earlier, we obtain
1 1
f ð ϱÞ = - = ð4:155Þ
h þ kϱ2 a2 þ b2 ϱ2
where we set -h = a2 and -k = b2. The two constants a and b are to be determined
from Eqs. (4.153) and (4.154) which become, given the value of f(ϱ),
ZP
ϱ2 dϱ P a bP
n = 4π = 4π - arctg ð4:156Þ
a 2 þ b2 ϱ2 b 2 b3 a
0
ZP
ϱ4 dϱ 4πP3 a2
L = 4π = - 2n ð4:157Þ
a2 þ b ϱ2
2
3b2 b
0
From the last equation, we get

4π 3 a2
2
= 3 Lþ 2n : ð4:158Þ
b P b
Substituting this equation into the first equation of (4.157) gives

L a2 a bP
n=3 2 þ 2 2 n 1- arctg : ð4:159aÞ
P b P Pb a
If however b2 is negative, we put -b2 instead of b2 and obtain
1
f ð ϱÞ = ,
a2 - b2 ϱ2

P a a þ bP
n = 4π - 2 þ 3 ln
b 2b a - Pb
4πP3 a2
L= -
þ 2n
3b2 b

L 2
an a a þ bP
n=3 2 - 2 2 1- ln ð4:159bÞ
P bP 2bP a - Pb
From Eqs. (4.159a) and (4.159b), one first has to calculate the ratio a/b; and by the
same means by which Eq. (4.143) was analyzed, we first determine whether bP/a is
infinitely small, finite, or infinitely large, the only difference being in Eq. (4.159a)
every infinitesimal variation of L/nP2 occurs. However, I will not discuss the point
further, except to note that as n and P grow larger; one also cannot get a result, which
depends only on the average kinetic energy. Also, I will not discuss in detail those
cases where there are other constraint equations besides the equation for kinetic
energy, as this would lead me too far afield. [C.B. Having other constraints would
not lead to any more permutations, but actually less number of permutations.]
To provide a demonstration of how general the concept of the most probable state
distribution of gas molecules is, here I supply another definition for it. Suppose again
that each molecule can only have a discrete number of values for the kinetic energy,
0, E, 2E, 3E, . . ., 1. The total kinetic energy is L = λE. We want to determine the
kinetic energy of each molecule in the following manner: We have in an urn just as
many identical balls (n) as molecules present. Every ball corresponds to a certain
molecule. We now make λ draws from this urn, returning the ball to the urn each
time. The kinetic energy of the first molecule is now equal to the product of E and the
number of times the ball corresponding to this molecule is drawn. The kinetic
energies of all other molecules are determined analogously. We have produced a
distribution of the kinetic energy L among the molecules (a complexion). We again
make λ draws from the urn and produce a second complexion, then a third, etc. many
times (J ), and produce J complexions. We can define the most probable state
distribution in two ways: First, we find how often in all J complexions a molecule
has kinetic energy 0, how often the kinetic energy is E, 2E, etc., and say that the ratios
of these numbers should provide the probabilities that a molecule has kinetic energy
0, E, 2E, etc. at the thermal equilibrium. Second, for each complexion, we form the
corresponding state distribution. If some state distribution is composed of P com-
plexions, we then denote the quotient P =J as the probability of the state distribution.
At first glance, this definition of a state distribution seems very plausible. However,
we shall presently see that this should not be used, because, under these conditions,
the distribution whose probability is the greatest would not correspond to thermal
equilibrium. [C. B.: Boltzmann does not justify or explain this statement. However,
later Planck does show that the highest probability is the equilibrium and uses P/J.] It
is easy to cast the hypothesis that concerns us into formulas. First of all, we want to
discuss the first method of probability determination. We consider the first molecule,
and assume that λ draws were made; the probability that the first molecule was
picked in the first draw is 1/n; however, the probability that another ball was drawn is
(n - 1)/n. Thus the probability that on the 1st, 2nd, 3rd, . . ., kth draws the molecule
corresponding to the first ball has been picked, and then a different ball for each of
the following is given by
λ-k λ
1 k
n-1 n-1 1 k
= ð4:160Þ
n n n n-1
Likewise, the probability that the ball corresponding to the first molecule is
picked on the 1st, 2nd, 3rd, . . ., (k - 1)th, and then (k + 1)th draws, etc. The
probability that the ball corresponding to the first molecule is picked for any arbitrary
k draws and not for the others is
λ
λ! n-1 1 k
wk = : ð4:161Þ
ðλ - k Þ!k! n n-1
This probability that a molecule has then kinetic energy kE is exactly the same for
all the other molecules. Using again the approximation formula for the factorial, we
obtain
rffiffiffiffiffi k
λ
1 n-1 1 λ-k
wk = λ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi ðλ - k Þ - λ , ð4:162Þ
1 - kλ k ðn - 1Þk
2π n
This shows that the probability of the larger kinetic energies is so disproportion-
ately important that the entire expression does not approach a clearly identifiable
limit with increasing k, λ, 1/E, and n. We will now proceed to the second possible
definition of the most probable state distribution. We need to consider all
J complexions that we have formed by J drawings of λ balls from our urn. One of
the various possible complexions consists of λ drawings of the ball corresponding to
the first ball. We want to express this complexion symbolically by mλ1 m02
m03 . . . m0n . A second complexion, with λ - 1 draws of the ball corresponding to
the first molecule, and one draw of the ball corresponding to the second molecule we
want to express as
mλ1 - 1 m12 m03 ⋯ m0n
We see that the different possible complexions are expressed exactly by various
components; the sum of these appears as the power series
ðm1 þ m2 þ m3 þ ⋯ þ mn Þλ ð4:163Þ
That is developed according to the polynomial theorem. The probability of each

such complexion is thus exactly proportional to the coefficient of the corresponding
power series term when you first form the product
0 ð λÞ ðλÞ
m1 þ m02 þ ⋯ þ m0n m001 þ m002 þ ⋯ þ m00n ⋯ m1 þ m2 þ ⋯ þ mðnλÞ
Finally omit from this product the upper indexes, which then generate a term
exactly proportional to the polynomial coefficient. Then the symbols m01 m003
000
m7 . . . we understand that the first pick corresponded to the first molecule, the
second pick corresponded to the third molecule, on the third pick the ball
corresponding to the seventh molecule was picked out, etc. All possible products
of the variables m01 ,m001 ,m02 etc. represent equi-probable complexions. We want to
know how often among all the terms of the power series (4.163) (whose total number
is nλ); there occur terms whose coefficients contain any one state distribution. For
example, consider the state distribution where one molecule has all the kinetic
energy, and all others have zero kinetic energy. This state distribution appears to
correspond to the following members of the power series (4.163)
mλ1 m02 m03 , . . . ,m01 mλ2 m03 , . . . ,m01 m02 mλ3 . . . etc:
with “undivided” λ. Similarly, for the state distribution in which w0 molecules have
kinetic energy zero, w1 molecules have kinetic energy E, w2 molecules have kinetic
energy 2E, etc., there are
λ!
w0 !w1 !w2 ! . . . wλ !
members of the power series (4.163). Each of these elements has the same polyno-
mial coefficient, and that is identical to
λ!
ð0!Þw0 ! ð1!Þw1 ! ð2!Þw2 ! . . . ðλ!Þwλ !
In summary, therefore, according to the now accepted definition, the probability

of this state distribution is
ðλ!Þ2 1 1

nλ w0 !w1 !w2 ! . . . wλ ! ð0!Þw0 ð1!Þw1 ð2!Þw2 . . . ðλ!Þwλ !
However, the maximization of this quantity also does not lead to the state
distribution corresponding to thermal equilibrium.
Relationship of the Entropy to that Quantity which I Have Called

the Probability Distribution
When considering this relationship, let us initially deal with the simplest and clearest
case, by first investigating a monoatomic gas on which no external forces act. In this
case, formula (4.101) of Section II applies. To give it full generality, however, this
formula must also include the x, y, z coordinates of the position of the molecule.
[Note by C.B., LB considers the new axis linearly independent of space-time.] The
maximum of such a generalized expression (4.101) then yields, not only the distri-
bution of the velocity components of the gas molecules, which was sufficient for the
case considered there, but also the distribution of the whole mass of gas in an
enclosing vessel, where there it was taken for granted that the gas mass fills the
vessel uniformly.
The generalization of (4.101) for the permutability measure can be easily
obtained from Eq. (4.124) by substituting x, y, z, u, v, w for p1, p2, . . ., qr and simply
omitting the terms with accented variables. It reads as follows:
Ω=
Z Z Z Z Z Z
- f ðx, y, z, u, v, wÞ ln f ðx, y, z, u, v, wÞdxdydzdudvdw,
ð4:164Þ
where f(x, y, z, u, v, w)dxdydzdudvdw is the number of gas molecules present for

which the six variables x, y, z, u, v, w lie between the limits x and x + dx, y and
y + dy, z and z + dz. . . etc. [u and u + du, v and v + dv] w and w + dw and the
integration limits for the velocity are between -1 and +1, and for the coordinates
over the dimensions of the vessel in which the gas exists. If the gas was not
previously in thermal equilibrium, this quantity must grow. We want to compute
the value this quantity has when the gas has reached the state of thermal equilibrium.
Let V be the total volume of the gas, T be the average kinetic energy of a gas
molecule, and N be the total number of molecules of the gas; finally, m is the mass of
a gas molecule. There is then for the state of thermal equilibrium
N - 3mðu2 þv2 þw2 Þ

f ðx, y, z, u, v, wÞ = 4πT 3=2 e 4T : ð4:165Þ
V 3m
Substituting this value into Eq. (4.164) gives

3
3N 4πT 2
Ω= þ N ln V - N ln N ð4:166Þ
2 3m
If by dQ we denote the differential heat supplied to the gas, where

dQ = NdT þ pdV ð4:167Þ
And
2N
pV = T: ð4:168Þ
3
p is the pressure per unit area, the entropy of the gas is then
Z
dQ 2 3
= N ln V T 2 þ C: ð4:169Þ
T 3
Since N is regarded as a constant, with a suitable choice of constant

Z
dQ 2
= Ω ð4:170Þ
T 3
It follows that for each so-called reversible change of state, wherein in the infinites-
imal limit the gas stays in equilibrium throughout the change
R of state, the increase in
the Permutability measure Ω multiplied by 23 is equal to dQ/T taken over the state
change, i.e., it is equal to the entropy increment. Whereas in fact when a very small
amount of heat dQ is supplied to a gas, its temperature and volume increase by dT
and dV. Then it follows from Eqs. (4.167) and (4.168)
2N
dQ = NdT þ TdV ð4:171Þ
3V
while from Eq. (4.166) it is found that
dV 3N dT
dΩ = þ N þ : ð4:172Þ
V 2 T
It is known that if in a system of bodies many reversible changes are taking place,
then the total sum of the entropy of all these bodies remains constant. On the other
hand, if among these processes are ones that are not reversible, then the total entropy
R
of all the bodies must necessarily increase, as is well known from the fact that dQ/T
integrated over a non-reversible cyclic process is negative. According to Eq. (4.170),
the sum of the permutability measures of the bodies ΣΩ and the total permutability
measure must have the same increase. Therefore, at thermal equilibrium the magni-
tude of the permutability measure times a constant is identical to the entropy, within
an additive constant; but it also retains meaning during a non-reversible process,
continually increasing.
We can establish two principles: The first refers to a collection of bodies in which
various state changes have occurred, at least some of which are irreversible, e.g.,
where some of the bodies were not always in thermal equilibrium. If the system was
in a state of thermal equilibrium before and after all these changes, then the sum of
the entropies of all the bodies can be calculated before and after those state changes
without further ado, and it is always equal to 2/3 times the permutability measure of
all the bodies. The first principle is that the total entropy after the state changes is
always greater than before it. The same is of course true of the permutability
measure. The second principle refers to a gas that changes state without requiring
that it begin and end in thermal equilibrium. Then the entropy of the initial and final
states is not defined, but one can still calculate the quantity, which we have called the
permutability measure; and that is to say that its value is necessarily larger after the
state change than before. We shall presently see that the latter proposition can be
applied without difficulty to a system of several gases, and it can be extended as well
to polyatomic gas molecules, and when external forces are acting. For a system of
several gases, the sum of the individual gas permutability measures must be defined
to be the permutability measure of the system; if one introduces on the other hand the
number of permutations itself, then the number of permutations of the system would
be the product of the number of permutations of the constituents. If we assume the
latter principle applies to anybody, then the two propositions just discussed are
special cases of a single general theorem, which reads as follows:
We consider any system of bodies that undergoes some state changes, without requiring the
initial and final states to be in thermal equilibrium. Then the total permutability measure for
the bodies continually increases during the state changes and can remain constant only so
long as all the bodies during the change of state remain infinitely close to thermal equilib-
rium (reversible state changes).
[C.B.: it is important to point out that here Boltzmann drops the term “gas,” in his theorem,
because there is nothing in his formulation that restricts it to gasses. The liquid and solid state
of the matter will introduce additional constraints but not less than gasses. As a result, his
permutability measure is an upper bound for all possibilities.]
To give an example, we consider a vessel divided into two halves by a very thin
partition. The remaining walls of the vessel should also be very thin so that the heat
they absorb can be neglected, and surrounded by a substantial mass of other gas.
One-half of the vessel should be completely filled with gas, while the other is
initially completely empty. Suddenly pulling away from the divider, which requires
no significant work, causes that gas to spread at once throughout the vessel.
Calculating the permutability measure for the gas, we find that this increases during
this process, without changes in any other body. Now the gas is compressed in a
reversible manner to its old volume by a piston. To achieve this manipulation, we
can if we want assume that the piston is created by a surrounding dense gas enclosed
in infinitely thin walls. This gas is unchanged except that it moves down in space.
Since the permutability measure does not depend on the absolute position in
space, the permutability measure of the gas driving the piston does not change. That
of the gas inside the vessel decreases to the initial value since this R gas has gone
through a cyclic process. However, since this was not reversible, dQ/T integrated
over this cycle is not equal to the difference between the initial and final values of the
entropy, but it is smaller due to the uncompensated transformation in the expansion.
In contrast, heat is transferred to the surrounding gas. So, for this process, the
permutability measure of the surrounding gas is increased by as much as that of
the enclosed gas in the vessel during the first process. Since the latter mass of gas
went through a cyclic process, its entropy decreasedR by as much during the second
process as it increased during the first, but not by dQ/T; and because the second
process was reversible, the entropy of the surrounding gas increased as much as the
enclosed gas’s decreased. The result is, as it has to be, that the sum of the
permutability measures of all bodies of gas has increased. For a gas that moves at
a constant speed in the direction of the x-axis9
4T ððu - αÞ þv þw Þ :
N - 3m 2 2 2
f ðx, y, z, u, v, wÞ = V qffiffiffiffiffiffiffiffiffiffiffiffi
e ð4:173Þ
4πT 3
3m
If we substitute this expression into Eq. (4.164), we get exactly Eq. (4.166) again.
Thus, the translational movement of a mass of gas does not increase its permutability
measure. And, the same is true for the kinetic energy arising from any other net mass
movement (molar movement), because it arises from the progression of the individ-
ual volume elements and their deformations and rotations which are of a higher
order—infinitely small—and therefore entirely negligible. Here we obviously ignore
the changes in permutability measure due to internal friction or temperature changes
connected with those molecular motions. The temperature T of the moving gas is
understood to mean half of the average value of m[(u - α)2 + v2 + w2]. So, if
frictional [bold emphasis by CB] and temperature changes are not present (e.g., if a
gas, together with its enclosing vessel falls freely) a net mass movement does not
affect the permutability measure until its kinetic energy is converted into heat, which
is why molar motion is known as the heat of infinite temperature.
Let us now move on to a mono-atomic gas on which gravity acts. The
permutability measure is represented by the same equation (4.124) but instead of
the generalized coordinates, we again introduce x, y, z, u, v, w. Equation (4.124) thus
gives us a value for Ω which is exactly the same as Eq. (4.164). In the case of thermal
equilibrium, one has
2

f ðx, y, z, u, v, wÞ = Ce - 2T
3
gzþmw2
ð4:174Þ
where ω2 = u2 + v2 + w2. The constant C is determined by the density of the gas. One
has, e.g., a prismatic-shaped vessel of height h, with a flat, horizontal bottom surface
with area =q. Further, let N be the total number of gas molecules in the vessel, and
z denote the height of a gas molecule from the bottom of the vessel, then
9
Translator note: Here V should appear in the denominator, cf the equation after (4.165).
N N
C= Z h = 3 ð4:175Þ
4πT 32 3gz 4πT 2 2T 3gh
q 3g 1 - e - 2T
3m q e - 2T dz 3m
0
3N 4πT 3=2 2T 3gh

Ω= þ N ln þ N ln q þ N ln 1 - e - 2T
2 3m 3g
0 1
3gh
3ghe - 2T A
þ N @1 - 3gh
- N ln N ð4:176Þ
2T 1 - e 2T
It is immediately apparent from the last formula that when a mass of gas falls a bit
lower, without otherwise changing, that does not change its Ω value a bit. (Of course,
gravity acts as a constant downward force, and its increase with the approach to the
earth’s center is neglected, which is always the case in heat theory problems.)
Let us go now to the most general case of an arbitrary gas on which any external
force acts, so we again apply Eq. (4.124). However, so that the formulas are not too
lengthy, let only one type of gas be present in the vessel. The permutability measure
of a gas mixture can then be found without difficulty since it is simply equal to the
sum of the permutability measures each component would have if present in the
vessel alone. For thermal equilibrium, then
- hðχþLÞ
Ne
f=Z Z do dw, ð4:177Þ
e - hðχþLÞ
where χ is the potential energy, L is the kinetic energy of a molecule, and N is the
number of molecules in the vessel, where
do = dp1 dp2 . . . dpr , dw = dq1 dq2 . . . dqr ð4:178Þ
It therefore follows
Z Z Z Z
Ω= - f ln fdo dw = - N ln N þ N ln e - hðχþLÞ do dw þ hN χ
rN
þ ð4:179Þ
2
In the penultimate term, χ is the average potential energy of a molecule, whose

magnitude is
Z Z
Z Z χe - hðχþLÞ dodw
1
χf dodw = Z Z ð4:180Þ
N
e - hðχþLÞ dodw
The last term can be found by taking into account that

Z Z
Le - hðχþLÞ dodw
r
L= Z Z = ð4:181Þ
2h
e - hðχþLÞ dodw
(In respect of the above, see the already cited book of Watson, pp 36 and 37). The
second term on the right of Eq. (4.179) can be further transformed if one introduces
instead of q1, q2, . . ., qr the variables s1, s2, . . ., sr, which have the property that the
term L is reduced to s21 þ s22 þ ⋯ þ s2r . We then denote by Δ the Jacobian
X dq1 dq2 dqr

± ⋯
ds1 ds2 dsr
So then
Z
Z Z r Z Δχe - hχ do
- hðχþLÞ π - hχ
do, χ = Z
2
e do dw = Δe ð4:182Þ
h
Δe - hχ do
And so also
Z
Nr rN
Ω = N ln Δe - hχ do - ln h þ hNχ þ ð1 þ ln π Þ - N ln N ð4:183Þ
2 2
To this expression, one can compare Equation (18) of my paper “Analytical proof of
the second law of thermodynamics from the approach to equilibrium in kinetic
energy” (Wiss. Abhand. Vol I, reprint 20), or Equation (95) of my “Further
Studies” (Wiss. Abhand. Vol I, reprint 22) by replacing p1, p2, . . . with x1, y2, . . .,
q1, q2, . . . with u1, v2, . . .; s1, s2, . . . with
rffiffiffiffi rffiffiffiffi
m m
u1 , v ...,
2 2 1
r
with 3r, whereby

3r
2 2
Δ= ð4:184Þ
m
Equation (4.183) is identical to 3N/2 times Equation (18) of the aforementioned

paper, except for an additive constant, wherein multiplying by N indicates that
Equation (18) applies to one molecule only. Equation (95) of “Further Studies” is
in an opposite manner denoted as Ω and thus also as the entropy. Taken with a
negative sign, it is however greater than Ω here by N ln N. The former is because in
the “Further Studies” I was looking for a value, which must decrease, because of
this, introducing the magnitude of f instead of using f; this was however less clear.
From this agreement, it follows that our statement about the relationship of entropy
to the permutability measure applies to the general case exactly as it does to a
mono-atomic gas [bold emphasis by this author].
Up to this point, these propositions may be demonstrated exactly using the theory
of gases. If one tries, however, to generalize to liquid drops and solid bodies, one
must dispense with an exact treatment from the outset, since far too little is known
about the nature of the latter states of matter, and the mathematical theory is barely
developed. But, I have already mentioned reasons in previous papers, in virtue of
which it is likely that for these two aggregate states, the thermal equilibrium is
achieved when Eq. (4.124) becomes a maximum, and that when thermal equilibrium
exists, the entropy is given by the same expression. It can therefore be described as
likely that the validity of the principle which I have developed is not just limited
to gases, but that the same constitutes a general natural law applicable to solid
bodies and liquid droplets, although the exact mathematical treatment of these
cases still seems to encounter extraordinary difficulties [bold emphasis by this
author].
4.3.2 Critic of Boltzmann’s Mathematical Derivation
Boltzmann assumes that particles are independent rather than an ensemble like in a
molecular chain or crystal. He also assumes that there is no group interdependent
interaction. However, when his formulation is used in conjunction with conservation
of energy, conservation of mass, and Newton’s laws, it automatically makes a solid
basis for statistical mechanics for all phases of matter. Because even if the molecules
he used in his virtual experiments were no single gas atoms and were interacting
chains, the number of microstates could not change for the ensemble. Essentially gas
atoms can be replaced by an ensemble of molecules in his formulation and his
statistical basis would not change because he does not make any assumptions to
preclude forces between molecules or external forces. However, internal friction and
inter-dependent group interaction between an ensemble of atoms would make his
formulation an upper bound for solids. Because of internal inter-dependent interac-

tions and frictions, there may be fewer microstates possible in solids compared to
gasses and liquids; however, the number of possible microstates cannot be more than
Boltzmann’s analytical derivation has. Finally, Boltzmann’s entropy equation is
given by
S = k ln Ω ð4:185Þ
where S is thermodynamic entropy, k is Boltzmann’s constants, and Ω is the number

of microstates corresponding to a macrostate. In 1901, Max Planck modified this
equation to be
S = k ln w þ constant ð4:186Þ
Here w is the probability (wahrscheinlichkeit) of a microstate.

While we discussed the difference between Boltzmann’s equation and Planck’s
version of Boltzmann’s equation earlier, unfortunately, except for the physics
textbook by Halliday and Resnick (1986), w term defined by Planck (1901) has
been misreported in the literature as the number of microstates. Therefore, we find it
necessary to include the entire paper by Max Planck (1901) in this chapter.
The most important contribution of Max Planck’s (1901) publication is that it
establishes the relationship between entropy and disorder. Even though his paper is
about the electromagnetic theory of radiation, the concept applies to solids, liquids,
and gasses.
The fact that Boltzmann does not account for inter-dependent ensemble interac-
tion makes his number of possible permutations bigger and the possibility of
complexions higher. Essentially, he calculates an upper bound for the number of
complexions. Of course, if we accounted for interaction between the molecules, as
spring constraints, the number of complexions may be smaller. Therefore, his
permutation gives the upper bound solution for the number of complexions. When
we calculate the complexions in solids, of course, we need to take into account the
ensemble interaction by all governing laws [Newton and Hooke’s], and conditions
imposed on the system. Therefore, Boltzmann’s solution, as he also believes, is
perfectly valid for systems with ensemble interactions in any state of the matter.
Moreover, Boltzmann assumes that molecules do not have any potential energy.
However, he maintains the conservation of energy. Therefore, if we assume that his
kinetic energy also includes the potential energy of the molecules, like strain energy,
then his mathematics is still accurate.
4.3.3 On the Law of Distribution of Energy in the Normal

Spectrum, [Max Planck, (1901) Annalen Der Physik,
Vol. 4, p. 553 Ff, 1901]
The recent spectral measurements made by O. Lummer and E. Pringsheim,10 and

even more notable those by H. Rubens and F. Kurlbaum,11 which together confirmed
an earlier result obtained by H. Beckmann,12 show that the law of energy distribution
in the normal spectrum, first derived by W. Wien from molecular-kinetic consider-
ations and later by me from the theory of electromagnetic radiation, is not valid
generally.
In any case, the theory requires a correction, and I shall attempt in the following to
accomplish this based on the theory of electromagnetic radiation, which I developed.
For this purpose, it will be necessary first to find in the set of conditions leading to
Wien’s energy distribution law that the term can be changed; thereafter it will be a
matter of removing this term from the set and making an appropriate substitution
for it.
In my last article,13 I showed that the physical foundations of the electromagnetic
radiation theory, including the hypothesis of “natural radiation,” withstand the most
severe criticism; and since to my knowledge there are no errors in the calculations,
the principle persists that the law of energy distribution in the normal spectrum is
completely determined when one succeeds in calculating the entropy S of an
irradiated, monochromatic, vibrating resonator as a function of its vibrational energy
U. Since one then obtains from the relationship dS/dU = 1/θ, the dependence of the
energy U on the temperature θ, and since the energy is also related to the density of
radiation at the corresponding frequency by a simple relation,14 one also obtains the
dependence of this density of radiation on the temperature. The normal energy
distribution is then the one in which the radiation densities of all different frequen-
cies have the same temperature.
Consequently, the entire problem is reduced to determining S as a function of U,
and it is to this task that the most essential part of the following analysis is devoted.
In my first treatment of this subject, I had expressed S, by definition, as a simple
function of U without further foundation, and I was satisfied to show that this form of
entropy meets all the requirements imposed on it by thermodynamics. At that time, I
believed that this was the only possible expression and that consequently Wien’s
law, which follows from it, necessarily had general validity. In a later, closer
10
O. Lummer and E. Pringsheim, Transactions of the German Physical Society 2 (1900), p. 163.
11
H. Rubens and F. Kurlbaum, Proceedings of the Imperial Academy of Science, Berlin, October
25, 1900, p. 929.
12
H. Beckmann, Inaugural dissertation, Tübingen 1898. See also H. Rubens, Weid. Ann.
69 (1899), p. 582.
13
M. Planck, Ann. d. Phys. 1 (1900), p. 719.
14
Compare with Eq. (4.205).
analysis,15 however, it appeared to me that there must be other expressions that yield
the same result and that in any case, one needs another condition to be able to
calculate S uniquely. I believed I had found such a condition in the principle, which
at the time seemed to me perfectly plausible, that in an infinitely small irreversible
change in a system, near thermal equilibrium [bold emphasis by this author], of
N identical resonators in the same stationary radiation field, the increase in the total
entropy SN = NS with which it is associated depends only on its total energy UN = NU
and the changes in this quantity, but not on the energy U of individual resonators.
This theorem leads again to Wien’s energy distribution law. But since the latter is not
confirmed by experience, one is forced to conclude that even this principle cannot be
generally valid and thus must be eliminated from the theory.16
Thus, another condition must now be introduced which will allow the calculation
of S, and to accomplish this it is necessary to look more deeply into the meaning of
the concept of entropy. Consideration of the untenability of the hypothesis made
formerly will help to orient our thoughts in the direction indicated by the above
discussion. In the following, a method will be described that yields a new, simpler
expression for entropy and thus provides also a new radiation equation that does not
seem to conflict with any facts so far determined.
Calculations of the Entropy of a Resonator as a Function of Its Energy
1. Entropy depends on the disorder and this disorder, according to the electromag-
netic theory of radiation for the monochromatic vibrations of a resonator when
situated in a permanent stationary radiation field, depends on the irregularity with
which it constantly changes its amplitude and phase, provided one considers time
intervals large compared to the time of one vibration but small compared to the
duration of a measurement. If amplitude and phase both remained absolutely
constant, which means completely homogeneous vibrations, no entropy could
exist, and the vibrational energy would have to be completely free to be converted
into work. The constant energy U of a single stationary vibrating resonator
accordingly is to be taken as time average, or what is the same thing, as a
simultaneous average of the energies of a large number N of identical resonators,
situated in the same stationary radiation field, and which are sufficiently separated
so as not to influence each other directly. It is in this sense that we shall refer to the
average energy U of a single resonator. Then the total energy
U N = NU ð4:187Þ
Of such a system of N resonators, there corresponds a certain total entropy
15
M. Planck, loc. cit., pp. 730 ff.
16
Moreover one should compare the critiques previously made of this theorem by W. Wien (Report
of the Paris Congress 2, 1900, p. 40) and by O. Lummer (loc. cit., 1900, p. 92).
SN = NS ð4:188Þ
In the same system, where S represents the average entropy of a single

resonator and the entropy SN depends on the disorder with which the total energy
UN is distributed among the individual resonators.
2. We now set the entropy SN of the system proportional to the logarithm of its
probability [bold emphasized by CB] w, within an arbitrary additive constant, so
that the N resonators together have the energy EN:
SN = k log w þ constant ð4:189Þ
In my opinion, this serves as a definition of the probability

[wahrscheinlichkeit] w [bold emphasized by CB], since in the basic assumptions
of electromagnetic theory there is no definite evidence for such a probability. The
suitability of this expression is evident from the outset, because of its simplicity
and close connection with a theorem from kinetic gas theory.17 [C.B. it is not
clear why Planck uses log versus ln used by Boltzmann] [C.B. it is important to
point out that the term “probability” defined by Planck is different than the
modern use of the term. This w is not limited to 0 and 1.] Later we define the
thermodynamic state index, which can only have values between 0 and 1.
3. It is now a matter of finding the probability w so that the N resonators together
possess the vibrational energy UN. Moreover, it is necessary to interpret UN not as
a continuous, infinitely divisible quantity, but as a discrete quantity composed of
an integral number of finite equal parts. Let us call each such part the energy
element E; consequently, we must set
UN = P E ð4:190Þ
where P represents a large integer generally, while the value of E is yet

uncertain.
(The above paragraph is in the original German)
Es kommt nun darauf an, die Wahrscheinlichkeit W dafür zu finden, dass die N Resonatoren
insgesamt die Schwingungsenergie UN besitzen. Hierzu ist es notwendig, UN nicht als eine
stetige, unbeschränkt teilbare, sondern als eine discrete, aus einer ganzen Zahl von
endlichen gleichen Teilen zusammengesetzte Grösse aufzufassen. Nennen wireinen solchen
Teil ein Energieelement E, so ist mithin zu setzen
U N = PE ð4:191Þ
wobei P eine ganze, im allgemeinen grosse Zahl bedeutet. . .
17
L. Boltzmann, Proceedings of the Imperial Academy of Science, Vienna, (II) 76 (1877), p. 428.
Now it is evident that any distribution of the P energy elements among the
N resonators can result only in a finite, integral, definite number. Every such form
of distribution we call, after an expression used by L. Boltzmann for a similar idea, a
“complex,” [actually Boltzmann calls this complexion]. If one denotes the resona-
tors by the numbers 1, 2, 3, . . ., N, and writes these side by side, and if one sets under
each resonator the number of energy elements assigned to it by some arbitrary
distribution, then one obtains for every complex a pattern of the following form:
ð4:192Þ
Here we assume N = 10, P = 100. The number R of all possible complexes is

obviously equal to the number of arrangements that one can obtain in this fashion for
the lower row, for a given N and P. For the sake of clarity, we should note that two
complexes must be considered different if the corresponding number patterns
contain the same numbers but in a different order.
From combination theory, one obtains the number of all possible complexes as
N ðN þ 1ÞðN þ 2Þ⋯ðN þ P - 1Þ ðN þ P - 1Þ!

R= = ð4:193Þ
1 2 3⋯P ðN - 1Þ!P!
Now according to Stirling’s theorem, we have in the first approximation
N! = N N ð4:194Þ
Consequently, the corresponding approximation is
ðN þ PÞNþP
R= ð4:195Þ
N N PP
4. The hypothesis which we want to establish as the basis for further calculation
proceeds as follows: for the N resonators to possess collectively the vibrational
energy UN, the probability w must be proportional to the number R of all
possible complexes formed by the distribution of the energy UN among the
N resonators; or in other words, any given complex is just as probable as any
other. Whether this actually occurs in nature one can, in the last analysis, prove
only by experience. But should experience finally decide in its favor it will be
possible to draw further conclusions from the validity of this hypothesis about
the particular nature of resonator vibrations, namely, in the interpretation put
forth by J. v. Kries18 regarding the character of the “original amplitudes,
18
Joh. v. Kries, The Principles of Probability Calculation (Freiburg, 1886), p. 36.
comparable in magnitude but independent of each other.” As the matter now

stands, further development along these lines would appear to be premature.
5. According to the hypothesis introduced in connection with Eq. (4.189), the
entropy of the system of resonators under consideration is, after suitable deter-
mination of the additive constant,
SN = k log R = k fðN þ PÞ log ðN þ PÞ - N log N - P log Pg ð4:196Þ
Moreover, by considering (4.190) and (4.187)

n o
U U U U
SN = kN 1þ log 1 þ - log ð4:197Þ
E E E E
Thus, according to Eq. (4.188), the entropy S of a resonator as a function of

its energy U is given by
n o
U U U U
S=k 1þ log 1 þ - log ð4:198Þ
E E E E
Introduction of Wien’s Displacement Law

6. Next to Kirchhoff’s theorem of the proportionality of emissive and absorptive
power, the so-called displacement law, discovered by and named after
W. Wien,19 which includes as a special case the Stefan-Boltzmann law of
dependence of total radiation on temperature, provides the most valuable con-
tribution to the firmly established foundation of the theory of heat radiation. In
the form given by M. Thiesen20 it reads as follows:
E dλ = θ5 ψ ðλθÞ dλ ð4:199Þ
where λ is the wavelength, E dλ represents the volume density of the “black-

body” radiation21 within the spectral region λ to λ + dλ, θ represents temperature,
and ψ(x) represents a certain function of the argument x only.
7. We now want to examine what Wien’s displacement law states about the
dependence of the entropy S of our resonator on its energy U, and its charac-
teristic period, particularly in the general case where the resonator is situated in
an arbitrary diathermic medium. For this purpose, we next generalize Thiesen’s
form of the law for the radiation in an arbitrary diathermic medium with the
velocity of light c. Since we do not have to consider the total radiation, but only
19
W. Wien, Proceedings of the Imperial Academy of Science, Berlin, February 9, 1893, p. 55.
20
M. Thiesen, Transactions of the German Physical Society 2 (1900), p. 66.
21
Perhaps one should speak more appropriately of a “white” radiation, to generalize what one
already understands by total white light.
the monochromatic radiation, it becomes necessary to compare different dia-

thermic media to introduce the frequency n instead of the wavelength λ.
Thus, let us denote by u dν the volume density of the radiation energy
belonging to the spectral region ν to ν + dν; then we write u dν instead of E dλ;
c/ν instead of λ, and c dν/ν2 instead of dλ. From which we obtain

c cθ
u = θ5 ψ ð4:200Þ
ν2 ν
Now according to the well-known Kirchhoff-Clausius law, the energy emit-

ted per unit of time at the frequency ν and temperature θ from a black surface in a
diathermic medium is inversely proportional to the square of the velocity of
propagation c2; hence the energy density u is inversely proportional to c3 and we
have

θ5 θ
u= 2 3 f ð4:201Þ
ν c ν
where the constants associated with the function f are independent of c. In

place of this, if f represents a new function of a single argument, we can write

ν3 θ
u= f ð4:202Þ
c3 ν
In addition, from this, we see, among other things, that as is well known, the
radiant energy u λ3 at a given temperature and frequency is the same for all
diathermic media.
8. To go from the energy density u to the energy U of a stationary resonator
situated in the radiation field and vibrating with the same frequency, ν, we use
the relation expressed in equation (34) of my paper on irreversible radiation
processes:22
ν2
K= U ð4:203Þ
c2
(K is the intensity of a monochromatic linearly, polarized ray), which

together with the well-known equation
8πK
u= ð4:204Þ
c
yields the relation
22
M. Planck, Ann. d. Phys. 1 (1900), p. 99.
8πν2
u= U ð4:205Þ
c3
From this and Eq. (4.202) follows

θ
U =ν f ð4:206Þ
ν
where now c does not appear at all. In place of this we may also write
U
θ=ν f ð4:207Þ
ν
9. Finally, we introduce the entropy S of the resonator by setting
1 dS
= ð4:208Þ
θ dU
We then obtain
dS 1 U
= f ð4:209Þ
dU ν ν
And integrated
U
S=f ð4:210Þ
ν
That is, the entropy of a resonator vibrating in an arbitrary diathermic

medium depends only on the variable U/ν, containing besides this only universal
constants. This is the simplest form of Wien’s displacement law known to me.
10. If we apply Wien’s displacement law in the latter form to Eq. (4.198) for the
entropy S, we then find that the energy element E must be proportional to the
frequency ν, thus:
ε = hν ð4:211Þ
And consequently
n o
U U U U
S=k 1þ log 1 þ - log ð4:212Þ
hν hν hν hν
Here h and k are universal constants [C. B. Planck’s constant and

Boltzmann’s constant]. By substitution into Eq. (4.207) one obtains

1 k hν
= log 1 þ ð4:213Þ
θ hν U
hν
U= ð4:214Þ
ekθ - 1
hν
and, from Eq. (4.205) there then follows the energy distribution law sought
for
8π h ν3 1
u= hν ð4:215Þ
c3 ekθ - 1
or, by introducing the substitutions given in (4.202), in terms of wavelength

λ instead of the frequency
8πch 1
E= ch ð4:216Þ
λ5 ekλθ - 1
I plan to derive elsewhere the expressions for the intensity and entropy of
radiation progressing in a diathermic medium, as well as the theorem for the
increase of total entropy in nonstationary radiation processes.
Numerical Values
11. The values of both universal constants h and k may be calculated rather precisely
with the aid of available measurements. F. Kurlbaum,23 designating the total
energy radiating into the air from 1 square centimeter of a black body at
temperature t °C in 1 s by St, found that
S100 - S0 = 0:0731 W=cm2 = 7:31 105 erg=cm2 s ð4:217Þ
From this one can obtain the energy density of the total radiation energy in
the air at the absolute temperature 1:
4 7:31 105
= 7:061 10 - 15 erg=cm3 deg4 ð4:218Þ
3 105 3734 - 2734
On the other hand, according to Eq. (4.215), the energy density of the total
radiant energy for θ = 1 is
23
F. Kurlbaum, Wied. Ann. 65 (1898), p. 759.
Z 1 Z 1
8πh v3 dv
u = u dv = 3 ð4:219aÞ
0 c 0 ehv=k - 1
Z 1
8πh
u = ν3 e - k þ e - þ e-
hν 2hν 3hν
k k þ . . . dν ð4:219bÞ
c3 0
And by termwise integration

4
8πh k 1 1 1
u = 6 1þ þ þ þ⋯
c3 h 24 34 44
ð4:219cÞ
48πk 4
= 3 3 1:0823
c h
If we set this equal to 7.061 10-15, then, since c = 3 1010 cm/s, we obtain
k4
= 1:1682 1015 ð4:220Þ
h3
12. O. Lummer and E. Pringswim24 determined the product λmθ, where λm is the
wavelength of maximum energy in the air at temperature θ, to be 2940
microndegree. Thus, in absolute measure
λm = 0:294 cm deg ð4:221Þ
On the other hand, it follows from Eq. (4.216), when one sets the derivative
of E with respect to θ equal to 0, thereby finding λ = λm

ch
1- ech=kλm θ = 1 ð4:222Þ
5kλm θ
And from this transcendental equation
ch
λm θ = ð4:223Þ
4:9651 k
Consequently
h 4:9561 0:294
= = 4:866 10 - 11 ð4:224Þ
k 3 1010
From this and Eq. (4.220) the values for the universal constants become
24
O. Lummer and E. Pringsheim, Transactions of the German Physical Society 2 (1900), p. 176.
h = 6:55 10 - 27 erg s ½today known as Planck0 s constant

k = 1:346 10 - 16 erg=deg ½today known as Boltzmann0 s constant
These are the same numbers that I indicated in my earlier communication.

[There are no references at the end of Planck’s paper.]
4.3.4 Thermodynamic State Index (TSI) in Unified

Mechanics Theory
We postulate that at any given temperature and pattern of internal rearrangement

within the matter, the rate at which any specific microstructural rearrangement
occurs is fully determined by the thermodynamic forces; this is the same concept,
postulated by Rice (1971). The thermodynamic state index (TSI) axis coordinate
defines, as the name implies, the thermodynamic state of any system between 0 and
1. The TSI axis coordinate value, which can be between 0 and 1, cannot be defined
by the Newtonian space-time coordinate system; therefore it is a linearly indepen-
dent axis. It is like what Boltzmann (1871) defined as the additional momentum axis.
Therefore, the unification of Boltzmann’s second law formulation and the laws of
Newton requires this additional linearly independent axis. However, in the unified
mechanics theory, the thermodynamic state index axis is in the normalized form of
Boltzmann’s second law formulation; therefore, it is unitless.
The coordinate value on the TSI axis is directly defined by the laws of thermo-
dynamics and according to the thermodynamic fundamental equation of the system/
matter. The thermodynamic fundamental equation must account for all irreversible
entropy generation mechanisms according to the “failure” definition chosen. In the
last sentence “failure” is in quotation because failure is a vague (subjective) term.
For example, if the color change in an acrylic particle-filled polymer is our failure
definition, then the fundamental equation will only have mechanisms that contribute
to color change, not mechanical failure. All processes, regardless of their result
effect, are governed by the laws of thermodynamics. The term “failure” here is used
to refer to change in order and increasing disorder, which can be any process when
there is irreversible entropy generation. Of course, if the physical collapse of a
building is our definition of failure for a problem at hand, the thermodynamic
fundamental equation must include all entropy generation mechanisms contributing
to physical collapse.
We have already presented the second law of thermodynamics in Boltzmann-
Planck formulation as
S = k ln w ð4:225Þ
Fig. 4.5 Schematic of a

solid continuum
undergoing internal
entropy change. A sample
under uniaxial tension W,S
are disorder parameter and
entropy, respectively
where S is entropy, k is Boltzmann’s constant, and w is the disorder parameter, which

is the probability that the system will exist in the state it is in, relative to all the
possible states it could be in.
In Fig. 4.5a W, S are the disorder parameter and entropy, respectively, and
subscript o is for the initial state and f is for the final state. The definition of “final
failure” is problem specific. The problem shown in Fig. 4.5b is identical to the
thermodynamic chamber explained in Fig. 3.1. Assume we have a composite
material as shown with a schematic in Fig 4.5b. Initially, each particle has a
coordinate, such as A3. As a result of external loading particles move to new location
and new coordinates. As a result, the disorder will increase proportional to entropy
generated. This process is identical to the virtual experiment described by Boltzmann
in Sect. 4.3.1. Boltzmann gives each particle a kinetic energy; in the example shown
in Fig. 4.5b, assuming a pseudo-static mechanical loading, particles will have strain
energy.
The second law of thermodynamics states that there is a natural tendency of any
isolated (closed) system, living or non-living, to degenerate into a more disordered
state. When the irreversible entropy generation rate becomes zero, the system
reaches a thermodynamic equilibrium point. This is usually a valley in the energy
landscape (Fig. 4.6). It is important to point out that not all specific entropy
Fig. 4.6 Definition of thermodynamic state index
generation mechanisms contribute to our “failure” definition. Those that do must be

identified and must be included in the thermodynamic fundamental equation.
Reaching absolute thermodynamic equilibrium is usually not necessary for our
formulation to be valid. Because the formulation assumes the closed system is
near thermodynamic equilibrium (Fig. 4.1).
In statistical mechanics, thermodynamic entropy is considered an intrinsic and
independent property of a system. Boltzmann-Planck equations relate entropy to the
number of microstates and the probability of a microstate that is consistent with the
macroscopic boundary conditions and initial conditions that characterize the system.
It is important to point out that the final form of the Boltzmann equation was
derived by Planck (1901). This equation is referred to as Boltzmann-Planck equation
or just the Boltzmann equation. Therefore, we will use Planck’s definition for
variable w (probability or disorder parameter). However, even if we use Ω, the
number of microstates, nothing would change in the formulation of TSI, because we
normalize the equation with respect to a reference state; as a result, it is a
non-dimensional axis.
In Eq. (4.225) entropy is for one atom. To convert it to unit-specific mass (g/mol),
we need to multiply it by Avogadro’s number NA = 6.022352 × 1023 atoms/mol and
divide it by the specific mass (g/mol) ms
NA
s= k ln w ð4:226Þ
ms
As a result, the unit for specific entropy s is J/kg K. Initially, let the probability of
a matter being in a completely “ordered ground state” [a reference state] be w0.
Under any external loading/disturbance, the system will move from the initial
[reference] configuration to a new microstate defined by s and w
s0 → s
ms S0 ms S
w0 = e N A k → w = e N A k ð4:227Þ
According to the second law, in the final stage, the system will reach maximum
entropy and maximum disorder (zero entropy generation rate) state.
s → s max
ms s ms s max
w=e NA k
→ w max = e NA k
ð4:228Þ
During this travel over the energy terrain, we can define the thermodynamics state
of the system at any point as a dimensionless variable that defines the distance from
the origin, [or any reference state] as follows:
w - w0
0≤Φ= ≤1 ð4:229Þ
w
ms s ms s0
h s - s0 i
eN A k - e N A k
= 1 - e NA ð k Þ
- ms
Φ= ms s ð4:230Þ
eN A k
Boltzmann constant k can also be given by = NRA , where R is the gas constant.
Finally, the thermodynamics state index (TSI) can be given in (Fig. 4.4)
h ms Δs
i
Φ = 1 - e- R ð4:231Þ
The thermodynamic state index imposes the laws of thermodynamics on New-

ton’s laws. We will explain this with the following simple examples. For example,
we have a soccer ball with an initial acceleration at A (Fig. 4.7).
Eventually, the ball will come to a stop and acceleration will be zero at some point
B. The energy of the ball between these 2 points will decrease according to the laws
of thermodynamics, which is represented by (1 - Φ).
Another example would be a simple one-dimensional linear-elastic spring shown
in Fig. 4.8.
If the spring is subjected to any cyclic or continuous monotonic loading, over
time the stiffness of the spring will degrade, and the displacement at the tip of the
spring will increase. Finally, the spring will break into two pieces due to fatigue. The
degradation of the strain energy capacity [referred to as resilience in some textbooks]
of the spring will follow the second law of thermodynamics
Fig. 4.7 Initial acceleration of the ball slows down according to laws of thermodynamics
Fig. 4.8 One dimensional

spring subjected to uni-axial
cycling loading
1 2
U= kδ ð1 - ΦÞ ð4:232Þ
2
Taking the first derivative of strain energy with respect to deformation and
ignoring dΦ
dδ for the sake of simplicity, we obtain
F
δ= ð4:233Þ
k ð1 - ΦÞ
Of course, the degradation of elastic strain energy storage capacity is due to the
degradation of stiffness of the spring material. However, we should point out that in
∂½12kδ2
Eq. (4.233) we assumed that ∂Φ ∂δ
is smaller than ∂δ which is true for high cycle
fatigue, but it is not true for low cycle fatigue.
It is important to remind readers that entropy is essentially energy unavailable for
work. Therefore, the total available energy of a closed-isolated system degrades
along the TSI axis between 0 and 1. Multiplication of stiffness k with (1 - Φ) is due
to the simplicity of the example chosen. Essentially, the TSI coordinate maps the
entropy generation rate [thermodynamic fundamental equation] of any system to a
linearly independent axis.
The question may be asked what happens to a system moving from one stable
equilibrium to another and getting an external energy boost during the travel due to
an external factor. Because entropy is an additive property, and we have to maintain
conservation of energy, we can include the new addition in the system in an
incremental form. Actually, in computational mechanics, we solve differential
equations in incremental format. Therefore, using the thermodynamic state index
works very well for incremental solution procedures.
The ramification of the TSI coordinate is the fact that entropy generation rate
becomes a nodal unknown in computational mechanics in addition to other nodal
unknowns, for example, displacements in a thermo-mechanical analysis. However,
to reduce the amount of computation, it is easier to calculate the TSI at Gauss
integration points based on the results of the previous step. The amount of error is
negligible if the increments are small.
4.4 Experimental Verification Example 215
4.4 Experimental Verification Example
Now that the Thermodynamic State Index (TSI) is defined, we will use two cases to
calculate TSI using the experimental data. First, a fully reversed uniaxial cyclic
loading on a steel sample is considered, and then a monotonic loading on a steel
sample. The experiments were performed in the same conditions and on identical
specimens (Fig. 4.1) for the two cases.
Now that the Thermodynamic State Index (TSI) is defined, we will use two
different experiments to calculate TSI using the measured data. First, a fully reversed
uniaxial cyclic
loading on a steel sample is considered, and then a monotonic loading case on a
steel sample. The experiments were performed in the same conditions and on
identical specimens (Fig. 4.9) for the two cases.
To evaluate the Thermodynamic State Index (TSI) evolution in both cases, only
the effects of plastic deformation are taken into account, for the sake of simplicity.
We assume that heat generated during cycling loading is insignificant, because the
frequency of loading is very small, and the temperature of the specimen remains
constant. Furthermore, because the load is uniaxial, the plasticity is reduced to one
dimension. The internal entropy generation is then reduced to
Z t
σ : ε_ p
Δs = dt ð4:234Þ
t0 ρT
For numerical computation, Eq. (4.234) is simplified to
X ðσ Δεp Þ
Δsi = : ð4:235Þ
ρT
4.4.1 Tension-Compression Cyclic Loading
The experiment consists of a displacement-controlled test conducted in a material

characterization unit, which applies uniaxial tension and compression in repeated
and fully reversed cycles to a sample. The experimental data obtained from the
digital output consists of force (F) versus displacements (u). Here the fatigue failure
Fig. 4.9 Specimen dimensions and properties

Fig. 4.10 Testing obtained stress–strain diagram of cyclic loading
Fig. 4.11 Thermodynamic state index evolution for cyclic loading
occurs after 80 cycles. Since we assume a small strain in the model, plastic strain is
calculated as follows:
σ
εp = ε - εe , εp = ε - : ð4:236Þ
E
Figure 4.10 shows the entire engineering stress and strain diagram obtained from
the experiment, which is a fully reversed uniaxial cyclic loading. Figure 4.11 depicts
the TSI evolution as a function of number cycles. As expected, TSI is initially zero
and finally approaches the value of one. Because TSI is an exponential function, it
never reaches exactly one. Readers can find more fatigue examples in the papers
cited in the references section.
4.5 Dynamic Equilibrium Equations in Unified Mechanics Theory 217
Fig. 4.12 Stress–strain diagram of cyclic loading
4.4.2 Monotonic Loading Test
For this case, the sample is identical to the one used for the cyclic loading. The only
difference is that this time it is monotonically loaded until failure [separation of the
sample into two pieces] occurs. The loading is uniaxial tension; the experimental
data is obtained similarly and treated in the same way as before. Figure 4.12 shows
the stress–strain diagram obtained from the monotonic tension test. Figure 4.13
presents the evolution of the TSI for the specimen. TSI starts from zero and
approaches one (asymptotically) at the end. We can see that the TSI remains very
small for a while before it starts to increase. This domain corresponds to the elastic
range of the material. Because we assumed that only plastic deformation could cause
degradation in the material, there is no entropy generated in the elastic range
according to our assumption. Of course, there is entropy generation due to elastic
response; however, it is smaller by orders of magnitude.
4.5 Dynamic Equilibrium Equations in Unified Mechanics

Theory
In this section, we will derive the dynamic equilibrium equations using the unified
mechanics theory. For further details, readers are referred to Bin Jamal et al. (2021).
Newtonian mechanics does not include a term for energy loss; an empirical
damping term “C” is used in the dynamic equilibrium equation. However, in unified
1
0.9
0.8
Damage Parameter
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
steps
Fig. 4.13 Thermodynamic state index evolution for monotonic loading
mechanics theory energy loss is automatically included because they are part of the
universal laws. One-dimensional free vibration analysis, with frictional dissipation,
is used to compare the unified mechanics theory results with that of the Newtonian
mechanics solution.
4.5.1 Derivation of the Dynamic Equilibrium Equation
The action of a material point, over a time increment, Δt can be defined by the
following integral:
Z t2
I= L dt ð4:237Þ
t1
where L represents the Lagrangian of the system. After we introduce TSI, the
Lagrangian of the system can be given by
_ Φi Þ
L = L ðt, u, u, ð4:238Þ
where u and u_ represent displacement and velocity of a material point, respectively.

The variable Φi is used to represent n number of different independent dissipation
[irreversible entropy generating] mechanisms within a given thermodynamic system
at a material point.
If the integral function in Eq. (4.237) attains a local minimum at a point, q, and let
ζ(t) be an arbitrary function that has at least one derivative and vanishes at the initial
time, t1, and final time, t2, then for any real number E, closer to zero, the following
condition is valid:
I ðqÞ ≤ J ðEÞ ð4:239Þ
where the real-valued function, J, is represented as follows:
J ð E Þ = I ðq þ E ζ Þ ð4:240Þ
Let’s assume the functional “I” has a local minimum at u = q, and the function
J has a local minimum at E = 0. Hence the following equation must hold:
Z t2
dJ dL
= dt = 0 ð4:241Þ
dE E=0 t1 dE E=0
Taking the total derivative of Lagrangian, L, yields a new additional term when
compared to Newtonian mechanics formulation,
dL ∂L du ∂L d u_ X ∂L dΦi
n
= þ þ ð4:242Þ
dE ∂u dE ∂u_ dE i = 1 ∂Φi dE
In Eq. (4.242) that the derivative of Lagrangian with respect to the new axis,
Thermodynamic State Index, is non-zero. The displacement is represented as
follows:
u=q þ E ζ ð4:243Þ
Hence velocity can be given by
u_ = q_ þ E ζ_ ð4:244Þ
Using Eqs. (4.243) and (4.244), we can write Eq. (4.242) as follows:
∂L _ X ∂L dΦi
n
dL ∂L
= ζþ ζþ ð4:245Þ
dE ∂u ∂u_ i=1
∂Φi dE
For example, a single degree of freedom mass-spring system, subjected to

Coulomb friction, is chosen. Hence it is assumed that friction is the only dissipation
mechanism. We can write the following total derivative of Φ with respect to E:
dΦ ∂Φ du ∂Φ du_
= þ ð4:246Þ
dE ∂u dE ∂u_ dE
Hence, by substituting Eq. (4.246) in Eq. (4.245), we can write


dL ∂L ∂L ∂Φ ∂L ∂L ∂Φ _
= þ ζþ þ ζ ð4:247Þ
dE ∂u ∂Φ ∂u ∂u_ ∂Φ ∂u_
Therefore, Eq. (4.241) takes the following form:

Z t2 Z t2
dL ∂L ∂L ∂Φ
dt = þ ζ dt
t1 dE E = 0 t1 ∂u ∂Φ ∂u E=0
Z t2
∂L ∂L ∂Φ _
þ þ ζ dt = 0 ð4:248Þ
t1 ∂u_ ∂Φ ∂u_ E=0
The second term on the right-hand side can be integrated by parts; we get the
following equation:
Z t2
∂L ∂L ∂Φ d ∂L ∂L ∂Φ
þ - þ ζ dt
t1 ∂u ∂Φ ∂u dt ∂u_ ∂Φ ∂u_ E=0
t 2
∂L ∂L ∂Φ
þ þ ½ζ E = 0 =0 ð4:249Þ
∂u_ ∂Φ ∂u_ t1
By the definition of ζ, the second term in Eq. (4.249) vanishes at the boundary of
the integral domain. Hence, using Eqs. (4.243) and (4.244) in Eq. (4.249) and by
applying the principle of local state, we can write

∂L ∂L ∂Φ d ∂L ∂L ∂Φ
þ - þ =0 ð4:250Þ
∂u ∂Φ ∂u dt ∂u_ ∂Φ ∂u_
Compared to the Newtonian mechanics-based Euler-Lagrange equation,

Eq. (4.250) has an additional derivative as a function of TSI. Using Lagrange-
d’Alembert’s principle, and variational calculus principles, for an external point
load, F, we can write the following equation:

∂L ∂L ∂Φ d ∂L ∂L ∂Φ
þ - þ þ F=0 ð4:251Þ
∂u ∂Φ ∂u dt ∂u_ ∂Φ ∂u_
Equation (4.251) is the dynamic equilibrium equation in unified mechanics

theory. Specific form for a given particular system is derived by writing the ther-
modynamic fundamental equation for the system and the Lagrangian of the system.
In the following section, a mass-spring system with friction as the sole source of
dissipation is considered.
Fig. 4.14 Block diagram of

a mass-spring system
subjected to friction
4.5.2 Vibration of a 1-DOF Mass-Spring System
The governing equation of motion for a mass-spring system, Fig. 4.14, is derived
using the unified mechanics theory and compared with that of a Newtonian mechan-
ics equation of motion.
Lagrangian of the system is given by
L=K -V ð4:252Þ
where K and V are the total kinetic energy and total potential energy of the system,
respectively. In this example, mass is constant as a function of time. For a mass m,
excited by an external force, the total kinetic energy of a conservative system
(without damping) is given by the following equation:
1 _2
K= m~u ð4:253Þ
2
where ~u_ represents the velocity of a mass-spring system in Newtonian mechanics

(with no dissipation). In the unified mechanics theory, we express the net kinetic
energy of the mass, with a final velocity of u_ as follows:
1 2 1 _2
mu_ = m~u ð1 - ΦÞ ð4:254Þ
2 2
In the same fashion, the potential energy is expressed, by introducing entropy

generation due to friction. Hence, the Lagrangian in the unified mechanics theory is
given by the following equation:
1 _2 1 2 1 1 2 1 1 2
L= m~u - k~u = mu_ - ku ð4:255Þ
2 2 ð1 - Φ Þ 2 ð1 - ΦÞ 2
In Eq. (4.255), entropy generation due to degradation of the spring stiffness is

ignored.
Derivatives in Eq. (4.251) can be calculated using Eq. (4.255):
∂L
= - ð1 - ΦÞ - 1 ku ð4:256Þ
∂u
∂L 1 1 2 1 1 2
= mu_ - ku ð4:257Þ
∂Φ ð1 - ΦÞ2 2 ð1 - ΦÞ2 2
∂Φ ms ð - ΔSmRs Þ ∂ΔS ms ∂ΔS
= e = ð1 - ΦÞ ð4:258Þ
∂u R ∂u R ∂u
Note: Eq. (4.258) is the derivative of entropy with respect to displacement. This
term, ∂Φ
∂u
, is taken as zero in Newtonian mechanics, even if it is used in a thermody-
namically consistent formulation, because there is no TSI axis.
∂L
= ð1 - ΦÞ - 1 mu_ ð4:259Þ
∂u_

d ∂L d 1
= mu_ ð1 - ΦÞ - 1 þ m€u ð4:260Þ
dt ∂u_ dt ð1 - ΦÞ
dΦ
d ∂L 1
= mu_ dt
þ m€
u ð4:261Þ
dt ∂u_ ð1 - ΦÞ 2 ð1 - ΦÞ
dΦ ms dΔS
= ð1 - ΦÞ ð4:262Þ
dt R dt

d ∂L ∂Φ d ∂L ∂Φ ∂L d ∂Φ
= þ ð4:263Þ
dt ∂Φ ∂u_ dt ∂Φ ∂u_ ∂Φ dt ∂u_
! !
d ∂L d 1 1 2 d 1 1 2
= mu_ - ku ð4:264Þ
dt ∂Φ dt ð1 - ΦÞ2 2 dt ð1 - ΦÞ2 2
!
1 2d 1 1 d 1 2 1 d
= mu_ þ mu_ - ku2
2 dt ð1 - ΦÞ2 ð1 - ΦÞ2 dt 2 2 dt
!
1 1 d 1 2
- ku ð4:265Þ
ð1 - Φ Þ2 ð1 - ΦÞ2 dt 2
!
d ∂L 1 2 1 2 1 ms dΔS 1
= mu_ - ku 2 þ
dt ∂Φ 2 2 ð1 - ΦÞ 2 R dt ð1 - ΦÞ2
ðm€u - kuÞu_ ð4:266Þ
∂Φ ms ∂ΔS
= ð1 - ΦÞ ð4:267Þ
∂u_ R ∂u_

d ∂Φ m d ∂ΔS
= s ð1 - ΦÞ ð4:268Þ
dt ∂u_ R dt ∂u_

ms ∂ΔS d ms d ∂ΔS
= ðð1 - ΦÞÞ þ ð1 - ΦÞ ð4:269Þ
R ∂u_ dt R dt ∂u_
or

d ∂Φ m d ∂ΔS m 2 ∂ΔS dΔS
= s ð1 - Φ Þ - s ð1 - ΦÞ ð4:270Þ
dt ∂u_ R dt ∂u_ R ∂u_ dt

d ∂L ∂Φ
dt ∂Φ ∂u_
( ! )
1 2 1 2 1 ms dΔS 1
= mu_ - ku 2 þ u - kuÞu_
ðm€
2 2 ð1 - ΦÞ2 R dt ð1 - ΦÞ2
( )
ms ∂ΔS 1 1 2 1 1 2
ð1 - ΦÞ þ mu_ - ku
R ∂u_ ð1 - ΦÞ2 2 ð1 - ΦÞ2 2

ms d ∂ΔS m 2 ∂ΔS dΔS
ð1 - ΦÞ - s ð1 - Φ Þ ð4:271Þ
R dt ∂u_ R ∂u_ dt
Finally, we can write the following equilibrium equation:

1 1 m 1 dΔS m 1
m€u þ ku þ m s u_ þ s
ð1 - ΦÞ ð1 - ΦÞ R ð1 - ΦÞ dt R ð1 - ΦÞ
n o
1 1 ∂ΔS ∂ΔS
ku2 - mu_ 2 þℊ þ F=0 ð4:272Þ
2 2 ∂u ∂u_
where the function, ℊ, is the summation of all the terms as a function of ∂ΔS
∂u_
.
Newtonian mechanics governing equation of motion for the 1-D sliding friction
system shown in Fig. 4.14 is given by
m€u þ μmg sgn ðu_ Þ þ ku = F ð4:273Þ
where μ is the Coulomb friction coefficient, m is the mass, g is the gravitational

constant, and sgn() is the signum function. The damping term “C” in Eq. (4.273) is
represented by the Coulomb friction force. Equation (4.273) does not include any
dissipation other than mechanical work due to friction. However, other dissipative
processes are accounted for in the unified mechanics equilibrium Eq. (4.272). These
additional entropy generation mechanisms are defined in terms of change in entropy,
ΔS, and represented along the thermodynamic state index axis, Φ. The entropy
generation function during the sliding friction is derived in the following section.
4.5.3 Entropy Generation in Sliding Friction Contact:

Thermodynamic Fundamental Equation
Mechanical work dissipated during friction sliding is given by the following

equation:
Z t2
Wm = μmg j du j ð4:274Þ
t1
The coefficient of friction, μ, is obtained empirically. It is defined by the critical force

that initiates the onset of sliding. However, it does not account for heat [continuous
friction leads to heat generation] and acoustic losses. The heat generation during
sliding friction is dependent on the velocity. A large increment in temperature leads
to diffusion of the material, and such process is applied in friction welding applica-
tions. This example, for simplicity, does not account for diffusion mechanism
entropy generation. Because the purpose of the example is just to derive the dynamic
equilibrium equation in unified mechanics theory for a dynamic event. Nevertheless,
we simply assume that dissipation due to all other ignored entropy generation
mechanisms is proportional to dissipation due to mechanical sliding, Hence, we
can write the following relation for total dissipation work:
W Td = ð1 þ κ d ÞW m ð4:275Þ
where W Td is the total dissipation work and κ d is the fraction of work dissipated as
heat, acoustic losses, diffusion, etc. The initial potential energy of the spring under
free vibration will be dissipated as work of friction, heat, acoustic losses, diffusion,
etc. However, this conversion cannot be perfect, and there will be energy lost to
unintended work. Hence, the specific entropy generated is given by the following
equation:
Z t2
1
ΔS = ð1 þ κd Þψ s mμg j du j ð4:276Þ
mT t1
where ψ s represents the inefficiency factor, as the ratio between the maximum
entropy at which TSI, Φ tends to unity (say, at Φ = 0.99) and the maximum entropy
that the mass-spring system could generate, for initial potential energy in the spring,
if it had zero inefficiencies. Hence, by substituting Eq. (4.276) in Eq. (4.272), we get
the following equation:
n o
1 1 m 1 1 2 1 2
m€u þ ku þ s ku þ mu_
ð1 - ΦÞ ð1 - ΦÞ R ð1 - ΦÞ 2 2
h i
1
ð1 þ κd Þψ s mμg sgn ðu_ Þ = F ð4:277Þ
mT
κ d is required to be able to solve Eq. (4.277). To calculate κ d, we assume the

following condition. The rate of heat generation, acoustic losses, and diffusion are
dependent on the state of contact surfaces and the impulse force generated due to
friction. Since friction is proportional to normal reaction, it will be impulsive when a
normal force is impulsive. Due to unevenness (asperity) of the contact surface, a
sudden change in the direction of the normal contact force generates an impulsive
friction force in addition to the sliding friction force. Impulse force due to interface
surface asperities increases the dissipation. It is shown in the literature that the rate of
dissipation is maximum in the initial state and is dependent on the TSI, Φ. Similarly,
if we represent the dependence of κd with Φ and propose the following relation for
the determination of parameter, κ d.

λmg Δt
κd = Φ ð4:278Þ
_
mjuj
where Δt represents the duration for which the momentum is calculated and λ is the
thermodynamic friction coefficient. The inefficiency factor, ψ s, represents the dissi-
pation of initial strain energy during a free vibration. Once the TSI, Φ, is close to
unity (say 0.99), the maximum entropy for a value of, assuming Φc = 1, can be
calculated from TSI as follows:
h i
Φ = Φcritical 1 - e - ΔS R
ms
ð4:279Þ
When ΔS0u = 4:605 mRs, Φcritical = 0.99. Keep in mind that Eq. (4.279) is an exponen-
tial function; hence, Φcritical = 1.0 only when ΔS = 1.
ΔS0u is the maximum entropy, for which TSI is close to unity. For any initial
potential energy, the maximum entropy generation cannot be more than the initial
potential energy; hence it is given by the following equation:
1 1 2
ΔSu = ku ð4:280Þ
mT 2 0
where u0 is the initial excitation displacement. Hence the inefficiency factor is given
by the following equation:
ΔS ′ u
ψs = ð4:281Þ
ΔSu
Of course, if we calculate until ΔS = 1 ψ s = 1 will be equal to one. It is less than

one ψ s < 1 because we are stopping calculations at Φ = 0.99
Equation (4.277) is solved together with Eqs. (4.276) and (4.281), and the results
are presented in the following section.
4.5.4 Comparison with Newtonian Mechanics Results
Free vibration analysis is done for a single degree of freedom system under frictional
damping. Equations (4.273) and (4.277) are solved by using Newmark’s method in
MATLAB. The model parameters used in the present study are listed in Table 4.2.
Simulation results are plotted in Figs. 4.15, 4.16, 4.17 and 4.18 for the input
parameters given in Table 4.2. The solution of the equation of motion in unified
mechanics theory and Newtonian mechanics for damped oscillation of the mass is
shown in Fig. 4.15 for displacement time history and Fig. 4.16 for velocity time
history, respectively. It can be readily noticed that both models capture damped
oscillation. However, the decay in amplitude is linear for the Newtonian mechanics
model, and in unified mechanics, the response is nonlinear. Experimental results on
similar models show that the amplitude decay is nonlinear. Since the response of the
mass-spring system in the unified mechanics equilibrium equation is dependent on
the entropy evolution within the system, a detailed study is required to get the
entropy generation function, also known as the thermodynamic fundamental equa-
tion. However, this simple example is limited to the derivation of the governing
equation of motion using the fundamental principles of the unified mechanic theory.
Figure 4.17 shows the response of the mass-spring oscillator subjected to friction,
along the Thermodynamic State Index axis. In the present context, the TSI axis
represents the thermodynamic state of oscillation. Hence, the interpretation of TSI in
this example is limited to the dissipation of initial strain energy stored in the spring.
As shown in Fig. 4.17, the oscillations dampen, and the strain energy and kinetic
energy of the mass-spring system decays. Even though the Newtonian equation of
motion given for the mass-spring system accounts for damping due to friction
utilizing a pseudo linear force, the equation does not directly deal with the
Table 4.2 List of parameters used to solve the proposed model and the Newtonian mechanics
equation
Parameter Value Unit
Spring stiffness, k 6.3 × 105 N/m
Mass, m 2 kg
Critical TSI, Φcri 0.99 Unitless
Friction coefficient, μ 0.2 Unitless
Thermodynamic friction coefficient, λ 0.003 Unitless
Molar mass for iron, ms 55.84 × 10-3 kg/mol
Initial spring displacement at t = 0, u0 6.25 × 10-4 m
0.0007 Displacement, u (UMT model)

0.0006 Displacement, u (Newton's model)
0.0005
0.0004
0.0003
Displacement, u (m)
0.0002
0.0001
0.0000
-0.0001
-0.0002
-0.0003
-0.0004
-0.0005
-0.0006
-0.0007
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
Time (s)
Fig. 4.15 Displacement time history for a damped oscillator in the presence of contact surface
friction
0.4
Velocity, v (UMT model)
Velocity, v (Newton's model)
0.2
Velocity, v (m/s)
0.0
-0.2
-0.4
0.0 0.1 0.2 0.3 0.4
Time (s)
Fig. 4.16 Velocity time history time for a damped oscillator in the presence of contact surface
friction
0.0006
Displacement, u (UMT model)
Velocity, v (UMT model) 0.4
0.0004
0.2
Displacement, u (m)
Velocity, v (m)
0.0002
0.0000 0.0
-0.0002
-0.2
-0.0004
-0.4
-0.0006
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
TSI
Fig. 4.17 The response of the mass-spring oscillator along with the Thermodynamic State Index
(TSI) axis
Fig. 4.18 Displacement amplitude versus time for various static friction coefficient values
thermodynamics of the energy dissipation. However, the unified mechanics theory

accounts for the thermodynamics of the system, in the equation of motion, as derived
in Eq. (4.272). The dynamic equation of motion in unified mechanics theory must
account for all the dissipative processes in the thermodynamic fundamental equa-
tions. Unfortunately, these equations are not readily available and must be derived.
Figure 4.18 shows the variations in displacement amplitude decay for different
frictional coefficients. The linear amplitude decay of the Newtonian equation of
motion for friction is not supported by experiments. Tangents to the curves in Fig
4.18 show that the rate of amplitude decay in the unified mechanics theory-based
models is high in the initial stage of oscillation when compared with the final stage.
This could be due to the macro-level model approximations for the entropy gener-
ation equation assumed in Eq. (4.276). A detailed study is required for the identifi-
cation of all the entropy generation sources that are active in the contact surfaces
of the friction block model. Such a study can help in reproducing the nonlinear
response of friction block experiments.
The unified mechanics theory captures the time at which oscillation stops, as
predicted by using the Newtonian mechanics equation of motion. However, the
thermodynamic fundamental equation used in this simple example is very crude with
several oversimplifications
4.5.5 Comparison with Experimental Data
Test data published in the literature by Feeny and Liang (1996) is used to compare
the predictions using the dynamic equilibrium of the unified mechanics’ theory
(UMT) and Newton’s equilibrium equation. Feeny and Liang (1996) experimental
setup consists of a mass-spring single degree of freedom system with dry friction
(Imanian & Modarres, 2015) subjected to an initial displacement. A linear voltage
differential transformer is used to sense the motion of mass. A combined effect of
Coulomb friction and viscous damping on a single degree of freedom system is
investigated by using a similar experimental setup Liang (2005). A list of the model
parameters used in this comparative study is given in Table 4.3.
Simulations and test data for displacements and displacement-amplitude of vibra-
tion are shown in Figs. 4.19 and 4.20, respectively. Both the unified mechanics
theory and Newtonian mechanics capture the damped oscillations under Coulomb
friction. The amplitude variations, as shown in Fig. 4.20, show a different trend for
the unified mechanics theory when compared with the predictions using Newton’s
equation of motion. Newton’s equation of motion predicts results closer to the test
data. However, it is to be noted that Newton’s model parameter involves experi-
mental curve-fit data, representing the frictional force. Nevertheless, the unified
mechanics theory-based model involves TSI, which evolves with the
Table 4.3 Parameters used to compare UMT, Newtonian mechanics, and experimental data
Parameter Value Unit
Spring stiffness, k 54 N/m
Critical TSI, Φc 1 –
Thermodynamic friction coefficient, λ 0.003 –
Mass, m 1.51 kg
Sliding friction coefficient, μ 4.23 × 10-4 –
Molar mass for iron, ms 55.84 × 10-3 kg/mol
Displacement at t = 0, u0 4.25 × 10-3 m
0.005
0.004 Displacement, u (UMT model)

Displacement, u (Newton's model)
Displacement, u (m) 0.003 Displacement, u (Test data)
0.002
0.001
0.000
-0.001
-0.002
-0.003
-0.004
-0.005
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Time (s)
Fig. 4.19 Comparison between simulations predictions and test data for the displacement time
history. (After Bin Jamal et al. (2020))
Fig. 4.20 Comparison 5

between the predictions and
Test data
Displacement amplitude (×10 m)
test data for the amplitude

time history. (After Bin 4 Newton's model
-3
Jamal et al. (2020)) UMT model
0
0 1 2 3 4 5 6 7 8 9 10 11 12
Time (s)
thermodynamics of dissipative mechanisms that are associated with friction along

the TSI axis. Hence, UMT tracks dissipated energy, not displacement. Therefore, it
is expected that empirical models based on Newtonian mechanics give a better
match to the experiment they are obtained from.
Spring stiffness is assumed to be isotropic under the compressive and tensile
loading. Newton’s equilibrium equation agrees with the initial positive peaks of half-
cycles of oscillation but does not predict the negative peaks of initial half-cycle data
References 231
for displacement. While comparing the duration of free vibration, UMT predicts a
closer result to the test data. Since the equilibrium equation in UMT is a function of
the thermodynamic fundamental equation, the model requires the computation of
entropy accurately to get a solution matching the test data.
Here we conclude the formulation of the unified mechanics theory. In the
following chapter applications of the theory are discussed in greater detail.
References
Amiri, M., & Khonsari, M. M. (2012). On the role of entropy generation in processes involving
fatigue. Entropy, 14(1), 24–31.
Basaran, C., & Chandaroy, R. (2002). Thermomechanical analysis of solder joints under thermal
and vibrational loading. Transactions of the ASME: Journal of Electronic Packaging, 124(1),
60–67.
Basaran, C., & Lin, M. (2007a). Electromigration induced strain field simulations for
nanoelectronics lead-free solder joints. International Journal of Solids and Structures, 44,
4909–4924.
Basaran, C., & Lin, M. (2007b). Damage mechanics of electromigration in microelectronics copper
interconnects. International Journal of Materials and Structural Integrity, 1(1/2/3), 16–39.
Basaran, C., & Lin, M. (2008). Damage mechanics of electromigration induced failure. Mechanics
of Materials, 40, 66–79.
Basaran, C., & Nie, S. (2004). An irreversible thermodynamic theory for damage mechanics of
solids. International Journal of Damage Mechanics, 13(3), 205–224.
Basaran, C., & Nie, S. (2007). A thermodynamics based damage mechanics model for particulate
composites. International Journal of Solids and Structures, 44, 1099–1114.
Basaran, C., & Tang, H. (2002). Implementation of a thermodynamic framework for damage
mechanics of solder interconnects in microelectronic packaging. International Journal of
Damage Mechanics, 11(1), 87–108.
Basaran, C., & Yan, C. Y. (1998). A thermodynamic framework for damage mechanics of solder
joints. Transactions of the ASME: Journal of Electronic Packaging, 120, 379–384.
Basaran, C., Lin, M., & Ye, H. (2003). A thermodynamic model for electrical current induced
damage. International Journal of Solids and Structures, 40(26), 7315–7327.
Basaran, C., Zhao, Y., Tang, H., & Gomez, J. (2005). A damage mechanics based unified
constitutive model for solder alloys. Transactions of the ASME: Journal of Electronic Packag-
ing, 127(3), 208–214.
Basaran, C., Li, S., & Abdulhamid, M. (2008a). Thermomigration induced degradation in solder
alloys. Journal of Applied Physics, 103, 123520.
Basaran, C., Nie, S., & Hutchins, C. (2008b). Time-dependent behavior of a particle filled
composite PMMA/ATH at elevated temperatures. Journal of Composite Materials, 42(19),
2003–2025.
Bazant, Z. P. (1972). Thermodynamics of interacting continua with surfaces and creep analysis of
concrete structures. Nuclear Engineering Structures and Design, 20, 477–505.
Beven, K. (2006). A manifesto for the equifinality thesis. Journal of Hydrology, 320(1), 18–36.
Bin Jamal, M. N., Kumar, A., Rao, C. L., & Basaran, C. (2020). Low cycle fatigue life prediction
using unified mechanics theory in Ti-6Al-4V alloys. Entropy, 22(1), 24.
Bin Jamal, M. N., Lee, H. W., Lakshmana Rao, C., & Basaran, C. (2021). Dynamic equilibrium
equations in unified mechanics theory. Applied Mechanics, 2, 63–80.
Callen, H. B. (1985). Thermodynamics and an Introduction to Thermostatistics, Wiley.
Chalmers, M. (2016). Second law of thermodynamics “broken”. New Scientist. Retrieved 2016-
02-09.
Chudnovsky, A. (1973). Fracture of solids. In Scientific papers on Elasticity and plasticity, N9
(pp. 3–43) (Russian).
Chudnovsky, A. (1984). Statistics and thermodynamics of fracture. International Journal of
Engineering Science, 22(8–10), 989–997.
Cuadras, A., Crisóstomo, J., Ovejas, V. J., & Quilez, M. (2015). Irreversible entropy model for
damage diagnosis in resistors. Journal of Applied Physics, 118, 165103.
Cuadras, A., Romero, R., & Ovejas, V. J. (2016). Entropy characterization of overstressed capac-
itors for lifetime prediction. Journal of Power Sources, 336, 272–278.
Cuadras, A., Yao, J., & Quilez, M. (2017). Determination of LEDs degradation with entropy
generation rate. Journal of Applied Physics, 122, 145702.
David Halliday and Robert Resnick (1966) Physics, John Wiley & Sons
De Groot, S. R., & Mazur, P. (1962a). Nonequilibrium thermodynamics. North-Holland.
Dehoff, R. T., “Thermodynamics in Materials Science,” 2nd Edition, Mc Graw-Hill, New York,
1993.
Ehrenfest, P., Trkal, V. Deduction of the dissociation-equilibrium from the theory of quanta and a
calculation of the chemical constant based on this. Ann. Phys., 1921, 65, 609–628.
Evans, D. J., Cohen, E. G. D., & Morriss, G. P. (1993). Probability of second law violations in
shearing steady states. Physical Review Letters, 71(15), 2401–2404.
Feeny, B., & Liang, J. (1996). A decrement method for the simultaneous estimation of coulomb and
viscous friction. Journal of Sound and Vibration, 195, 149–154.
Gerstner, E. Second law broken. Nature (2002). https://doi.org/10.1038/news020722-2
Getling, A. V. (1998). Rayleigh–Bénard convection: Structures and dynamics. World Scientific.
ISBN 978-981-02-2657-2.
Glansdorff, P., & Prigogine, I. (1971a). Thermodynamic theory of structure, stability and fluctua-
tions. Wiley.
Glansdorff, P., & Prigogine, I. (1971b). Thermodynamics theory of structure, stability, and fluctu-
ations. Wiley-Interscience.
Gomez, J., & Basaran, C. (2005). A thermodynamics based damage mechanics constitutive model
for low cycle fatigue analysis of microelectronics solder joints incorporating size effect.
International Journal of Solids and Structures, 42(13), 3744–3772.
Gomez, J., & Basaran, C. (2006). Damage mechanics constitutive model for Pb/Sn solder joints
incorporating nonlinear kinematic hardening and rate dependent effects using a return mapping
integration algorithm. Mechanics of Materials, 38, 585–598.
Gomez, J., Lin, M., & Basaran, C. (2006). Damage mechanics modeling of concurrent thermal and
vibration loading on electronics packaging. Multidiscipline Modeling in Materials and Struc-
tures, 2(3), 309–326.
Gunel, E. M., & Basaran, C. (2010). Stress whitening quantification in thermoformed of mineral
filled acrylics. ASME Journal of Engineering Materials and Technology, 132, 031002-11.
Gunel, E. M., & Basaran, C. (2011a). Damage characterization in non-isothermal stretching of
acrylics: Part I theory. Mechanics of Materials, 43(12), 979–991.
Gunel, E. M., & Basaran, C. (2011b). Damage characterization in non-isothermal stretching of
acrylics: Part II experimental validation. Mechanics of Materials, 43(12), 992–1012.
Guo, Q., Zaõri, F., & Guo, X. (2018). An intrinsic dissipation model for high-cycle fatigue life
prediction. International Journal of Mechanical Sciences, 140, 163–171.
Haddad, W. M. (2017). Thermodynamics: The unique universal science. Entropy, 19, 621.
Haddad, W. M. (2019). A dynamical systems theory of thermodynamics. Princeton University Press.
Haddad, W. M., Chellaboina, V., & Nersesov, S. G. (2005). Thermodynamics: A dynamical systems
approach. Princeton University Press.
Halliday, D. and Resnick R., Physics, John Wiley & Sons, Inc. 1986, New York, NY.
Hsiao, C.-C., & Liang, B.-H. (2018). The generated entropy monitored by pyroelectric sensors.
Sensors, 18, 3320.
References 233
Imanian, A., & Modarres, M. (2015). A thermodynamic entropy approach to reliability assessment
with applications to corrosion fatigue. Entropy, 17(10), 6995–7020.
Imanian, A., & Modarres, M. (2018). A thermodynamic entropy-based damage assessment with
applications to prognosis and health management. Structural Health Monitoring, 17(2),
240–254.
Jang, J. Y., & Khonsari, M. M. (2018). On the evaluation of fracture fatigue entropy. Theoretical
and Applied Fracture Mechanics, 96, 351–361.
Jaynes, E. T. (1957). Information theory and statistical mechanics. Physics Review, 106, 620–615.
Jaynes, E. (1992). The Gibbs paradox. In C. Smith, G. Erickson, & P. Neudorfer (Eds.), Maximum
entropy and Bayesian methods (pp. 1–22). Kluwer Academic Publishers.
Kestin, J., & Rice, J. R. (1970). A critical review of thermodynamics (E. B. Stuart et al., Eds.,
p. 275). Mono Book Corp.
Kijalbaev, D., & Chudnovsky, A. (1970). On fracture of deformable solids. Journal of Applied
Mechanics and Technical Physics, 3, 105.
Klamecki, B. E. (1980a). A thermodynamic model of friction. Wear, 63, 113–120.
Klamecki, B. E. (1980b). Wear – An entropy production model. Wear, 58, 325–330.
Klamecki, B. E. (1984). An entropy-based model of plastic deformation energy dissipation in
sliding. Wear, 96, 319–329.
Koschmieder, E. L. (1993). Bénard cells and Taylor vortices. Cambridge University Press. ISBN
0521-40204-2.
Lebowitz, J. (1993). Boltzmann’s entropy and time’s arrow. Physics Today, 46, 32–38.
Li, S., & Basaran, C. (2009). A computational damage mechanics model for thermomigration.
Mechanics of Materials, 41(3), 271–278.
Li, S., Abdulhamid, M., & Basaran, C. (2008). Simulating damage mechanics of electromigration
and thermomigration. Simulation: Transactions of the Society for Modeling and Simulation
International, 84(8/9), 391–401.
Liakat, M., & Khonsari, M. M. (2015). Entropic characterization of metal fatigue with stress
concentration. International Journal of Fatigue, 70, 223–234.
Liang, J.-W. (2005). Identifying Coulomb and viscous damping from free-vibration acceleration
decrements. Journal of Sound and Vibrations, 282, 1208–1220.
Lin, M., & Basaran, C. (2005). Electromigration induced stress analysis using fully coupled
mechanical-diffusion equations with nonlinear material properties. Computational Materials
Science, 34(1), 82–98.
fracture. Proceedings of the Royal Society A, 466, 423–438.
Onsager, L. (1931). Reciprocal relations in irreversible processes, I. Physical Review, 37, 405–426.
Onsager, L. (1932). Reciprocal relations in irreversible processes, II. Physical Review, 38,
2265–2279.
Osara, J. A., & Bryant, M. D. (2019a). Thermodynamics of fatigue: Degradation-entropy genera-
tion methodology for system and process characterization and failure analysis. Entropy,
21(7), 685.
Osara, J. A., & Bryant, M. D. (2019b). A thermodynamic model for lithium-ion battery degradation:
Application of the degradation-entropy generation theorem. Inventions, 4, 0023.
Ostoja-Starzewski, M. (2016). Second law violations, continuum mechanics, and permeability.
Continuum Mechanics and Thermodynamics, 28(1–2), 489–501.
Ostoja-Starzewski, M., & Raghavan, B. V. (2016). Continuum mechanics versus violations of the
second law of thermodynamics. Journal of Thermal Stresses, 39(6), 734–749.
Planck, M. (1901). “Ueber das Gesetz der Energieverteilung im Normalspektrum”. Annalen der
Physik. 309 (3): 553–563. Bibcode:1901AnP...309..553P. doi:10.1002/andp.19013090310
Prigogine, I. (1955). Introduction to thermodynamics of irreversible processes. Charles C. Thomas
Publisher.
Prigogine, I. (1957). The molecular theory of solutions. North-Holland Publishing Company.
Prigogine, I. (1961). Introduction to thermodynamics of irreversible processes (2nd ed.).

Interscience. OCLC 219682909.
Prigogine, I. (1968). Introduction to thermodynamics of irreversible processes. Wiley-Interscience.
Prigogine, I., & Defay, R. (1954). Chemical thermodynamics. Longmans Green and Co..
Prigogine, I., & Herman, R. (1971). Kinetic theory of vehicular traffic. American Elsevier. ISBN
0-444-00082-8.
Prigogine, I., & Nicolis, G. (1977). Self-organization in non-equilibrium systems. Wiley. ISBN
0-471-02401-5.
Rice, J. R. (1971). Inelastic constitutive relations for solids: An internal-variable theory and its
application to metal plasticity. Journal of the Mechanics and Physics of Solids, 19, 433–455.
Rice, J. R. (1977). Thermodynamics of the quasi-static growth of Griffith cracks. Journal of the
Mechanics and Physics of Solids, 26, 61–78.
Richard Swalin (1972). Thermodynamics of Solids, John Wiley & Sons, New York NY
Rivas, Á., & Martin-Delgado, M. A. (2017). Topological heat transport and symmetry-protected
boson currents. Scientific Reports, 7(1), 6350. https://doi.org/10.1038/s41598-017-06722-x
Rivli, R. S. (1981). Some comments on the endochronic theory of plasticity. International Journal
of Solids and Structures, 17(2), 231–248.
Searles, D. J., & Evans, D. J. (2004). Fluctuations relations for nonequilibrium systems. Australian
Journal of Chemistry, 57(12), 1119–1123. https://doi.org/10.1071/ch04115
Sharp, K.; Matschinsky, F. Translation of Ludwig Boltzmann’s Paper “On the Relationship
between the Second Fundamental Theorem of the Mechanical Theory of Heat and Probability
Calculations Regarding the Conditions for Thermal Equilibrium” Sitzungberichte der
Kaiserlichen Akademie der Wissenschaften. Mathematisch-Naturwissen Classe. Abt. II,
LXXVI 1877, pp 373–435 (Wien. Ber. 1877, 76:373–435). Reprinted in Wiss. Abhandlungen,
Vol. II, reprint 42, p. 164–223, Barth, Leipzig, 1909. Entropy 2015, 17, 1971–2009. https://doi.
org/10.3390/e17041971
Sherbakov, S. S., & Sosnovskiy, L. A. (2010). Mechanics of tribo-fatigue systems (p. 407). BSU.
Sosnovskiy, L. A. (1987). Statistical mechanics of fatigue damage (p. 288). Nauka i Tekhnika. (In
Russian).
Sosnovskiy, L. A. (1999). Tribo-fatigue: The dialectics of life (2nd ed., p. 116). BelSUT Press. (In
Russian).
Sosnovskiy, L. A. (2004). L-risk (Mechanothermodynamics of irreversible damages) (p. 317).
BelSUT Press. (In Russian).
Sosnovskiy, L. A. (2005). Tribo-fatigue: Wear-fatigue damage and its prediction (Foundations of
engineering mechanics) (p. 424). Springer.
Sosnovskiy, L. A. (2007). Mechanics of wear-fatigue damage (p. 434). BelSUT Press.
Sosnovskiy, L. A. (2009). Life field and golden proportions. Nauka i Innovatsii, 79, 26–33. (In
Russian).
Sosnovskiy, L. A., & Sherbakov, S. S. (2012). Mechanothermodynamical system and its behavior.
Continuum Mechanics and Thermodynamics, 24, 239–256.
Sosnovskiy, L. A., & Sherbakov, S. S. (2015). Mechanothermodynamics. Springer.
Sosnovskiy, L. A., & Sherbakov, S. S. (2016). Mechanothermodynamic entropy and analysis of
damage state of complex systems. Entropy, 18(7), 268.
Sosnovskiy, L. A., & Sherbakov, S. S. (2017). A model of mechanothermodynamic entropy in
tribology. Entropy, 19, 115.
Sosnovskiy, L. A., & Sherbakov, S. S. (2019). On the development of mechanothermodynamics as
a new branch of physics. Entropy, 21(12), 1188.
Suhir, E. (2019). Failure oriented accelerated testing (FOAT) Boltzmann Arrnhenius Zhurkov
equation (BAZ) and their application in aerospace microelectronics and photonics reliability
engineering. International Journal of Aeronautical Science and Aerospace Research, 6(3),
185–191.
Swendsen, R. H. (2006). Statistical mechanics of colloids and Boltzmann’s definition of the
entropy. American Journal of Physics, 74, 187–190.
References 235
Tang, H., & Basaran, C. (2003). A damage mechanics based fatigue life prediction model. Trans-
actions of the ASME: Journal of Electronic Packaging, 125(1), 120–125.
Temfack, T., & Basaran, C. (2015). Experimental verification of a thermodynamic fatigue life
prediction model. Materials Science and Technology, 31(13), 1627–1632.
Turing, A. M. (1952). The chemical basis of morphogenesis. Philosophical Transactions of the
Royal Society of London B, 237(641), 37–72. Bibcode:1952RSPTB.237...37T. https://doi.org/
10.1098/rstb.1952.0012.%20JSTOR%2092463
Valanis, K. C. (1971). Irreversibility and existence of entropy. International Journal of Non-Linear
Mechanicsm, 6(3), 337–360.
Van Kampen, N. G. (1984). The Gibbs paradox. In W. E. Parry (Ed.), Essays in theoretical physics
in honour of Dirk ter Haar (pp. 303–312). Pergamon.
Valanis and Komkov (1980). Irreversible thermodynamics from the point of view of internal
variable theory /A Lagrangian formulation, Archiwum Mechaniki Stosowanej, vol. 32, no. 1,
1980, p. 33–58.
Wang, G. M., Sevick, E. M., Mittag, E., Searles, D.J. & Evans, D.J. Experimental Demonstration of
Violations of the Second Law of Thermodynamics for Small Systems and Short Time Scales.
Physical Review Letters 89, 050601, (2002)
Wang, J., & Yao, Y. (2017). An entropy based low-cycle fatigue life prediction model for solder
materials. Entropy, 19, 503.
Wang, J., & Yao, Y. (2019). An entropy-based failure prediction model for the creep and fatigue of
metallic materials. Entropy, 21(11), 1104.
Wang, T., Samal, S. K., Lim, S. K., & Shi, Y. (2019). Entropy production based full-chip fatigue
analysis: From theory to mobile applications. IEEE Transactions on Computer-Aided Design of
Integrated Circuits and Systems, 38(1), 84–95.
Whaley, P. W. (1983). A thermodynamic approach to metal fatigue. In Proceedings of ASME.
International conference on advances in life prediction methods, Albany, NY, pp. 18–21.
Yao, W., & Basaran, C. (2012). Electromigration analysis of solder joints under ac load: A mean
time to failure model. Journal of Applied Physics, 111(6), 063703.
Yao, W., & Basaran, C. (2013a). Computational damage mechanics of electromigration and
thermomigration. Journal of Applied Physics, 114, 103708.
Yao, W., & Basaran, C. (2013b). Electrical pulse induced impedance and material degradation in IC
chip packaging. Electronic Materials Letters, 9(5), 565–568.
Yao, W., & Basaran, C. (2013c). Electromigration damage mechanics of lead-free solder joints
under pulsed DC loading: A computational model. Computational Materials Science, 71,
76–88.
Young, C., & Subbarayan, G. (2019a). Maximum entropy models for fatigue damage in metals with
application to low-cycle fatigue of aluminum 2024-T351. Entropy, 21(10), 967.
Young, C., & Subbarayan, G. (2019b). Maximum entropy models for fatigue damage aluminum
2024-T351. Entropy, 21, xx.
Zhang, M.-H., Shen, X.-H., He, L., & Zhang, K.-S. (2018). Application of differential entropy in
characterizing the deformation inhomogeneity and life prediction of low-cycle fatigue of metals.
Materials, 11, 1917.
Chapter 5
Unified Mechanics of Thermomechanical
Analysis
5.1 Introduction
This chapter describes the implementation of the unified mechanics theory (UMT)
for inelastic thermomechanical analysis.
5.2 Unified Mechanics Theory-Based Constitutive

Modeling
5.2.1 Flow Theory and Yield Criteria
Newtonian Mechanics: Elastic Stress-Strain Relationship (Hooke’s Law)

For classical von Mises rate-independent plasticity, the elastic constitutive relation-
ship is given by Hooke’s law in the rate form as follows:
σ_ = C : ε_ e ð5:1aÞ
Assuming a small strain case, we can write the following equation:

ε_ e = ε_ - ε_ p - ε_ θ ð5:1bÞ
where ε_ ,_εp , and ε_ θ are total strain rate, inelastic strain rate, and thermal strain rate
vectors, respectively, and C is the elastic constitutive tensor for the virgin material.
In Eq. (5.1) “:” represents the inner product between the fourth-order constitutive
tensor C and the elastic strain rate vector, ε_ e .
https://doi.org/10.1007/978-3-031-18621-9_5
238 5 Unified Mechanics of Thermomechanical Analysis
There is no uniformity in terminology in the plasticity literature; “yield function,”

“yield surface,” “plastic potential function,” and “plastic flow function” are all used
interchangeably in different publications, essentially referring to the same thing, a
boundary between elastic and inelastic behavior of a material, but with different
formulations.
Von Mises yield criterion is the most basic function originally developed for
metals. According to this yield criterion, plastic deformation initiates when the
second invariant of the deviatoric stress tensor reaches a critical value. It is surmised
that yielding at a stress point begins when the distortional energy reaches a value that
is equal to distortional energy at yield in simple uniaxial tension.
J2D < k2 if the material is in the elastic state
J2D = k2 if the material is yielding
where k is a material constant proportional to the initial yield stress in uniaxial
tension.
In a three-dimensional state of stress, the von Mises criterion can be given by
h i
1 2 2
σ xx - σ yy þ σ yy - σ zz þ ðσ zz - σ xx Þ2 þ σ 2xy þ σ 2yz þ σ 2xz = k2
6
In the case of uniaxial tension, this equation becomes
1 2
2σ xx = k2
6
In a simple uniaxial tension case, yielding starts when the applied stress, σ xx, is
σ
equal to the yield stress of the material, σ xx = σ y. Hence, pyffiffi3 = k:
Yield stress of most metals during loading and unloading cycles increases, called
isotropic hardening, and the origin of the yield surface moves in the stress space,
called kinematic hardening. Considering the general hardening case, for most
metals, an elastic-plastic domain can be defined according to the following yield
function of the von Mises type:
rffiffiffi
2
F ðσ, αÞ = kS - Xk - K ðαÞ kS - Xk - RðαÞ ð5:2Þ
3
where F(σ, α) is a yield surface separating the elastic from the inelastic domain, σ is
the second-order stress tensor, α is a hardening parameter that specifies the evolution
of the radius of the yield surface, X is the deviatoric component of the back-stress
tensor describing the position of the center of the yield surface in stress space, S is
the deviatoric component of the stress tensor given by S = σ - 13 pI whereqpffiffi is the
hydrostatic pressure and I is the second-order identity tensor, and RðαÞ 2
3 K ðαÞ
is the radius of the yield surface in stress space and k k represents the norm operator.
A schematic representation of the yield surface is described in Fig. 5.1.
5.2 Unified Mechanics Theory-Based Constitutive Modeling 239
qffiffiffi
Fig. 5.1 Yield surface in the principal stress space. 2
3 σ y is the yield stress coordinate on the π
plane
Flow Rule
The evolution of the plastic strain is represented by a general associative flow rule of
the form
∂F
ε_ p = γ γb
n ð5:3Þ
∂σ
with bn = ∂F∂σ
being a vector normal to the yield surface in the stress space and
specifying the direction of plastic flow [this is also referred to as the normality
rule], ε_ p are the plastic strain rate vector, and γ is a nonnegative consistency
parameter, which is derived later in the chapter.
Isotropic Hardening
Isotropic hardening describes the increasing yield stress of the material, which is
represented by the radius of the yield surface defined in Eq. (5.2). Chaboche (1989)
proposed the following evolution empirical function for isotropic hardening in
metals:
rffiffiffi
2
K ðαÞ = σ þ R1 ð1 - e - cα Þ ð5:4aÞ
3 y0
where α is the plastic strain trajectory evolving according to Eq. (5.4b), σ y0 is the
initial yield stress in uniaxial tension, R1 is an isotropic hardening saturation value,
and c is an isotropic hardening rate material parameter:
rffiffiffiffiffiffiffiffiffiffiffiffiffi
2 p p
α_ = ε_ ε_ ð5:4bÞ
3
Using Eqs. (5.3) and (5.4b), we can write the standard definition of equivalent
plastic strain trajectory as follows:
Z t1
rffiffiffiffiffiffiffiffiffiffiffiffiffi
2 p p
α= ε_ ε_ dt ð5:4cÞ
t0 3
The Nonlinear Kinematic Hardening (NLK) Rule

For implementation, the NLK rule describing the movement of the center of the yield
surface in stress space proposed by Chaboche (1989) is used. The model is based on
the work of Armstrong and Frederick (1966). Nonlinearities are introduced as a
recall term to the Prager (1955) linear hardening rule given in Eq. (5.5) and where c1
and c2 are material parameters:
X_ = c1 ε_ p - c2 Xα_ ð5:5Þ
In Eq. (5.5) the first term represents the linear kinematic hardening rule as defined
by Prager (1955). The second term is a recall term, often called a dynamic recovery
term, which introduces the nonlinearity between the back-stress X and the actual
plastic strain. When c2 = 0, Eq. (5.5) reduces to the Prager (1955) linear kinematic
hardening rule. The NLK equation describes the rapid changes in the yield stress due
to the plastic flow during cyclic loadings and plays an important role even under
stabilized conditions (after saturation of cyclic hardening). The NLK rule given by
Eq. (5.5) considers the transient hardening effects in each stress-strain hysteresis
loop during cycling loadings, and after unloading, dislocation remobilization is
implicitly described due to the back-stress effect and the larger plastic modulus at
the beginning of the reverse plastic flow.
Consistency Parameter, γ
In Eqs. (5.3) and (5.4b), γ is a nonnegative consistency parameter representing the
irreversible character of plastic flow. The consistency parameter must satisfy the
following requirements:
1. For a rate-independent plasticity material model, γ obeys the so-called loading/

unloading and consistency condition:
γ ≥ 0 and F ðσ, αÞ ≤ 0 ð5:6Þ

γ F_ ðσ, αÞ = 0 ð5:7Þ
5.2 Unified Mechanics Theory-Based Constitutive Modeling 241
2. For a rate-dependent plasticity, viscoplasticity, material model conditions spec-

ified by Eqs. (5.6) and (5.7) are replaced by a constitutive equation of the form
h∅ðF Þi
γ= ð5:8Þ
η
where η represents a viscosity material parameter, h i is Macaulay bracket,

and ∅(F) is a material-specific function defining the character of the
viscoplastic flow. When η → 0 the constitutive model approaches the rate-
independent case (Simo & Hughes, 1997).
In the case of a rate-independent plasticity, material model F satisfies conditions
specified by Eqs. (5.6) and (5.7), and additionally stress states such as F(σ, α) > 0 are
ruled out. On the other hand, in the case of a rate-dependent plasticity material
model, the magnitude of the viscoplastic flow is proportional to the distance of the
stress state to the surface defined by F(σ, α) = 0.
Viscoplastic Strain Rate
The relation between γ and η expressed in Eq. (5.8) is a general constitutive equation,
and different forms of this constitutive relationship describing the material-specific
viscoplastic strain rate are used. There are many creep models, and mainly they can
be separated into two categories: diffusional creep models and dislocation creep
models. In the example here, the creep law proposed by Kashyap and Murty (1981)
and extended to the multiaxial case by Basaran et al. (2005) is used in this chapter for
the implementation:
n p
AD0 Eb hF i b - Q=Rθ ∂F
ε_ =
vp
e ð5:9Þ
kθ E d ∂σ
where A is a dimensionless material parameter, which is temperature- and rate-

dependent, Di = D0e-Q/Rθ is a diffusion coefficient with D0 representing a frequency
factor, Q is the creep activation energy, R is the universal gas constant, θ is the
absolute temperature in Kelvin, E(θ) is a temperature-dependent Young’s modulus,
b is the characteristic length of crystal dislocation (magnitude of Burger’s vector), k
is Boltzmann’s constant, d is the average grain size, p is a grain size exponent, and n
is a stress exponent for viscoplastic deformation rate. In Eq. (5.9) we can identify
p
kθ d
h∅ðFÞi = hFin and η = eQ=Rθ ð5:10Þ
AD0 En - 1 b b
For a more in-depth study of different viscoplastic models and a comparison of

most creep rate models published in the literature, readers are referred to Lee and
Basaran (2011).
5.2.2 Effective Stress Concept and Strain Equivalence

Principle
Rabotnov (1969) introduced the effective stress concept, and in the following, we
introduce the formulation of Lemaitre and Chaboche (1990) and Lemaitre (1996).
To introduce the effective stress concept, it is useful to consider a representative
volume element (RVE) of a material loaded by force F. At a point M oriented by a
→ →
plane defined by its normal direction n and its abscissa x along the direction n
→
(Fig. 5.2), the nominal uniaxial stress is σ = F/δS, where F = n F; δS is the area of
the intersection of the plane with the RVE.
The effective area of all micro-cracks and micro-cavities that lie in δS is
represented by δSDx. It is assumed that no forces are carried by the micro-cracks
and micro-cavities. It is convenient to introduce an effective stress concept related to
the surface that effectively carries the load, (δS - δSDx), namely, for the n axis, the
equation can be given by Rabotnov (1969):
F
σ~x = ð5:11Þ
δS - δSDn
The effective stress σ~ is higher than the nominal stress, because stresses are
carried by the undamaged [intact] material only.
Thermodynamics enables us to define the state at any point by a set of continuous
state variables. This postulate means that the constitutive equations written for the
surface of (δS - δSDn) are not modified by the damage in the material. Hence, the
true stress is the effective stress and not the nominal stress. As a result, we can
postulate the following strain equivalence principle by Lemaitre and Chaboche
(1990): “Any strain constitutive equation for a damaged material may be derived
in the same way as for an intact material except that the usual stress is replaced by the
effective stress.”
Fig. 5.2 Representative

volume element. (After
Lemaitre and Chaboche
(1990))
5.3 Return Mapping Algorithms 243
5.2.3 Unified Mechanics Theory Implementation
Based on the third law of the unified mechanics theory, the strain energy density rate
can be given by
1 1
U_ = σ_ ε_e = ð1 - ΦÞC : ½ε_ e 2 ð5:12Þ
2 2
where Φ is the thermodynamic state index, C is the tangential constitutive tensor,

and ε_ e is the elastic strain rate vector. Taking the derivative with respect to strain rate
and ignoring the derivative of strain energy rate with respect to thermodynamic state
index, Φ, for the simplified version of formulation, yields
σ_ = ð1 - ΦÞ C : ε_ e ð5:13Þ
Of course, ignoring the derivative of Φ with respect to strain rate is not accurate.
However, for any incremental procedure, the ignored portion is smaller by orders of
magnitude because of the square over strain rate. Moreover, our objective here is to
introduce the simplest implementation of unified mechanics theory. For small strain
problems, the elastic strain rate can be calculated from the total strain rate vector by
subtracting the inelastic and thermal components:

σ_ = ð1 - ΦÞC : ε_ - ε_ vp - ε_ θ ð5:14Þ
rffiffiffi
2
F = S - XΦ - ð1 - ΦÞ K ðαÞ S - X Φ - ð1 - ΦÞRðαÞ ð5:15Þ
3
The evolution of the back-stress, including degradation effects, is given by
Φ
X_ = ð1 - ΦÞðc1 ε_ vp - c2 Xα_ Þ ð5:16Þ
It is important to point out that in unified mechanics theory, we only care about
degradation along the thermodynamics state index axis. The kinematic hardening
function given in Eq. (5.16) is empirical and hence can be defined in many other
forms.
5.3 Return Mapping Algorithms
Material nonlinear finite element method implementation requires more sophisti-

cated solution methods compared to linear elastic analysis. A return mapping
algorithm is a particular form of a solution algorithm based on a combination
between an explicit method and an implicit method. The explicit method makes an
initial approximation to the solution. The approximated solution is then used in the
implicit method to improve the prediction. The following trial (elastic predictor)
state can be written as
Strnþ1 = Sn þ ð1 - ΦÞ 2GΔenþ1 ð5:17Þ
where G is the shear modulus and Δen+1 is the deviatoric strain increment vector.
The increment of the back-stress can then be computed using Eq. (5.16):
Φ
dX Φ vp Φ
nþ1 = ð1 - ΦÞ c1 dεnþ1 - c2 Δγ βX n þ ð1 - βÞX nþ1 ð5:18Þ
qffiffi
where c02 = 2
3 c2 and a generalized midpoint rule for the recall term with the
extreme values of β = 0 and β = 1 corresponds to the backward and forward
Euler methods, respectively. Utilizing Eqs. (5.3) and (5.14), incremental form of
the viscoplastic strain can be written as
Snþ1 - X Φ
dεvp
nþ1 = Δγ
nþ1
ð5:19Þ
Snþ1 - XΦnþ1
Substitution of Eq. (5.19) into Eq. (5.18) yields any value of the integration
parameter β between 0 and 1:
" #
Snþ1 - X Φ c02 Φ
dX Φ
nþ1 = anþ1 Δγ
nþ1
- X ð5:20Þ
Snþ1 - X Φ
nþ1
c1 n
c1 ð1 - ΦÞ
with anþ1 = 1þc02 ð1 - ΦÞð1 - βÞΔγ .
Using the flow rule expressed in Eq. (5.3) allows us to express Eq. (5.17) as
Snþ1 - X Φ
Snþ1 = Strnþ1 - Δγ ð1 - ΦÞ2G nþ1
ð5:21Þ
Snþ1 - X Φ
nþ1
Φ
And introducing the relative stress tensor in ξD
nþ1 = Snþ1 - X nþ1 yields
Snþ1 - X Φ
ξΦ Φ
nþ1 = Snþ1 - X nþ1 Snþ1 - Δγ ð1 - ΦÞ2G
tr nþ1
- XΦ Φ
n - dX nþ1 ð5:22Þ
Snþ1 - X Φ
nþ1
Substituting Eq. (5.20) into Eq. (5.22) yields

Snþ1 - XΦ
Snþ1 - X Φ
nþ1 þ Δγ ½ð1 - ΦÞ2G þ anþ1
nþ1
= Bn ð5:23Þ
Snþ1 - XΦ
nþ1
c0
where Bn = Strnþ1 - X Φ n þ bnþ1 ΔγX n and bnþ1 = c1 anþ1 .
2
The normal to the yield surface can be expressed in terms of the initial values of
the stress; the state variables and the strain increment at each step can be defined as
follows:
Snþ1 - X Φ Bn
n_ nþ1 nþ1
= ð5:24Þ
Snþ1 - X Φ
nþ1
k Bnk
Taking the trace product of Eq. (5.22) yields
Snþ1 - X Φ
nþ1 þ Δγ ½ð1 - ΦÞ2G þ anþ1
n
Sn - X Φ
2
þ ð1 - ΦÞ2GΔenþ1 þ bnþ1 ΔγX Φ
2
n n ð5:25Þ
=

þ 2 Sn - X Φ Φ 1=2
n : ð1 - ΦÞ2GΔenþ1 þ bnþ1 ΔγX n g
Using Eq. (5.7) for the rate-independent case or Eq. (5.8) for the rate-dependent
case results in the following nonlinear scalar equation for the consistency parameter
γ which can be solved by a local Newton method:
gðΔγ Þ
n
2 2
Sn - X Φ
n þ ð1 - ΦÞ2GΔenþ1 þ bnþ1 ΔγX Φn
Φ

Φ 1=2
= þ 2 Sn - Xrn : ð1 - ΦÞ2GΔenþ1 þ bnþ1 ΔγX n g
ffiffiffi rffiffiffi !
2 2 Δγη
- ð1 - ΦÞ K αn þ Δγ - Δγ ½ð1 - ΦÞ2G þ anþ1 - Θ
3 3 Δt
ð5:26Þ
After solving Eq. (5.26) for Δγ, the following updating schemes are needed:
pffiffiffiffiffi
αnþ1 = αn þ 2=3Δγ ð5:27Þ
Bn
εvp
nþ1 = εn þ Δγ
vp
ð5:28Þ
kBn k
" #
Snþ1 - XΦ c0
XΦ Φ
nþ1 = X n þ anþ1 Δγ
nþ1
Φ - 2 XΦ ð5:29Þ
Snþ1 - Xnþ1 c1 n
Bn
ξΦ
nþ1 = ð1 - ΦÞK ðαnþ1 Þ ð5:30Þ
kBn k
Snþ1 = ξΦ Φ
nþ1 þ X nþ1 ð5:31Þ

Bn θ
σ nþ1 = ҡð1 - ΦÞtrðεnþ1 Þ I þ 2Gð1 - ΦÞ enþ1 - εvp - γ - e ð5:32Þ
n nþ1
kB n k nþ1
5.3.1 Linearization (Consistent Jacobian)
Differentiating Eq. (5.32) with respect to the total strain but not entropy at the end of
the step yields
O ∂Δγ O ∂b
nnþ1
dσ nþ1 = ð1 - ΦÞ C - 2G b
nnþ1 - 2GΔγ : dεnþ1 ð5:33Þ
∂εnþ1 ∂εnþ1
Ignoring differentiation with respect to entropy is not correct. If the strain

increments are small, then the ignored part of the differentiation is smaller by an
order of magnitude. However, our objective here is to present an earlier and simpler
version of the UMT
N formulation.
The operator represents the ordered (dyadic) outer product and bnnþ1 is the
normal vector to the yield surface as d4efined in Eq. (5.3) evaluated at the end of the
∂Δγ
increment, and ∂ε nþ 1
can be calculated from Eq. (5.25):
∂Δγ b
n
= nþ1 ð5:34Þ
∂εnþ1 K3
Using the following definitions:
K3 = K1 þ K2
K0 anþ1
K1 = 1 þ þ
3G 2Gð1 - ΦÞ
a0nþ1 Δγ b
n bnþ1 1 ∂Θ
K2 = þ nþ1 ½b ð1 - βÞΔγ - 1 : XΦ
n þ
2Gð1 - ΦÞ 2Gð1 - ΦÞ nþ1 2Gð1 - ΦÞ ∂Δγ
∂b
nnþ1
In addition, ∂εnþ1
can be obtained from Eq. (5.24) as
∂b
nnþ1 ∂bnnþ1 ∂Bn 1 O ∂Bn
= → I -b
nnþ1 b
nnþ1 : ð5:35aÞ
∂εnþ1 ∂Bn ∂ε nþ1 k Bnþ1 k ∂ε nþ1
and
∂Bn O O ∂Δγ
b - 1I
= 2μð1 - ΦÞ Π I þ b0nþ1 Δγ þ bnþ1 Xn
∂εnþ1 3 ∂εnþ1
where Π b is fourth-order unit tensor mapped into the unit matrix and I second-order
identity tensors mapped into unit column vector and K′ denotes the derivative with
respect to the argument in K(α).
Defining K 4 = b0nþ1 Δγ þ bnþ1 and substituting ∂Bn in Eq. (5.35a) yields
∂εnþ1
∂b
nnþ1 ∂bnnþ1 ∂Bn 2Gð1 - ΦÞ b O 1 O
= Π-b
nnþ1 b
nnþ1 - I I
∂εnþ1 ∂Bn ∂εnþ1 kBn k 3
1
þ
kBn k
O O
K4
I-b nnþ1 b
nnþ1 : b
nnþ1 XΦ
n ð5:35bÞ
K3
Using Eqs. (5.34) and (5.35b) into Eq. (5.33) results in the following algorithmic
version of the material Jacobian coupling the effects of damage and rate dependency:
O O
b - 1I
nþ1 = ð1 - ΦÞҡI
CEVPD I þ 2Gð1 - ΦÞδnþ1 Π
3
I
O 2Gð1 - ΦÞ O
- 2Gð1 - ΦÞθnþ1bnnþ1 b
nnþ1 - Δγ Πb -b
nnþ1 b
nnþ1
kBn k
O
4
K
: b
nnþ1 XΦn ð5:35cÞ
K3
where ҡ is the bulk modulus and
Δγ2Gð1 - ΦÞ 1 Δγ2Gð1 - ΦÞ
δnþ1 = 1 - and θnþ1 = -
kB n k K3 kBn k
The pseudocode corresponding to the above algorithm is detailed in Table 5.1.

Moreover, Table 5.2 shows the local Newton-Raphson algorithm used to solve for
the consistency parameter, which preserves the quadratic rate of convergence of the
Newton scheme.
Table 5.1 Return mapping algorithm-classical theory rate-dependent model

Update strain εn + 1 = εn +— Δμ
nþ1 = Sn þ ð1 - Φn Þ2GΔenþ1
Str
Compute trial state
qffiffi
Compute trial yield function
nþ1 = Snþ1 - X n - ð1 - ΦÞ 3 K ðα n Þ
tr D
Ftr 2
nþ1 > 0 THEN

IF Ftr
n local and solve g(Δγ) = 0 for Δγ
Call Newton

2 2
gðΔγ Þ = Sn - X Φn þ ð1 - ΦÞ2GΔenþ1 þ bnþ1 ΔγX Φ
n þ 2 Sn - X Φ
n :
1=2
ð1 - ΦÞ2GΔenþ1 þ bnþ1 ΔγXΦ
n
qffiffi qffiffi
- ð1 - ΦÞ 23K αn þ 23Δγ - Δγ ½ð1 - ΦÞ2G þ anþ1 - Θ Δγη Δt
c1 ð1 - ΦÞ
where anþ1 = 1þc02 ð1 - ΦÞð1 - βÞΔγ , bnþ1 = c2
c1 anþ1
Update
Snþ1 - X Φ Φ Φ
nnþ1 nþ1
= Bn
where Bn = Str
nþ1 - X n þ bnþ1 ΔγX n
kSnþ1 -
qffiffinþ1 k
XΦ kBn k
αnþ1 = αn þ 3Δγ
2
εvp = εvp þ Δγ kBBnn k

nþ1 n

Snþ1 - X Φ c02
XΦ Φ
nþ1 = X n þ anþ1 Δγ
nþ1
- XΦ
k Snþ1 - X Φ
nþ1 k c1 n
ξΦ
nþ1 ð1 - ΦÞK ðα nþ1 Þ Bn
kBn k
Snþ1 = ξΦ Φ
nþ1 þ X nþ1
σ nþ1 = ҡð1 - ΦÞtrðεnþ1 ÞbI þ 2Gð1 - ΦÞ enþ1 - εvp θ
n - γ nþ1 kBn k - enþ1
Bn
Compute consistent Jacobian

N
b bI þ 2Gð1 - ΦÞδnþ1
Q 1 Ob
nþ1 = ð1 - ΦÞҡI
CEVPD - bI I
3
N
nnþ1 b
- 2Gð1 - ΦÞθnþ1 b nnþ1
Q N N
- 2Gkð1B-
nk
ΦÞ
Δγ ð - bnnþ1 b nnþ1 X Φ
nnþ1 Þ : KK4 b n 3
Δγ2Gð1 - ΦÞ Δγ2Gð1 - ΦÞ
Where δnþ1 = 1 - kBn k and θnþ1 = 1
K3
- kBn k with
K3 = K1 þ K2
K0
K 1 = 1 þ 3G þ 2Gða1nþ1
- ΦÞ
0
anþ1 Δγ bnnþ1 bnþ1
K 2 = 2Gð1 - ΦÞ þ 2Gð1 - ΦÞ ½bnþ1 ð1 - βÞΔγ - 1 : X Φ
n þ 2Gð1 - ΦÞ
1 ∂Θ
∂Δγ
K 4 = b Δγ þ bnþ1
0
nþ1
ELSE
Elastic step ð Þnþ1 = ð Þtr nþ1 (Exit)
END IF
EXIT
5.4 Thermodynamic Fundamental Equation in Thermomechanical Problems 249
Table 5.2 Local Newton iteration for the consistency parameter-classical theory
Let Δγ (0) ← 0
αn + 1(0) ← αn
Start Iterations
DO_UNTIL |g(Δγ)| < tol
k←k+1
Compute nΔγ (k + 1)
2 2
gðΔγ Þ = Sn - X Φn þ ð1 - ΦÞ2GΔenþ1 þ bnþ1 ΔγX Φ
n þ 2 Sn - X Φ
n :
1=2
ð1 - ΦÞ2GΔenþ1 þ bnþ1 ΔγXΦ
n
qffiffi qffiffi
- ð1 - ΦÞ 23K αn þ 23Δγ - Δγ ½ð1 - ΦÞ2G þ anþ1 - Θ Δγη Δt
∂gðΔγ ðkÞ Þ
dg Δγ ðkÞ = ∂Δγ ðkÞ
ðkþ1Þ ðkÞ gðΔγ ðkÞ Þ
Δγ ← Δγ - dgðΔγðkÞ Þ
END DO_UNITL

in Thermomechanical Problems
We have defined thermodynamic fundamental relations [equation] in the earlier

chapter on thermodynamics. However, we feel it is necessary to restate some basics
again to make it easier to understand the derivation for the readers. We find it helpful
to quote these basics directly from DeHoff (1993).
In irreversible thermodynamics, the so-called balance equation for the entropy
generation plays a central role. This equation expresses the fact that the entropy of a
volume element changes with time for two reasons. First, it changes because entropy
flows into the volume element, and second it changes because there is an entropy
generation due to irreversible phenomena inside the matter. The internal entropy
generation is always a nonnegative quantity, since entropy can only be created, but
never destroyed. This contrasts with energy that cannot be created.
We find it helpful to quote DeHoff (1993):
Thermodynamic entropy is the measure of how much energy is unavailable for work.
Imagine an isolated and closed system with some hot objects and some cold objects.
Work can be done as heat is transferred from the hot to the cooler objects; however, once
this transfer has occurred, it is impossible to extract additional work from them alone.
Energy is always conserved, but when all objects have the same temperature, the energy is
no longer available for conversion into work.
The entropy [energy unavailable for work or disorder] of the universe increases or remains
constant in all-natural processes. It is possible to find a system for which entropy decreases
but only due to a net increase in a related system. For example, the originally hot objects and
cooler objects reaching thermal equilibrium in an isolated system may be separated, and
some of them put in a refrigerator. The objects would again have different temperatures after
some time, but now the system of the refrigerator would have to be included in the analysis
of the complete system. No net decrease in entropy of all the related systems occurs. This is
yet another way of stating the second law of thermodynamics. (DeHoff, 1993)
The concept of entropy has far-reaching implications that tie the order/disorder of our
universe to probability and statistics. Imagine a new deck of cards in order by suits, with
each unit in numerical order. As the deck is shuffled, no one would expect the original order
to return. There is a probability that the randomized order of the shuffled deck would return
to the original format, but it is exceedingly small. An ice cube melts, and the molecules in the
liquid form have less order than in the frozen form. An infinitesimally small probability
exists that all of the slower-moving molecules will aggregate in one space so that the ice cube
will reform from the pool of water. The entropy, or disorder, of the universe increases as hot
bodies cool and cold bodies warm. Eventually, the entire universe will be at the same
temperature so the energy will be no longer usable. (DeHoff, 1993)
To relate the entropy generation directly to irreversible processes that occur in a

system, one needs the macroscopic conservation laws of mass, momentum, and
energy in local form, i.e., differential form. The conservation laws are quantified
utilizing information from processes such as diffusion, heat flow, and velocity,
which are related to the transport of mass, exchange of energy, and exchange of
momentum, respectively. Then, the entropy generation can be calculated by using
the thermodynamics Gibbs relation, which connects the rate of the change in entropy
in the medium to the rate of the change in energy and work. Entropy generation rate
has a relatively simple formula: it is a sum of all entropy generating micro-
mechanism terms, each being a product of a flux characterizing an irreversible
process, and a quantity called thermodynamic force, which is related to the gradient
[non-uniformity] in the system (Mazur & De Groot, 1962). We discussed this topic
earlier in Chap. 3 in detail in Onsager relations, as well. The complete entropy
generation rate equation can thus serve as a basis for the systematic description of the
irreversible processes occurring in a system. We refer to this equation as the
thermodynamic fundamental equation.
“As yet, the set of conservation laws, together with the entropy balance equation
and the equations of state are to a certain extent empty, since this set of equations
contains the irreversible fluxes as unknown parameters and can therefore not be
solved with the given initial and boundary conditions for the state of the system”
(Mazur & De Groot, 1962). At this point, we must therefore supplement the
equations with an additional set of phenomenological relationships [Onsager recip-
rocal relations], which relate the irreversible fluxes and the thermodynamic forces
appearing in the entropy source strength. Irreversible thermodynamics, in its present
form, is mainly restricted to the study of the linear relationship between the fluxes
and the thermodynamic forces as well as possible cross-effects between various
phenomena. This is not a very serious restriction, however, since even rather extreme
physical situations are still described by linear laws.
Together with the phenomenological equations, the original set of conservation laws may be
said to be complete in the sense that one now has a consistent set of partial differential
equations for the state parameters of a material system, which may be solved with the proper
initial and boundary conditions. (Mazur & De Groot, 1962)
5.5 Conservation Laws 251
5.5 Conservation Laws
Thermodynamics is based on two fundamental laws: the first law of thermodynamics

or the law of conservation of energy and the second law of thermodynamics or the
entropy law. A systematic macroscopic scheme for the description of irreversible
processes must also be built upon these two laws. However, it is necessary to
formulate these laws in a suitable way for use in continuum mechanics. Since we
wish to develop a theory applicable to systems in which the properties are contin-
uous functions of space coordinates and time, we need a local formulation of the law
of conservation of energy. As the local momentum and mass densities may change in
time, we will also need local formulations of the laws of conservation of momentum
and conservation of mass. In solid mechanics, the thermodynamic system is usually
chosen as a collection of continuous matter, i.e., the system is closed and does not
interchange matter with its surroundings, and the bounding surface of the system in
general moves with the flow of matter.
5.5.1 Conservation of Mass
Consider an arbitrary volume V fixed in space, bounded by surface Ω. The rate of

change of the mass within the volume V is given by, using the notation used by
Malvern (1969)
Z Z
d ∂ρ
V ρdV = V dV ð5:36Þ
dt ∂t
where ρ is the density (mass per unit volume). If no mass is created or destroyed
inside V, this quantity must be equal to the rate of the matter flow into the volume
V through its surface Ω:
Z Z
∂ρ
V dV = Ω ρ ~v dΩ ð5:37Þ
∂t
where ~v is the velocity and dΩ is a vector with magnitude |dΩ| normal to the surface
and defined as positive from the inside toward the outside. The quantities ρ and v are
all functions of time and space coordinates. Applying Gauss’s theorem to the surface
integral in equation (5.37), we obtain
∂ρ
= - ρ div ~ν ð5:38Þ
∂t
Equation (5.38) is valid for any arbitrary volume V, which expresses the fact that
the total mass is conserved, i.e., the total mass in any volume element of the system
can only change if matter flows into (or out of) the volume element. This equation
has the form of a so-called balanced equation: the local change of the density is equal
to the negative divergence of the flow of mass. The continuity equation in the vector
form of equation (5.38) is independent of any choice of coordinate system.
The conservation of mass equation can also be written in an alternative form by
introducing the substantial time derivative (Mazur & De Groot, 1962):
d ∂
= þ ~ν grad ð5:39Þ
dt ∂t
The substantial time derivative is the rate of change of a quantity (i.e., mass) as
∂
measured by an observer moving along with the flow of matter. The first term ∂t
accounts for the time rate of change according to a stationary observer, and the
second term ð~ν gradÞ accounts for the movement of the observer.
With the help of Eq. (5.39), we can rewrite Eq. (5.38) as follows:
dρ
= - ρ div ~v ð5:40Þ
dt
With the specific volume ~v = ρ - 1 , formula (5.40) may also be written as
d~v
ρ = div ~v ð5:41Þ
dt
Finally, the following substantial time derivative relation is valid for an arbitrary
local property a that may be a scalar or a component of a vector or tensor:
da ∂aρ
ρ = þ div aρ~v ð5:42Þ
dt ∂t
which is a consequence of Eqs. (5.38) and (5.39). We can verify Eq. (5.42) directly.
According to Eq. (5.39), the left-hand side of Eq. (5.42) is
da ∂a
ρ =ρ þ ρ~v grad a ð5:43Þ
dt ∂t
According to Eq. (5.38), the right side of Eq. (5.42) is
∂aρ ∂ρ ∂a
þ div aρev = a þρ þ a div ρev þ ρev grad a
∂t ∂t ∂t
∂a ð5:44Þ
= að - div ρevÞ þ ρ þ a div ρev þ ρev grad a
∂t
∂a
=ρ þ ρev grad a
∂t
Therefore, Eq. (5.42) is true.

5.5.2 Momentum Principle in Newtonian Mechanics
The momentum principle for a collection of particles states that the time rate of the
change in the total momentum for a given set of particles equals the vector sum of all
the external forces acting on the particles of the set provided Newton’s Third Law of
action and reaction governs the initial forces (Malvern, 1969). Consider a given mass
of the medium, instantaneously occupying a volume V bounded by surface A and
acted upon by external surface tractions σ (n) and body force b. Then the momentum
principle can be expressed as follows:
Z Z Z
d
A σ ðnÞ dAþ V ρ b dV = V ρ ~v dV ð5:45Þ
dt
or in rectangular coordinates:
Z Z Z
d
A σ i dAþ V ρbi dV = V ρ v~i dV ð5:46Þ
dt
Substituting σ i = σ jinj and transforming the surface integral to a volume integral

by using the divergence theorem, we obtain
Z
∂σ ji dv~i
V þ ρbi - ρ dV = 0 ð5:47Þ
∂xj dt
Equation (5.47) must be true for any arbitrary volume V. At each point, we have
d~vi ∂σ ji
ρ = þ ρbi ð5:48Þ
dt ∂xj
where nj is the component of the normal unit vector n, v~i ði = 1, 2, 3Þ is a Cartesian

component of velocity ~v , and xj ( j = 1, 2, 3) is the Cartesian coordinates. The
quantities σ ji (i, j = 1, 2, 3) and bi (i = 1, 2, 3) are the Cartesian components of the
stress tensor σ and body force b, respectively. For a nonpolar case, the stress tensor σ
is symmetric, namely,
σ ij = σ ji ði, j = 1, 2, 3Þ ð5:49Þ
In tensor notation, equation (5.48) is written as
d~v
ρ = div σ þ ρb ð5:50Þ
dt
From a microscopic point of view, the stress tensor σ results from the short-range
interactions, [strain] between the particles of the system, whereas b contains the
external forces as well as a possible contribution from long-range interactions in the
system.
Using Eq. (5.46), the equation of motion Eq. (5.50) can also be written as
∂ρ~v
= - div ðρ~v~v - σÞ þ ρb ð5:51Þ
∂t
N
where ~v~v = ~v ~v is an ordered (dyadic) product. This equation also has the form of a
balance equation for the momentum density ρ~v . One can interpret the quantity
ðρ~v~v - σÞ as a momentum flow with a convective part ρ~v~v , and the quantity ρb as
a source of momentum, but there is no entropy generation part.
It is also possible to derive from Eq. (5.48) a balance equation for the kinetic
energy of the center of gravity motion by multiplying both members with the
components of ~v and then summing over i:
d 12 ~v2 X3 ∂ X3 ∂
ρ = i,j = 1 ∂xj
σ ji ~
v i - σ v~ þ ρbi v~i
i,j = 1 ji ∂xj i
ði = 1, 2, 3Þ ð5:52Þ
dt
or in tensorial notation:
d 12 ~v2
ρ = divðσ ~vÞ - σ : L þ ρb ~v ð5:53Þ
dt
where L = grad~v is the spatial gradient of the velocity. L can be written as the sum of
a symmetric tensor D, called the rate of deformation tensor and a skew-symmetric
tensor W called the spin tensor (or the vorticity tensor) as follows:
L=D þ W ð5:54Þ

where D = 12 L þ LT , and W = 12 L - LT .
Since W is skew-symmetric, while σ is symmetric, it follows that
σ : grad ~v = σ ij Lij = σ ij Dij = σ : D ð5:55Þ
We can also establish the relationship between the strain rate dε/dt and the rate of
the deformation tensor D:
dε
= FT D F ð5:56Þ
dt
where F is the deformation gradient tensor referring to the undeformed configura-

tion. When the displacement gradient components are small compared to unity,
equation (5.56) is reduced to
dε
≈D ð5:57Þ
dt
With the help of equations (5.42) and (5.53) become
∂ 12 ρ~v2 1
= - div ρ~v2 ~v - σ ~v - σ : D þ ρb ~v ð5:58Þ
∂t 2
For the conservative body forces that can be derived from a potential Ψ indepen-
dent of time, we can write (Mazur & De Groot, 1962)
∂Ψ
b = - gradΨ, =0 ð5:59Þ
∂t
We can now establish an equation for the rate of change of the potential energy
density ρΨ. It follows from Eqs. (5.38) and (5.59) that
∂ρΨ ∂ρ ∂Ψ
=Ψ þρ = Ψð - div ρevÞ
∂t ∂t ∂t ð5:60Þ
= - div ρΨev þ ρev gradΨ = - div ρΨev - ρb ev
Adding Eqs. (5.59) and (5.60) for the rate of change of the kinetic energy 12 ρ~v2
and the potential energy ρΨ:
n o
∂ρ 12 ~v2 þ Ψ 1
= - div ρ ~v2 þ Ψ ~v - σ ~v - σ : D ð5:61Þ
∂t 2
This equation shows that the sum of kinetic and potential energy is not conserved,
since an entropy source term appears on the right-hand side. However, this entropy
source term is one of many. There are far more entropy source terms in
thermomechanical loading. It is covered in more detail in the following chapter.
Moreover, assuming the potential to be time-independent significantly simplified the
formulation, albeit unrealistically.
5.5.3 Conservation of Energy
The first law of thermodynamics relates the work done on the system and the heat
energy flowing into the system to the change in total energy of the system. Suppose
that the only energy transferred to the system is by mechanical work done on the
system by surface tractions and body forces, by heat exchange through the boundary
surface, and the heat generated within the system by external agencies (e.g., induc-
tive heating). According to the principle of conservation of energy [the first law of
thermodynamics], the total energy within an arbitrary volume V in the system can
only change if energy flows into (or out of) the volume considered through its
boundary A, which can be expressed as
Z V Z V Z A Z V
d ∂ρe
ρ e dV = dV = - Je dA þ ρ r dV ð5:62Þ
dt ∂t
where e is the energy per unit mass, Je is the energy flux per unit surface and unit
time, and r is the distributed internal heat source of strength per unit mass. We will
refer to e as the total specific energy because it includes all forms of energy in the
system. Similarly, we will call Je the total energy flux. With the help of Gauss’s
theorem, we can obtain the differential [local form] of the law of conservation of
energy as follows:
∂ρe
= - div Je þ ρr ð5:63Þ
∂t
To relate this equation to the previously obtained Eq. (5.61) for the kinetic energy
and potential energy, we must specify what are the various contributions to the total
specific energy e and the total energy flux Je. The total specific energy e includes the
specific kinetic energy 12 v2 , the specific potential energy Ψ, and the specific internal
energy u (Mazur & De Groot, 1962):
1
e = ~v2 þ Ψ þ u ð5:64Þ
2
From a macroscopic point of view, this relation can be considered as the defini-
tion of internal energy, u. From a microscopic point of view, u represents the energy
of thermal agitation [atomic vibrations] as well as the energy due to the short-range
atomic interactions, such as the strain energy.
Similarly, the total energy flux includes a convective term ρe~v , an energy flux
ðσ ~vÞ due to the mechanical work performed on the system, and finally a heat flux
Jq (Mazur & De Groot, 1962):
Je = ρe~v - σ ~v þ Jq ð5:65Þ
This equation may be also considered as defining the heat flux Jq. Then the heat
flow rate per unit mass is given by
dq
ρ = - div Jq ð5:66Þ
dt
where q is the heat flowing into the system per unit mass. If we subtract Eq. (5.61)
from Eq. (5.63) and use Eqs. (5.64) and (5.65), we get the balance equation for the
internal energy u:
∂ρu
= - div ρu~v þ Jq þ σ : D þ ρr ð5:67Þ
∂t
5.6 Entropy Law: Second Law of Thermodynamics 257
It is apparent from equation (5.67) that the internal energy u is not conserved. A
dissipation source term appears which is equal but of opposite sign to the source term
of the balance equation (5.61) for kinetic and potential energy.
With the help of Eqs. (5.42), (5.67) may be written in an alternative form as
follows:
du
ρ = - div Jq þ σ : D þ ρr ð5:68Þ
dt
The total stress tensor σ can be split into a scalar hydrostatic pressure part p and a
deviatoric stress tensor S:
σ = S - pI ð5:69Þ
where I is the unit matrix with element δij(δij = 1, if i = j; δij = 0, if i ≠ j) p = - 13 σ kk.

With the help of Eqs. (5.69), (5.68) becomes
du
ρ = - div Jq - p div ~v þ S : D þ ρr ð5:70Þ
dt
where the following equality is utilized:
X3 ∂ X3 ∂
I : D = I : grad ~v = δ
i:j = 1 ij
~vi = v~ = div ~v
i = 1 ∂xi i
ð5:71Þ
∂xj
Utilizing Eq. (5.41), the time rate of change of internal energy can be written in
the following form:
du 1 1
= σ : D þ r - div Jq ð5:72Þ
dt ρ ρ
5.6 Entropy Law: Second Law of Thermodynamics
Historically, thermodynamics in the traditional sense was concerned with the study
of reversible transformations [which are imaginary]. For an irreversible process in
which the thermodynamic state of a matter changes from some initial state to a
current state, it can be assumed that such a process can occur along an imaginary
reversible isothermal path. The processes defined in this way will be thermodynam-
ically admissible if, at any instant of evolution, the Clausius-Duhem inequality is
satisfied. According to the principles of thermodynamics, two more new variables,
temperature T and entropy S, are introduced for any macroscopic system. The
entropy [disorder/energy unavailable for work] of the universe, taken as a system
plus whatever surroundings are involved in producing the change within the system,
can only increase. Everything that happens in the real world is always an irreversible
process, which results in the production of positive entropy and thus a permanent/
irreversible change in the universe (DeHoff, 1993).
The variation of the entropy dS may be written as the sum of two and only two
terms for a closed system (Mazur & De Groot, 1962):
dS = dSe þ dSi ð5:73Þ
where dSe is the entropy generated by the transfer of heat from external sources
across the boundary of the system and dSi is the entropy produced inside the system
and by the matter. The second law of thermodynamics states that dSi must be zero for
any reversible process and positive for the irreversible transformation of the system
from one state to any other state (Mazur & De Groot, 1962):
dSi ≥ 0 ð5:74Þ
The entropy supplied externally, dSe, on the other hand, may be positive, zero, or
negative, depending on the interaction of the system with its surroundings.
For an irreversible process in which the thermodynamic state of a matter changes
from some initial state to a different state, it is assumed that such a process can occur
along an imaginary reversible isothermal path which consists of a two-step sequence
(Krajcinovic, 1996). This is the so-called local equilibrium assumption, which
postulates that the thermodynamic state of a material at a given point and instant is
completely defined by the knowledge of the values of a certain number of state
variables at that instant. The method of local state implies that the laws which are
valid for the macroscopic system remain valid for infinitesimally small parts of it,
which agrees with the point of view currently adopted in the macroscopic description
of a continuous system. However, we need to clarify what we mean by “infinites-
imally small,” which is relative. We mean that small enough for laws of continuum
mechanics to be applicable but not smaller. This method also implies, on a micro-
scopic model, that the local macroscopic measurements performed on the system are
measurements of the properties of small parts of the system, which still contain many
of the constituting particles. In general, it is assumed that representative volume
element (RVE) is the smallest unit of a continuum. This hypothesis of “local
equilibrium” can, from a macroscopic point of view, only be justified by the validity
of the conclusions derived from it. Ultrarapid phenomena for which the time scale of
the evolutions is in the same order as the atomic relaxation time for a return to
thermodynamic equilibrium are excluded from this theory’s field of application
(Lemaitre & Chaboche, 1990). As we discussed in Chap. 3, all physical processes
can be described with precision by utilizing the proper number of thermodynamic
state variables. The processes defined in this way will be thermodynamically admis-
sible if, at any instant of evolution, the Clausius-Duhem inequality is satisfied.
In irreversible thermodynamics, one of the important objectives is to relate the

dSi, the internal entropy generation, to the various irreversible phenomena which
occur inside the matter. Before calculating the entropy production in terms of
quantities that characterize the irreversible phenomena, we can rewrite Eqs. (5.73)
and (5.74) in a form that is more suitable for the mathematical description of the
systems in which the densities of the extensive properties (such as mass and energy,
which are needed in conservation laws) are continuous functions of spatial coordi-
nates (Mazur & De Groot, 1962):
Z V
S= ρsdV ð5:75Þ
Z A
dSe
=- JS,tot dA ð5:76Þ
dt
Z V
dSi
= γ dV ð5:77Þ
dt
where s is the specific entropy per unit mass, JS, tot is the total entropy flux which is a
vector that coincides with the direction of entropy flow and has a magnitude equal to
the entropy crossing unit area perpendicular to the direction of flow per unit time,
and γ is the entropy generation rate per unit volume and per unit time.
Utilizing Eqs. (5.75)–(5.77), and Gauss’s theorem, Eq. (5.73) may be written in
the following form (Mazur & De Groot, 1962):
Z V
∂ρs
þ div JS,tot - γ dV = 0 ð5:78Þ
∂t
where the divergence of JS, tot simply represents the net entropy leaving unit volume
per unit time. Equation (5.78) must be true for any arbitrary volume V; hence, we can
write
∂ρs
= - div JS,tot þ γ ð5:79Þ
∂t
γ≥0 ð5:80Þ
These last two equations are the local forms of Eqs. (5.73) and (5.74), i.e., the
local mathematical expressions for the second law of thermodynamics. Equation
(5.79) is formally a balance equation for the entropy density s with an entropy
generation rate γ which satisfies the important inequality (5.80). With the help of
equation (5.42), equation (5.79) can be rewritten in a slightly different form as
follows:
ds
ρ = - div JS þ γ ð5:81Þ
dt
where the entropy flux JS is the difference between the total entropy flux JS, tot and a
convective term ρs~v:
JS = JS,tot - ρs~v ð5:82Þ
For applications in continuum mechanics, we must relate the changes in the

system to the entropy generation rate. This requires us to obtain explicit expressions
for the entropy flux JS and the entropy generation rate γ that appears in equation
(5.81). This explicit equation is called the thermodynamic fundamental equation.
We postulate the existence of a thermodynamic potential from which the state
laws can be derived. Let us assume that a function, with a scalar value, concave with
respect to temperature and convex with respect to other variables, allows us to satisfy
a priori the conditions of thermodynamic stability imposed by Clausius-Duhem
inequality. The specific Helmholtz free energy, φ, is defined as the difference
between the specific internal energy density u and the product between the absolute
temperature T and specific entropy s:
φ = u - Ts ð5:83aÞ
Differentiating this and with the help of the law of conservation of energy, we
have the following relations:
dφ = du - Tds - sdT ð5:83bÞ

= ðδq þ δwÞ - Tds - sdT ð5:83cÞ

= δq þ δwd þ δwe - Tds - sdT ð5:83dÞ

= δq þ δwd - Tds þ ðδwe - sdT Þ ð5:83eÞ
where q is the total heat flowing into the system per unit mass, including the
conduction through the surface and the distributed internal heat source, w is the
total work done on the system per unit mass by external loads and body forces, wd is
the dissipated energy associated with the total work, which may be dissipated in the
form of heat, and we is the elastic energy associated with the total work. For the
quantitative treatment of entropy for irreversible processes, let us introduce the
definition of entropy for irreversible processes:
δq þ δwd
ds = ð5:84Þ
T
With the help of Eq. (5.84), we can write the following relation:
Tds = du - dwe ð5:85Þ

This is the Gibbs relation which combines the first and second laws of thermo-
dynamics. From the definition of entropy, we also have
dwe = dφ þ sdT or dφ = dwe - sdT ð5:86Þ
The Helmholtz free energy, φ, is the isothermal recoverable elastic energy available
for work. The specific elastic energy we, namely, the work stored in the system per
unit mass during a process, is not a function of the process path [path independent]. It
depends only on the end state of a process for a given temperature. The elastic energy
is frequently also referred to as the available energy of the process. The elastic
energy is the maximum amount of work that could be produced by a system between
any given two states if the entropy generation is zero, which is not possible in any
process.
To find the explicit form of the entropy balance equation, Eq. (5.81), we insert
Eq. (5.72) for du
dt into Eq. (5.85) with the time derivatives given by Eq. (5.39):
ds div Jq 1 ρ dwe ρr
ρ =- þ σ : D- þ ð5:87aÞ
dt T T T dt T
Noting that
div Jq Jq 1
= div þ 2 Jq gradT ð5:87bÞ
T T T
it is easy to cast equation (5.87) into the form of a balance equation, Eq. (5.81):
ds Jq 1 1 ρ dwe ρr
ρ = - div - 2 Jq gradT þ σ : D - þ ð5:88Þ
dt T T T T dt T
From comparison with equation (5.81), it follows that the expressions for the
entropy flux and the entropy production rate are given by
Jq
JS = ð5:89Þ
T
1 ρ dwe 1 ρr
γ= σ : D- - 2 Jq gradT þ ð5:90Þ
T T dt T T
Equation (5.90) represents the entropy generation by the internal dissipations.

The sum of the first two terms is called intrinsic dissipation or mechanical dissipa-
tion. It consists of plastic dissipation plus the dissipation associated with the evolu-
tion of other internal variables; it is generally dissipated in the form of heat, but not
always. The last two terms are thermal dissipation due to the conduction of heat and
internal heat generation. The structure of the expression for γ is that of a
bilinear form: it consists of a sum of products of two factors. One of these factors
in each term is a flux quantity (heat flow Jq, σ stress tensor) already introduced in the
conservation laws. The other factor in each term is related to a gradient of an
intensive state variable (gradients of temperature and velocity). These quantities

which multiply the fluxes in the expression for the entropy production are called
thermodynamic forces. As we discussed earlier, actually assignment of flux and
thermodynamic force is rather arbitrary. However, their multiplication always must
yield the entropy generation rate.
According to the second law of thermodynamics, the entropy generation rate γ
must be zero if the thermodynamic equilibrium conditions are satisfied within the
system. Another requirement that Eq. (5.90) must satisfy is that it is invariant under
the transformation of different reference frames since the notions of reversible and
irreversible behavior must be invariant under a reference frame transformation.
Equation (5.90) satisfies these requirements. Finally, it may be noted that equation
(5.90) also satisfies the Clausius-Duhem inequality:

dφ dT gradT
σ : D-ρ þs - Jq ≥0 ð5:91Þ
dt dt T
Between two particles that are at different temperatures, heat is transferred by

conduction, a process that takes place at the molecular and atomic levels. The law of
heat conduction for isotropic bodies is given by
Jq = - k gradT ð5:92Þ
where k is the thermal conductivity with units of Btu/fthr°F and Jq is the heat flux.
This law of heat conduction was stated first by Fourier who based it on experi-
mental observation. Equation (5.92), also known as Fourier’s law, expresses a linear
relation between the heat flux vector Jq and its dual variable gradT. Since solid,
opaque bodies are of primary interest in this chapter, heat is transferred from point to
point within this body solely by conduction. The field equation of the boundary
value problem will, therefore, always be some form of the Fourier heat conduction
equation. Of course, heat may be transferred to the surface of the body by other
modes of heat transfer. Then the expression for the internal entropy generation rate
for thermomechanical problems can be simplified as
1 ρ dwe k ρr
γ= σ : D- þ 2 jgradT j2 þ ð5:93aÞ
T T dt T T
Total entropy generation is obtained by time integration of Eq. (5.93a):

Z t2
1
s= γ dt ð5:93bÞ
ρ t1
The specific entropy production rate for small deformation and small strain
thermomechanical problems in metals may be simplified as
dsi γ σ : ε_ p k r
= = þ 2 jgradT j2 þ ð5:93cÞ
dt ρ Tρ T ρ T
5.7 Fully Coupled Thermomechanical Equations 263
However, this simplification assumes that entropy generation due to elastic

deformation and all other mechanisms are negligible. Of course, this would be not
true for high strain rate loading, elastic fatigue, and most composite materials where
significant entropy is generated by internal relative elastic deformations of
constituents.
5.7 Fully Coupled Thermomechanical Equations
The formalism of continuum mechanics and thermodynamics requires the existence

of a certain number of state variables. For thermomechanical problems in metals at
low strain rates of loading, there are two observable variables: the temperature T and
the total strain ε. For dissipative phenomena, the current state also depends on the
history (loading trajectory) that is represented, in the local state, by the values of
internal variables at each instant. Plasticity and viscoplasticity require the introduc-
tion of the plastic (or viscoplastic) strain εp (or density of dislocations) as a state
variable. Other phenomena, such as softening, hardening, degradation, and fracture,
require the introduction of other internal variables. These variables represent the
internal state of matter (density of dislocations, the crystal structure of lattice,
material phase, polycrystalline grain size, the configuration of micro-cracks and
cavities, etc.) Lemaitre and Chaboche (1990) state that “There is no objective way
to choose the internal state variables best suited to the study of a phenomenon.”
However, this is only true if Newtonian mechanics is used in conjunction with an
empirical damage potential, curve fit to a test data. In the unified mechanics theory,
internal state variables are defined by the entropy-generating mechanism, which all
must be considered. Thermodynamic fundamental equation dictates which state
variables must be included.
For the general case, state variables will be denoted by Vk(k = 1, 2, . . .)
representing either a scalar or a tensorial variable.
For small strain formulation, total strain can be written as a summation of elastic
and plastic components:
ε = ε e þ εp ð5:94Þ
The relations between the energy, stress tensor, and strain tensor can be obtained
using the formalism of thermodynamics. Here we choose the specific Helmholtz free
energy, φ, which depends on observable variables and internal state variables:
φ = ðε, T, εe , εp , V k Þ ð5:95Þ
For small strain formulation, the strain appears only in the form of their additive
decomposition, so that
φððε - εp Þ, T, V k Þ = φðεe , T, V k Þ ð5:96Þ
which shows that (Lemaitre & Chaboche, 1990)

∂φ ∂φ ∂φ
= =- p ð5:97Þ
∂εe ∂ε ∂ε
and the following expressions define the relation between Helmholtz free energy and
thermodynamic state variables:
∂φ
σ=ρ ð5:98Þ
∂εe
∂φ
s= - ð5:99Þ
∂T
∂φ
Ak = ρ ð5:100Þ
∂V k
where Ak is a thermodynamic force associated with the internal variables Vk. The
vector formed by the variables is the gradient of the function φ in the space of the
variables T, εe, and Vk. This vector is normal to the surface φ = constant.
Let us return to the equation of the conservation of energy for small strains
(utilizing Eqs. (5.57) and (5.68)):
ρu_ = - div Jq þ σ : ε_ þ ρr ð5:101Þ
and replace ρu_ by the expression derived from Eqs. (5.83a–5.83e):
ρu_ = ρφ_ þ ρ_sT þ ρsT_ ð5:102Þ
And utilizing φ_ and s_ by their expression as a function of the state variables with
the help of equations (5.98–5.100):
∂φ e ∂φ _ ∂φ _ 1
φ_ = : ε_ þ Tþ V k = σ : ε_ e - sT_ þ Ak V_ k ð5:103Þ
∂εe ∂T ∂V k ρ
2 2 2
∂ φ ∂ φ_ ∂ φ _ 1 ∂σ e ∂s _ 1 ∂Ak _
s_ = - : ε_ e - T- Vk = - : ε_ þ T- V
∂εe ∂T ∂T 2 ∂V k ∂T ρ ∂T ∂T ρ ∂T k
ð5:104Þ
We obtain

∂s _ ∂σ e ∂Ak _
- div Jq = ρT T - σ : ðε_ - ε_ e Þ þ Ak V k - ρr - T : ε_ þ V k ð5:105Þ
∂T ∂T ∂T
By introducing the specific heat defined by

5.7 Fully Coupled Thermomechanical Equations 265
∂s
C=T ð5:106Þ
∂T
and considering Fourier’s law for isotropic materials
div Jq = - k div ðgrad T Þ = - k∇2 T ð5:107Þ
where ∇2 denotes the Laplacian operator. Using ε_ p = ε_ - ε_ e : we obtain

_ ∂σ e ∂Ak _
k∇ T = ρCT - σ : ε_ þ Ak V k - ρr - T
2 p
: ε_ þ Vk ð5:108Þ
∂T ∂T
This is the fully coupled thermomechanical equation, which can simulate the
evolution of temperature due to mechanical work with properly imposed boundary
conditions. Ak V_ k represents the non-recoverable energy in the matter corresponding
to other dissipation mechanisms. If we simplify the problem to thermomechanical
loading at small strain and small strain rates on pure metals, we may be able to ignore
other dissipation terms, and then we can write
Ak V_ k ≈ 0 ð5:109Þ
which results in the fully coupled elastic-plastic thermomechanical equation:
∂σ e
k∇2 T = ρC T_ - σ : ε_ p - ρr - T : ε_ ð5:110Þ
∂T
Equation (5.110) also allows us to calculate heat flux Jq generated due to elastic
and/or inelastic work in a solid body. This will be covered in more detail later in the
chapter on fatigue.
For the isotropic linear thermoelastic materials, the stress-strain relationship in
Newtonian mechanics is given by
σ ij = λδij εkk þ 2Gεij - ð3λ þ 2GÞδij αðT - T 0 Þ ð5:111Þ
where T0 is the reference temperature, α is the isotropic thermal expansion coeffi-

cient, λ and G are given by
vE E
λ= ; G= ð5:112Þ
ð1 þ vÞð1 - 2vÞ 2ð1 þ vÞ
If the internal heat generation is neglected, equation (5.108) defines the response
of isotropic linear thermoelastic materials:
k∇2 T = ρC T_ þ ð3λ þ 2GÞαT ε_ kk ð5:113Þ
The last term represents the interconvertibility of thermal and mechanical energy.
With the help of Eqs. (5.87) and (5.108), the total specific entropy generation rate
for small strains is given by

ds c ∂T 1 ∂σ e ∂Ak
= - : ε_ þ : Vk ð5:114Þ
dt T ∂t ρ ∂T ∂T
where it is assumed σ : ε_ - ρw_ e = σ : ε_ p - Ak V_ k holds, which represents the total

mechanical dissipation rate and Ak V_ k is assumed to be zero. Of course this assump-
tion is not true for elastic fatigue loading or for any problem where plastic work
dissipation is not the dominant entropy generation mechanism. In the presence of
plastic dissipation, elastic dissipation is small for quasi-static loading.
With the help of Eqs. (5.93a), (5.93b), and (5.93c), the specific entropy produc-
tion rate for small strain and quasi-static loading becomes
dsi γ σ : ε_ p r k
= = þ - 2 jgrad T j2 ð5:115Þ
dt ρ Tρ T T ρ
The equations governing the temperature, stress, deformation, and entropy pro-
duction rate in a continuum have been derived in the previous chapters. However,
these quantities are all interrelated and must be determined simultaneously in
incremental form.
In most quasi-static problems, the effect of stresses and deformations on the
temperature distribution is quite small and may be neglected. This procedure allows
the determination of the temperature distribution resulting from prescribed thermal
boundary conditions to become the first step of a thermal stress analysis. The second
step is then the determination of the stresses, deformations, and thermodynamic
stress index field due to this temperature distribution. Of course, performing
thermomechanical analysis in one step is possible, but it is far more computationally
intensive.
Entropy change caused by the heat transfer between the system and its surround-
ings does not influence the degradation of the material if the temperature field in the
entire body is uniform because of this exchange. If this heat exchange leads to a
nonuniform temperature field, then heat coming from outside can lead to irreversible
entropy generation in the material as in bending, thermomigration, and other mech-
anisms. Only the internal entropy generation, namely, the entropy created in the
system, should be used as a basis for the systematic description of the irreversible
processes, which can be given by
5.8 Numerical Validation of the Thermomechanical Constitutive Model 267
Z Z t Z t
t
σ : ε_ p k r
Δs = Δsi = dt - 2
jgrad T j dt þ dt ð5:116Þ
t0 Tρ t0 T ρ
2
t0 T
Equation (5.116) shows that the entropy generation is a function not only of the
loading or straining process but also of the temperature. However, a uniform increase
in temperature in a stress-free field does not cause any change in TSI. While this
fundamental equation accounts for strain rate, it does not account for all mechanisms
under thermomechanical loading. If the strain rate is very high, there are far more
entropy generation mechanisms such as phase change, elastic dissipation, melting,
and other mechanisms that can be significant entropy generation sources that this
equation does not account for.
Thermodynamic state index earlier was given by the following equation:
h ms ðΔsÞ
i
Φ = 1 - e- R ð5:117Þ
5.8 Numerical Validation of the Thermomechanical

Constitutive Model
Two problems are selected to validate the model described in this chapter. First, the
model results are compared with testing performed on specimens of thin layer solder
joints of Pb37/Sn63 under monotonic and fatigue shear testing at different strain
rates and temperatures.
5.8.1 Thin Layer Solder Joint: Monotonic and Fatigue Shear

Simulations
Table 5.3 lists the strain rate and temperature ranges for testing and simulations.
Monotonic and cycling shear tests were performed on a thin layer of the solder joint
(Pb37/Sn63) in pure shearing mode. Figure 5.3 shows the specimen attached to
copper plates. Material parameters used for the numerical simulation are shown in
Table 5.4. Figure 5.4 shows a comparison between numerical simulation results and
test data under monotonic shear testing for different temperatures at a strain rate of
1.67 × 10-3/s. Figure 5.5 shows monotonic shear testing under strain rate of
1.67 × 10-3/s, at 22 °C and inelastic strain range (ISR) = 0.005.
Figures 5.6, 5.7, 5.8, and 5.9 show a comparison between the cyclic shear
simulation results and test data at 22 °C and strain rate of 1.67 × 10-3/s. Figure 5.10
shows simulation results for isothermal fatigue testing at strain rate 1.67 × 10-4/s at
22 °C with an inelastic strain range of 0.022.
Table 5.3 Loading scheme used in the analysis

Case I Monotonic shear loading
1 Strain rate 1.67e-3/s Temperature (°C)
a -40
b 22
c 60
d 100
2 Strain rate 1.67e-1/s 22
Case II Cyclic shear loading
1 Temp = 22 °C
Strain rate 1.67e-3/s ISR
a 0.005
Fig. 5.3 Thin layer solder joint attached to copper plates
Table 5.4 Material Elastic (θ temper. in °K)

parameters
Young’s modulus (GPa) 52.10-0.1059θ
Shear modulus (GPa) 19.44-0.0395θ
Isotropic hardening
R00 (MPa) 37.47-0.0748θ
c (Dimensionless) 383.3
σ y (MPa) 60.069-0.140θ
Kinematic hardening
c1 (MPa) 2040
c2 (Dimensionless) 180
Creep strain rate
A (Dimensionless) 7.60E+09
D0 (mm/s2) 48.8
b (mm) 3.18E-07
d (mm) 1.06E-02
n 1.67
p 3.34
Q (mJ/mol) 4.47E+07
Φcritical 1
Fig. 5.4 Monotonic shear testing under strain rate 1.67 × 10-3/s at different temperature
Fig. 5.5 Monotonic shear

testing under strain rate
1.67 × 10-3/s at 22 °C
Fig. 5.6 Cyclic shear

simulation vs. test
data at 22 °C, strain
rate 1.67 × 10-3/s, and
ISR = 0.005

simulation vs. test
data at 22 °C, strain
rate 1.67 × 10-3/s, and
ISR = 0.012

simulation vs. test
data at 22 °C, strain rate
1.67 × 10-3/s, and
ISR = 0.02

simulation vs. test
data at 22 °C, strain rate
1.67 × 10-3/s, and different
inelastic strain range
Fig. 5.10 Isothermal

fatigue at strain rate
1.67 × 10-4/s at 22 °C with
inelastic strain range 0.022
Figure 5.11 shows the evolution of the thermodynamic state index under cycling
fatigue loading with an inelastic strain range of 0.022.
The computational simulations are in good agreement with the experimental
results. The differences are due to the imperfect microstructure (such as voids) of
the solder joint, while the numerical model assumes a perfect continuum with no
defects. Moreover, entropy generation due to internal heat generation was ignored
for the sake of simplicity in this example.
More examples of unified mechanics theory for thermomechanical problems are
provided in detail in the references listed at the end of the chapter.
5.9 Thermomechanical Analysis of Cosserat Continuum:

Length Scale Effects
Classical continuum mechanics formulation does not include a term for the size
effect. We should clarify here what is meant by “size.” It is not the size of the system,
such as 150 stories high building versus a very small bunker. “Size” refers to the
dimensions of the microstructure of the material. When the size of the microstructure
becomes a dominant factor in the response of the material, in addition to strains,
gradients of strains are also included in the continuum mechanics formulation. The
size effect is known to be a big factor, especially in the plastic zone of the material
response. The incremental theory of plasticity is called strain gradient plasticity
theory when strain gradients are also included in the yield surface formulation in
addition to strains. Strain gradient plasticity is especially needed when traditional
5.9 Thermomechanical Analysis of Cosserat Continuum: Length Scale Effects 273
Fig. 5.11 Evolution of the

damage parameter per
Eq. (4.37) for Φcr = 1.0
under fatigue loading with
ISR = 0.022
continuum mechanics formulation is unable to represent the material stress-strain

behavior due to microstructure size and deformation gradient effects, because
traditionally, continuum mechanics formulations are independent of size effect.
However, in some mechanics problems, this is not true, and material response is
size-dependent, especially low melting point metals such as solder alloys. Class of
plasticity theories where material size effect is considered is usually referred to as
strain gradient plasticity theories.
In strain gradient plasticity theories, it is assumed that classic plastic behavior is
due to the slip of statistically stored dislocations, and length [size] scale effects are
due to the slip of geometrically necessary dislocations. Geometrically necessary
dislocations are due to having a very large strain gradient in the material, like in a
thin film subjected to bending. As a result, the dislocation term in Taylor’s flow
stress equation is modified to add the geometrically necessary dislocation density:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
τ=α G b ρSSD þ ρGND ð5:118Þ
where α is a geometrical factor that depends on the type and arrangement of the
interacting dislocations, G is the shear modulus, b is Burger’s vector, ρSSD is the
density of statistically stored dislocations, and ρGND is the density of geometrically
necessary dislocations.
There are different methods to account for the size effect in continuum mechan-
ics. One of these methods is the use of the Cosserat continuum. In this section, we
will discuss (Cosserat continuum) couple stress-based strain gradient theory without
restrictions to any constitutive model. Next, we describe a constitutive model

incorporating coupling between degradation and strain rate dependency and consid-
ering size effects utilizing a couple of stress theory material length scales. We
provide two distinct theories within the same framework, namely, a general couple
stress theory where translational and rotational degrees of freedom are treated as
independent kinematic variables and a more restrictive reduced couple stress theory,
where translational and rotational degrees of freedom are constrained to satisfy a
given kinematic relation. The kinematic constraint allows the implementation of the
reduced theory into a displacement-based finite element method formulation. The
use of this approach relaxes the strong C1 shape function (approximation function)
continuity requirements in finite element method and allows the use of C0 approx-
imation functions, thus keeping the convergence features of the element. Here the
kinematic constraint is enforced via a penalty function/reduced integration scheme
that allows treatment of the reduced theory based on ideas motivated by the general
theory. The current description progressively arrives at this formulation after explor-
ing different possibilities within the context of the variational calculus. Throughout
all the discussion, the objective should be clear that the final goal is to implement the
reduced couple stress theory framework in the form of a user element subroutine
within the finite element code ABAQUS which imposes additional restrictions.
5.9.1 Cosserat Couple Stress Theory
The motivation behind Cosserat’s couple stress theory also referred to as the
Cosserat continuum, Fig. 5.12, stems from the following arguments. In classical
continuum mechanics, it is assumed that a material point occupies no space (has zero
volume) and then the size of the material microstructure is not part of the formula-
tion. To include the effects of the finite size of a material’s microstructure while
retaining a mathematical continuum, classical continuum mechanics have been
extended or generalized. In the Cosserat continuum, each material point is enriched
with higher-order kinematic degrees of freedom. In particular, the specification of
the position requires also the definition of a rotation. In the more general Cosserat
theory, this new kinematic quantity, the micro rotation, is independent of the
classical theory rotation vector θk = 12 eijk uj,k where eijk is the alternating pseudo-
tensor and a coma represents a derivative with respect to rectangular Cartesian
coordinates. This type of theory has been termed as “indeterminate couple stress
theory” by Eringen (1968) and “generalized couple stress theory” by Fleck and
Hutchinson (1997). The following discussion elaborates further upon this theory.
There are different interpretations of the Cosserat continuum. In Cosserat and
Cosserat (1909) couple stress theory, a differential material element is subjected to
not only normal and shear stresses but also couple stress components as shown in
Fig. 5.12. For linear elastic behavior, the traditional stress components are functions
of the strains and the couple stresses are functions of the strain gradients. Two
distinct theories are identified in the work of Cosserat and Cosserat (1909). First,
Fig. 5.12 Cosserat continuum stress point
there is a reduced couple stress theory with the kinematic degrees of freedom being
the displacement ui and an associated material rotation θi tied to the displacements
by the kinematic constraint in Eq. (5.119) and with the definitions of strain and
curvatures expressed in Eqs. (5.120) and (5.121):
1
θi = e u ð5:119Þ
2 ijk k,j
1
εij = ui,j þ uj,i ð5:120Þ
2
xij = θi,j ð5:121Þ
Second, there is a general couple stress theory in terms of a micro rotation, ωi,
which is regarded as an independent kinematic degree of freedom. The traditional
continuum mechanics rotation and Cosserat continuum micro rotation are related by
a relative rotation tensor αij = eijkωk - eijkθk. For the choice of ωk = θk, the general
couple stress theory reduces to the reduced couple stress theory. Equilibrium of the
differential element shown in Fig. 5.12, after neglecting body forces and body
couples [in Newtonian mechanics], yields
σ ji,j þ τji,j = 0 ð5:122Þ

1
τjk þ ejik mpi,p = 0 ð5:123Þ
2
where σ ij and τ ij are the symmetric and antisymmetric components of the Cauchy
stress tensor, respectively, and mij is the couple’s stress tensor. Surface stress
tractions and surface couple stress tractions are given by
ðnÞ
σ i = σ ij þ τij nj ð5:124Þ
qi = mij nj ð5:125Þ
ð nÞ
where σ i is the traction vector, qi is the couple traction vector, and ni is a surface
outward normal vector. For a linear elastic solid, the strain energy density is given by
Eq. (5.126) where l is a length scale associated with elastic behavior and may be
regarded to be of the order of inter-atomic distances (Koiter, 1964).
E v
W= ε ε þ εij εij þ l2 xij xij ð5:126Þ
2ð1 þ vÞ 1 - 2v ii jj
A constitutive relationship can be directly derived out of Eq. (5.126) as follows:
∂W
σ ij = C ijkl εkl ð5:127Þ
∂εij
∂W
mij = Dijkl xkl ð5:128Þ
∂xij
where Cijkl and Dijkl are elastic constitutive tensors relating stress and couple stress
components to elastic strains and elastic curvatures, respectively. The principle of
virtual work for the reduced Cosserat theory can be given by [in Newtonian
mechanics]
ZZZ ZZZ ZZ ZZ
ðnÞ
σ ij δεij dV þ mij δxij dV = σ i δui dA þ qi δθi dA ð5:129Þ
V V A A
The concept of “reduced Cosserat” will be discussed later in this chapter.
5.9.2 Toupin-Mindlin Higher-Order Stress Theory
The Cosserat couple stress theory introduces curvatures as gradients of rotation.

Toupin (1962) and Mindlin and Tiersten (1962) and Mindlin (1964, 1965) extended
the original Cosserat formulation by considering also gradients of the normal strain
components, therefore including all the components of the second gradient of
displacement. The additional gradients of strain as variables give rise to new work
conjugates in the form of force couples per unit area or higher-order stresses. The
physical interpretation of these force couples and higher-order stresses is explained
in Fig. 5.12. In this type of solid, Newtonian mechanics equilibrium (neglecting
body forces) is given by
σ ik,i - τijk,ij = 0 ð5:130Þ

Fig. 5.13 Schematic of a

generally solid body
where σ ij is the symmetric Cauchy stress tensor and τijk is the higher-order stress
tensor for a generally solid body (Fig. 5.13). Surface tractions and surface couple
tractions (moments) are given in Eqs. (5.124) and (5.125), respectively. The higher-
order stresses correspond to force couples per unit area:
ðnÞ
σ i = σ ik - τijk:j ni þ ni nj τijk Dp np - Dj ni τijk ð5:131aÞ
r k = ni nj τijk ð5:131bÞ
where the operators Dj and D are defined as Dj = (δjk - njnk)∂k and D = nk∂k.
Generalized strain-displacement and strain gradient-displacement relations can be
given by
1
εij = ui:j þ uj,i ð5:132aÞ
2
ηijk = uk,ij ð5:132bÞ
A strain energy density function can now be given by
1
W= λε ε þ μεij εij þ a1 ηijj ηikk þ a2 ηiik ηkjj þ a3 ηiik ηjjk þ a4 ηijk ηijk þ a5 ηijk ηkji
2 ii jj
ð5:133Þ
where λ and μ are Lamè constants and ai are material constants with units of force. A
constitutive relationship can be directly derived out of Eq. (5.133) such that
∂W
σ ij = C ijkl εkl ð5:134aÞ
∂εij
∂W
τijk = Dijklmn ηlmn ð5:134bÞ
∂ηijk
where Cijkl and Dijklmn are elastic constitutive tensors relating stress and higher-order
stress components to elastic strains and elastic strain gradients, respectively. The
principle of virtual work can be written by equating an external work increment and
an internal increment of strain energy and can be given by [in Newtonian mechanics]
ZZZ ZZZ ZZ ZZ
ðnÞ
σ ij δεij dV þ τijk δηijk dV = σ i δui dA þ r i ðDδui ÞdA ð5:135Þ
V V A A
5.9.3 Equilibrium Equations
In this section, we derive the equilibrium equations for two different classes of the
Cosserat continuum. Traditionally, the notation used to define a solid body occupy-
ing a volume Ω and bounded by a surface ∂Ω defined by an outward normal vector
n as schematically is shown in Fig. 5.13. In our formulation, we will use dV for
volume differentials and dA for surface area differentials.
There are several different interpretations of the Cosserat continuum. In the
Cosserat and Cosserat (1909) couple stress theory, a differential material element
has not only normal and shear stresses but also couple stress (moment) components
as shown in Fig. 5.12. For linear elastic behavior, the usual stress components are
functions of the strains and the couple stresses are functions of the strain gradients.
Two distinct theories are identified in the Cosserat and Cosserat (1909) as described
by Aero and Kuvshinsky (1961), Mindlin (1964), de Borst and Muhlhaus (1992), de
Borst (1993), and Shu and Fleck (1999). First, there is a reduced Cosserat couple
stress theory with the kinematic degrees of freedom being the displacement ui and an
associated material rotation θi tied to the displacements by a kinematic constraint
given in Eq. (5.119) and with the definitions of strain and curvatures expressed in
Eqs. (5.120) and (5.121), respectively.
According to the reduced Cosserat continuum theory, the continuum is assumed
to possess bending stiffness allowing for the introduction of additional stress mea-
sures in the form of moments per unit area. The presence of the couple stresses
(moments) renders the Cauchy stress tensor asymmetric; however, only the sym-
metric component generates work upon deformation. Second, there is a general
Cosserat continuum couple stress theory in terms of a micro rotation ωi which is
regarded as an independent kinematic variable. The rotation and micro rotation are
related by a relative rotation tensor αij = eijkωk - eijkθk. For the case of ωk = θk, the
general Cosserat continuum couple stress theory reduces to the reduced Cosserat
continuum couple stress theory. This general theory assumes that at a material point,
there is also an embedded micro-volume giving rise to the micro rotation ωi.
Different interpretations have been postulated depending on the deformation prop-
erties assumed for the embedded micro-volume (see Mindlin and Tiersten (1962),
Toupin (1962), and Mindlin (1964, 1965) for a review of the different approaches).
Fig. 5.14 Relative deformation in the couple stress theory by Mindlin (1964)
For instance, Fig. 5.14 shows the deformation state in a material point including the
micro-volume for the case of pure shear in the solid proposed by Mindlin (1964).
Figure 5.15 shows the case of reduced Cosserat continuum couple stress theory
where the micro-volume is assumed rigid.
Newtonian mechanics equilibrium of the differential element shown in Fig. 5.12
(which is valid for both theories of the Cosserat continuum), after neglecting body
forces and body couples, yields
σ ji,j þ τji,j = 0 ð5:136Þ

1
τjk þ eijk mpi,p = 0 ð5:137Þ
2
where σ ij and τ ij are the symmetric and antisymmetric components of the Cauchy
stress tensor, respectively, and mij is the couple’s stress tensor. Surface stress
tractions and surface couple stress (moment) tractions are given by

σ ðnÞ i = σ ij þ τij nj ð5:138Þ
qi = mij nj ð5:139Þ
Fig. 5.15 Reduced Cosserat continuum stress point
where σ (n)i is the surface traction vector, qi is the surface couple tractions vector, and
ni is a surface normal vector. To establish a general framework for elastic material
behavior, Cosserat couple stress continuum can be obtained after postulating a strain
energy density function dependent on strains and curvatures (rotation gradients).
Toupin (1962) and Mindlin (1965) extended this theory to include also stretch
gradients. In particular, they considered invariants of the strain and strain gradients
in the strain energy density function in terms of the following generalized von Mises
strain invariant which is given by Fleck and Hutchinson (1997):
2 0 0
E2 = ε ε þ c1 η0iik η0jjk þ c2 η0ijk η0ijk þ c3 η0ijk η0kij ð5:140Þ
3 ij ij
where prime superscript denotes deviatoric component, ηijk = uk, ij are the strain
gradients and the c1,c2,c3 are additional material constant with dimensions of length
squared (L2). Equation (5.140) is the basis for the most general strain gradient
plasticity theory, and it essentially reveals the phenomenological coupling between
the densities of statistically and geometrically stored dislocations. The reduced
couple stress theory is just a special case of Eq. (5.140) where only rotation gradients
are considered.
5.9.4 Finite Element Method Implementation
The following elastic constitutive relationships can be written for the symmetric
Cauchy stress tensor, the asymmetric stress tensor, and the couple stress tensor:
σ ij = C ijkl εkl ð5:141Þ

τij = Dijkl αkl ð5:142Þ
-1
l mij = Dijkl l xkl ð5:143Þ
where Cijkl is a tangential constitutive tensor relating strains to Cauchy stresses, Dijkl
is a constitutive tensor relating curvatures to couple stresses, and Dijkl is a constitu-
tive tensor relating relative rotations to the antisymmetric component of the Cauchy
stress tensor, and l is characteristic length. The total potential energy functional ∏ for
the general elastic Cosserat continuum can be written by considering separately the
contributions from the symmetric and antisymmetric stress tensors and the couple
stress tensor as follows:
Z Z
Q 1 1
ðui , ωi Þ = C ijkl εkl ðui Þεij ðui ÞdV þ Dijkl xij ðωi Þxkl ðωi ÞdV
2 2
VZ V Z Z ð5:144aÞ
1 ðnÞ
þ Dijkl αij ðui , ωi Þαkl ðui , ωi ÞdV - σ i ui dA - qi ωi dA
2 A A
V
where translational and rotational degrees of freedom are independent degrees of

freedom. For the discussion that follows, it is convenient to write in the equivalent
form of
Y
ðu, ωÞ = F ðuÞ þ GðωÞ þ H ðu, ωÞ
where
1
aðu, uÞ - f ðuÞ
F ð uÞ =
2
1
GðωÞ = bðω, ωÞ - gðωÞ
2
Z
1
H ðu, ωÞ = Dijkl αij ðui , ωi Þαkl ðui , ωi Þ dV ð5:144bÞ
2
V
R R
with aðu, uÞ C ijkl εij εkl dV and bðω, ωÞ Dijkl xij xkl dV being symmetric bilinear
Ω V
forms and f(u), g(ω) corresponding to the boundary terms in Eq. (5.144a). For the
particular case of the reduced couple stress theory where the stress tensor is sym-
metric and satisfies the constraint in Eq. (5.119), αij (or equivalently H(u, ω))
vanishes and Eq. (5.144b) reduces to Eq. (5.145a):
Z Z
1 1
Π ð ui Þ = C ijkl εkl ðui Þ εij ðui Þ dV þ Dijkl xij ðui Þ xkl ðui Þ dV
2 2
V V
Z Z
ðnÞ
- σ i ui dA - qi ωi dA ð5:145aÞ
A A
where now the strain energy contribution from the curvatures becomes a function of
the translational degrees of freedom only. Note that Eq. (5.145a) is analogous to the
total potential energy functional for the particular case of the so-called Timoshenko
beam theory. Using the alternative notation, (5.145a) can be written as
ΠðuÞ = F ðuÞ þ G0 ðuÞ ð5:145bÞ

R
where G′(u) = b′(u, u) - g′(u) with b0 ðu, uÞ Dijkl xij ðui Þxkl ðui ÞdV and g′(u)
V
corresponding to the boundary condition associated with the rotation. The prime
superscript notation ()′ has been introduced to clarify the fact that in the reduced
theory, the curvatures are kinematically constrained to the displacements. Equations
(5.144a) and (5.145a) are the basis for the finite element implementation. The
reduced theory continuum can also be formulated in terms of independent rotational
degrees of freedom with the constraint to the translational degrees of freedom
considered as the limit when the term H(u, ω) → 0 in the general Cosserat
continuum. This implies that Eq. (5.145a) is written in terms of additional indepen-
dent rotational degrees of freedom but with the additional requirement imposed by
the constraint between translational and rotational degrees of freedom. In this case,
the reduced theory can also be described by the functional Πb ðu, ωÞ that follows:
Z Z
b ð ui , ω i Þ = 1
Π C ijkl εkl ðui Þεij ðui ÞdV þ
1
Dijkl xij ðωi Þxkl ðωi ÞdV
2 2
V V
Z Z
ðnÞ
- σ i ui dA - qi ωi dA ð5:145cÞ
A A
A comparison of Eqs. (5.144b) and (5.145c) reveals that Πðu, ωÞ → Π b ðu, ωÞ

when H(u, ω) → 0. This means that the general couple stress theory approaches
the reduced couple stress theory in the limit of vanishing relative rotation αij.
5.9.5 General Couple Stress Theory: Variational

Formulation
Consider the case of the general couple stress continuum where there is no constraint
set but instead, there is a space of admissible functions with translational and
rotational degrees of freedom. The rotational degrees of freedom are linearly inde-
→
pendent and belong to the space Q alternatively, and we can define displacements
→ → → →
and rotations as elements of the single space V × Q . The product V × Q results in a
→ → → →
third space with elements being all the ordered pairs of the form u , ω 2 V × Q
because translational and rotational degrees of freedom are linearly independent. In
terms of the introduced notation, this is equivalent to the following variational
→ → → → → → → →
problem; find u , ω 2 V × Q such that for any v , φ 2 V × Q , Π assumes
→ → → →
its minimum value at u, ω where V × Q is now the corresponding space of
admissible functions.
The generalized form of the principle of virtual displacements for the general
couple stress theory can be given by
Z Z Z Z
ðnÞ
σ ij εij ðvi ÞdV þ mij xij ðφi ÞdV þ τij αij ðvi , φi ÞdV - σ i vi dA
A
V V V
Z
- qi φi dA = = 0 ð5:146Þ
A
where vi and φi denote virtual displacement and rotation, respectively. Displacement

→ →
and rotation u , ω are regarded as linearly independent kinematic degrees of
freedom. This is consistent with the general couple stress theory and the resulting
variational problem is thus unconstrained.
5.9.6 Reduced Couple Stress Theory: Variational

Formulation
The principle of virtual displacements for a reduced couple stress theory can be
given by
Z Z Z Z
ðnÞ
σ ij εij ðvi ÞdV þ mij χ ij ðvi ÞdV - σ i vi dA - qi φi ðvi ÞdA = 0 ð5:147Þ
V V A A
In contrast to the case defined in Eq. (5.146), the present principle of virtual
→
displacements shows that the only kinematic variable is displacement u which
defines the strains εij and the curvatures χ ij. This is the definition of the reduced
couple stress theory. There are no linearly independent rotational degrees of freedom
and there is no constraint to be enforced. However, the displacement approximation
(shape) function needs to be C1 continuous in the finite element implementation.
However, when there is no kinematic constraint between displacements and rota-
tions, mesh-dependent results are obtained in finite element analysis. This is a major
obstacle.
5.9.7 Reduced Couple Stress Theory: Mixed Variational

Principle
The principle of virtual displacements with the kinematic constraints imposed in a

weak sense leads to
Z Z Z Z
ðnÞ
σ ij εij ðvi ÞdV þ mij χ ji ðφi ÞdV þ τij αij ðvi , φi ÞdV = σ i i vi dA
V V V A
Z
þ qi φi dA ð5:148aÞ
A
The kinematic constraint is given by

Z
ρij αij ðui , ωi ÞdV = 0 ð5:148bÞ
V
The first two terms on the right-hand side of Eq. (5.148a) correspond to virtual
work done by the stresses and couple stresses, respectively. The third term corre-
sponds to the virtual work done by the asymmetric component of the stress tensor.
Equation (5.148b) is the weak enforcement of the constraint of vanishing relative
rotation. This formulation was proposed by Xia and Hutchinson (1996) to solve the
mesh dependency problem that happens in the displacement-based finite element
method when Eq. (5.147) is used.
Finite Element Method Implementation Details
This section discusses the discretization of the equations derived above starting from
their corresponding variational equations. In each case, we will schematically show a
typical finite element with its associated vector of nodal point degrees of freedom.
The nodal degrees of freedom may be translational only, translational and rotational
degrees of freedom, or translational and rotational degrees of freedom with addi-
tional Lagrange multipliers. To distinguish between the values of the degrees of
freedom at any point within the element and its nodal point value, we will use the
following notation. For instance, if displacement is being considered, the value at
any point within the element will be denoted by the vector u, and its corresponding
nodal point value will be denoted by b u. The value of any degree of freedom at any
point within the element is obtained via interpolation functions from the known
nodal point values. Strains and curvatures are usually calculated at Gauss integration
points from the nodal point displacements and rotations values using interpolation
functions. We will denote such a function by a subscripted symbol where the
subscript indicates the variable that is being interpolated. For instance, we will
denote the displacement-curvature transformation matrix by Bχ where the curvature
at any point within a given element is obtained out of the nodal point displacement
vector as χ = Bχ b
u where b u may have translational or translational and rotational
degrees of freedom.
5.9.8 General Couple Stress Theory Implementation
In the case of the general couple stress theory, the continuum is free of constraints as
→ →
a result, and u , ω are linearly independent degrees of freedom. In the finite element
formulation, C0 continuity of shape functions is enough to satisfy the displacement
compatibility requirement. The nodal point displacement degrees of freedom for an
n-nodded two-dimensional quadric-lateral element have the following general form:
uTe = ½u1 v1 ω1 . . . un vn ωn as shown in Fig. 5.16. To describe the finite element
b
discretization, we recall that for a general couple stress theory, principle of virtual
work is given by
Z Z Z Z
ðnÞ
σ ij εij ðvi ÞdV þ mij χ ij ðφi ÞdV þ τij αij ðvi , φi ÞdV - σ i vi dA
V V V A
Z
- qi φi dA = 0 ð5:149Þ
A
Fig. 5.16 2-D finite

element for the general
couple stress theory
u
u = Nu
ðnÞ b bT
uLetting i,, ε = Bεui,, χ = Bχ ui,, α = Bαui, and σ i = N t where t =
t 1 t v1 t ω1 . . . t un t vn t ωn , where subscript n represents node number. Using the constitutive
relationships given by Eqs. (5.141)–(5.143), we can write the following discretized
form after eliminating the virtual variables:
2 3
Z Z Z Z
4 BTε CBε dV þ BTχ DBχ dV þ BTα DBα dV 5ui = N Tbt dA ð5:150Þ
V V V A
or using the stiffness matrices, the equilibrium equation can be given by

ε
K e þ K χe þ K αe ui = F ð5:151Þ
Reduced Couple Stress Theory Implementation: Displacement Formulation

In the case of the reduced couple stress theory, u , !
→
ω are related through the
constraint relationship given in Eqs. (5.119)–(5.121) and are not linearly indepen-
dent of each other. In terms of a formulation based on translational degrees of
freedom only, this constraint implies that approximation functions with C1 continu-
ity are needed to satisfy displacement compatibility requirements. The nodal point
displacement degrees of freedom for an n-nodded quadric-lateral element have the
following general form: buTe = ½u1 v1 . . . un vn , where subscript n is the node number as
shown in Fig. 5.17.
To describe the finite element discretization for reduced couple stress theory, we
start with the variational formulation:
Z Z Z Z
ðnÞ
σ ij εij ðvi ÞdV þ mij χ ij ðvi ÞdV - σi vi dA - qi φi ðvi Þ dA = 0 ð5:152Þ
V V A A
Fig. 5.17 Typical finite

element for the case of
reduced couple stress theory
using translational degrees
of freedom only
Letting u = Nui, εe = Bεui, χ e = Bχ ui, and q = NTq bt and introducing the

constitutive relationships introduced in Eqs. (5.141)–(5.143), we can write
2 3
Z Z Z
4 BTε CBε dV þ BTχ DBχ dV 5ui = NT þ N Tq bt dA ð5:153Þ
V V A
or equivalently using stiffness matrices, we can write the following equilibrium

equation:
ε
K e þ K χe u = F ð5:154Þ
Reduced Couple Stress Theory: Lagrange Multiplier Formulation

→
In the case of the reduced couple stress theory, translational u and rotational degrees
→
of freedom ω are related through a Lagrange multiplier constraint relationship
given in Eqs. (5.119)–(5.121) and are not independent of each other.
The elements have nodal degrees of freedom shown in Fig. 5.18 [for a 2-D
quadric-lateral element]. The nodal displacement degrees of freedom vector
uTi = ½u1 v1 ω1 . . . un vn ωn and an additional nodal point vector of Lagrange mult
τ Te = ½τ1 . . . τn are shown in Fig. 5.18.
\ipliers b
To describe the finite element discretization, we recall that reduced couple stress
theory with the kinematic constraints defined via Lagrange multipliers is given by
Z Z Z Z
ðnÞ
σ ij εij ðvi ÞdV þ mij χ ji ðφi ÞdV þ τij αij ðvi , φi ÞdV = σ i vi dA
V V V A
Z
þ qi φi dA ð5:155aÞ
A
Fig. 5.18 Typical finite

element for the case of
reduced couple stress theory
using Lagrange multipliers u
Z
ρij αij ðui , ωi ÞdA = 0 ð5:155bÞ
A
Using u = Nuui, τ e = Nτbτe ε = Bεui, χ = Bχ ui, and α = Bαui t = Nbt where

bt = t u t v t ω . . . t u t v t ω and after introducing the constitutive relationships, we can
T
1 1 1 n n n
write in the following matrix form:
2Z Z Z 3
BTε CBε dVþ BTχ DBχ dV BTα Nτ dV 2Z 3
6 7h i
6 Z 7 bu T
6 7 4 N tdA
5
7 bτ = A ð5:155cÞ
V V V
6
4 N Tτ Bα dV 0 5 0
V
or using stiffness matrices

h ih i h i
K uu K uτ bu = F ð5:155dÞ
K Tuτ 0 bτ 0
5.10 Cosserat Continuum Implementation in Unified

Mechanics Theory
Up to this point, all our derivations have been based on Newtonian mechanics. In the
Cosserat continuum, the side effects are introduced by enhancing the definition of
equivalent plastic strain with the addition of an equivalent plastic curvature, which is
a gradient of strain. This approach leads to a straightforward extension of the classic
flow theory (also known as the theory of incremental plasticity) which allows the
treatment of cyclic loading and monotonic loading. First, we will start with the
formulation that presents the flow theory equations for the undamaged rate-
independent case. A key feature is the coupling between the Cauchy stress and a
5.10 Cosserat Continuum Implementation in Unified Mechanics Theory 289
couple of stress components. This coupling is not present in the initial elastic
material but progressively appears with the accumulation of plastic curvatures. In
the case of degradation, additional terms appear in the expression for the Helmholtz
free energy determination. On the other hand, consideration of rate-dependent
effects assumes the validity of the same rules for the plastic curvatures as for the
plastic strains. This assumption allows for using a single creep law for both compo-
nents (strain and curvature) and is justified by the kinematic constraint present in the
reduced couple stress continuum.
5.10.1 Rate-Independent Plasticity Without Degradation
The relationship between the symmetric part of the Cauchy stress tensor and the
elastic strains and the couple stress tensor and the elastic curvatures can be written in
rate form as follows:
σ_ ij = C ijkl ε_ elkl ð5:156Þ

ℓ - 1 m_ ij = Dijkl ℓ χ_ elkl ð5:157Þ
where Dijkl = Gδikδjl, where G is the shear modulus. The strains and curvatures are
decoupled into elastic and inelastic components which results in
ε_ ij = ε_ elij þ ε_ pl
ij ð5:158Þ
ℓ χ_ ij = ℓ χ_ elij þ ℓ χ_ pl
ij ð5:159Þ
Considering the following definition of generalized deviatoric stress norm kΣk:

1=2
Σ0 = Sij Sij þ ℓ - 1 mij ℓ - 1 mij ð5:160Þ
where Sij is the deviatoric stress tensor and mij is the couple stress tensor, which is
deviatoric (it does not have a spherical component). Similarly, the generalized strain
tensor norm can be given by
1=2
Ε = εij εij þ ℓ χ ij ℓ χ ij ð5:161Þ
Next, we need to introduce a yield surface separating the elastic and inelastic
domains per classical plasticity theory. To account for the Bauschinger
effect observed in metals, a back-stress tensor βij and a couple of back-stress tensors
ℓ -1ηij defining the movement of the yield surface in stress space need to be defined.
The difference fij = Sij - βij between the back-stress and the deviatoric component of
the symmetric part of the Cauchy stress tensor is the relative stress tensor fij. In an
analogous form for the couple back-stress, there follows that the relative couple
b ij which is defined by
back-stress ℓ - 1 C
b ij = ℓ - 1 mij - ℓ - 1 ηij :
ℓ - 1C
The generalized relative stress kξk can be described in terms of the relative
stresses and given by
→ h i1=2
b ij C
ξ = f ij f ij þ ℓ - 2 C b ij ð5:162Þ
We can now define a yield surface to distinguish between elastic domain and
inelastic domain. In the classical theory of incremental plasticity, the yield surface is
defined in terms of a hardening parameter that can be shown to be proportional to the
equivalent plastic strain or plastic work. In the couple stress-based strain gradient
plasticity theory introduced here, the hardening parameter also incorporates the size
effects via the equivalent plastic curvatures. The generalized yield criteria can
therefore be given by
rffiffiffi
-1
2
F σ, ℓ m, α = ξ - K ðαÞ ð5:163Þ
3
where α is the generalized isotropic hardening parameter and K(α) represents the
radius of the yield surface. To complete the constitutive model, it is necessary to
define the rate-independent plastic flow rules (i.e., evolution equations for the plastic
strains and curvatures) and hardening laws (i.e., the evolution of the hardening
parameter and back-stress components). Here it is assumed that the flow rule
obeys associative plasticity. Considering a more general stress space with normal
stresses and couple stresses, the yield surface can be considered as a hypersphere in a
stress space with normal N b to the yield surface, F, defined as
b = ∂F ½b
N n, bv ð5:164Þ
→
∂Σ
b
ℓ-1
∂F f ij ∂F C ij
where b
n= ∂σ
and bv = ∂l - 1 m
.
ξij ξij
And the plastic flow rules are given by
∂F f ij
ε_ pl
ij = γ γ γb
n ð5:165aÞ
∂σ ij ξij
∂F b ij
C
ℓ χ_ pl
ij = γ -1
γℓ - 1 γb
ν ð5:165bÞ
∂ℓ mij ξij
γ is the consistency parameter which is defined from the loading/unloading condi-

tions and is related to the evolution of the generalized equivalent plastic strain
trajectory defined by
rffiffiffi
2
α_ = γ ð5:166Þ
3
The constitutive model is completed with the evolution equations for the back-
stresses:
2
β_ ij = H 0 γb nij ð5:167aÞ
3
2
ℓ η_ ij = H 0 γb
νij ð5:167bÞ
3
where H′ represents a kinematic hardening modulus which may be a linear or a

nonlinear function of the hardening parameter α. For instance, the assumption of a
constant kinematic hardening modulus leads to the so-called Prager-Ziegler rule
(Fung & Tong, 2001). Introducing Eqs. (5.165a) and (5.165b) into the generalized
strain norm for the plastic quantities yields
h i1=2 h i1=2
γ b ij C
b ij
Ε_ = ε_ pl
pl
_ pl
ij ε
2 pl pl
ij þ ℓ χ_ ij χ_ ij f ij f ij þ ℓ - 2 C ð5:168Þ
ξij
which implies Ε_ = γ
pl
Using this result in Eq. (5.166) and integrating yields
Z t rffiffiffi
2 _ pl
α ðt Þ = Ε ðτÞ dτ ð5:169Þ
3
0
This is the generalized version of equivalent plastic strain (trajectory) including

the gradients of plastic strains. The evolution equations are complemented by the
loading-unloading conditions which allow the determination of the consistency
parameter. In terms of the yield function, the following loading/unloading condition
must be satisfied:

γ≥0 F σ, ℓ - 1 m, α ≤ 0 ð5:170aÞ

γ≥0 γF σ, ℓ - 1 m, α = 0 ð5:170bÞ
And the consistent condition

γ F_ σ, ℓ - 1 m, α = 0 ð5:171Þ
Determination of Consistency Parameter

Using the definition of the yield function, we can write
∂F ∂F ∂F ∂F
F_ = : σ_ þ - 1 : ℓ - 1 m
_ þ p : ε_ pl þ : χ_ pl ð5:172Þ
∂σ ∂ℓ m ∂ε ∂ℓχ pl
Using Hooke’s law together with the plastic flow rule yields
F_ = ½b
n : C : ε_ þ bv : D : ℓχ_

∂F ∂F
b b b
- γ ½n : C : n þ v : D : v - b b
:nþ : bv ð5:173Þ
∂εpl ∂ℓχ pl
And imposing the consistency condition yields the consistency parameter γ in the
following form:
n : C : ε_ þ bv : D : ℓ χ_
b
γ= ð5:174Þ
b ∂F ∂F
n:C:b
n þ bv : D : bv - ∂εpl
:b
n þ ∂ℓχ pl : b
v
Using the following relations

rffiffiffi rffiffiffi
∂F 2 0 2 εpl ∂F 2 0 2 ℓχpl
=- K and =- K ð5:175Þ
∂εpl 3 3 α ∂ℓχpl 3 3 α
yields
rffiffiffi rffiffiffi
∂F 2 0 2 εpl S ∂F 2 0 2 ℓχ pl ℓ - 1 m
:b
n= - K : and : bv = - K :
∂εpl 3 3 α Σ ∂ℓχ pl 3 3 α Σ
ð5:176Þ
and then the term within brackets in the denominator reduces to

∂F ∂F
:b nþ
∂εpl ∂ℓχ pl
rffiffiffi
2 0 2 1 pl 2
: bv = - K ε : S þ ℓχ pl : ℓ - 1 m = - K 0 ð5:177Þ
3 3 α Σ 3
after introducing εpl : S + ℓχ pl : ℓ -1m = Kα.
The consistency parameter γ reduces to
b
n : C : ε_ þ bv : D : ℓ χ_
γ= ð5:178Þ
b
n:C:b n þ bv : D : bv þ 23 K 0
Using
S:S
b
n : C = 2Gb
n and b
n:C:b
n = 2G
k Σk 2
and
ℓ - 1m : ℓ - 1m
bv : D = 2Gbv and bv : C : bv = 2G
kΣ k2
where G is shear modulus yields
1S : ε_ þ ℓ - 1 : ℓ χ_ 1 S : ε_ þ ℓ - 1 : ℓ χ_
γ= = ð5:179Þ
Σ
K0
1 þ 3G Σ Kb
Introducing this result into Hooke’s law yields
2G O 2G O
σ_ = C : ε_ - b
n b
n : ε_ - b
n bv : ℓχ_ ð5:180aÞ
b
K Kb
2G O 2G O
ℓ - 1m
_ = D : ℓ χ_ - b
n bv : ε_ - bv bv : ℓ χ_ ð5:180bÞ
Kb Kb
Using
O a
c 1 O a
c
C = κI I þ 2G - I I and D = 2G ð5:181Þ
3
and after simplifying yields

2 O N N 3
N ‘
b - 1I b b
n n b
n bv
6 κI I- - 2G
bN 7
Iþ2G
3 N b
M ep = 6
4
K K 7 ð5:182Þ
b
n bv ‘
b b
v v 5
- 2G 2G -
b
K b
K
Alternatively, it can be written as

2 N N 3
C:b
n b
n b
n D : bv
6 C - C:
b
- C:
b 7
M =6
ep
4
N K NK 7 ð5:183Þ
bv C : b
n bv D : bv 5
- D: D - D:
b
K Kb
The plastic flow theory just described can be written in the following alternative
form, which is useful in the numerical treatment of the problem. Using the general-
ized vector and matrix notation, for the simple case of isotropic hardening, we make
the following definitions:
h i h i h i
σ ε
Σ= ℓ -1
m
, Ε= ℓχ
, M= C
0
0
D
ð5:184Þ
and then the constitutive model equations can be written as Σ = M : Εel. The yield
criteria and flow rule follow:
rffiffiffi
2
F σ, ℓ - 1 m, α = Σ0 - K ðαÞ ð5:185Þ
3
E_ = γ N
b
pl
ð5:186Þ
then
∂F _ ∂F _ pl
F_ = Σþ Ε ð5:187Þ
∂Σ ∂Εpl
Using Hooke’s law and the flow rule results in
F_ = N b þ ∂F γ N
b : M : Ε_ - γN : M : N b ð5:188Þ
∂Εpl
Utilizing the consistency conditions γ F_ = 0 gives us the following relations:
N : M : Ε_
pl
∂F b 2
γ= or after using N = - K0 ð5:189Þ
b
N:M:N b - ∂Fpl N
b ∂Ε pl 3
∂Ε
we have
b : M : Ε_ pl
N
γ= ð5:190Þ
b :M:N
N b þ 2 K0
3
Using
N bT, N
b : M = 2GN b :M:N
b = 2G, b : M : Ε = 2GN
N b Ε
yields
b Ε_
N
γ= ð5:191Þ
b
K
Using this result in the generalized Hooke’s law yields

O
_Σ = M - 2G N
b b
N : Ε_ ð5:192Þ
Kb
The constitutive tensor given by Eq. (5.192) is equivalent to Eq. (5.183) where
the coupling between strains and curvatures becomes evident in the off-diagonal
terms in the generalized constitutive tensor.
5.10.2 Rate-Dependent Plasticity (Viscoplasticity) Without

Degradation
In classical continuum mechanics, in the case of a rate-dependent material, the

conditions established by Eqs. (5.170a), (5.170b), and (5.171) are replaced by a
constitutive equation of the form:
hφðF Þi
γ= ð5:193Þ
η
where η represents viscosity. A rate-independent material F satisfies conditions

given by Eqs. (5.170a), (5.170b), and (5.171), and additionally stress states such
F(σ, ℓ-1m, α) > 0 are ruled out. In other words, the state of stress cannot be outside
the yield surface. In the case of a rate-dependent material, on the other hand, the
intensity of the viscoplastic flow is proportional to the distance of the state of stress
to the yield surface defined by F(σ, ℓ -1m, α) = 0. Therefore, the yield condition for a
rate-dependent material can be given by
F = ΘðγηÞ ð5:194Þ
whereas Θ(γη) = φ-1(γη)

5.10.3 Introducing the Thermodynamic State Index
In the following formulation, for the sake of simplicity in presentation, we will

ignore the derivatives with respect to entropy. In the Cosserat continuum, TSI is
introduced into the formulation using the same approach as in the classical contin-
uum; stress-strain and couple stress-curvature relations can be given by the third law
of the unified mechanics theory as follows:
σ_ ij = ð1 - ΦÞC ijkl ε_ ekl ð5:195aÞ

ℓ - 1m
_ ij = ð1 - ΦÞDijkl ℓχ_ ekl ð5:195bÞ
And the yield function now becomes

rffiffiffi
2
F σ, ℓ - 1 m, α, Φ = ξΦ - K ðαÞ ð5:196Þ
3
where superscript Φ indicates modification due to entropy generation at each loading

step. The thermodynamic state index can be anisotropic, Φ. The rate-dependent
constitutive model equations are summarized in Table 5.5. The constitutive model
can also be written in the equivalent alternative form given in Table 5.6, which is
more convenient for the numerical implementation of the algorithm using a return
mapping scheme. Using the Cosserat continuum generalized stress and strain defi-
nitions, for small strain problems, can be given by
θ
Σ_ = ð1 - ΦÞM e : Ε_ - Ε_ - Ε_
vp
ð5:197Þ
Table 5.5 UMT rate-dependent-strain gradient formulation 1

Third Law of UMT
σ_ ij = ð1 - ΦÞCijkl ε_ ekl
ℓ - 1m _ ij = ð1 - ΦÞDijkl ℓχ_ ekl
Yield function
qffiffi
F σ, ℓ - 1 m, α = ξΦ - 23K ðαÞ
Normality rule
∂F f ij ∂F -1 C ij
ε_ pl
ij = γ ∂σ ij γ n and ℓχ_ pl
γb ij = γ ∂ℓ - 1 m γℓ γb
ν
ij
ξij ξij
Isotropic and kinematic hardening laws

qffiffi
α_ = 23γ
Φ
β_ ij = ð1 - ΦÞ 32 H 0 γb
nij and ℓ η_ Φ 2 0
ij = ð1 - ΦÞ 3 H γb
νij .
Consistency parameter
γ = hφðηFÞi
Table 5.6 UMT rate depen- Third law of UMT

dent model-strain gradient
σ_ ij = ð1 - ΦÞCijkl ε_ ekl ℓ - 1m
_ ij = ð1 - ΦÞDijkl ℓχ_ ekl
formulation 2
Yield function
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffi
FðΣ, αÞ = ξΦ PξΦ - 23K ðαÞ
T
Flow rule
Ε_ = γ ð1 Pξ
vp
- ΦÞ
Hardening laws
qffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Φ
α_ = 23γ ð1 - ΦÞ ξΦ PξΦ b
X_ = γ 32 H 0 ðαÞð1 - ΦÞN
T
Consistency parameter
γ = hφðηFÞi
θ
where E_ is the total strain rate, Ε_ is the viscoplastic strain rate, and Ε_ is the
vp
thermal strain rate. The yield function can be given by

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffi
2
F ðΣ, αÞ = ξΦ P ξΦ -
T
K ðαÞ ð5:198Þ
3
where ξΦ = ΣΦ - XΦ is the generalized

h relative
i stress previously introduced with
Φ Φ -1 Φ
the generalized back-stress X βij , ℓ ηij , where the following definitions
are used:
Φ 2
β_ ij = ð1 - ΦÞ H 0 γ b
nij ð5:199Þ
3
2 0
ℓη_ Φ
ij = ð1 - ΦÞ 3 H γ b
νij ð5:200Þ
In Eq. (5.198) P is a constant matrix; hence, PξΦ gives the deviatoric component
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
of ξΦ and 23 ξΦ P ξΦ is the von Misses equivalent stress for the generalized relative
T
stress vector. K(α) represents the radius of the yield surface. The direction of plastic
flow and the hardening laws are given for the associative plasticity case as follows:
Pξ
Ε_ = γ
vp
ð5:201Þ
ð1 - ΦÞ
rffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
γ ð1 - ΦÞ ξΦ PξΦ
T
α_ = ð5:202Þ
3
Φ 2 b
X_ = γ H 0 ðαÞð1 - ΦÞN ð5:203Þ
3
In Eq. (5.203) the hardening modulus H′ can be a nonlinear function of the

hardening parameter α. In this formulation, Eq. (5.203) assumes that the couple
stress continuum back-stresses evolve in the same manner as the symmetric stress
tensor back-stress. However, this is not essential for the formulation, because they
can be different. The couple back-stress is needed to allow for the uniform move-
ment of the yield surface in the enhanced (Cosserat continuum) stress space.
Viscoplastic Creep Law
The evolution of the generalized viscoplastic strain rate is specified by a function
representative of the viscoplastic flow micro-mechanisms in the material. In the
following formulation, the same creep law is assumed to hold for both the strains and
the strain gradients. The following viscoplastic strain rate function is one of the most
generalized models for low melting point solder alloys. It accounts for multiple creep
micro-mechanisms in the polycrystalline alloys as well as accounting for coarsening
of the crystals:
n p
AD0 Eb hF i b ∂F
ε_ vp
ij = e - Q=Rθ ð5:204Þ
kθ E d ∂σ ij
n p
AD0 Eb hF i b ∂F
ℓ χ_ vp
ij = e - Q=Rθ - 1 ð5:205Þ
kθ E d ∂ℓ mij
n p
_Ε vp = AD0 Eb hF i b
e - Q=Rθ
∂F
ð5:206Þ
kθ E d ∂Σ
where A, n, p are material parameters, D0 is the temperature-independent diffusivity

constant, k is Boltzmann’s constant, b is Burger’s vector, R is the gas constant, E is
the elastic modulus, d is the initial grain size, Q is the creep activation energy, and θ
is the temperature in Kelvin.
The fluidity parameter η can be summarized as
p
kθ d
η= eQ=Rθ ð5:207Þ
AD0 E n - 1 b b
5.10.4 Entropy Generation Rate in Cosserat Continuum
In the Cosserat continuum couple stress theory, the internal energy can be given by
de
ρ = σ Sij Dij þ mji χ ij þ ρr - qi,i ð5:208Þ
dt
where σ Sij is the symmetric part of the Cauchy stress tensor, Dij is the rate of
deformation tensor, mji is the couple stress tensor, and χ ij are the corresponding
curvatures. The rate of change of the Helmholtz free energy φ is written in terms of
the symmetric part of the Cauchy stress tensor; thus,
dφ
ρ = σ Sij Deij ð5:209Þ
dt
Combining Eqs. (5.208) and (5.209), the difference between the changes in the
internal energy and the Helmholtz free energy with respect to a reference state in the
presence of couple stresses can be given by
Zt2 Zt2 Zt2 Zt2

1 1 1
Δe - Δφ = σ Sij Dpij dt þ mji χ_ pl dt þ r dt - qi,i dt ð5:210Þ
ρ ρ ij ρ
t1 t1 t1 t1
5.10.5 Integration Algorithms
In the material nonlinear finite element method, the solution is achieved in small
increments with at every increment Newton-Raphson iteration method is used to find
the equilibrium. At every increment, the problem may be regarded as analogous to a
strain-controlled test. At the beginning of the time step, the total and viscoplastic
strain fields and the internal state variables are considered to be known. Assuming
that displacement increment is known, the basic problem is to update the field
variables to their new values at the end of the time step in a manner consistent
with the constitutive model. Below we present the integration algorithm. First, the
case of a rate-independent model without degradation using both a radial return and
return mapping scheme is presented. Then the rate-dependent model with degrada-
tion case with the return mapping algorithm is presented in Table 5.6.
Radial Return Algorithm: Rate-Independent Model with Linear Isotropic
Hardening
Using Eqs. (5.148a) and (5.148b) and the backward Euler finite difference integra-
tion scheme, we can write
b nþ1
Εpnþ1 = Εpn þ Δγ N
rffiffiffi
2 ð5:211Þ
αnþ1 = αn þ Δγ
3
Now consider the following trial state obtained after freezing the plastic flow:
Σtrnþ1 = Σn þ 2GΔΕnþ1
ð5:212Þ
Σnþ1 = Σtr - 2GΔγ Nb nþ1
nþ1
from which it is concluded that
Σnþ1 þ 2GΔγ = Σtrnþ1 ð5:213aÞ
and
b nþ1 = Σnþ1
tr
N ð5:213bÞ
Σtrnþ1
Assuming linear isotropic hardening yields,

rffiffiffi
2 2
Σtrnþ1 - K ðαn Þ - K 0 Δγ - 2GΔγ = 0 ð5:214Þ
3 3
which is solved for the consistency parameter Δγ

Linearization of the Incremental Solution Process
Consider
Σ_ nþ1 = M : Ε_ nþ1 - 2GΔγ N

b nþ1 ð5:215Þ
and then it follows that

" #
dΣnþ1 O dΔγ d b nþ1
N
= M - 2G Nb nþ1 þ Δγ ð5:216Þ
dΕnþ1 dΕnþ1 dΕnþ1
using
dΔγ 1b
= N ð5:217Þ
b nþ1
dΕnþ1 K
and
b nþ1
dN 2G
= H nþ1 ð5:218Þ
dΕnþ1 Σtrnþ1
where
2‘ O O 3
b - 1I I-b
n b
n 0
H nþ1 = 4 3 nþ1 5 ð5:219Þ
nþ1 ‘
d N
0 -b
v bv
nþ1 nþ1
thus,
2 3
N
dΣnþ1 b
6N b nþ1
N Δγ2G 7
= M - 2G4 nþ1 þ H nþ1 5 ð5:220Þ
dΕnþ1 b
K Σtrnþ1
Expanding Eq. (5.220) yields
M ep
nþ1
2 O O 3
N ‘b - 1I N 2G
6 κI Iþ2Gδnþ1 I - 2Gθnþ1bnnþ1 bnnþ1 - b
n bvnþ1 7
=4 3 O b nþ1
K 5
2G ‘
b N
- b
nnþ1 bvnþ1 2Gδnþ1 - 2Gθnþ1 v bv
b
K
nþ1 nþ1
ð5:221Þ
When Δt → 0 Eq. (5.221) approaches (5.183). The coupling between the

curvature and strain in the tangent stiffness matrix is apparent
Return Mapping Algorithm: Rate-Dependent Model-Combined
Isotropic/Kinematic Hardening with Degradation.
The rate-dependent model, Table 5.6, is integrated using a return mapping algorithm
as presented in Simo and Hughes (1997). A backward Euler integration scheme
yields the following set of algorithmic equations:
Εnþ1 = Εn þ ΔΕnþ1 ð5:222Þ

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 T
αnþ1 = αn þ Δγ ð1 - ΦÞ ξ Pξ ð5:223Þ
3 nþ1 nþ1
2
Xnþ1 = Xn þ Δγ H 0 ð1 - ΦÞξnþ1 ð5:224Þ
3
The standard operator split technique defines the following trial state:
Σtr
nþ1 = Σn þ ð1 - ΦÞMΔΕ nþ1 ð5:225Þ
Using Eq. (5.225), we can write the following relations:
Σnþ1 = Σtr
nþ1 - MΔγPξnþ1 ð5:226Þ
nþ1 = Σnþ1 - X n
ξtr tr
ð5:227Þ
From (5.226) and (5.227), updated relative stress can be obtained in terms of the
algorithmic consistency parameter Δγ:
1
ξnþ1 = ΞðΔγ Þ M - 1 ξtr ð5:228Þ
1 þ 23 H 0 Δγ ð1 - ΦÞ nþ1
where
" #-1
ΔγP
ΞðΔγ Þ = M - 1 þ 2 0
ð5:229Þ
1 þ 3 H ð1 - ΦÞΔγ
To complete the above algorithm, it is still necessary to compute the consistency

parameter Δγ, which can be obtained from the yield condition and the constitutive
model. Thus
rffiffiffi
ηΔγ 2 ηΔγ
F ðΔγ Þ - Θ bf nþ1 - K ðαnþ1 Þ - Θ =0 ð5:230Þ
Δt 3 Δt
1=2
where b f nþ1 = ξTnþ1 Pξnþ1 and Θ ηΔγ
Δt = φ
- 1 ηΔγ
Δt as defined in the previous
section.
Equation (5.230) is a scalar nonlinear equation in the consistency parameter Δγ,
which can be solved by a local iteration. In the Newton-Raphson iteration scheme,
the elastoplastic tangent modulus consistent with the integration scheme is needed to
preserve the convergence. Linearizing the above set of equations yields
!
dΣnþ1 1 O
= ð1 - ΦÞ ΞðΔγ Þ - Ν Ν ð5:231Þ
dΕnþ1 1 þ β~
where
ΞðΔγÞPξnþ1
Ν = qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi ð5:232Þ
ξTnþ1 PΞðΔγÞPξnþ1

~β = ð1 - ΦÞ 2θ1 b 2 θ2 dΘ b 1
f nþ1 ðK 0 θ1 þ H 0 θ2 Þ þ 1 f nþ1 T ð5:233Þ
3θ2 θ2 dΔγ ξnþ1 PΞðΔγ ÞPξnþ1
2
θ1 = 1 þ H0 ð1 - ΦÞΔγ
3
2 0
θ2 = 1 - K ð1 - ΦÞΔγ
3
Finding the Consistency Parameter

Recalling Eq. (5.198) and updating the formula given by Eq. (5.228) yields
rffiffiffi
2
F ðΔγ Þ = bf nþ1 - K ðαnþ1 Þ = 0 ð5:234Þ
3
1
ξnþ1 = ΞðΔγ Þ M - 1 ξtr ð5:235Þ
1 þ 23 H 0 Δγ ð1 - ΦÞ nþ1
In Newton-Raphson iteration to find the solution x of a nonlinear equation of the

general form F(x) = 0, the ith correction to the solution is computed as Δxi =
FðΔxi - 1 Þ
Δxi - 1 - JðΔxi - 1 Þ where J Δxi - 1 = ∂Δx
∂F
i - 1 represents the “Jacobian” of F(x).
Solution of Eq. (5.230) with a local iteration requires the “Jacobian” J(Δγ) to be
defined. In the Newton-Raphson iteration, the Jacobian represents the derivative of
the function; hence,
rffiffiffi
dF dΘ dbf 2 dK dΘ
J ðΔγ Þ = - nþ1 - - ð5:236Þ
dΔγ dΔγ dΔγ 3 dΔγ dΔγ
Equation (5.235) can be given by
h i-1
2
ξnþ1 = I þ H 0 Δγ ð1 - ΦÞI þ ΔγPM ξtr
nþ1 ð5:237Þ
3
db
f nþ1
To find dΔγ , the symmetric matrices P and M are diagonalized. Since both
matrices share the same characteristic space, the following relationships can be
written:
P = QΛP QT ð5:238Þ
M = QΛM QT ð5:239Þ
and then
PM = QΛP ΛM QT
where QT = Q-1. The matrices for a plane strain idealization are

h i
Q= Q1
0
0
I3 × 3
with ð5:240aÞ
" pffiffi pffiffi #
0ffiffi
p 6 ffiffi
2= p 1=p3ffiffi
Q1 = -pffiffi 2=2 - 1=p6ffiffi 1=p3ffiffi ð5:240bÞ
2=2 - 1= 6 1= 3
h i
P= P1
0
0
P2
with ð5:241aÞ
" #
2=3 - 1=3 - 1=3
P1 = - 1=3 2=3 - 1=3 and P2 = DIAG½2, 1, 1 ð5:241bÞ
- 1=3 - 1=3 2=3
ΛP = DIAG½1, 1, 0, 2, 1, 1 and ΛM = DIAG½2G, 2G, 3λ þ 2G, G, 2G, 2G
Now the update formula Eq. (5.237) can be written as
h i-1
2
ξnþ1 = 1 þ ΔγH0 I þ ΔγQΛP ΛM QT ξtr
nþ1 ð5:242Þ
3
or after using the following definition:

-1
2
ΓðΔγÞ = 1 þ ΔγH0ð1 - ΦÞ I þ ΔγΛP ΛM ð5:243Þ
3
As a result, this expression becomes
ξnþ1 = QΓðΔγ ÞQT ξtr

nþ1 ð5:244Þ
where
ΓðΔγ Þ
2 1 1 1 3
, , ,
6 1þ
2 2 2
ð1 - ΦÞH 0 þ 2G Δγ 1 þ ð1 - ΦÞH 0 þ 2G Δγ 1 þ ð1 - ΦÞH 0 Δγ 7
6 7
= DIAG6 3 3 3 7
6 1 1 1 7
4 , , 5
2 2 2
1 þ ð1 - ΦÞH 0 þ 2G Δγ 1 þ ð1 - ΦÞH 0 þ 2G Δγ 1 þ ð1 - ΦÞH 0 þ 2G Δγ
3 3 3
ð5:245Þ
Making use of Eq. (5.214), bf nþ1 becomes

h i1=2
bf nþ1 = ξtrT QT ΓðΔγ ÞQPQT ΓðΔγ Þξtr ð5:246Þ
nþ1 nþ1
db
f nþ1
Using Eqs. (5.244), (5.245), and (5.246), dΔγ can be defined as
dbf nþ1 b
g
= nþ1 ð5:247Þ
dΔγ b
f nþ1
where
T
b
gnþ1 = ξtr
nþ1 QGðΔγ ÞQ PQΓðΔγ ÞQ ξnþ1
T T tr
ð5:248Þ
dΓðΔγ Þ
GðΔγ Þ = = DIAG½G11 , G22, G33 , G44 , G55 , G66 with
dΔγ
2 0

3 ð1 - ΦÞH þ 2G
G11 = G22 = G44 = G55 = G66 = - 2
1 þ 23 ð1 - ΦÞH0 þ 2G Δγ
0
3 ð1 - ΦÞH
2
G33 = - 2
1 þ 23 ð1 - ΦÞH0Δγ
Using the definition of K(α), Eq. (5.247) and θ2 from Eq. (5.231) yield the
Jacobian needed for the local Newton iteration:
dF ðΔγ Þ b
g 2
= θ2 nþ1 - K 0bf nþ1 ð5:249Þ
dΔγ bf 3
nþ1
The complete Newton-Raphson integration algorithm is shown in Table 5.7. The

local Newton iteration for the determination of the consistency parameter is shown in
Table 5.8.
Table 5.7 Return mapping integration algorithm

Update strain Εn + 1 = Εn +— Δu
θ

Compute trial state nþ1 = ð1 - ΦÞM Ε nþ1 - Ε n - Εnþ1
Σtr vp
nþ1 = Σnþ1 - X n
ξtr tr
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffi
Compute trial yield function
nþ1 = ξtrT
nþ1 Pξ nþ1 - 3K ðαn Þ
tr
Ftr 2
nþ1 > 0 THEN

IF Ftr
Call Newton local and solve f(Δγ) = 0 for Δγ
h i-1
Compute
ΞðΔγ Þ = M - 1 þ 1þ2H0 ðΔγP
1 - ΦÞΔγ
3
Update ξnþ1 = ΞðΔγ Þ 1þ2H0 ð11- ΦÞΔγ M - 1 ξtr

nþ1
3
X nþ1 = X n þ Δγ 23 H0 ð1 - ΦÞξnþ1
Σn + 1 = ξ n + 1 + Xn + 1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
αnþ1 = αn þ Δγ ð1 - ΦÞ 23 ξTnþ1 Pξnþ1
Pξ
Εvp
nþ1 = Ε n þ Δγ ð1 - ΦÞ
vp nþ1
Εenþ1 = Εnþ1 - Εvp

nþ1 - Εθnþ1
Compute consistent Jacobian

N
dΣnþ1
= ð1 - ΦÞ ΞðΔγ Þ - 1
Ν Ν
dΕnþ1 ð1þ~βÞ
ELSE
Elastic step (EXIT)
END IF
EXIT
Table 5.8 Local Newton iteration to determine the consistency parameter

Let Δγ (0) ← 0
α(0)n + 1 ← αn
Start Iterations
DO_UNTIL jF(Δγ j < tol
k←k+1
Compute Δγ (k + 1)
ðkÞ pffiffiffiffiffi
F Δγ ðkÞ = bf Δγ
nþ1 - 2=3K ðαnþ1 Þ - Θ ηΔγ ðkÞ
Δt
bg ðΔγðkÞ Þ
J Δγ ðkÞ = θ2 nþ1 - 23 K 0 ð1 - ΦÞbf nþ1 Δγ ðkÞ - dΘ
bf nþ1 ðΔγðkÞ Þ dΔγ
FðΔγ ðkÞ Þ
Δγ ðkþ1Þ ← Δγ ðkÞ - JðΔγðkÞ Þ
pffiffiffiffiffiffiffiffi
αnþ1 ðkþ1Þ ← αnþ1 ðkÞ þ 2=3bf nþ1 ð1 - ΦÞ Δγ ðkÞ
END DO_UNITL
References
Aero, E., & Kuvshinsky, E. (1961). Fundamental equations of the theory of elastic media with
rotationally interacting particles. Soviet Physics Solid State, 2, 1272.
Armstrong, P., & Frederick, C. (1966). A mathematical representation of the multiaxial
Bauschinger effect. CEGB Report RD/N731.
Basaran, C., Zhao, Y., Tang, H., & Gomez, J. (2005). A damage mechanics based unified
constitutive model for Pb/Sn solder alloys. ASME Journal of Electronic Packaging, 127(3),
208–214.
Chaboche, J. (1989). Constitutive equations for cyclic plasticity and viscoplasticity. International
Journal of Plasticity, 3, 247–302.
Cosserat, E., & Cosserat, F. (1909). Theorie des corps deformables. A Hermann & Fils.
De Borst, R. (1993). A generalization of J2-flow theory for polar continua. Computer Methods in
Applied Mechanics and Engineering, 103, 347–362.
De Borst, R., & Muhlhaus, H. (1992). Gradient dependent plasticity: Formulation and algorithmic
aspects. International Journal for Numerical Methods in Engineering, 35, 521–539.
DeHoff, R. T. (1993). Thermodynamics in materials science. McGraw Hill.
Eringen, A. C. (1968). Theory of micropolar elasticity. In H. Leibowitz (Ed.), Fracture, and
advanced treatise (pp. 621–729). Academic Press.
Fleck, N., & Hutchinson, J. (1997). Strain gradient plasticity. Advances in Applied Mechanics, 33,
295–361.
Fung, Y. C., & Tong, P. (2001). Classical and computational solid mechanics. Advanced series in
engineering science (Vol. 1). World Scientific.
Jean-Louis Chaboche. (1990). Mechanics of Solid Materials, Cambridge University Press; 1st
edition
Kashyap, B., & Murty, G. (1981). Experimental constitutive relations for the high-temperature
deformation of a Pb-Sn eutectic alloy. Material Science and Engineering, 50, 205–213.
Koiter, W. T. (1964). Couple of stresses in the theory of elasticity. Proceedings of the Koninklijke
Nederlandse Akademie van Wetenschappen, Series B: Physical Sciences, 67, 17–44.
Krajcinovic, D. (1996). Damage mechanics. North-Holland series in applied mathematics and
mechanics. Elsevier.
Lee, Y., & Basaran, C. (2011). A creep model for solder alloys. Transactions of the ASME: Journal
of Electronic Packaging, 133(4), 044501-1.
References 307
Lemaitre, J. (1996). A course on damage mechanics. Springer-Verlag.

Lemaitre, J., & Chaboche, J. L. (1990). Mechanics of solid materials. Cambridge University Press.
Mazur, P., & De Groot, S. (1962). Non-equilibrium thermodynamics. Dover Publications, Inc..
Mindlin, R. (1964). Micro-structure in linear elasticity. Archive for Rational Mechanics and
Analysis, 16, 51–78.
Mindlin, R. (1965). Second gradient of strain and surface tension in linear elasticity. International
Journal of Solids and Structures, 1, 417–438.
Mindlin, R., & Tiersten, H. F. (1962). Effects of couple-stresses in linear elasticity. Communicated
by C. Truesdell. Archive for Rational Mechanics and Analysis, 11, 415–448.
Prager, W. (1955). The theory of plasticity: A survey of recent achievements (James Clayton
Lecture). Proceedings of the Institution of Mechanical Engineers, 169, 41.
Rabotnov, Y.N. (1969) Creep Problems in Structural Members, Wiley
Shu, J. Y., & Fleck, N. A. (1999). Strain gradient plasticity: Size-dependent deformation of
bicrystals. Journal of the Mechanics and Physics of Solids, 47, 297–324.
Simo, J., & Hughes, T. (1997). Computational inelasticity. Interdisciplinary applied mathematics.
Springer.
Toupin, R. (1962). Elastic materials with couple-stresses. Archive for Rational Mechanics and
Analysis, 11, 385–414.
Xia, Z., & Hutchinson, J. (1996). Crack tip fields in strain gradient plasticity. Journal of the
Mechanics and Physics of Solids, 44(10), 162–1648.
Chapter 6
Thermomechanical Analysis
of Particle-Filled Composites
6.1 Introduction
Modeling the macroscopic constitutive response of composite materials starting

from the local description of the microstructure is necessary. Composite materials
are heterogeneous. However, continuum mechanics formulation requires homoge-
nization of the medium. The homogenization method used to go from the local level
to the macro-level and the localization from macro-level quantities to the
corresponding local micromechanics variables must be well defined.
Using the Eshelby method (Eshelby, 1957; Ju & Chen, 1994a, b) formulated the
effective mechanical properties of elastic multiphase composites containing many
randomly dispersed ellipsoidal inhomogeneities with perfect bonding. Within the
context of the representative volume element (RVE), four governing
micromechanical ensemble-volume averaged field equations utilized to relate
ensemble-volume averaged stresses, strains, volume fractions, Eigen strains, particle
shapes, and orientations, and elastic properties of constituent phases of the particu-
late composites. Ju and Tseng (1996) formulated combining a micromechanical
interaction approach and the continuum plasticity to predict effective elastoplastic
behavior of two-phase particulate composite containing many randomly dispersed
elastic spherical inhomogeneities. Explicit pairwise interparticle interactions are
considered in both the elastic and plastic responses. Furthermore, the ensemble-
volume averaging procedure is employed and the formulation is of complete second
order.
Let us consider a perfectly bonded two-phase composite consisting of an elastic
matrix (named “phase 0”) with bulk modulus k0 and shear modulus μ0, and randomly
dispersed elastic spherical particles (named “phase 1”) with bulk modulus k1 and
shear modulus μ1. For the effective bulk modulus k and effective shear modulus μ,
this two-phase composite for the noninteracting solution, neglecting the interparticle
interaction effects, was derived by Ju and Chen (1994a) as follows:
https://doi.org/10.1007/978-3-031-18621-9_6
310 6 Thermomechanical Analysis of Particle-Filled Composites

3ð1 - v0 Þðk 1 - k 0 Þφ
k = k0 1þ ð6:1Þ
3ð1 - v0 Þk0 þ ð1 - φÞð1 þ v0 Þðk 1 - k 0 Þ

15ð1 - v0 Þðμ1 - μ0 Þφ
μ = μ0 1þ ð6:2Þ
15ð1 - v0 Þμ0 þ ð1 - φÞð8 - 10v0 Þðμ1 - μ0 Þ
where φ is the particle volume fraction and v0 is Poisson’s ratio of the matrix.
If the effects due to interparticle interactions are included for the two-phase elastic
composites with randomly located spherical particles, the solutions for the effective
bulk modulus k and shear modulus μ are provided by Ju and Chen (1994b):

30ð1 - v0 Þφð3γ 1 þ 2γ 2 Þ
k = k0 1þ ð6:3Þ
3α þ 2β - 10ð1 þ v0 Þφð3γ 1 þ 2γ 2 Þ

30ð1 - v0 Þφγ 2
μ = μ0 1 þ ð6:4Þ
β - 4ð4 - 5v0 Þφγ 2
with

k0 μ0
α = 2ð5v0 - 1Þ þ 10ð1 - v0 Þ - ð6:5Þ
k1 - k0 μ1 - μ0
μ0
β = 2ð4 - 5v0 Þ þ 15ð1 - v0 Þ ð6:6Þ
μ1 - μ0
And

5φ 8α
γ1 = ð 13 - 14v Þv - ð 1 - 2v Þ ð1 þ v Þ ð6:7Þ
8β2
0 0
3α þ 2β 0 0

1 5φ 6α
γ2 = þ 25 - 34v þ 22v 2
- ð 1 - 2v Þ ð 1 þ v Þ ð6:8Þ
2 16β2 0 0 3α þ 2β 0 0
Alternatively, effective Young’s modulus E and Poisson’s ratio v of particulate

composites are easily obtained through the following relationships:
9k μ
E = ð6:9Þ
3k þ μ
3k - 2μ
v = ð6:10Þ
6k þ 2μ
The experimental validation of the Ju and Chen model was provided by Nie and
Basaran (2005), by comparing the analytical predictions with the experimental data
on the particulate composite prepared using lightly cross-linked poly-methyl meth-
acrylate (PMMA) filled with alumina trihydrate (ATH). All filler particles are
assumed spherical and both the matrix and the filler are isotropic elastic. The
material properties involved are as follows:
6.1 Introduction 311
12
Composite [test data]

Young’s Modulus (GPa) 10
Pairwise interacting solution

8
6
Noninteracting solution
4
PMMA [test data]

2
0
0 0.1 0.2 0.3 0.4 0.5 0.6
Particles Volume Fraction
Fig. 6.1 Effective Young’s modulus as a function of particle volume fraction. (After Nie (2005))
ATH: E1 = 70 GPa, v1 = 0.24.

PMMA: E0 = 3.5 GPa, v0 = 0.31
Particle volume fraction: φ = 0.48
The experimental mean value of Young’s modulus for this composite is about
10.2 GPa at room temperature. Figure 6.1 shows the comparisons among the
analytical solution (including pairwise interacting solution and noninteracting solu-
tion) and experimental results. It is observed that agreement between the pairwise
interacting prediction and experimental data is very good for effective Young’s
modulus. Based on the foregoing preliminary analytical and experimental compar-
isons, it appears that Ju and Chen’s analytical micromechanical approach offers a
simple, approximate, yet sufficiently accurate framework for the prediction of
effective elastic moduli of two-phase particle-filled composites.
Particle-filled composites consist of the bulk matrix, filler particles, and interfa-
cial transition zone around particles, which often have very different properties such
as coefficient of thermal expansion (CTE) and stiffness. Within the particulate
composite microstructure, there are micro-stresses due to the coefficient of the
thermal expansion (CTE) mismatch between the matrix and the filler particles. For
the composite prepared using lightly cross-linked poly-methyl methacrylate
(PMMA) filled with alumina trihydrate (ATH), these micro-stresses can be imaged
using the fact that PMMA is stress optically birefringent. The micro-stresses imaged
using this technique are indeed due to CTE mismatch as they dissipate at tempera-
tures close to the glass transition temperature Tg of the PMMA. These stresses can be
quantified by calculation or by direct measurement and are of the order between 15%
and 75% of the tensile strength of the composite. Therefore, the thermal stress
associated with the CTE mismatch between the matrix and particles is an important
factor in the failure of particulate composites subjected to thermomechanical loads.
6.2 Ensemble-Volume Averaged Micromechanical Field

Equations
To obtain effective constitutive equations and properties of random heterogeneous

composites, one typically performs the ensemble-volume averaging process within a
mesoscopic representative volume element (RVE). The volume-averaged stress
tensor is defined as
2 3
Z Z Xn Z
1 16 7
σ= σðxÞdx = 4 σðxÞdx þ q=1
σðxÞdx5 ð6:11Þ
V V
V V0 Vq
where V is the volume of an RVE, V0 is the volume of the matrix, Vq is the volume of
the qth-phase particles, and n denotes the number of particulate phases [fillers] of
different materials (excluding the matrix).
Similarly, the volume-averaged strain tensor is defined as
2 3
Z Z Xn Z h Xn i
1 16 71
ε= εðxÞdx = 4 εðxÞdx þ εðxÞdx
5 V 0 ε0 þ V ε
V V q=1 V q=1 q q
V V0 Vq
ð6:12Þ
According to Eshelby’s equivalence principle, the perturbed strain field ε′(x)

induced by inhomogeneity can be related to the specified Eigen strain ε(x) by
replacing the inhomogeneities with the matrix material. Sometimes the inhomoge-
neities may also involve their Eigen strain caused by, i.e., phase transformation,
precipitation, plastic deformation, or CTE mismatch between different constituents
of the composites. However, it is not necessary to attribute the Eigen strain to any
specific source. Then the total perturbed stress is the sum of the two parts, one caused
by the inhomogeneity, and the other one is the Eigen stress caused by the Eigen
strain. For the domain of the qth-phase particles with elastic stiffness tensor Cq, Mura
(1987) provided the following stress-strain relation:
h i h i
Cq : ε0 þ ε0 ðxÞ - εTq ðxÞ = C0 : ε0 þ ε0 ðxÞ - εTq ðxÞ - εq ðxÞ ð6:13Þ
where C0 is the stiffness tensor of the matrix and ε0 is the uniform elastic strain field
induced by far-field loads for a homogeneous matrix material only. εTq is its Eigen
6.2 Ensemble-Volume Averaged Micromechanical Field Equations 313
strain associated with the qth particle, and εq is the fictitious equivalent Eigen strain
by replacing the qth particles with the matrix material. ε′(x) is the perturbed strain due
to distributed Eigen strain εT and ε is associated with all the particles in the RVE.
The stress tensor for the matrix is given in Newtonian mechanics by
σ0 = C 0 : ε 0 ð6:14Þ
The strain at any point within an RVE is decomposed into two parts, the uniform
strain and the perturbed strain due to the distributed Eigen strain. It is emphasized
that the Eigen strain ε and εT are nonzero in the particle domain and zero in the
matrix domain, respectively. The perturbed strain field induced by all the distributed
Eigen strains ε and εT can be expressed by (Mura, 1987)
Z

ε0 ðxÞ = Gðx - x0 Þ : εT ðx0 Þ þ ε ðx0 Þ dx0 ð6:15Þ
V
where x, x′ 2 V, and G is Green’s function in a linear elastic homogeneous matrix.

For a linear elastic isotropic matrix, the fourth-rank tensor Green’s function can be
given by (Ju & Chen, 1994a)
Gijkl ðx - x0 Þ
1
= F ð - 15, 3v0 , 3, 3 - 6v0 , - 1 þ 2v0 , 1 - 2v0 Þ ð6:16Þ
8π ð1 - v0 Þr 3 ijkl
where r x - x′, r j rj, and v0 is Poisson’s ratio of the matrix. The components of
the fourth-rank tensor F are defined by

F ijkl ðBm Þ = B1 ni nj nk nl þ B2 δik nj nl þ δil nj nk þ δjk ni nl þ δjl ni nk þ B3 δij nk nl
ð6:17Þ
þ B4 δkl ni nj þ B5 δij δkl þ B6 δik δjl þ δil δjk
with the unit normal vector n r/r and index m = 1–6.

From Eqs. (6.13) and (6.15), we arrive at
Z

- Aq : εq ðxÞ = ε0 - εTq ðxÞ þ Gðx - x0 Þ : εT ðx0 Þ þ ε ðx0 Þ dx0 x0 2 V ð6:18Þ
V
where
-1
Aq = Cq - C0 C0 ð6:19Þ
Furthermore, the total local strain field ε(x) can be expressed as

Z

ε ð xÞ = ε 0 þ ε 0 ð xÞ = ε0 þ Gðx - x0 Þ : εT ðx0 Þ þ ε ðx0 Þ dx0 ð6:20Þ
V
Using the renormalization procedure of the volume-averaged strain tensor given

by Ju and Chen (1994a), as
Z Z
1
ε = ε0 þ Gðx - x0 Þ : εT ðx0 Þ þ ε ðx0 Þ dx0 dx = ε0 þ s
V
V V
h Xn i
: φ εTq þ εq
q=1 q
ð6:21Þ
where s is a constant tensor for unidirectionally aligned and similar ellipsoidal filler
particles. If the linear elastic matrix material is isotropic and all inclusions are
spherical, then the s takes the form of the Eshelby tensor S:
1
Sijkl = ð5v0 - 1Þδij δkl þ ð4 - 5v0 Þ δik δjl þ δil δjk ð6:22Þ
15ð1 - v0 Þ
where δij signifies the Kronecker delta.

Similarly, using Eqs. (6.11)–(6.13), the ensemble-volume averaged stress field
can be given by
2 3
Z Xn Z h i
16 7
σ= 4 C0 : εðxÞdx þ q=1
C0 : εðxÞ - εTq - εq dx5
V
V0 Vq
h Xn h ii ð6:23Þ
1
= V 0 C0 : ε0 þ =
V q C0 : εq - εTq - εq
V h
q 1
i
Xn

= C0 : ε - q=1
φ q ε T
q þ ε q
The effective elastic properties can be obtained, in principle, from Eqs. (6.18),
(6.21), and (6.23) since the variables are σ,ε,ε0 ,εq. Essentially, one needs to solve the
relationship between ε and εq, which involves the solution of the integral Eq. (6.18).
εq depends on interparticle interactions, particle-matrix interactions, and microstruc-
ture (i.e., particle sizes, orientation, shapes, volume fractions, locations, configura-
tions, and probability functions) of a composite system. For randomly dispersed
particles, one needs to obtain the ensemble-volume averaged relation between ε and
εq by averaging all possible solutions of the integral Eq. (6.18) for any particle
configurations generated according to the specified probability function.
Taking the ensemble-volume average of Eq. (6.18) overall qth-phase particles, we
obtain
6.2 Ensemble-Volume Averaged Micromechanical Field Equations 315
- Aq : εq = ε0 - εTq þ εq 0 ð6:24Þ
where
Z Z
1
ε0q = Gðx - x0 Þ : εT ðx0 Þ þ ε ðx0 Þ dx0 dx ð6:25Þ
Vq
Vq V
If all particles in the qth phase have the same ellipsoidal shape and orientation,
then εq 0 can be given by
ε0q = ε0p
q þ Sq : εq þ εq
T
ð6:26Þ
with
8 9
Z >
< Z >
=
1 XN q
ε0p
q = Gðx - x0 Þ : εT ðx0 Þ þ ε ðx0 Þ dx0 dx ð6:27Þ
Vq i=1 >
: >
;
Ωiq V - Ωiq
representing the interparticle interaction effects, where Ωiq is the domain of the qth
filler particle in the qth-phase domain Vq and Nq is the number of the [filler] phase
q particles dispersed in V. Sq is the Eshelby tensor associated with the qth particle.
From Eqs. (6.24) and (6.26), we arrive at

- Aq - Sq : εq = ε0 - εT þ Sq : εT þ ε0p
q ð6:28Þ
In summary, the three basic governing micromechanical ensemble-volume aver-

aged field equations can be given as follows:
h Xn i

σ = C0 : ε - q=1
φ q ε q þ ε T
q ð6:29Þ
Xn
ε = ε0 þ φs
q=1 q
: εTq þ εq ð6:30Þ

- Aq þ Sq : εq = ε0 - εT þ Sq : εTq þ ε0p
q ð6:31Þ
To solve Eqs. (6.29)–(6.31) and obtain the effective elastic properties of the
composite, it is essential to express the qth-phase average Eigen strain εq in terms
of the volume average strain ε . Namely, one must solve the integral Eq. (6.18)
exactly for each phase, which involves details of random microstructure.
6.3 Noninteracting Solution for Two-Phase Composites
Let us consider a perfectly bonded two-phase composite consisting of a viscoplastic

matrix (phase 0) with elastic bulk modulus k0 and elastic shear modulus μ0, and
randomly dispersed elastic spherical particles (phase 1) with bulk modulus k1 and
shear modulus μ1. For the sake of simplicity, the von Mises yield criterion is used for
the matrix. The extension of the present framework to a general yield criterion and a
general hardening law, however, is straightforward. Accordingly, at any matrix
material point, the deviatoric stress σ and the equivalent plastic strain e p must satisfy
the following yield function:
pffiffiffiffiffiffiffiffiffiffiffiffiffi
F σD , e p = H ð σ D Þ - K ð e p Þ ð6:32Þ
where K(e p) is the isotropic hardening function of the matrix-only material. Further-
more, H(σD) signifies the square of the deviatoric stress norm in the matrix only:

H σD = σ : I D : σ ð6:33Þ
where ID denotes the deviatoric part of the fourth-rank identity tensor:
D 1 1
I ijkl = - δij δkl þ δik δjl þ δil δjk ð6:34Þ
3 2
To solve the elastoplastic response exactly, the stress at any local point must be
known and used to determine the plastic response through the local yield criterion for
all possible configurations. This approach is in general infeasible due to the com-
plexity of the statistical and microstructural information. Therefore, a framework in
which an ensemble-averaged yield criterion is constructed for the entire composite.
6.3.1 Average Stress Norm in Matrix
Total stress σ(x) at any point x in the matrix can be given by the superposition of the
far-field stress σ0 and the perturbed stress σ′ induced by the inclusion particles:
σ ð xÞ = σ 0 þ σ 0 ð xÞ ð6:35Þ
According to the Eshelby theory, the perturbed stress σ ′ at any point in the matrix
due to the presence of the particles can be given by
Z
0

σ ðxÞ = C0 : Gðx - x0 Þ : εT þ ε ðx0 Þ dx0 ð6:36Þ
V
6.3 Noninteracting Solution for Two-Phase Composites 317
where ε(x′) denotes the fictitious elastic Eigen strain in the particle-induced by
replacing the inhomogeneity [filler] with the matrix material. εT is the Eigen strain of
the inhomogeneity. Assuming εT is uniform in the particles, G(x - x′) is the fourth-
rank tensor of Green’s function defined by Eq. (6.16).
According to the Eshelby theory, the Eigen strain ε(x′) due to a single ellipsoidal
inclusion is uniform for the interior points of an isolated (noninteracting) inclusion.
Therefore, the perturbed stress for any matrix point x due to a typically isolated
inhomogeneity centered at x1 takes the form

σ0 ðxjx1 Þ = C0 Gðx - x1 Þ : εT þ ε0 ð6:37Þ
where
Z
Gðx - x1 Þ = Gðx - x0 Þdx0 For x=
2 Ω1 ð6:38Þ
Ω1
Here Ω1 is the particle domain centered at x1. Alternatively, we can derive
ρ3 1
G ð x - x1 Þ = H þ ρ2 H2 ð6:39Þ
30ð1 - v0 Þ
where the components of H1 and H2 are given by
H 1ijkl ðrÞ = 5F ijkl ð - 15, 3v0 , 3, 3 - 6v0 , - 1 þ 2v0 , 1 - 2v0 Þ ð6:40Þ
H 2ijkl ðrÞ = 3F ijkl ð35, - 5, - 5, - 5, 1, 1Þ ð6:41Þ
where r x - x1, r j rj, ρ = a/r, and a is the radius of a spherical particle. The
components of the fourth-rank tensor F are given by Eq. (6.17). In Eq. (6.37),
ε0 denotes the solution of the Eigen strain ε for the single inclusion problem,
which is given (from Eq. (6.31) when ε0p
q is dropped):
ε0 = - ðA þ SÞ - 1 : ε0 þ ðA þ SÞ - 1 ðI - SÞ : εT ð6:42Þ
where
A = ðC1 - C0 Þ - 1 C0 ð6:43Þ
We define H(x| ℘) as the square of the current deviatoric stress norm at the local
point x, which determines the plastic strain in a particulate composite for a given
phase configuration ℘. Since there is no plastic strain in the elastic particles or voids,
H(x| ℘) can be written as (Ju & Tseng, 1997)
(
σ ðxj℘Þ : ID : σðxj℘Þ x in the matrix
H ðxj℘Þ = ð6:44Þ
0 otherwise
In addition, hHim(x) is defined as the ensemble average of H(x| ℘) over all

possible realizations where point x is in the matrix phase. Matrix point receives
the perturbations from particles. Therefore, the ensemble-average stress norm for
any matrix point x can be evaluated by collecting and summing up all the current
stress norm perturbations produced by any typical particle centered at x1 in the
particle domain and averaging over all possible locations of x1, namely, for point x in
the matrix:
Z

hH im ðxÞ = H 0 þ H ðxjx1 Þ - H 0 Pðx1 Þdx1 ð6:45Þ
r>a
Here a is the radius of the particle, P(x1) denotes the probability density functions
for finding a particle centered at x1, and H0 corresponds to the far-field deviatoric
stress norm in the matrix:
H 0 = σ0 : I D : σ0 ð6:46Þ
where ID signifies the deviatoric part of the fourth-rank identity tensor.

Assuming that P(x1) is statistically homogeneous, isotropic, and uniform, and
P(x1) takes the form
N
Pðx1 Þ = ð6:47Þ
V
where N is the total number of filler particles dispersed in volume V. According to

the assumption of statistical isotropy and uniformity, Eq. (6.45) can be recast into a
more convenient form:
Z Z
N
h H i m ð xÞ ffi H 0 þ dr H ðrÞ - H 0 dA ð6:48Þ
V
r>a AðrÞ
with the help of Eqs. (6.35), (6.37), (6.44), and (6.46), plus using the following
identities:
Z
4πr2
ni nj dA = δ ð6:49Þ
3 ij
Aðr Þ
Z
4πr2
ni nj nk nl dA = δij δkl þ δik δjl þ δil δjk ð6:50Þ
15
Aðr Þ
Z Z
dr H 1 ðrÞdA = 0 ð6:51Þ
r>a Aðr Þ
Z Z
dr H 2 ðrÞdA = 0 ð6:52Þ
r>a Aðr Þ
Z Z
dr GðrÞdA = 0 ð6:53Þ
r>a Aðr Þ
Equation (6.48) can be recast as

Z Z
N
h H i m ð xÞ = H þ
0
dr σ0 ðxÞ : ID : σ0 ðxÞ dA ð6:54Þ
V
r>a Aðr Þ
After some lengthy yet straightforward derivations, one can obtain the ensemble-
averaged current stress norm at any matrix point x as:
hH im ðxÞ = σ0 : T : σ0 þ σT : T : σT - 2σ0 : T : σT ð6:55Þ
where
σT = AC1 : εT ð6:56Þ
The components of the positive definite fourth-rank tensors T and T are

T ijkl = T 1 δij δkl þ T 2 δik δjl þ δil δjk ð6:57Þ

with
φ
3T 1 þ 2T 2 = 200ð1 - 2v0 Þ2 ð6:59Þ
ð3α þ 2βÞ2
1 φ
T2 = þ 23 - 50v0 þ 35v20 2 ð6:60Þ
2 β
φ
3T 1 þ 2T 2 = 200ð1 - 2v0 Þ2 ð6:61Þ
ð3α þ 2βÞ2
φ
T 2 = 23 - 50v0 þ 35v20 2 ð6:62Þ
β
α and β are given by Eqs. (6.5) and (6.6), and φ is the filler particle volume fraction.
The ensemble-averaged current stress norm at a matrix point must be established

in terms of the macroscopic stress σ to express the effective loading function in terms
of the macroscopic stress. In the special case of uniform dispersions of identical
elastic spheres in a homogeneous matrix, the macroscopic stress and the far-field
stress take the form

σ = C0 : ε - φ ε0 þ εT ð6:63Þ

σ0 = C0 : ε - φS : ε0 þ εT ð6:64Þ
Using Eqs. (6.63), (6.64), and (6.42), the relation between the far-field stress
σ0 tensor and the macroscopic stress σ takes the form
σ = P : σ0 - Q : σT ð6:65Þ
where
P = I þ φðI - SÞðA þ SÞ - 1 ð6:66Þ
Q = φðI - SÞðA þ SÞ - 1 ð6:67Þ
with the components of P and Q given by

Pijkl = P1 δij δkl þ P2 δik δjl þ δil δjk ð6:68Þ

Qijkl = Q1 δij δkl þ Q2 δik δjl þ δil δjk ð6:69Þ
where
3P1 þ 2P2 = aφ þ 1 ð6:70Þ

1
P2 = ðbφ þ 1Þ ð6:71Þ
2
3Q1 þ 2Q2 = aφ ð6:72Þ
1
Q2 = bφ ð6:73Þ
2
20ð1 - 2v0 Þ
a= ð6:74Þ
3α þ 2β
ð7 - 5v0 Þ
b= ð6:75Þ
β
With the help of Eq. (6.65), we arrive at the alternative expression for the
ensemble-averaged current stress norm at a matrix point x as:

hH im ðxÞ = σ : T : σ þ 2σ : T : σT þ σT : T : σT ð6:76Þ

where the positive definite fourth-rank tensors T, T , and T are defined as follows:
T
T = P-1 T P-1 ð6:77Þ
T
T = P - 1 T P - 1 Q - T P - 1 ð6:78Þ
T
T = P - 1 Q T P - 1 Q þ T - 2T P - 1 Q ð6:79Þ

where the components of T,T , and T are


Tijkl = T 1 δij δkl þ T 2 δik δjl þ δil δjk ð6:81Þ

with
3T 1 þ 2T 2
3T 1 þ 2T 2 = ð6:83Þ
ðaφ þ 1Þ2
T2
T2 = ð6:84Þ
ðbφ þ 1Þ2
3T 1 þ 2T 2
3T 1 þ 2T2 = - ð6:85Þ
ðaφ þ 1Þ2
T2 1
T2 = - þ ð6:86Þ
ðbφ þ 1Þ 2 2 ð bφ þ 1Þ
3T 1 þ 2T 2
3T 1 þ 2T 2 = ð6:87Þ
ðaφ þ 1Þ2
T2 bφ - 1
T2 = þ ð6:88Þ
ðbφ þ 1Þ2 2ðbφ þ 1Þ
Because σT = AC1 : εT is spherical stress, Eq. (6.76) of the ensemble-averaged

loading function can be simplified as

hH im ðxÞ = σ - σT : T : σ - σT ð6:89Þ
It should be noted that the effective loading function is pressure-dependent now

and not of the von Mises type anymore, because spherical stress influences plastic
flow resistance of polymers. In other words, the yield stress is a function of
hydrostatic stress. Therefore, the particles have significant effects on the viscoplastic
behavior of the matrix materials. Plastic yielding and plastic flow occur only in the
matrix because the filler particles are assumed elastic. Since the two-phase
composite is in the plastic deformation range when the ensemble-volume averaged

current stress norm in the matrix reaches a critical level. The magnitude of the
current equivalent stress norm is utilized to determine the possible viscoplastic strain
for any point in the composite.
6.3.2 Average Stress in Filler Particles
If the particle interaction for the two-phase composite is ignored, Eq. (6.30) becomes

ε = ε0 þ φS : εT þ ε0 ð6:90Þ
With the noninteracting solution ε0 of the Eigen strain given by Eq. (6.42), we
arrive at
h i
ε = I - φSðA þ SÞ - 1 : ε0 þ φSðA þ SÞ - 1 ðA þ IÞ : εT ð6:91Þ
The volume-averaged stress tensor for the particles is defined by

Z Z
1 1
σ1 = σ1 ðxÞdx = C1 : ε0 þ ε0 ðxÞ - εT ðxÞ dx = C1 : ε0 þ ε01 - εT1
V1 V1
V1 V1
ð6:92Þ
where ε01 can be recast as

8 9
< Z =
1 T 0
ε01 = S : ε ðx Þ þ ε0 dx0 = S : εT1 þ ε0 ð6:93Þ
:V 1 ; 1
V1
Therefore, we have
σ 1 = C1 AðA þ SÞ - 1 : ε0 - C1 AðI - SÞðA þ SÞ - 1 : εT ð6:94Þ
As a result of Eq. (6.91), the average internal stresses of particles can be expressed
as
h i-1
σ1 = C1 AðA þ SÞ - 1 I - φSðA þ SÞ - 1 :ε
h i-1
- C1 AðA þ SÞ - 1 φSðA þ SÞ - 1 ðA þ IÞ I - φSðA þ SÞ - 1 þ ðI - SÞ : εT
= U : ε - V : εT
ð6:95Þ
where
6.4 Pairwise Interacting Solution for Two-Phase Composites 323
h i-1
U = C1 AðA þ SÞ - 1 I - φSðA þ SÞ - 1 ð6:96Þ
h i-1
V = C1 AðA þ SÞ - 1 φSðA þ SÞ - 1 ðA þ IÞ I - φSðA þ SÞ - 1 þ ðI - S Þ ð6:97Þ
By carrying out the lengthy algebra, the components of the positive definite
fourth-rank tensor U and V are explicitly given by

U ijkl = U 1 δij δkl þ U 2 δik δjl þ δil δjk ð6:98Þ

V ijkl = V 1 δij δkl þ V 2 δik δjl þ δil δjk ð6:99Þ
where
ð3α þ 2βÞ - 10ð1 þ v0 Þ

3U 1 þ 2U 2 = 3k1 ð6:100Þ
ð3α þ 2βÞ - 10ð1 þ v0 Þφ
β - ð8 - 10v0 Þ
U2 = μ ð6:101Þ
β - ð8 - 10v0 Þφ 1

30ð1 þ v0 Þk 21 ð3α þ 2βÞ - 10ð1 þ v0 Þ
3V 1 þ 2V 2 = φ - 1 þ 3k 1
ðk 1 - k 0 Þð3α þ 2βÞ ð3α þ 2βÞ - 10ð1 þ v0 Þφ
ð6:102Þ

ð8 - 10v0 Þμ21 β - ð8 - 10v0 Þ
V2 = φ - 1 þ μ1 ð6:103Þ
ðμ1 - μ0 Þβ β - ð8 - 10v0 Þφ
6.4 Pairwise Interacting Solution for Two-Phase

Composites
In this section, we extend the noninteracting solution for two-phase composites

developed in Sect. 6.3 to account for the interparticle interactions.
6.4.1 Approximate Solution of Two-Phase Interaction
If we neglect the interparticle interaction effects, then the ensemble-volume averaged

perturbed strain ε0p
q can be dropped. The resulting noninteracting approximation for
the particles becomes
- A : ε0 = ε0 - εT þ S : εT þ S : ε0 ð6:104Þ
Within the present two-sphere context, the integral equation governing the
distributed Eigen strain ε for a given filler particle configuration and remote strain
field ε0 can be written as
Z h i
- A : εðiÞ ðxÞ = ε0 - εT ðxÞ þ Gðx - x0 Þ : εT ðx0 Þ þ εðiÞ ðx0 Þ dx0
Ωi
Z h i
þ Gðx - x0 Þ : εT ðx0 Þ þ εðjÞ ðx0 Þ dx0 ði ≠ j, i, j = 1, 2 Þ
Ωj
ð6:105Þ
assuming εT is uniform in the particles.

By subtracting the noninteracting solution (Eq. 6.104), the effect of the
interparticle interaction can be founded by solving the following integral equation:
Z Z
ðiÞ 0 ðiÞ 0 0
-A : d ðxÞ = G ðx - x Þ : d ðx Þdx þ Gðx - x0 Þ : dðjÞ ðx0 Þdx0
Ωi Ωj
Z

þ Gðx - x0 Þ : εT ðx0 Þ þ ε0 dx0 ði ≠ j, i, j = 1, 2 Þ
Ωj
ð6:106Þ
where d(i) is given by
dðiÞ = εðiÞ - ε0 ð6:107Þ
Following the procedure given by Ju and Chen (1994b), the approximate

ðiÞ
equations for d for the two-sphere interaction problem can be written as
ðiÞ ðiÞ ðjÞ

- A : d = S : d þ G1 xi - xj : d þ G2 xi - xj

: εT þ ε0 þ 0 ρ8 ð6:108Þ
where
ρ3 1
G 1 xi - xj = H þ ρ2 H2 ð6:109Þ
30ð1 - v0 Þ
ρ3 1
G 2 xi - xj = H þ 2ρ2 H2 ð6:110Þ
30ð1 - v0 Þ
In addition, 0(ρ6) denotes the terms, which are higher than the order of ρ6, where
r xi - xj, r j rj, ρ = a/r, and a is the radius of a spherical filler particle. The
components of H1 and H2 are given by Eqs. (6.40) and (6.41).
Furthermore, we observe that
ðiÞ ðjÞ
d =d =d ð6:111Þ
Therefore, the solution of Eq. (6.107) is given by

d = - 30ð1 - v0 ÞT - 1 G2 : εT þ ε0 þ 0 ρ8 ð6:112Þ
where

T xi - xj = 30ð1 - v0 Þ A þ S þ G1 xi - xj ð6:113Þ
The corresponding expression to the order of 0(ρ3) is
T - 1 = K - 1 þ ρ3 L þ ⋯ ð6:114Þ
where
K ijkl = F ijkl ð0, 0, 0, 0, α, βÞ ð6:115Þ

5 6αð1 - 2v0 Þ 6αð1 þ v0 Þ 2αð2 - v0 Þ
Lijkl = F ijkl - 15, 3v0 , , , - , 1 - 2v0
4β2 3α þ 2β 3α þ 2β 3α þ 2β
ð6:116Þ
with α and β given in Eqs. (6.5) and (6.6).

The final expression for d is:

d = - ρ3 K - 1 H1 þ 2ρ2 H2 : εT þ ε0 - ρ6 L H1

: εT þ ε0 þ 0 ρ8 ð6:117Þ

To obtain the ensemble-average solution of d within the context of approximate
pairwise particle interaction, Eq. (6.117) must be integrated over all possible posi-
tions x2 of the second filler particle for a given location of the first filler particle x1.
The ensemble-averaging process can be expressed by
Z

d = d ðx1 - x2 ÞPðx2 jx1 Þdx2 ð6:118Þ
V - Ω1
Assuming that the two-point conditional probability function is statically isotro-

pic, is uniform, and obeys the following rule:
(
N
Pðx2 jx1 Þ = if jx2 - x1 j ≥ 2a ð6:119Þ
V
0 otherwise
where N/V is the density of filler particles in a composite. Accordingly, the ensemble
integration of Eq. (6.118) can be written as
8 2 39
>
< Z1 Z Z1 Z >
=
N 6 27

d =- K-1 4 ρ3 H1 dAdr þ 2 ρ5 H dAdrH 5 : εT þ ε0
2
V>: >
;
2a A 2a A
2 3
Z1 Z
N6 7
- 4 ρ 6
L H1 dAdr5 : εT þ ε0 þ 0 ρ8
V
2a A
ð6:120Þ
where A denotes the spherical surface of a particle with radius r. After a lengthy
but straightforward mathematical manipulation, the final expression takes the
following form:
hε i = Γ : ε0 þ Γ : εT ð6:121Þ
The components of the positive definite fourth-rank tensor Γ and Γ read

Γijkl = γ 1 δij δkl þ γ 2 δik δjl þ δil δjk ð6:122Þ

Γijkl = γ 1 δij δkl þ γ 2 δik δjl þ δil δjk ð6:123Þ
where

5φ 8α
γ1 = ð 13 - 14v Þv - ð 1 - 2v Þ ð 1 þ v Þ ð6:124Þ
8β2
0 0
3α þ 2β 0 0

5φ 6α
γ2 = 25 - 34v0 þ 22v0 - 2
ð1 - 2v0 Þð1 þ v0 Þ ð6:125Þ
16β2 3α þ 2β
with γ 1 and γ 2 given in Eqs. (6.7) and (6.8).

6.4.2 Ensemble-Average Stress Norm in the Matrix
The total stress at any point x in the matrix is given by the superposition of the
far-field stress σ0 and the perturbed stress σ′ induced by the filler particles. Assuming
the elastic Eigen strain ε in the filler particle is uniform, the perturbed stress at
matrix point x takes the form (Ju & Chen, 1994a)

σ0 ðxjx1 Þ = C0 Gðx - x1 Þ : εT þ ε ð6:126Þ
where εT is the Eigen strain caused by the CTE mismatch between the matrix and the
filler particle. Using the ensemble-volume averaged Eigen strain given in
Eq. (6.121), the stress perturbation can be given by the following relation:

σ0 ðxjx1 Þ = C0 Gðx - x1 Þ Γ : εT þ ε0 ð6:127Þ
where ε0 denotes the solution of the Eigen strain for the single inclusion problem,
which is given by Eq. (6.42). Therefore,
h i
σ0 ðxjx1 Þ = - C0 G ΓðA þ SÞ - 1
h i
: ε0 þ C0 G ΓðA þ SÞ - 1 ðA þ IÞ : εT ð6:128Þ
Following the same procedure used for the noninteracting solution, we obtain the
ensemble-averaged current stress norm at any matrix point x as:
where σT is given by Eq. (6.56), and the components of the positive definite fourth-
rank tensors T and T read:


with
ð3γ 1 þ 2γ 2 Þ2
3T 1 þ 2T 2 = 200ð1 - 2v0 Þ2 φ ð6:132Þ
ð3α þ 2βÞ2
1 4γ 2
T2 = þ 23 - 50v0 þ 35v20 22 φ ð6:133Þ
2 β
ð3γ 1 þ 2γ 2 Þ2
3T 1 þ 2T 2 = 200ð1 - 2v0 Þ2 φ ð6:134Þ
ð3α þ 2βÞ2
4γ 2
T 2 = 23 - 50v0 þ 35v20 22 φ ð6:135Þ
β
In addition, α and β are given in Eqs. (6.5) and (6.6).

The ensemble-averaged current stress norm at a matrix point can also be
expressed in terms of the macroscopic stress σ. Following the same procedure as
in the former section, the relation between the far-field stress σ0 and the macroscopic
stress σ takes the form
σ = P : σ0 - Q : σT ð6:136Þ
where
P = I þ φðI - SÞΓðA þ SÞ - 1 ð6:137Þ
Q = φðI - SÞΓðA þ SÞ - 1 ð6:138Þ
where the components of P and Q are given:


with
3P1 þ 2P2 = aφ þ 1 ð6:141Þ

1
P2 = ðbφ þ 1Þ ð6:142Þ
2
3Q1 þ 2Q2 = aφ ð6:143Þ
1
Q2 = bφ ð6:144Þ
2
3γ þ 2γ 2
a = 20ð1 - 2v0 Þ 1 ð6:145Þ
3α þ 2β
2γ 2
b = ð7 - 5v0 Þ ð6:146Þ
β
Using Eq. (6.136), we arrive at the alternative expression for the ensemble-
averaged current stress norm in a matrix point x:


where the positive definite fourth-rank tensors T, T , and T are defined as
T
T = P-1 T P-1 ð6:148Þ
T
T = P - 1 T P - 1 Q - T P - 1 ð6:149Þ
T
T = P - 1 Q T P - 1 Q þ T - 2T P - 1 Q ð6:150Þ

where the components of T, T , and T are



with
3T 1 þ 2T 2
3T 1 þ 2T 2 = ð6:154Þ
ðaφ þ 1Þ2
T2
T2 = ð6:155Þ
ðbφ þ 1Þ2
3T 1 þ 2T 2
3T 1 þ 2T 2 = - ð6:156Þ
ðaφ þ 1Þ2
T2 1
T2 = - þ ð6:157Þ
3T 1 þ 2T 2
3T 1 þ 2T 2 = ð6:158Þ
ðaφ þ 1Þ2
T2 bφ - 1
T2 = þ ð6:159Þ
σT = AC1 : εT is spherical stress in nature and caused by CTE mismatch; hence,

Eq. (6.147) of the ensemble-averaged loading function can be simplified as follows:

h H i m ð xÞ = σ - σ T : T : σ - σ T ð6:160Þ
6.4.3 Ensemble-Average Stress in the Filler Particles
The ensemble-volume averaged strain for two-phase composites can be given by


ε = ε0 þ φS : εT þ ε ð6:161Þ
Substituting Eq. (6.121),

ε = ε0 þ φSΓ : εT þ ε0 ð6:162Þ
With the noninteracting solution Eigen strain ε0 given by Eq. (6.42), we arrive at
h i h i
ε = I - φSΓðA þ SÞ - 1 : ε0 þ φ SΓðA þ SÞ - 1 ðA þ IÞ : εT ð6:163Þ
The volume-averaged stress tensor for the particles is defined as

Z Z
1 1
σ1 = σ1 ðxÞdx = C1 : ε0 þ ε0 ðxÞ - εT ðxÞ dx = C1 : ε0 þ ε0 - εT
V1 V1
V1 V1
ð6:164Þ
where ε0 can now be given by

8 9
< Z =
1
ε0 = S : εT ðx0 Þ þ ε ðx0 Þ dx0 = SΓ : εT þ ε0 ð6:165Þ
:V 1 ;
V1
Using Eq. (6.165), we have
σ1 = C1 : ε0 þ ½C1 ðSΓ - IÞ : εT þ ðC1 SΓÞ : ε0 ð6:166Þ
Utilizing the noninteracting solution Eigen strain ε0 , we arrive at

h i h i
σ1 = C1 I - SΓðA þ SÞ - 1 : ε0 þ C1 SΓðA þ SÞ - 1 ðA þ IÞ - I : εT ð6:167Þ
Combining with Eq. (6.163), the average internal stress of filler particles can be
expressed as
h ih i-1
σ1 = C1 I - SΓðA þ SÞ - 1 I - φSΓðA þ SÞ - 1 :
h ih i-1
ε - C1 φSΓ I - SΓðA þ SÞ - 1 I - φSΓðA þ SÞ - 1 ðA þ SÞ - 1 ðA þ IÞ
h i
-1
- SΓðA þ SÞ ðA þ IÞ - I : εT = U : ε - V : εT
ð6:168Þ
6.5 Noninteracting Solution for Three-Phase Composites 331
where
h ih i-1
U = C1 I - SΓðA þ SÞ - 1 I - φSΓðA þ SÞ - 1 ð6:169Þ
h ih i-1
V = C1 φSΓ I - SΓðA þ SÞ - 1 I - φSΓðA þ SÞ - 1 ðA þ SÞ - 1 ðA þ IÞ
h i
- SΓðA þ SÞ - 1 ðA þ IÞ - I
ð6:170Þ
By carrying out the lengthy algebra, the components of the positive definite
fourth-rank tensor U and V can be obtained as


where
ð3α þ 2βÞ - 10ð1 þ v0 Þð3γ 1 þ 2γ 2 Þ

3U 1 þ 2U 2 = 3k 1 ð6:173Þ
ð3α þ 2βÞ - 10ð1 þ v0 Þφð3γ 1 þ 2γ 2 Þ
β - 4ð4 - 5v0 Þγ 2
U2 = μ ð6:174Þ
β - 4ð4 - 5v0 Þφγ 2 1
30ð1 þ v0 Þð3γ 1 þ 2γ 2 Þk21
3V 1 þ 2V 2 =
ðk1 - k0 Þð3α þ 2βÞ

ð3α þ 2βÞ - 10ð1 þ v0 Þð3γ 1 þ 2γ 2 Þ
φ - 1 þ 3k1 ð6:175Þ
ð3α þ 2βÞ - 10ð1 þ v0 Þφð3γ 1 þ 2γ 2 Þ

4ð4 - 5v0 Þγ 2 μ21 β - 4ð4 - 5v0 Þγ 2
V2 = φ - 1 þ μ1 ð6:176Þ
ðμ1 - μ0 Þβ β - 4ð4 - 5v0 Þγ 2 φ
6.5 Noninteracting Solution for Three-Phase Composites
Let us consider an initially perfectly bonded two-phase composite consisting of an

elastic matrix (phase 0) with bulk modulus k0 and shear modulus μ0, and randomly
dispersed elastic spherical filler particles (phase 1) with bulk modulus k1 and shear
modulus μ1. Subsequently, as loadings or deformations are applied, some filler
particles could experience complete interfacial debonding from the matrix. These
completely debonded filler particles can be regarded as spherical voids (phase 2).
6.5.1 Effective Elastic Modulus of Multiphase Composites
Effective elastic moduli of multiphase composites containing randomly located,

unidirectionally aligned elastic ellipsoids were explicitly derived by Ju and Chen
(1994a). For a multiphase composite, the effective (noninteracting) elasticity tensor
C reads
h i
C = C0 I þ B ðI - S BÞ - 1 ð6:177Þ
where B takes the form

Xn -1
B= φ
q=1 q
S þ Aq ð6:178Þ
Since phase 2 contains spherical voids, we have A2 = - I. The effective bulk

modulus k and shear modulus μ for the three-phase composite (n = 2) can be
explicitly expressed as
0 P2 φq
1
30ð1 - v0 Þ q = 1 3αq þ2βq
k = k 0 @1 þ P φq
A ð6:179Þ
1 - 10ð1 þ v0 Þ 2q = 1 3αq þ2β
q
0 φq P2 1
15ð1 - v0 Þ
q = 1 βq
μ = μ 0 @1 þ P φ
A ð6:180Þ
1 - 2ð4 - 5v0 Þ 2q = 1 β q
q
with

k0 μ0
αq = 2ð5v0 - 1Þ þ 10ð1 - v0 Þ - ð6:181Þ
kq - k0 μq - μ0
μ0
βq = 2ð4 - 5v0 Þ þ 15ð1 - v0 Þ ð6:182Þ
μq - μ0
6.5.2 Ensemble-Average Stress Norm in the Matrix
According to the Eshelby theory, the perturbed stress σ ′ at any point in the matrix
due to the presence of the particles and voids can be written as
Z Z
0 0

σ ð xÞ = C 0 : G ðx - x Þ : ε1 þ εT1 0
dx þ C0 : G ð x - x0 Þ
V V
0
: ε2 þ εT2 dx ð6:183Þ
The Eigen strain in a single ellipsoidal inclusion is uniform for the interior points
of an isolated (no interacting) inclusion. Therefore, the perturbed stress for any
matrix point x due to a typical isolated q-phase inhomogeneity centered at xq takes
the following form:

σ0 xjxq = C0 G x - xq : ε0
q þ εq
T
ð6:184Þ

where G x - xq is given by Eq. (6.39), and
-1 0 -1
ε0
q = - Aq þ S : ε þ Aq þ S ðI - SÞ : εTq ð6:185Þ
We denote by H(x| ℘) the square of the current stress norm at the local point x,
which determines the plastic strain for a given phase configuration ℘. Since there is
no plastic strain in the elastic particles or voids, H(x| ℘) can be written as

σðxj℘Þ : I D : σðxj℘Þ if x is in the matrix
H ðxj℘Þ = ð6:186Þ
0 otherwise
In addition, hHim(x) is defined as the ensemble average of H(x| ℘) over all

possible points where x is in the matrix phase. Matrix point receives perturbations
from filler particles and voids. Therefore, the ensemble-average stress norm for any
matrix point x can be evaluated by collecting and summing up all the current stress
norm perturbations produced by any typical particle centered at x1 in the particle
domain and any typical void centered at x2 in the void domain, and averaging over
all possible locations of x1 and x2,
Z

h H i m ð xÞ = H 0 þ H ðxjx1 Þ - H 0 Pðx1 Þdx1
r1 > a
Z

þ H ðxjx2 Þ - H 0 Pðx2 Þdx2 ð6:187Þ
r2 > a
where r1 = j x - x1j, r2 = j x - x2j P(x1) and P(x2) denote the probability density
functions for finding a particle centered at x1 and a void centered at x2, respectively.
For simplicity, P(x1) and P(x2) are assumed to be statistically homogeneous,
isotro-

pic, and uniform. Using the properties of the fourth-order tensor G x - xq , we
obtain the ensemble-averaged current stress norm at any matrix point x as:
Z Z
N1
h H im ð xÞ = H 0 þ dr 1 ðσ0 ðxjx1 Þ : Id : σ0 ðxjx1 ÞÞdA
V
r1 > a Aðr1 Þ
Z Z ð6:188Þ
N2
þ dr 1 ðσ0 ðxjx2 Þ : Id : σ0 ðxjx2 ÞÞdA
V
r2 > a Aðr2 Þ
By carrying out the lengthy but straightforward algebra, we arrive at
where
σT = A1 C1 : εT ð6:190Þ
The components of the positive definite fourth-rank tensor T and T read


with
X2 φq
3T 1 þ 2T 2 = 200ð1 - 2v0 Þ2 q=1
2 ð6:193Þ
3αq þ 2βq
1 X 2 φq
T2 = þ 23 - 50v0 þ 35v20 ð6:194Þ
2 q=1 β2q
φ1
3T 1 þ 2T 2 = 200ð1 - 2v0 Þ2 ð6:195Þ
ð3α1 þ 2β1 Þ2
φ
T 2 = 23 - 50v0 þ 35v20 2 ð6:196Þ
β1
The ensemble-averaged current stress norm at a matrix point must be established

in terms of the macroscopic stress σ to express the effective loading function in terms
of the macroscopic stress. In the special case of uniform dispersions of identical
elastic spheres in a homogeneous matrix, the macroscopic stress and the far-field
stress take the following forms:
h X2 i
0
σ = C0 : ε - φ
q=1 q
ε q þ ε T
q ð6:197Þ
h X2 i
0
σ0 = C 0 : ε - q=1
φ q S ε q þ ε T
q ð6:198Þ
Using Eqs. (6.197), (6.198), and (6.185), the relation between the far-field stress
σ0 and the macroscopic stress σ is given by
σ = P : σ0 - Q : σT ð6:199Þ
where
X2 -1
P=I þ φ ðI - SÞ
q=1 q
Aq þ S ð6:200Þ
Q = φ1 ðI - SÞðA1 þ SÞ - 1 ð6:201Þ
The components of P and Q are given by


where
X2
3P1 þ 2P2 = aφ
q=1 q q
þ1 ð6:204Þ
1 X2
P2 = b φ þ1
q=1 q q
ð6:205Þ
2
3Q1 þ 2Q2 = a1 φ1 ð6:206Þ
1
Q2 = b φ ð6:207Þ
2 1 1
20ð1 - 2v0 Þ
aq = ð6:208Þ
3αq þ 2βq
ð7 - 5v0 Þ
bq = ð6:209Þ
βq
Using Eqs. (6.199) and (6.189), we arrive at the alternative expression for the
ensemble-averaged current stress norm in a matrix point x as:


where the positive definite fourth-rank tensors T, T , and T are defined as
T
T = P-1 T P-1 ð6:211Þ
T
T = P - 1 T P - 1 Q - T P - 1 ð6:212Þ
T
T = P - 1 Q T P - 1 Q þ T - 2T P - 1 Q ð6:213Þ

where the components of T ,T , and T are



with
3T þ 2T 2
3T 1 þ 2T 2 = P 1 2
ð6:217Þ
2
q = 1 a q φq þ 1
T2
T2 = P 2
ð6:218Þ
2
q = 1 bq φq þ 1
3T þ 2T 2
3T 1 þ 2T 2 = - P 1 2
ð6:219Þ
2
q = 1 a q φq þ 1
T2 1
T2 = - P 2
þ P2 ð6:220Þ
q = 1 b q φq þ1
2
q = 1 b q φq þ1 2
3T þ 2T 2
3T 1 þ 2T 2 = P 1 2
ð6:221Þ
2
q = 1 a q φq þ 1
P2
T2 q = 1 bq φ q -1
T2 = P 2
þ P2 ð6:222Þ
q = 1 b q φq þ 1
2
q = 1 bq φ q þ 1
2
Because σT = AC1 : εT is spherical stress caused by the CTE mismatch,

Eq. (6.147) of the ensemble-averaged loading function can be simplified as

h H i m ð xÞ = σ - σ T : T : σ - σ T ð6:223Þ
6.5.3 Ensemble-Average Stress in Filler Particles
The ensemble-volume averaged strain for three-phase composites takes the

following form:
X2
ε = ε0 þ φS
q=1 q
ε0
q þ εq
T
ð6:224Þ
With the noninteracting solution Eigen strain ε0 given by Eq. (6.185), we can
write
h X2 - 1i 0
ε= I- q=1
φ q S Aq þ S : ε þ φ1 SðA1 þ SÞ - 1 ðA1 þ IÞ : εT1 ð6:225Þ
The volume-averaged stress tensor for the filler particles is defined as

Z Z
1 1
σ1 = σ1 ðxÞdx = C1 : ε0 þ ε0 ðxÞ - εT ðxÞ dx = C1
V1 V1
V1 V1

: ε0 þ ε01 - εT1 ð6:226Þ
where ε1 0 can be recast as

8 9
< Z =
1 T 0
ε′1 =S : ε ðx Þ þ ε0 dx0 = S : εT1 þ ε0 ð6:227Þ
:V 1 ; 1
V1
With the help of Eqs. (6.227) and (6.185), Eq. (6.226) can be simplified as
σ1 = C1 A1 ðA1 þ SÞ - 1 : ε0 - C1 A1 ðI - SÞðA1 þ SÞ - 1 : εT1 ð6:228Þ
Finally, the average internal stress of filler particles can be expressed as

h X2 - 1i - 1
σ 1 = C 1 A 1 ðA 1 þ S Þ - 1 I - q=1
φ q S Aq þ S :ε
h X2
-1 - 1i - 1
- C 1 A 1 ðA 1 þ S Þ φ 1 S ðA 1 þ S Þ - 1 ðA 1 þ I Þ I - q=1 q
φ S A q þ S þ ð I - SÞ : εT1
= U : ε - V : εT1
ð6:229Þ
where
h X2 - 1i - 1
U = C1 A1 ðA1 þ SÞ - 1 I - q=1
φ q S Aq þ S ð6:230Þ
n
C1 A1 ðA1 þ SÞ - 1 φ1 SðA1 þ SÞ - 1 ðA1 þ IÞ½I
V= X2 -1 -1 ð6:231Þ
- φ S Aq þ S
q=1 q
þ ðI - SÞg
By carrying out the lengthy but simple algebra, the components of the positive
definite fourth-rank tensors U and V are explicitly obtained as follows:


where
ð3α1 þ 2β1 Þ - 10ð1 þ v0 Þ

3U 1 þ 2U 2 = h P i 3k1 ð6:234Þ
φq
ð3α1 þ 2β1 Þ 1 - 10ð1 þ v0 Þ 2q = 1 3αq þ2β
q
β1 - 2ð4 - 5v0 Þ
U2 = h P i μ1 ð6:235Þ
φ
β1 1 - 2ð4 - 5v0 Þ 2q = 1 β q
q
ð3α1 þ 2β1 Þ - 10ð1 þ v0 Þ

3V 1 þ 2V 2 =
ð3α1 þ 2β1 Þ
8 9
>
> >
>
< =
10ð1 þ v0 Þφ1 k 1 2ð1 - 2v0 Þ
þ 3k1
>
> P φ 3ð 1 - v 0 Þ >
>
:ð3α1 þ 2β1 Þ 1 - 10ð1 þ v0 Þ 2q = 1 q
ðk 1 - k 0 Þ ;
3αq þ 2βq
ð6:236Þ
β1 - 2ð4 - 5v0 Þ
V2 =
β1
2 3
2ð4 - 5v0 Þφ1 μ1 7 - 5v0 5
4 h P2 φ q i þ μ ð6:237Þ
β1 1 - 2ð4 - 5v0 Þ q = 1 β ðμ1 - μ0 Þ 15 ð 1 - v0 Þ 1
q
6.6 Effective Thermomechanical Properties
In particle-filled composites, there is a thin layer of interfacial layer (interphase)

between a filler particle and the matrix. The imperfect interface bond may be due to a
very compliant thin interfacial layer that is assumed to have perfect boundary
conditions with the matrix and the filler particle. This defines a three-phase com-
posite that includes the filler particle, the thin interphase, and the matrix as shown in
Fig. 6.2. Once the effective mechanical and thermal properties of the inner compos-
ite sphere assemblage (CSA), which consists of the filler particle and interphase layer
of thickness δ, are computed, the composite models that are used for the perfect
interface composite model can be applied to this case of composites with imperfect
interface conditions.
6.6 Effective Thermomechanical Properties 339
Fig. 6.2 Three-phase

composite system. Note:
filler particle is
polycrystalline
Fig. 6.3 Schematic

illustration of composite
spherical assemblage (CSA)
The composite sphere assemblage (CSA) model was first proposed by Kerner
(1956) and Van der Poel (1958) as shown in Fig. 6.3. Smith (1974, 1975),
Christensen and Lo (1979), and Hashin and his co-workers (1962, 1963, 1968,
1990, 1991a, b) improved on the CSA model. The CSA model assumes that the
filler particles are spherical and that the action on the particle is transmitted through a
spherical interphase shell. The overall macro-behavior is assumed isotropic and is
thus characterized by two effective moduli: the bulk modulus k and the shear
modulus μ. In this section, we summarize the theoretical solution for effective
thermomechanical properties of the CSA model consisting of an elastic spherical
filler particle and an elastic interphase layer with a thickness of δ. In the following
formulas, k represents the bulk modulus, μ represents the shear modulus, and α
represents the coefficient of thermal expansion. The subscripts i, f, and m refer to the
interphase, filler particle, and matrix, respectively.
6.6.1 Effective Bulk Modulus
The effective bulk modulus k for the CSA model as obtained by Hashin (1962) is
given below:
φ
k = ki þ ð6:238Þ
1
kf - ki þ 33kð1i þ4μ
- φÞ
i
where
r 3
φ= ð6:239Þ
rþδ
r is the radius of the filler particle

δ is the thickness of interphase
Hashin and Shtrikman’s formulas (Hashin & Shtrikman, 1963) and Walpole’s
formulas (1966) provide upper and lower bounds for the bulk modulus of the CSA
model. Equation (6.238) is the same as the highest lower bound value.
6.6.2 Effective Coefficient of Thermal Expansion (ECTE)
The effective coefficient of thermal expansion α for the CSA model as obtained by
Levin (1967) is given as

αf - αi 1 1
α = αi þ 1 - ð6:240Þ
kf - ki
1 k ki
6.6.3 Effective Shear Modulus
Based on the generalized self-consistent scheme (GSCS) model proposed by Hashin

(1962), Christensen and Lo (1979) have given the condition for determining the
effective shear modulus of the CSA model as follows:
2
μ μ
A þB þ D=0 ð6:241Þ
μi μi
where

A = 8 μf =μi - 1 ð4 - 5vi Þη1 φ10=3 - 2 63 μf =μi - 1 η2 þ 2η1 η3 φ7=3

þ252½μf =μi - 1η2 φ5=3 - 50 μf =μi - 1 7 - 12vi þ 8vi η2 φ þ 4ð7 - 10vi Þη2 η3
2
ð6:242Þ

B = - 4 μf =μi - 1 ð1 - 5vi Þη1 φ10=3 þ 4 63 μf =μi - 1 η2 þ 2η1 η3 φ7=3
ð6:243Þ
- 504½μf =μi - 1η2 φ5=3 þ150 μf =μi - 1 ð3 - vi Þvi η2 φ þ 3ð15vi - 7Þη2 η3

D = 4 μf =μi - 1 ð5vi - 7Þη1 φ10=3 - 2 63 μf =μi - 1 η2 þ 2η1 η3 φ7=3
2 ð6:244Þ
þ252½μf =μi - 1η2 φ5=3 þ25 μf =μi - 1 vi - 7 η2 φ - ð7 þ 5vi Þη2 η3
with

η1 = μf =μi - 1 49 - 50vf vi þ 35 μf =μi vf - 2vi þ 35 2vf - vi ð6:245Þ

η2 = 5vf μf =μi - 8 þ 7 μf =μi þ 4 ð6:246Þ

η3 = μf =μi ð8 - 10vi Þ þ ð7 - 5vi Þ ð6:247Þ
φ is given in Eq. (6.239).

Based on Van der Poel’s formula for the shear modulus of a particulate compos-
ite, Smith (1974, 1975) provided the condition for determining the effective shear
modulus of the CSA model as follows:
2
μ μ
α -1 þ β -1 þ γ=0 ð6:248Þ
μi μi
where
h i
α = 4Pð7 - 10vi Þ - Sφ7=3 ½Q - ð8 - 10vi ÞðM - 1Þφ
2
- 126PðM - 1Þφ 1 - φ2=3 ð6:249Þ
β = 35ð1 - vi ÞP½Q - ð8 - 10vi ÞðM - 1Þφ - 15ð1 - vi Þ

h i
4Pð7 - 10vi Þ - Sφ7=3 ðM - 1Þφ ð6:250Þ
γ = - 525Pð1 - vi Þ2 ðM - 1Þφ ð6:251Þ
with
M = μf =μi ð6:252Þ

P = 7 þ 5vf M þ 4 7 - 10vf ð6:253Þ
Q = ð8 - 10vi ÞM þ ð7 - 5vi Þ ð6:254Þ

S = 35 7 þ 5vf M ð1 - vi Þ - Pð7 þ 5vi Þ ð6:255Þ
φ is given in Eq. (6.239).

Equations (6.241) and (6.248) are verified as equivalent in the determination of
the effective shear modulus of the CSA model. Solving the above equations, one can
determine the exact solution for the effective shear modulus of the CSA model. One
of the roots is negative and extraneous. The positive root provides the value of the
effective shear modulus.
6.6.4 Effective Young’s Modulus and Effective Poisson’s

Ratio
Effective Young’s modulus and Poisson’s ratio for the CSA model can be calculated
from the well-known expressions:
9k μ
E = ð6:256Þ
3k þ μ
3k - 2μ
v = ð6:257Þ
6k þ 2μ
6.6.5 Numerical Examples
To illustrate the effect of imperfect interface and interphase thickness on the overall
effective mechanical and thermal properties, we consider a special case of CSA that
consists of spherical alumina trihydrate (ATH) particle and interphase with the
following properties:
For particles:
E f = 70 GPa
vf = 0:24
αf = 13 × 10 - 6 = ° C
For interphase material:
vi = 0:31
αi = 70 × 10 - 6 = ° C
δ=r = 0:01 and δ=r = 0:001
The nondimensional interface parameter is defined as q = Ef/Ei. A zero value of

q implies that there is no abrupt jump in displacement at the interface and the
perfectly bonded interface condition exists. At the other extreme, the infinite value
of q implies that the interface tractions do not exist, and the filler particle is
completely debonded from the matrix. Finite positive values for q define an imper-
fect interface, which lies between two extreme cases mentioned above (Fig. 6.4).
Figures 6.5, 6.6, and 6.7 show the effect of the interphase bond modulus and
thickness on the effective shear modulus, Young’s modulus, and bulk modulus of
the CSA. The stiffness of the bond has a strong effect on the degradation of the shear
15
14.5
δ/ r = 0.01
14
α*
13.5
δ/r = 0.001
13
12.5
12
100 102 104 106
Interface Parameter q
Fig. 6.4 Variation of effective CTE with the interface parameter, q, Nie (2005)
0.8
δ/r = 0.01
0.6
m*/mf
δ/ r = 0.001
0.4
0.2
0
100 102 104 106
Fig. 6.5 Variation of effective shear modulus with the interface parameter q, Nie (2005)
0.8
δ/r = 0.01
0.6
E*/Ef
0.4 δ/ r = 0.001
0.2
0
100 102 104 106
Fig. 6.6 Variation of effective Young’s modulus with the interface parameter q, Nie (2005)
0.8
δ/r = 0.01
0.6
k*/kf
δ/ r = 0.001
0.4
0.2
0
100 102 104 106
Fig. 6.7 Variation of effective bulk modulus with the interface parameter q, Nie (2005)
6.7 Micromechanical Constitutive Model of the Particulate Composite 345
0.4
0.35
0.3
ν* δ/r = 0.01
0.25
0.2
δ/r = 0.001
0.15
0.1
0.05
0
100 102 104 106
Fig. 6.8 Variation of effective Poisson’s ratio with the interface parameter q, Nie (2005)
modulus, Young’s modulus, and bulk modulus. As the interface becomes thinner for
the same Ei value, the effective shear modulus increases. On the other hand, for the
same interphase thickness, as the elastic modulus value Ei of the interphase
decreases, the effective shear modulus decreases. This is also true for both the
effective bulk modulus and effective Young’s modulus. Figure 6.8 shows the effect
of the interphase bond modulus and thickness on effective Poisson’s ratio. Effective
Poisson’s ratio seems to be independent of the interphase stiffness when q ≤ 102 and
q ≥ 104.
Analytical and numerical evaluation of the effective elastic properties and the
thermal expansion coefficients of the composite sphere assemblage show that the
bulk and shear moduli are insensitive to the value of Poisson’s ratio of interphase.
Numerical results show that the nature of the interphase has a significant effect on the
stiffness and the thermal expansion coefficient of the CSA.
6.7 Micromechanical Constitutive Model of the Particulate

Composite
The nature of the bond between particles and the matrix has a significant effect on the
mechanical behavior of particulate composites. Most analytical and numerical
models assume that the bond between the filler and matrix is perfect and can be
modeled using the continuity of tractions and displacements across the interface.
However, internal defects and imperfect interfaces are well-known to exist in

composites, and the incorporation of such phenomena into the general theory
requires modification and relaxation of the continuity of displacements between
the constituents. The imperfect interface bond may be due to the compliant interfa-
cial layer or interface damage, which may have been created deliberately by coating
the particles with a debonding agent. It may also develop during the manufacturing
process due to chemical reactions between the particles and the matrix or due to
interface damage from cyclic loading. Most importantly, the strength of the bond at
the interface controls the mechanical response and fatigue life of the composite
(Basaran & Nie, 2004). By controlling the stiffness of the interphase, it is possible to
control the overall behavior of the composite.
In the following section, a micromechanical model for particulate composites
with the imperfect interface between the filler particles and the matrix and also a
CTE mismatch between the filler particle and matrix is presented, based on the work
of Ju and Chen (1994a, b) and Ju and Tseng (1996), for perfectly bonded particle-
matrix interfaces with no CTE mismatch between the filler particles and the matrix.
In the micromechanical model, particulate composites are treated as three-phase
composites consisting of an agglomerate of filler particles, the bulk matrix, and the
interfacial transition zone around the agglomerate as shown in Fig. 6.2. The com-
pliant interfacial transition zone is assumed to have perfect bonding with matrix and
filler particles. The inner composite sphere assemblage (CSA), consisting of the
particle and the interfacial transition zone, is regarded as an equivalent spherical
particle with the effective mechanical and thermal properties derived in the previous
section.
6.7.1 Modeling Procedures for Particulate Composites
In this section, we discuss the modeling procedures for a particulate composite,

including how the interface properties, CTE mismatch between the particle and the
matrix, and an isotropic damage parameter are introduced.
The modeling process consists of four steps. The first step is the simplification of
the actual particulate composite, which is shown in Fig. 6.9. A is a representation of
the actual microstructure for a particulate composite, where filler particles have
different sizes and shapes. To get an analytical expression for the effective behavior
of the particulate composite, we must make some assumptions. For simplicity, we
assume that the particles are uniform in shape and spherical. The interfacial layer is
used to model the imperfect interface condition. The displacement jumps at the
interface are defined by the deformation of the interphase. The particulate composite
is defined as a three-phase system that includes particles, the thin interphase, and the
matrix. The simplified microstructure of a particulate composite is shown in Fig. 6.9.
The simplified microstructure for the particulate composites includes the particle,
the interphase, and the matrix. For the sake of simplicity, the particle and the
interphase are regarded as one, namely, a particle-interphase assemblage. The
Fig. 6.9 Modeling procedures—step 1: simplification, Nie (2005)
Fig. 6.10 Modeling procedures—step 2: equivalence, Nie (2005)
microstructure of composite is now simplified as C shown in Fig. 6.10. The

microstructure of C includes the particle-interphase assemblage and the matrix.
The thermomechanical properties are calculated for the new two-phase particle-
interphase assemblage.
Finding the effective thermomechanical properties for the two-phase (Fig. 6.11)
spherical particle-interphase assemblage is essential.
The particulate composite is simplified as a two-phase system with perfect
interfacial bonding between the particle-interphase assemblage and the matrix.
This final modeling step is shown in Fig. 6.12.
Fig. 6.11 Modeling procedures—step 3: thermomechanical properties, Nie (2005)
Fig. 6.12 Modeling

procedures—step 4:
micromechanics, Nie (2005) Equivalent
C Continuum
Medium
6.7.2 Elastic Properties of Particulate Composites
Based on Ju and Chen’s (1994a, b) solution for noninteracting particles, the effective
properties of a two-phase composite with imperfect bonding can be obtained by
substituting the properties of the particle with the properties of composite sphere
assemblage:
( )
3ð1 - vm Þðk - k m Þφf
k = km 1þ ð6:258Þ
3ð1 - vm Þkm þ 1 - φf ð1 þ vm Þðk - km Þ
( )
15ð1 - vm Þðμ - μm Þφf
μ = μm 1þ ð6:259Þ
15ð1 - vm Þμm þ 1 - φf ð8 - 10vm Þðμ - μm Þ
Similarly, the pairwise interacting solution for the effective properties of the
two-phase composite with the imperfect bond can be given by

30ð1 - vm Þð3γ 1 þ 2γ 2 Þφf
k = km 1þ ð6:260Þ
3β1 þ 2β2 - 10ð1 þ vm Þð3γ 1 þ 2γ 2 Þφf

30ð1 - vm Þγ 2 φf
μ = μm 1 þ ð6:261Þ
β2 - 4ð4 - 5vm Þγ 2 φf
where

k μ
β1 = 2ð5vm - 1Þ þ 10ð1 - vm Þ m - m ð6:262Þ
k - k m μ - μm
μm
β2 = 2ð4 - 5vm Þ þ 15ð1 - vm Þ ð6:263Þ
μ - μm
and

5φf 8β1
γ1 = ð13 - 14vm Þvm - ð1 - 2vm Þð1 þ vm Þ ð6:264Þ
8β2 2 3β1 þ 2β2

1 5φf 6β1
γ2 = þ 25 - 34vm þ 22vm -
2
ð1 - 2vm Þð1 þ vm Þ ð6:265Þ
2 16β2 2 3β1 þ 2β2
where k and μ are the effective bulk modulus and shear modulus of the composite
spherical-filler assemblage and φf is the filler particle volume fraction.
Example
Nie (2005) studied lightly cross-linked poly-methyl methacrylate (PMMA) filled
with alumina trihydrate (ATH). The following phase properties were used for the
numerical analysis:
Particle: Ef = 70GPa, vf = 0.24
Matrix: Em = 3.5GPa, vm = 0.31
The particle volume fraction is φf = 0.48
Nondimensional interface parameter is defined as q = Ef/Ei.
Three kinds of composites with different interfacial properties were prepared in
the lab. Composite A has the strongest interfacial adhesion due to the addition of a
special adhesion-promoting additive, where the value of the nondimensional inter-
face parameter q = Ef/Ei can be assumed one. The interfacial adhesion strength of
composite C is the weakest due to the addition of a debonding promoting additive.
The value of the nondimensional interface parameter q for composite C can be
assumed to be a very large number. The interfacial strength of composite B is
moderate. The value of nondimensional interface parameter q for composite B can
be assumed to be some value between that for composite A and composite C.
12
Composite A [Test Data]
Composite B [Test Data]

Effective Young’s Modulus (Gpa) 10
Pairwise Interacting Solution
4 Noninteracting Solution
Composite C [Test Data]

2
0
100 101 102 103 104 105 106
Fig. 6.13 Effective Young’s modulus vs. interface parameter at a volume fraction of 48% (the
thickness of interphase is 1/10 of the diameter of particles), Nie (2005)
The effects of the interphase properties on effective Young’s modulus of the

two-phase composite are shown in Figs. 6.13, 6.14, and 6.15, where the elastic
properties of composites with different interphase properties obtained from test data
were also plotted for comparison. These results indicate that the pairwise interacting
solution yields a better approximation of the overall elastic modulus. Figures 6.13,
6.14, and 6.15 indicate that as the thickness of the interphase gets larger, the effective
elastic modulus value decreases for the same q value.
6.7.3 A Viscoplasticity Model
Viscoplastic flow occurs only in the matrix because filler particles have a very high
melting temperature, and they are brittle and deform linear elastic. The magnitude of
the current equivalent stress norm of the matrix can be used to determine viscoplastic
behavior. When the ensemble-volume average stress norm in the matrix reaches a
certain level, the effective yield function for the composite can be given by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

f ðσ, αÞ = σ - σ T : T : σ - σT
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð6:266Þ
- T 1 þ 2T 2 σ y ðαÞ
12
Effective Young’s Modulus (Gpa) 10
6
Noninteracting Solution

2
0
100 101 102 103 104 105 106
12
Effective Young’s Modulus (Gpa)
10
Noninteracting Solution
6
4
0
100 101 102 103 104 105 106
where σ is the average stress in the particulate composite; σT is the stress caused by
the CTE mismatch between the matrix and the particle which was derived earlier in
Eq. (6.190); T is the fourth-order tensor which is given by Eq. (6.211); T1 and T2 are
given by Eqs. (6.217) and (6.218), respectively; σ y ðαÞ is the isotropic hardening
function of the composite materials; and α is the equivalent viscoplastic strain that
defines the isotropic hardening of the composite. Its rate is given by
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
-1
α_ = T 1 þ 2T 2 ε_ : T : ε_
vp vp
ð6:267Þ
The factors in the effective yield and the effective plastic strain increment
equations are chosen so that the effective stress and effective plastic strain incre-
ments are equal to the uniaxial stress and uniaxial plastic strain increment in a
uniaxial monotonic tensile test. It should be noted that the effective yield function
is spherical pressure-dependent now and not of the von Mises type anymore. So the
particles and hydrostatic pressure have a significant effect on the viscoplastic
behavior of the matrix.
Unified mechanics theory provides a basic framework to introduce degradation
evolution intrinsically without an empirical evolution function. According to the
strain equivalence principle, the effective yield stress function can be given by
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
σ σ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
f ðσ, αÞ = - σT : T : - σT - T 1 þ 2T 2 σ y ðαÞ ð6:268Þ
1-Φ 1-Φ
The degradation increases the equivalent stress norm of the composite, which
tends to amplify the viscoplastic flow.
If the kinematic hardening behavior is included, let β define the center of the yield
surface
of the composite in the stress space, and the relative stress can be defined as
σ - β . Assume the filler particles have the same effect on the stress norm of the
matrix as on the kinematic behavior of the matrix; the stress norm defining the
viscoplastic behavior of the matrix can be updated as

hH im ðxÞ = σ - β - σT : T : σ - β - σT ð6:269Þ
Therefore, the effective yield function for the composite can be given as
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
σ σ
f ðσ, qÞ = - β - σT : T : - β - σT
1-Φ 1-Φ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
- T 1 þ 2T 2 σ y þ K ðαÞ ð6:270Þ

where q = α, β is chosen as the internal viscoplastic variable.
For simplicity, the Perzyna-type viscoplasticity model is employed to character-

ize the rate (viscosity) in the matrix. Therefore, the effective ensemble-volume
average viscoplastic strain rate for the composite can be expressed as
∂f 1
ε_ vp = γ = γn ð6:271Þ
∂σ 1 - Φ
where

T : 1 -σ Φ - β - σT
n = qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ð6:272Þ
σ
1-Φ - β - σ
T
: T : 1 -σ Φ - β - σT
and γ denotes the plastic consistency parameter
hf i hf i
γ= = ð6:273Þ
η 2μτ
where η is the viscosity coefficient and τ is called the relaxation time.

And the effective equivalent viscoplastic strain rate is defined as
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi γ
α_ = T 1 þ 2T 2 ð6:274Þ
1-Φ
The evolution of back-stress, β , depends on the plastic strain or plastic work

history. The earliest model for the evolution of back-stress is due to Prager (1956)
and was subsequently modified by Ziegler (1959). The Prager kinematic hardening
assumption is that during yielding the back-stress increment dβ is equal to the
component of dσ in the direction normal to the yield surface. Since the plastic strain
increment is also normal to the yield surface, the increment dβ can be written as

dεp dεp
dβ = dσ p ð6:275Þ
jdε j jdεp j
In the case of linear kinematic hardening, we have

1:5
T 1 þ 2T 2
dβ = 2 2
H 0α_ n ð6:276Þ
T 1 þ 2T 2 þ 2T 1
where H′ is the kinematic hardening modulus. The factor in Eq. (6.276) is chosen so
that the effective stress and effective viscoplastic strain increments are equal to the
uniaxial stress and uniaxial plastic strain increment in a uniaxial monotonic
ensile test.
6.7.4 Thermodynamic State Index
The thermodynamic state index (TSI) is given by
Φ = Φcr 1 - e - R Δs
ms
ð6:277Þ
where Φcr is the critical value of the TSI; however, this term can also be used as a
function to map elastic modulus degradation, onto the TSI axis [however, using Φcr
is not necessary, and can be taken as one, especially if all entropy generation
mechanisms are included in the fundamental equation. In the other examples in
this book, it is taken as 1], and R is gas constant, ms is molar mass, and Δs is the
change in entropy and must be calculated from the thermodynamic fundamental
equation at each Gauss integration point. For thermomechanical analysis of partic-
ulate composites, the following thermodynamic fundamental equation represents the
majority, but not all, of the entropy generation mechanisms:
Z Z t0 Z t
σ : ε_
t vp
k 2 r
Δs = dt þ jgradT j dt þ dt ð6:278Þ
t0 Tρ t T ρ
2
t0 T
where ρ is the mass density, T is temperature, k is the thermal conductivity of the

composite, and r is the distributed internal heat generation rate. Unfortunately,
Eq. (6.278) ignores entropy generation due to aging in PMMA and other long-
term chemical reactions in the PMMA molecular chains, which are not trivial.
6.7.5 Solution Algorithm
A general return mapping algorithm is used to solve Eqs. (6.270)–(6.278). The

general return mapping algorithm was proposed by Simo and Taylor (1985) and
summarized in detail by Simo and Hughes (1998). To minimize confusion, the
symbol Δ is used to denote an increment over a time step or an increment between
successive iterations. For the rate-of-slip γ, we adhere to the following conventions:
Δγ = γΔt denotes the increment of γ over a time step Δt, and Δ2γ denotes the
increment of Δγ between iterations.
Let C be the elastic consistent tangent moduli of the particulate composite;
ignoring the derivative of TSI with respect to displacement for the sake of simplicity,
the overall stress-strain relationship for small strain formulation can be given by

σnþ1 = ð1 - ΦÞC : εnþ1 - εvp
nþ1 - εnþ1
th
ð6:279Þ
Since εn+1 and εth

nþ1 are fixed during the return mapping stage, it follows that
1
Δεvp = - C - 1 Δσ ð6:280Þ
1-Φ
From Eq. (6.277), we have
_ = Φcr ms exp - ms Δs Δ_s

Φ ð6:281Þ
R R
Assuming that irreversible entropy generation is only due to plastic work and heat
generation, from Eq. (6.278) for an isothermal process at each increment, we can
write
σ : ε_
vp
r
Δ_s = þ ð6:282Þ
Tρ T
If the heat generated within the system is negligible, the distributed internal heat
source of strength per unit mass is zero [which is not true, especially if the rate of
loading is high]. Using Eqs. (6.280) and (6.282), Eq. (6.281) becomes
1 Φcr ms m
ΔΦ = - exp - s Δs σC - 1 Δσ ð6:283Þ
1 - Φ TρR R
We should point out that in Eq. (6.282) we ignored many prominent irreversible
entropy generation mechanisms, such as the relative motion between the filler
particles and the matrix, the entropy generation in the filler polycrystals, and aging
in PMMA molecular chains. Moreover, if the strain rate of the loading is very high,
there will be additional entropy generation mechanisms. However, here the goal is to
demonstrate the process as simple as possible.
From Eqs. (6.271), (6.274), and (6.276), we also have
1
εvp
nþ1 = εn þ
vp
Δγ n ð6:284Þ
1-Φ
1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
αnþ1 = αn þ T 1 þ 2T 2 Δγ ð6:285Þ
1-Φ
1:5
T 1 þ 2T 2
βnþ1 = βn þ 2 2
ΔHn ð6:286Þ
T 1 þ 2T 2 þ 2T 1
Let
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
m1 = T 1 þ 2T 2 ð6:287Þ
1:5
T 1 þ 2T 2
m2 = 2 2
ð6:288Þ
T 1 þ 2T 2 þ 2T 1
Φcr ms m
m3 = - exp - s Δs ð6:289Þ
TρR R
Also, we have
∂f 1
= n ð6:290Þ
∂σ 1 - Φ
2
∂ f 1
= N ð6:291Þ
∂σ 2
ð1 - ΦÞ2
∂f
= -n ð6:292Þ
∂β
2
∂ f
2
=N ð6:293Þ
∂β
∂f n:σ
= ð6:294Þ
∂D ð1 - ΦÞ2
∂f
= - m1 K 0 ð6:295Þ
∂α
∂n
= -N ð6:296Þ
∂β
∂n 1
= N ð6:297Þ
∂σ 1 - Φ
∂n 1
= N:σ ð6:298Þ
∂D ð1 - ΦÞ2
∂σ y
where K 0 = ∂α
is the isotropic hardening modulus.
N
T-n n
N = qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
σ ffi ð6:299Þ
σ
1-Φ - β - σ : T 1 - Φ - β - σT
T
We define the following residual functions as follows:
m1
Rα = - αnþ1 þ αn þ Δγ ð6:300Þ
1-Φ
2μτΔγ
Rγ = - hf i ð6:301Þ
Δt
1
Rε = - εvp
nþ1 þ εn þ 1 - Φ Δγn
vp
ð6:302Þ
Rβ = - βnþ1 þ βn þ m2 ΔHn ð6:303Þ

Linearizing these residual functions yields
m1 m1 m3
Rα - Δαnþ1 þ Δ2 γ þ ΔγσC - 1 Δσ = 0 ð6:304Þ
1-Φ ð1 - ΦÞ 3
2 0
0 0Rα m1 K τ
n Δσ - nΔβ = Rf þ m1 K þ þ 2μ Δ2 γ ð6:305Þ
1-Φ Δt
Δ2 γ
P : Δσ - Q : Δβ þ n þ Rε = 0 ð6:306Þ
1-Φ
m1 m2 0Δ2
R : Δσ - S : Δβ þ H γn þ m2 H 0 Rα n þ Rβ = 0 ð6:307Þ
1-Φ
where
" #
1 Δγ Δγm3 Nσ O
-1 -1
P= C þ Nþ þn σC ð6:308Þ
1-Φ 1-Φ ð1 - ΦÞ2 1 - Φ
Δγ
Q= N ð6:309Þ
1-Φ
" #
m2 m3 O
0 -1
R= ΔHN þ ðΔHNσ þ m1 H ΔγnÞ σC ð6:310Þ
1-Φ ð1 - ΦÞ2
S = I þ m2 ΔHN ð6:311Þ
" #
1 m3
n0 = nþ nσ - m1 2 K 0 Δγ σC - 1 ð6:312Þ
1-Φ ð1 - ΦÞ 2
Combining Eqs. (6.306) and (6.307) yields
h in o n o n o
Δσ Δ2 γ
P Q
þ n m m1 H 0 þ m H 0 RRεnþR = 0 ð6:313Þ
R S - Δβ 1-Φ 1 2 2 α β
From Eqs. (6.305) and (6.313), we have

n o
- Rf - m1 K 0Rα - ½ n0 n A
Rε
0
m2 H Rα nþRβ
Δ2 γ = n o ð6:314Þ
m1 2 K 0
1-Φ þ 2μ Δtτ þ 1 -1 Φ ½ n0 n A
n
m1 m2 H 0 n
where
h i
A-1 = P
R
Q
S
ð6:315Þ
h~ ~
i
A= P
~
R
Q
~
S
ð6:316Þ

~ = P - Q S-1 R -1
P ð6:317Þ

~ = - P - Q S-1 R -1 Q S-1
Q ð6:318Þ

~ = - S-1 R P - Q S-1 R -1
R ð6:319Þ

~ = S-1 þ S-1 R P - Q S-1 R -1 Q S-1
S ð6:320Þ
So we have
8 9
n o >
< Δ2 γ >
=
nþ Rε
Δσ
= -A 1 - Φ ð6:321Þ
> m1 Δ γ
2
- Δβ
: þ Rα m2 H 0 nþRβ >
;
1-Φ
1
Δεvp = - C - 1 Δσ ð6:322Þ
1-Φ
σΔεvp
Δ2 s = ð6:323Þ
Tρ
Δsnþ1 = Δsn þ Δ2 s ð6:324Þ
ΔΦnþ1 = - m3 ðσ nþ1 Δε Þ vp
ð6:325Þ
m1 m1 m3
Δαnþ1 = Rα þ Δ2 γ - ΔγσΔεvp ð6:326Þ
1-Φ ð1 - ΦÞ2
Then, the viscoplastic strain, consistency parameter, stress, TSI and entropy
production rate, and effective equivalent viscoplastic strain are updated at each
Gauss integration point and iterated until the norms of residual functions
(Eqs. 6.300–6.303) are smaller than a predefined tolerance. Normally, the conver-
gence is achieved when the norms of these residual functions are all equal to or less
than 1 × 10-5. The value of the tolerance is problem-dependent.
The procedure summarized above is simply a systematic application of Newton’s
method to the system of Eqs. (6.270)–(6.278) that results in the computation of the
closest point projection from the trial state onto the yield surface.
It should be noted that the general return mapping algorithm is unconditionally
stable, and the convergence of the algorithms toward the final value of the state
variable is obtained at a quadratic rate. Further information on the general return
mapping algorithm is given in Simo and Hughes (1998) and Ortiz and
Martin (1989).
It is necessary to point out that the formulation presented above calculates the TSI
at Gauss integration points. In the finite element formulation, it is possible to define
TSI at nodal points and solve for TSI and other nodal unknowns. However, this
would significantly increase the computational cost. It is much more cost-effective to
calculate TSI at Gauss integration points from thermodynamic variables obtained in
the previous step. Because the solution process is incremental, this simplification
introduces very little error.
6.7.6 Consistent Elastic-Viscoplastic Tangent Modulus
An important advantage of the algorithm lies in the fact that it can be exactly
linearized in closed form. This leads to the notion of consistent elastic-viscoplastic
tangent moduli.
Let C be the consistent elastic-viscoplastic tangent moduli, and then the incre-
mental stress-strain relationship based on the third law of the unified mechanics
theory can be given by
dσ = ð1 - ΦÞC : ðdε - dεvp Þ ð6:327Þ
Differentiating Eqs. (6.277) and (6.278), we have
dΦ = - m3 σdεvp ð6:328Þ
where m3 is given by Eq. (6.289).

Differentiating Eq. (6.284), we have
Δγ Δγ dΔγ
dεvp = WNdσ - WNdβ þ Wn ð6:329Þ
ð1 - Φ Þ2 1-Φ 1-Φ
where
Δγm3 Nσ O
W-1 = I þ 2 1-Φ
þn σ ð6:330Þ
ð1 - ΦÞ
m1 m1 m3
dα = dΔγ - Δγσdεvp ð6:331Þ
1-Φ ð1 - ΦÞ2
m2 m3 ΔH O m ΔH
dβ = - Nσ σ dεvp þ 2 Ndσ - m2 ΔHNdβ
ð1 - ΦÞ 2 1-Φ
þ m2 nH 0 dα ð6:332Þ
Differentiating Eq. (6.273) yields

1 m3
ndσ - ndβ - nσ - m1 2 K 0 Δγ σdεvp
1-Φ ð1 - ΦÞ 2

τ m 2K0
= 2μ þ 1 dΔγ ð6:333Þ
Δt 1 - Φ
From Eqs. (6.327) and (6.329)
Wn
p : dσ - q : dβ þ dΔγ = dε ð6:334Þ
1-Φ
where
1 Δγ
p= C-1 þ WN ð6:335Þ
1-Φ 1-Φ
Δγ
q= WN ð6:336Þ
1-Φ
From Eqs. (6.329), (6.331), and (6.332), we have
m1 m2 H 0
r : dσ - s : dβ þ WndΔγ = 0 ð6:337Þ
1-Φ
where
1 Δγ
r= m ΔHN þ VWN ð6:338Þ
1-Φ 2 1-Φ
Δγ
s = I þ m2 ΔHN þ VWN ð6:339Þ
1-Φ
h O i
m mm Nσ
V = - 1 2 32 H 0 Δγ þn σ ð6:340Þ
ð1 - ΦÞ 1-Φ
Combining Eqs. (6.334) and (6.337)

h in o n o n o
Wn
p q dσ
þ 1
dΔγ = dε
ð6:341Þ
r s - dβ 1-Φ m1 m2 H 0 0
Combining Eqs. (6.329) and (6.333) yields
n
dσ - n dβ = m4 dΔγ ð6:342Þ
1-Φ
where
6.8 Verification Examples 361
m3 Δγ
n = n - 3
nσ - m1 2 K 0 Δγ ðσWNÞ ð6:343Þ
ð1 - ΦÞ
m3 τ m 2K0
m4 = nσ - m1 2 K 0 Δγ ðσWnÞ þ 2μ þ 1 ð6:344Þ
ð1 - ΦÞ 3 Δt 1 - Φ

n 1 ~
1-Φ p þ ~r dε
dΔγ = h i n o ð6:345Þ
n
m4 þ 1 -1 Φ n a
Wn
0
1-Φ m 1 m2 H Wn
where
h i
a-1 = p
r
q
s
ð6:346Þ
h i
~ ~
a= p
~r
q
~s
ð6:347Þ
1
~dε -
dσ = p ~ þ ~r WndΔγ
p ð6:348Þ
1-Φ

1 N 1
dσ ~ ~
1 - Φ p þ r Wn n 1-Φ p ~ þ ~r
C= ~
=p- h i n o ð6:349Þ
dε n
m4 þ 1 -1 Φ n a
Wn
0
1-Φ m m
1 2 H Wn
6.8 Verification Examples
6.8.1 Material Properties of ATH
ATH (particle filler) is modeled as an isotropic elastic material. The

thermomechanical properties of ATH as provided by the manufacturer are as
follows:
Poisson’s ratio of ATH: νf = 0.24, CTE: αf = 1.47 × 10-6/ ° C
Elastic modulus of ATH: Ef = 70 GPa
The average diameter of a filler particle is 35 μm.
Fig. 6.16 Young’s 5
Young's Modulus (GPa)

modulus of PMMA as a
function of temperature 4
(Celsius) (Cheng et al.,
1990a, b)
3
0
-50 0 50 100 150 200
Temperature
6.8.2 Properties of PMMA
The matrix poly-methyl methacrylate (PMMA) is a very common polymer, which

has been extensively studied. Young’s modulus of PMMA as a function of temper-
ature is as shown in Fig. 6.16. (Cheng et al., 1990a, b, c):
Em = - 0:0234 T þ 4:124 ðGPAÞ
where T is the temperature in Celsius. Poisson’s ratio of PMMA is νm = 0.31.
6.8.3 Properties of Matrix-Filler Interphase
The interphase around the particle is also regarded as an isotropic elastic material. In
addition, it is reasonable to assume that Poisson’s ratio and CTE of the interphase are
the same as that of the matrix (PMMA); however, Young’s modulus of the inter-
phase is less than that of PMMA. For simulations, the thickness of the interphase is
taken as 1% of the diameter of filler particles; Young’s modulus is half of that of
PMMA, except where specified differently.
6.8.4 Properties of Particulate Composites
The following thermomechanical properties of the particulate composite A were

provided by the material manufacturer (Basaran & Nie, 2007).
The average specific mass for the composite is ms = 85 g/mole, and the density of
the composite is ρ = 1750 kg/m3 .The volume fraction of particle in the composite
is φ = 0.48
Fig. 6.17 Poisson’s ratio of 0.4

composite A as a function of
temperature 0.35
Poisson’s ratio
0.3
0.25
0.2
0 20 40 60 80 100
Temperature
Fig. 6.18 Coefficient of 1.E-04

thermal expansion (CTE) of
temperature 8.E-05
6.E-05
CTE
4.E-05
2.E-05
0.E+00
0 20 40 60 80 100 120 140
Temperature
Poisson’s ratio of the composite A [strong interfacial adhesion] is a function of

temperature as shown in Fig. 6.17:
ν = 0:008 T þ 0:334
The coefficient of thermal expansion of composite A is given as a function of

temperature as shown in Fig. 6.18.
When T ≤ 90 ° C:
α = 3:035 × 10 - 7 T þ 2:347 × 10 - 5
When T ≥ 90 ° C:
α = 1:0992 × 10 - 6 T - 5:012 × 10 - 5
Fig. 6.19 Relaxation time 1.6

of composite A as a function
of temperature
Relaxtion Time (s)

1.2
0.8
0.4
0
0 20 40 60 80 100 120
Temperature
Viscosity η is the ratio of the loss modulus to the angular frequency, which is
determined during the forced harmonic oscillation test (Nielsen & Landel, 1994).
The viscosity relaxation time τ is defined by Simo and Hughes (1998):
η
τ=
2μ
where μ is the shear modulus. It is important to realize that the controlling factor in
the relaxation process is the relative time t/τ. The absolute clock time t is regarded as
short or long only when compared with τ. The concept of relaxation time is
explained in greater detail by Simo and Hughes (1998). From dynamic mechanical
testing, the viscosity relaxation time τ is determined as shown in Fig. 6.19.
When T ≤ 90 ° C:
τ = 1:12406 × 10 - 6 T 3 - 1:67823 × 10 - 4 T 2 þ 7:91134 × 10 - 3 T þ 7:35 × 10 - 3
When T ≥ 90 ° C:
τ = - 2:6348 × 10 - 6 T 4 1:08452 × 10 - 3 T 3 - 1:64629 × 10 - 1 T 2 þ 10:9673T

- 271:11
The elastic modulus of composite A can be determined from the properties of the
filler particle, the matrix, and the interphase according to the micromechanical model
presented earlier in this chapter. The elastic modulus of composite A is also
determined by uniaxial tensile tests for strain rates and temperatures as shown in
Fig. 6.20:
Em = - 0:0005T 2 - 0:021T þ 13:33 þ 0:6 log ðε_ Þ ðGPaÞ
The gas constant is R = 8.3145 Joule/mole/K.

Fig. 6.20 Elastic modulus 12 Strain rate=1E-3

of composite A as a function
Elastic Modulus (GPa)

of temperature, Nie (2005) 10 Strain rate=1E-4
Strain rate=1E-5
8
0
0 50 100 150
Temperature
Fig. 6.21 Critical TSI of 80

temperature
60
Parameter Φcr
40
20
0
0 20 40 60 80 100
Temperature
The critical TSI is a function of temperature for composite A is determined from

the uniaxial tensile tests at various temperatures as seen in Fig. 6.21:
Φcr = 0:1972 T þ 50:22 T ≤ 100 °
where T is the temperature in Celsius. Φcr is not an essential parameter. As long as all
entropy generation mechanisms are included in the thermodynamic fundamental
equation and no simplifications are made in derivation, if not used, nothing in the
formulation changes. In this specific example [some entropy generation mecha-
nisms, such as polymer aging and heat generation, are not included], an Φcr is
determined using the statistical methods by matching the elastic modulus degrada-
tion determined from test data with the thermodynamic state index (TSI) evolution
calculated from the thermodynamic fundamental equation during uniaxial tensile
tests. This process allows us to map elastic modulus degradation onto TSI axis. If
Φcr = 1 is used, nothing changes in the formulation or the results. In the unified
mechanics theory, we only care about the dissipation along the thermodynamic state
index axis, which provides the entropy generation. The system will always have
many entropy generation mechanisms and degrade along the TSI axis, and the
entropy generation rate eventually always goes to zero.
6.8.5 Finite Element Simulation Results
The stress-strain response obtained from the simulations was compared with exper-
imental data as shown in Figs. 6.22 and 6.23 at 24 and 75 °C, respectively. These
simulation results are obtained based on the assumption that the thickness of
interphase is 1% of the diameter of the particle and the elastic modulus of interphase
is 50% of that of PMMA. Furthermore, the CTE mismatch effects are also included,
where the temperature when residual stresses begin to build up is assumed to be
100 °C. It is seen that the degradation effects must be accounted for at relatively large
strains.
The comparison of the TSI from the simulations with elastic modulus degradation
that was measured during experiments is given in Fig. 6.22.
6.8.6 Cyclic Stress-Strain Response
Mapping the elastic modulus degradation onto the TSI axis is not necessary if the
user is not seeking the exact stress-strain data cycle to cycle from simulations.
However, if it is needed to map the elastic modulus degradation onto the TSI axis,
the following process can be used. However, according to UMT, only the degrada-
tion along the TSI axis is what matters not the degradation along the elastic modulus
or any other material property that is a human construct.
Fig. 6.22 Comparison of 40

stress-strain relationship
among viscoplastic model Newtonian mechanics
based on Newtonian
mechanics, viscoplastic viscoplasticity model
model based on UMT, and 30
experiments at 750 °C UMT visco-
Stress (Mpa)
plasticity model
20
Experimental data
10
0
0.00 0.005 0.01 0.015 0.02
Strain
Fig. 6.23 Comparison of

TSI obtained from Experimental Measurement
simulations versus that Simulation
measured in experiments in
terms of elastic modulus
degradation at 24 °C 0.24
0.20
T.S.I
0.16
0.12
0.08
0.04
0.00
0.000 0.002 0.004 0.006 0.008
Strain
Fig. 6.24 Applied 1

displacement profile for
tension-compression fatigue
test with a strain amplitude
of 0.6% 0.5
Displacement(mm)
0
0 10 20 30 40 50
Time (s)
-0.5
-1
The critical TSI Φcr is temperature-dependent and can be determined from the test
data as follows: First, determine the elastic modulus degradation in cyclic tests.
Second, calculate the dissipated plastic strain energy during cyclic testing,
(Fig. 6.24) using the stress-strain hysteresis loop, (Fig. 6.25) and then determine
the TSI. This process is also equivalent to calculating the fatigue fracture entropy of
the material. From cyclic test data, Φcr can be determined as a function of temper-
ature as follows:
Fig. 6.25 Uniaxial stress- 60

strain hysteresis loop from
Stress (MPa)
simulation vs. experimental Test
40
data for cycle 5 at room Simulation
temperature with a strain
amplitude of 0.6% 20
0
-0.009 -0.006 -0.003 0.000 0.003 0.006 0.009
-20 Strain
-40
-60
-80

Stress (MPa)
simulation vs. experimental Test 40

data for cycle 50 at room
temperature with a strain Simulation
0
-0.009 -0.006 -0.003 0.000 0.003 0.006 0.009
-20 Strain
-40
-60
-80
Φcr = 0:00545T þ 0:659 T ≤ 100 ° C
Φcr (T) is not necessary, if all material properties are defined as a function of
temperature and all entropy generation mechanisms are formulated properly in the
thermodynamic fundamental equation, because the entropy generation is always
according to the second law of thermodynamics and defined by the TSI coordinate.
Other examples of thermomechanical loading applications where Φcr = 1 are given
by Bin Jamal et al. (2020, 2021a, b, c, 2022), Lee and Basaran (2021a, b), and Lee
et al. (2022, b, c, d) (Figs. 6.26, 6.27, 6.28, 6.29, 6.30, 6.31, and 6.32).

Stress (MPa)
simulation vs. experimental 40
data for cycle 104 at room
Test
temperature with a strain Simulation
0
-0.010 -0.005 0.000 0.005 0.010
-20
Strain
-40
-60
-80
Fig. 6.28 Uniaxial stress- Stress (MPa) 30

data for cycle 5 at 75 °C
with a strain amplitude of Simulation
0.6% 10
0
-0.009 -0.006 -0.003 0.000 0.003 0.006 0.009
-10 Strain
-20
-30
-40

Stress (MPa)

0.6% 10
0
-0.009 -0.006 -0.003 0.000 0.003 0.006 0.009
-10 Strain
-20
-30
-40

Stress (MPa)
0.6% 10
0
-0.009 -0.006 -0.003 0.000 0.003 0.006 0.009
-10 Strain
-20
-30
-40
Fig. 6.31 Comparison of 0.3

TSI at room temperature
with a strain amplitude of 0.25
0.6%
0.2
T.S.I
0.15
0.1 Simulation
0.05 Experimental Measurement
0
0 30 60 90 120
Number of Cycles
Fig. 6.32 Comparison of 0.4

TSI at 75 °C with strain
amplitude of 0.6%
0.3
T.S.I
0.2
Simulation
0.1
Experimental Measurement
0
0 100 200 300 400 500 600
Number of Cycles
References 371
Here, we conclude the presentation of modeling of particle-filled composites with

the unified mechanics theory for thermomechanical analysis.
References
Basaran, C., & Nie, S. (2004, July). An irreversible thermodynamic theory for damage mechanics of
solids. International Journal of Damage Mechanics, 13(3), 205–224. #41.
Basaran, C., & Nie, S. (2007). A thermodynamics based damage mechanics model for particulate
composites. International Journal of Solids and Structures, 44, 1099–1114. #63.
Bin Jamal, M. N., Kumar, A., Lakshmana Rao, C., & Basaran, C. (2020). Low cycle fatigue life
prediction using unified mechanics theory in Ti-6Al-4V alloys. Entropy, 22, 24.
Bin Jamal, N., Rao, L., & Basaran, C. (2021a). A semi-infinite edge dislocation model for the
proportionality limit stress of metals under high strain rate. Computational Mechanics, 68(3),
545–565. https://doi.org/10.1007/s00466-020-01959-2
Bin Jamal, N., Rao, L., & Basaran, C. (2021b). A unified mechanics theory-based model for
temperature and strain rate dependent proportionality limit stress of mild steel. Mechanics of
Materials, 155, 103762.
Bin Jamal, N., Lee, H. W., Rao, L., & Basaran, C. (2021c). Dynamic equilibrium equations in
unified mechanics theory. Applied Mechanics, 2(1), 63–80.
Bin Jamal, N. M., Rao, L., & Basaran, C. (2022). Unified mechanics theory based flow stress model
for the rate-dependent behavior of bcc metals. Materials Today Communications, 31, 103707.
Cheng, W.-M., Miller, G. A., Manson, J. A., Hertzberg, R. W., & Sperling, L. H. (1990a).
Mechanical behavior of poly(methyl methacrylate)–Part 1: Tensile strength and fracture tough-
ness. Journal of Materials Science, 25, 1917–1923.
Cheng, W.-M., Miller, G. A., Manson, J. A., Hertzberg, R. W., & Sperling, L. H. (1990b).
Mechanical behavior of poly(methyl methacrylate)–Part 2: Temperature and frequency effects
on the fatigue crack propagation behavior. Journal of Materials Science, 25, 1924–1930.
Cheng, W.-M., Miller, G. A., Manson, J. A., Hertzberg, R. W., & Sperling, L. H. (1990c).
Mechanical behavior of poly(methyl methacrylate)–Part 3: Activation processes for fracture
mechanism. Journal of Materials Science, 25, 1931–1938.
Christensen, R. M., & Lo, K. H. (1979). Solutions for effective shear properties in three-phase
sphere and cylinder models. Journal of the Mechanics and Physics of Solids, 27, 315–330.
Eshelby, J. D. (1957). The deformation of the elastic field of an ellipsoidal inclusion, and related
problems. Proceedings of the Royal Society of London, Series A, Mathematical Science,
241(1226), 376–396.
Hashin, Z. (1962). The elastic moduli of heterogeneous materials. Transaction of the ASME,
Journal of Applied Mechanics, 29, 143–150.
Hashin, Z. (1968). Assessment of the self consistent scheme approximation: Conductivity of
particulate composites. Journal of Composite Materials, 2(3), 284–301.
Hashin, Z. (1990). Thermoelastic properties and conductivity of carbon/carbon fiber composites.
Mechanics of Materials, 8, 293–308.
Hashin, Z. (1991a). The spherical inclusion with imperfect interface. Transaction of the ASME,
Journal of Applied Mechanics, 58, 444–449.
Hashin, Z. (1991b). Thermoelastic properties of particulate composites with imperfect interface.
Journal of the Mechanics and Physics of Solids, 39(6), 745–762.
Hashin, Z., & Shtrikman, S. (1963). A variational approach to the theory of the elastic behavior of
multiphase materials. Journal of the Mechanics and Physics of Solids, 11, 127–140.
Ju, J. W., & Chen, T. M. (1994a). Micromechanics and effective moduli of elastic composites
containing randomly dispersed ellipsoidal inhomogeneities. Acta Mechanica, 103, 103–121.
Ju, J. W., & Chen, T. M. (1994b). Effective elastic moduli of two-phase composites containing
randomly dispersed spherical inhomogeneities. Acta Mechanica, 103, 123–144.
Ju, J. W., & Tseng, K. H. (1996). Effective elastoplastic behavior of two-phase ductile matrix
composites: A micromechanical framework. International Journal of Solids & Structures,
33(29), 4267–4291.
Ju, J. W., & Tseng, K. H. (1997). Effective elastoplastic algorithms for ductile matrix composites.
Journal of Engineering Mechanics, 123(3), 260–266.
Kerner, E. H. (1956). The elastic and thermoelastic properties of composite media. The Proceedings
of Physical Society, 69B, 808–813.
Lee, H. W., & Basaran, C. (2021a). A review of damage, void evolution and fatigue life prediction
models. Metals, 11, 609.
Lee, H. W., & Basaran, C. (2021b). Predicting high cycle fatigue life with unified mechanics theory.
Mechanics of Materials, 104, 116, ISSN 0167-6636.
Lee, H. W., Basaran, C., Egner, H., Lipski, A., Piotrowski, M., Mroziński, S., Noushad Bin Jamal,
M., & Rao, C. L. (2022a). Modeling ultrasonic vibration fatigue with unified mechanics theory.
International Journal of Solids and Structures, 236–237, 111313.
Lee, H. W., & Basaran, C. (2022). Predicting high cycle fatigue life with unified mechanics theory.
Mechanics of Materials, 164, 104116.
Lee, H. W., Fakhri, H., Ranade, R., Basaran, C., Egner, H., Lipski, A., Piotrowski, M., &
Mroziński, S. (2022c). Modeling corrosion fatigue with unified mechanics theory. Material &
Design, 224, 111383.
Lee, H. W., Fakhri, H., Ranade, R., Basaran, C., Egner, H., Lipski, A., Piotrowski, M., &
Mroziński, S. (2022d). Modeling fatigue of pre-corroded bcc metals with unified mechanics
theory. Materials & Design, 111383. https://doi.org/10.1016/j.matdes.2022.111383
Levin, V. M. (1967). Thermal expansion coefficients of heterogeneous materials. Mechanics of
Solids, 2(1), 58–94.
Mura, T. (1987). Mechanics of elastic and inelastic solids: Micromechanics of defects in solids (2nd
ed.). Martinus Nijhoff Publishers.
Nie, S., & Basaran, C. (2005). A micromechanical model for effective elastic properties of
particulate composites with imperfect interfacial bonds. International Journal of Solids &
Structures., 42(14), 4179–4191. #62.
Nie, S., Basaran, C., Hutchins, S., & Ergun, H. (2006). Failure mechanisms in PMMA/ATH acrylic
casting dispersion. Journal of Mechanical Behavior of Materials, 17(2), 79–95. #39.
Nielsen, L. E., & Landel, R. F. (1994). Mechanical properties of polymers and composites (2nd
ed.). Marcel Dekker, Inc.
Ortiz, M., & Martin, J. E. (1989). Symmetry-preserving return-mapping algorithms and incremen-
tally extremal paths: A unification of concepts. International Journal for Numerical Methods in
Engineering, 28, 1839–1853.
Prager, W. (1956) A New Method of Analyzing Stresses and Strains in Work-Hardening Plastic
Solids. Journal of Applied Mechanics, 23, 493–496.
Simo, J. C., & Hughes, T. J. R. (1998). Interdisciplinary applied mathematics, mechanics and
materials, computational inelasticity. Springer.
Simo, J. C., & Taylor, R. L. (1985). Consistent tangent operators for rate-independent
elastoplasticity. Computer Methods in Applied Mechanics and Engineering, 48, 101–119.
Smith, J. C. (1974). Correction and extension of van der Poel’s method for calculating the shear
modulus of a particulate composite. Journal of Research of the National Bureau of Standards-A.
Physics and Chemistry, 78A(3), 355–361.
References 373
Smith, J. C. (1975). Simplification of van der Poel’s formula for the shear modulus of a particulate
composite. Journal of Research of the National Bureau of Standards-A. Physics and Chemistry,
79A(2), 419–423.
Van der Poel, C. (1958). On the rheology of concentrated suspensions. Rheologica Acta, 1, 198.
Walpole, L. J. (1966). On bounds for the overall elastic moduli of inhomogeneous systems-I.
Journal of the Mechanics and Physics of Solids, 14, 151.
Ziegler, H. (1959). A Modification of Prager’s Hardening Rule. Quarterly of Applied Mathematics,
17, 55–65.
Chapter 7
Unified Micromechanics of Finite
Deformations
7.1 Introduction to Finite Deformations
In this chapter, finite (large) deformation micromechanics is discussed. The formu-

lation is more complicated than the small deformation theory; therefore, it is easier to
explain with an example. Micromechanical modeling of a polymer is used as an
example. A dual-micro-mechanism rate-dependent constitutive model is used to
describe the thermomechanical response of amorphous polymers below and above
glass transition temperatures. Material property definitions, the evolution of internal
state variables, and plastic flow rules are revisited to provide a smooth and contin-
uous transition in material response around glass transition temperature, θg.
To formulate the large deformation mechanics, it is necessary to describe the
kinematics of the constitutive model based on finite deformation tensors. Consider a
body with an initial volume of Vo in undeformed (original or initial) configuration
(Σo) at time to which deforms into a volume V in the current (deformed) configura-
tion (Σ) at time t, as shown in Fig. 7.1. This body can be uniquely defined with a
continuous one-to-one mapping (χ) of position vectors r and x in the original
configuration and current configuration, respectively. r is the material (local) coor-
dinate of a particle, that is, the coordinates of its location in the reference coordinate
system. x is the spatial coordinates, defining location at time t. Both material
coordinates and spatial coordinates were discussed earlier in Chap. 2.
Deformation gradient tensor F describes the transformation of a line element dr at
the position of r in the original configuration to a deformed line element dx at the
position of x in the current configuration (Eq. (7.1)). Unique transformation ensures
a non-singular, nonnegative determinant of deformation gradient F or Jacobian
J (Eq. (7.3)). The spatial gradient of velocity tensor L is the gradient of the velocity
field ~v and relates deformation gradient tensor to its material time derivative F: _
https://doi.org/10.1007/978-3-031-18621-9_7
376 7 Unified Micromechanics of Finite Deformations
Fig. 7.1 Deformation from

the original configuration to
the current configuration
x = χðr, t Þ ð7:1Þ
∂x
F = ∇xðχ Þ = ð7:2Þ
∂r
J = detðFÞ > 0 ð7:3Þ
~vðx, t Þ = x_ ðx, t Þ ð7:4Þ
∂x_
L = gradð~vÞ = ð7:5Þ
∂r
F_ = L F ð7:6Þ
Deformation gradient tensor F can be decomposed into an elastic part Fe and

plastic part Fp, and information related to thermal deformation is included in the
elastic part of the deformation gradient tensor. Plastic deformation gradient tensor F p
defines a deformation with respect to the initial configuration Σo to an intermediate,
“relaxed,” or “natural” configuration Σ′ which is followed by an elastic deformation
tensor Fe to the final configuration (Σ). The symmetric part of the spatial gradient of
velocity tensor L is defined as stretching rate tensor D, and the asymmetric part of
velocity gradient tensor is defined as spin tensor W:
F = Fe Fp ; J = J e J p ð7:7Þ
J e = detðFe Þ > 0 ð7:8Þ
J = detðF Þ > 0
p p
ð7:9Þ
symmetric ðLÞ = D; asymmetricðLÞ = W; L = D þ W ð7:10Þ
7.1 Introduction to Finite Deformations 377
Using Eq. (7.7), the elastic and plastic part of the spatial gradient of velocity
tensor L can be defined as
-1
L = Le þ Fe Lp Fe ð7:11Þ
where
Le = F_ Fe
e -1
ð7:12Þ
Lp = F F _ p p-1 ð7:13Þ
Stretching rate tensor D and spin tensor W for elastic and plastic parts can be
derived similarly from Eq. (7.10), where superscripts “e” and “p” represent elastic
and plastic parts of corresponding quantity, respectively:
symðLe Þ = De ; asymðLe Þ = We ; Le = De þ We ð7:14Þ

symðL Þ = D ; asymðL Þ = W ; L = D þ W
p p p p p p p
ð7:15Þ
Right stretch tensor U and left stretch tensors V and rotation tensor R can be
found from the right and left polar decompositions of the deformation gradient
tensor F as follows:
F = RU ð7:16Þ
F = VR ð7:17Þ
U and V stretch tensors are positive definite symmetric tensors and R are proper
orthogonal tensors. Cauchy tensor C and Almansi tensor B can be formulated as
follows:
C = FT F = UU ð7:18Þ
B - 1 = F - TF - 1 = V - 1V - 1 ð7:19Þ
In the same fashion, the following relations can be written for elastic and plastic
parts of the deformation gradient tensor F:
Fe = Re Ue ð7:20Þ
F =V R
e e e
ð7:21Þ
T
Ce = Fe Fe = Ue Ue ð7:22Þ
-1 -T -1 -1 -1
Be = Fe Fe = Ve Ve ð7:23Þ
Fp = Rp Up ð7:24Þ
F =V R
p p p
ð7:25Þ
T
Cp = Fp Fp = Up Up ð7:26Þ
-1 -T -1 -1 -1
Bp = Fp Fp = Vp Vp ð7:27Þ
Multiplicative split of deformation gradient tensor F produces nonunique, locally

defined intermediate (relaxed) configurations. One way to solve this problem is to
introduce co-rotational rate definitions of elastic and plastic deformation gradients
based on (director) orthonormal vectors which was proposed by Mandel (1972). The
resulting elastic and plastic parts of the deformation rate tensors are invariant upon
any superposed rigid body rotations. However, the arbitrariness of intermediate
configuration can be also removed by setting the plastic spin tensor Wp equal to
zero (Eq. (7.28)). In this case, elastic and plastic deformation gradients will include
rotations, which can be handled with proper selection of stress and rate measures
establishing a frame indifferent model:
Wp = 0 ð7:28Þ
Jp = 1 ð7:29Þ
Assuming that plastic flow is irrotational (Eq. (7.28)) and incompressible

(Eq. (7.29)), Jacobian of total deformation gradient J and material time derivative
of the plastic deformation gradient F_ will be
p
J = Je ð7:30Þ
F_ = Dp Fp
p
ð7:31Þ
Because of the incompressible plastic flow and irrotational plastic flow assump-
tions, we have the following relations:
Lp = Dp ð7:32Þ
trðL Þ = trðD Þ = 0
p p
ð7:33Þ
-1
L = Le þ F e D p F e ð7:34Þ
The system is assumed to be at rest initially (t = 0) which provides the following

initial conditions:
Fp ðr, 0Þ = I ð7:35Þ
F ðr, 0Þ = I
e
ð7:36Þ
7.2 Frame of Reference Indifference 379
7.2 Frame of Reference Indifference
The material constitutive model must be invariant with respect to the frame of
reference (coordinate system). Therefore, a quantity or an equation is frame indif-
ferent (or objective) if it is invariant to changes in the frame of reference.
Let the rigid body motion of any material point χ(r, t) be defined by a proper
orthogonal rotation tensor Ω(t) and a vector ro(t) for any duration t as follows:
χ ðr, t Þ = Ωðt Þ½χ ðr, t Þ - O þ ro ðt Þ ð7:37Þ
where the orthogonal characteristic of rotation tensor is defined by
ΩT Ω = I ð7:38Þ
Here we define a stress tensor as objective if it does not depend on the frame of
reference. In Eq. (7.37), Ω(t) represents rigid body rotation and ro(t) represents rigid
body translation. In this context, frame indifference of a vector a and a second-order
tensor A are defined in terms of transformation rules as follows:
a = Ωa ð7:39Þ
A = ΩAΩT ð7:40Þ
Transformation of deformation gradient tensor F with respect to a change in the

frame of reference can be obtained from Eq. (7.31) as follows:
F = ΩF ð7:41Þ
Therefore, the deformation gradient tensor is not objective (it is not the frame of
reference indifferent) according to Eq. (7.40). Using Eq. (7.41) Cauchy tensor C in
the new transformed configuration can be given by
T
C = F F = FT ΩT ΩF = FT F = C ð7:42Þ
Since the Cauchy tensor is a Lagrangian tensor referring to the original frame of
reference, the Cauchy tensor should not be expected to change with the current
coordinate system (frame of reference) according to Eq. (7.40). Therefore, Cauchy
tensor C is not objective, but still invariant with respect to changes in the current
frame of reference in the deformed configuration. Using Eq. (7.41) and the right
polar decomposition rule in Eq. (7.16), we can write
F = RU = ΩF = ΩRU ð7:43Þ
Since right polar decomposition is unique, Eq. (7.43) implies that
R = ΩR ð7:44Þ
U=U ð7:45Þ
Therefore, right stretch tensor U and rotation tensor R are not objective. How-
ever, similar to the Cauchy tensor, the right stretch tensor refers to the original frame
of reference, and hence right stretch tensor is invariant to changes in the current
frame of reference. Similarly, using Eq. (7.19), Almansi tensor B in the transformed
configuration, we can write
T
B = FF = ΩFFT ΩT = ΩBΩT ð7:46Þ
Therefore, Almansi tensor B is objective. Using Eq. (7.41) and the left polar
decomposition rule in Eq. (7.17) yields
F = VR = ΩF = ΩVR ð7:47Þ
According to the transformation of rotation tensor in Eqs. (7.44) and (7.47) can be
rewritten as
F = VR = ΩVΩT ΩR ð7:48Þ
Since left polar decomposition is also unique, Eq. (7.48) implies that
V = ΩVΩT ð7:49Þ
Therefore, the left stretch tensor V is objective. Taking the time derivative of
Eq. (7.41), it can be shown that the material time derivative of deformation gradient
tensor F is not objective:
:
_ þ ΩF_
F = ΩF ð7:50Þ
Using Eq. (7.50) and the definition of velocity gradient tensor given in Eq. (7.6),
we can write
:
_ þ ΩF_ = ΩFF
F = LF = ΩF _ -1 T _ - 1 ΩT ΩF
Ω þ ΩFF ð7:51Þ
Equation (7.51) implies that the velocity gradient tensor is not objective as shown
below:
_ T
L = ΩLΩT þ ΩΩ ð7:52Þ
7.2 Frame of Reference Indifference 381
Using definitions of stretch rate tensor and spin tensor given in Eq. (7.10)
_ T
L = D þ W = ΩðD þ WÞΩT þ ΩΩ ð7:53Þ
which implies that
D = ΩDΩT ð7:54Þ
_ T
W = ΩWΩT þ ΩΩ ð7:55Þ
Therefore, stretch rate tensor D is objective, while spin tensor W is not objective.
Using the multiplicative decomposition definition of deformation gradient given in
Eq. (7.7), we can write
F = Fe Fp = ΩFe Fp ð7:56Þ
Hence, we can write the following relations:
Fe = ΩFe ð7:57Þ
Fp = Fp ð7:58Þ
Therefore, elastic and plastic deformation gradients are not objective. Since
plastic deformation gradient refers to the original and intermediate frame of refer-
ence, plastic deformation gradient is not objective but invariant to changes in the
current frame reference. Using definitions for elastic and plastic rotation tensors, left
stretch tensor, right stretch tensor, Cauchy tensor, and Almansi tensor in Eqs. (7.20),
(7.21), (7.22), (7.23), (7.24), (7.25), (7.26), and (7.27), the following relations can be
written:
Fe = Re Ue = ΩRe Ue ð7:59Þ
Re = ΩRe ð7:60Þ
Ue = Ue ð7:61Þ
Fe = Ve Re = ΩVe ΩT Ω Re ð7:62Þ
Ve = ΩVe ΩT ð7:63Þ
Ce = Ue Ue = Ue Ue = Ce ð7:64Þ
Be = Ve Ve = ΩVe ΩT Ω Ve ΩT = ΩBe ΩT ð7:65Þ
Fp = Rp Up = ΩRp Up ð7:66Þ
Rp = ΩR p
ð7:67Þ
Up = Up ð7:68Þ
Fp = Vp Rp = ΩVp ΩT Ω Rp ð7:69Þ
Vp = ΩVp ΩT ð7:70Þ
Cp = Up Up = Up Up = Cp ð7:71Þ
Bp = Vp Vp = ΩVp ΩT ΩVp ΩT = ΩBp ΩT ð7:72Þ
Therefore, elastic and plastic rotation tensors Re, Rp are not objective. Cauchy
tensors, elastic, and plastic right stretch tensors and Ce, Cp, Ue, Up are also not
objective, but invariant to changes in the current frame of reference. Almansi tensors,
elastic, and plastic left stretch tensors Be, Bp, Ve, Vp are all objective. Using elastic
and plastic velocity gradient definitions in Eqs. (7.12) and (7.13) with transformation
rules for elastic and plastic deformation gradient in Eqs. (7.57) and (7.58) yields
:
-1
Le = Fe Fe _ e Fe - 1 ΩT = ΩLe ΩT þ ΩΩ
= ΩFe þ ΩF _ T ð7:73Þ
:
-1
= F_ Fp = Lp
p -1
Lp = Fp Fp ð7:74Þ
Equation (7.73) shows that the elastic component of the velocity gradient is not
objective. According to (7.74) plastic component of the velocity, the gradient is not
objective either, but it is invariant to changes in the current frame of reference,
because the plastic velocity gradient refers to original and intermediate reference
frames, but does not refer to the current frame of reference. Using definitions of
elastic stretch rate tensor and spin tensor given by Eq. (7.14), we can write
_ T
Le = De þ We = ΩðDe þ We ÞΩT þ ΩΩ ð7:75Þ
which implies that
De = ΩDe ΩT ð7:76Þ
_ T
We = ΩWe ΩT þ ΩΩ ð7:77Þ
Therefore, elastic stretch rate tensor De is objective, while elastic spin tensor We
is not objective. Similarly, using the plastic stretch rate definition given in Eq. (7.15)
and applying the irrotational plastic flow assumption in Eq. (7.28), we can write the
following relations:
Lp = Dp = Dp ð7:78Þ
which implies that:

7.3 Unified Mechanics Theory Formulation for Finite Strain 383
Dp = Dp ð7:79Þ
Therefore, plastic stretch rate tensor Dp is not objective but invariant to changes
in the current reference frame. To summarize, it has been shown that
_
F,R,F,L,W,F ,R ,R ,L ,We are not objective; C,U, Fp, Ue, Ce, Up, Cp, Lp, Dp are
e e p e
not objective but invariant to changes in the current reference frame; and B,V,D, Ve,
Be, Vp, Bp, De are objective.
7.3 Unified Mechanics Theory Formulation for Finite

Strain
A thermodynamic potential is concave with respect to temperature (θ) and convex

with respect to all other internal state variables. It provides a basis for satisfying
conditions of thermodynamic stability imposed by Clausius-Duhem inequality.
Specific Helmholtz free energy (ψ) forms such a basis. Helmholtz free energy (ψ)
is defined as the difference between specific internal energy (u) and the product of
absolute temperature (θ) and specific entropy (s):
ψ = u - θs ð7:80Þ
The first law of thermodynamics (also known as the conservation of energy law)
states that energy can be transported, or converted from one form to another, but
cannot be destroyed or created. Accordingly, the internal energy of a system can
change by heat flow into or out of the system (δq), heat generated within the system
due to internal scattering and friction (r), mechanical work done on the system by the
external load (δw), or all other kinds of work done on the system (δw′) during any
process, including but not limited to electrical, thermal, chemical, etc. Considering
only the thermomechanical effects, rate of change in internal energy, u,_ can be given
by
u_ = Q þ l þ r ð7:81Þ
where Q is the rate of net heat flow into the system and l is the rate of net work done
on the system and r is the heat generated in the system, due to internal friction,
scattering, and chemical reactions. The second law of thermodynamics states that
there exists a thermodynamic state function, entropy, which increases in the universe
for all types of real processes due to irreversible entropy production. Unlike energy,
entropy (s) is not only transferred across boundaries of a system, but it may also be
created in the system which is called entropy production. Clausius-Duhem inequality
describes a condition imposed by the second law of thermodynamics in terms of a
nonnegative entropy production rate (γ) per unit volume for any kind of irreversible
process which is defined as
γ = ρ_s þ divðJs Þ > 0 ð7:82Þ
Js is net entropy flux into the system. Heat flows into the body can be defined in
terms of heat flux as follows:

ρQ = - div Jq ð7:83Þ
Substituting the time derivative of Eq. (7.80), internal energy definition given by
Eq. (7.81), and heat flow equation in Eq. (7.83) into Eq. (7.82), internal entropy
production rate density can be rewritten as

div Jq ρ r ρ
_ >0
γ = divðJs Þ - þ þ l - ψ_ - θs ð7:84Þ
θ θ θ
r is the internal heat generation source strength and ρ is the mass density.
7.3.1 Thermodynamic Restrictions
The main premise of the unified mechanics theory is the derivation of the thermo-
dynamic fundamental equation, which must contain all entropy generation terms for
all active mechanisms. Of course, the amount of entropy generated in each micro-
mechanism will be different. After a comparison of all mechanisms, it is reasonable
to ignore mechanisms that generate entropy in smaller amounts by orders of mag-
nitude. For dual-mechanism rate-dependent constitutive modeling of polymers, the
material response can be resolved into two components which necessitate multi-
mechanism generalization of multiplicative decomposition in Eq. (7.7) and descrip-
tion of different Helmholtz free energy functions and associated fundamental equa-
tion assuming that linear addition is applicable for Eq. (7.80) Accordingly,
Eqs. (7.12), (7.13), (7.14), (7.15), (7.20), (7.27), and (7.31) hold for each component
of the material’s resistance mechanism. Subscripts “I” and “M” will be used to
designate the component of a quantity in the Intermolecular mechanism and Molec-
ular network mechanism, respectively. For a description of dissipation inequality,
total Helmholtz free energy density in (original) reference configuration is written as
a summation of defect energy (ΨD) and elastic energy stored in an intermolecular
structure (ΨI) and molecular network structure(ΨM):

Ψ CeI , CeM , A, θ = ΨI EeI , θ þ ΨM CeM , θ þ ΨD ðA, θÞ ð7:85Þ
Defect energy (ΨD) is assumed to depend on a stretch like tensor A and temper-
ature (θ), elastic energy stored in an intermolecular structure (ΨI) is assumed to
depend on logarithmic elastic strain in the intermolecular structure EeI and temper-
ature (θ), and elastic energy stored in molecular network structure (ΨM) is assumed
to depend on elastic Cauchy tensor in molecular network structure CeM and temper-
ature (θ). If a similar decomposition also holds for specific entropy and specific
Helmholtz free energy, we can write the following relation:

s CeI , CeM , A, θ = sI EeI , θ þ sM CeM , θ þ sD ðA, θÞ ð7:86Þ

ψ CeI , CeM , A, θ = ψ I EeI , θ þ ψ M CeM , θ þ ψ D ðA, θÞ ð7:87Þ
Helmholtz free energy density in reference configuration (Ψ) can be simply

related to specific Helmholtz free energy function (ψ) through Eq. (7.88) and in
rate form, as follows:

ρo ψ CeI , CeM , A, θ = Ψ CeI , CeM , A, θ ð7:88Þ

_ Ce , Ce , A, θ
ρψ_ CeI , CeM , A, θ = J - 1 Ψ ð7:89Þ
I M
where ρo and ρ are mass densities in reference configuration and deformed config-
uration, respectively. Note that since eigenvalues of elastic Cauchy tensor CeI and
logarithmic elastic strain tensor EeI corresponding to the intermolecular structure are
related through Eq. (7.90) and eigenvectors of these tensors are identical, it is
possible to consider Helmholtz free energy density (ΨI) and specific entropy (sI)
associated with intermolecular structure as functions of temperature (θ) and elastic
Cauchy tensor CeI ; hence, we can write the following relations:
1
eigenvalue EeI = ln eigenvalue CeI ð7:90Þ
2
e
ΨI EI , θ ΨI CeI , θ ð7:91Þ

sI EeI , θ sI CeI , θ ð7:92Þ
The time derivative of Helmholtz free energy can be given by

2 3
∂ΨI EeI , θ _ e ∂ΨM CeM , θ _ e
6 : CI þ : CM þ 7
∂CeI ∂C
eMe
e e 6 7
_ C , C , A, θ = 6
Ψ 6 ∂ΨD ðA, θÞ _ ∂ΨI EI , θ 7
ð7:93Þ
I M
6 þ :Aþ :θþ 7_
7
4 ∂A ∂θ
∂ΨM CeM , θ _ ∂ΨD ðA, θÞ _ 5
þ :θþ :θ
∂θ ∂θ
According to the principle of virtual work, rate of work done per unit volume of
the deformed body (external work) is balanced with internal work, while total work
done on the system is stored as elastic strain energy (represented by the first two
terms in Eq. (7.95) and dissipated as plastic work (represented with the last two terms
in Eq. (7.95)):
ρl = w_ int ð7:94Þ
w_ int = ΓeI : LeI þ ΓeM : LeM þ J - 1 ΓpI : LpI þ J - 1 ΓpM : LpM ð7:95Þ
where ΓeI , ΓeM , ΓpI , and ΓpM are stress measures conjugate to the rate of deformation
measures LeI, LeM, LpI, and LpM , which were defined in Eqs. (7.12) and (7.13). Noting
-1
that J - 1 = J e , the J-1 the multiplier in front of the last two terms in Eq. (7.95)
recovers work definitions from intermediate configuration to deformed configura-
tion. The requirement of frame indifference for internal work can be given by the
following relation:
:
w_ int = w ð7:96Þ
int
which implies that

e e e e
ΓeI :LeI þΓeM :LeM þ ΓI :LI þΓM :LM þ
þJ - 1 ΓpI :LpI þJ - 1 ΓpM :LpM
= -1 p p -1 p
þJ ΓI :LI þJ ΓM :LM
p ð7:97Þ
Using transformation rules for elastic and plastic velocity gradients in Eqs. (7.73)
and (7.74), we can write the following relations:

ΓI :ðΩLeI ΩT þΩΩ
_ T ÞþJ - 1 Γ :Lp þ
e p
ΓeI :LeI þJ - 1 ΓpI :LpI þ
þΓeM :LeM þJ - 1 ΓpM :LpM
= þΓM :ðΩLeM ΩT þΩΩ
e I I
_ T ÞþJ - 1 Γp :Lp ð7:98Þ
M M
or
" #
ðΩT ΓI ΩÞp:LeI þ
e
ðΩT ΓMpΩÞp:LeM þ
e
ΓeI :LeI þJ - 1 ΓpI :LpI þ
þΓeM :LeM þJ - 1 ΓpM :LpM
= þJ ΓI :LI þJ - 1 ΓM :LM þ
-1 p
ð7:99Þ
e
_ T þΓe :ΩΩ
þΓI :ΩΩ _ T
M
The first two terms on the right-hand side of Eq. (7.99) indicate that stress
measures corresponding to elastic work ΓeI , ΓeM are objective as shown in

Eqs. (7.100) and (7.101). Since ΩΩ _ T is a skew-symmetric tensor (Eqs. (7.55)
and (7.77)), the last two terms on the right-hand
side
of Eq. (7.99) imply that stress
measures corresponding to elastic work ΓeI , ΓeM are symmetric (Eqs. (7.102) and
(7.103)). The third and fourth terms on the right-hand side of Eq. (7.99) show that
stress measures corresponding to plastic work ðΓpI , ΓpM Þ are not objective but
invariant to changes in the current frame of reference as shown in Eqs. (7.104) and
(7.105). Hence,
e
ΓI = ΩΓeI ΩT ð7:100Þ
e
ΓM = ΩΓeM ΩT ð7:101Þ
e eT T
ΓI = ΓI ; ΓeI = ΓeI ð7:102Þ
T
e e T
ΓM = ΓM ; ΓeM = ΓeM ð7:103Þ
p
ΓI = ΓpI ð7:104Þ
p
ΓM = ΓpM ð7:105Þ
Using symmetry property of stress tensor in Eqs. (7.102) and (7.103), and
irrotational plastic flow definition given in Eq. (7.28), total internal work over the
whole volume of a system can be written as
Z Z

_ int =
W w_ int dV = ΓeI : DeI þ ΓeM : DeM þ J - 1 ΓpI : DpI þ J - 1 ΓpM : DpM dV
V V
ð7:106Þ
Total external work acting on the system can be described in terms of surface
tractions on boundaries of the system and body forces acting on the system as
follows:
Z Z Z
_ ext =
W ρ l dV = σ ðnÞ
χ_ dA þ b χ_ dV ð7:107Þ
V A V
where A represents the surface and V represents volume. Consider the principle of
virtual work for a special case defined in Eq. (7.34) as
L = gradðχ_ Þ = LeI = LeM ð7:108Þ
where it is assumed that
DpI = DpM = 0 ð7:109Þ
The principle of virtual work can then be rewritten for this special case from
Eqs. (7.106) and (7.107) as
Z Z Z
ðnÞ
σ _
χdA þ _
b χdV = ½ΓeI : DeI þ ΓeM : DeM dV ð7:110Þ
A V V
Z Z Z
σðnÞ χdA
_ þ _
b χdV = _
½ðΓeI þ ΓeM Þ : gradðχÞdV ð7:111Þ
A V V
Z Z Z
σðnÞ χdA
_ þ _
b χdV = div½χ_ ðΓeI þ ΓeM ÞdV
A V V
Z
- divðΓeI þ ΓeM Þ χ_ dV ð7:112Þ
V
Since Eq. (7.112) is true for any choice of V and gradðχ_ Þ, from the first terms on
the left- and right-hand side, we can write

σ ðnÞ = ΓeI þ ΓeM n ð7:113Þ
which is essentially Cauchy’s stress theorem describing the relation between stress
tensor and surface tractions. From the second terms on the left- and right-hand side
of Eq. (7.112), we can write

div ΓeI þ ΓeM þ b = 0 ð7:114Þ
which represents Cauchy’s equation of motion for stationary systems. Therefore,

stress measures in Eqs. (7.113) and (7.114) correspond to Cauchy stress
T components in the intermolecular mechanism TI and molecular network mecha-
nism TM, respectively. Hence, we can write the following relations:
T = ΓeI þ ΓeM ð7:115Þ

ΓeI = TI = TT I ð7:116Þ
ΓeM = TM = TM T ð7:117Þ
Consider the principle of virtual work for a second special case defined from
Eq. (7.34) such that
-1 -1
L = gradðχ_ Þ = LeI þ FeI DpI FeI = LeM þ FeM DpM FeM = 0 ð7:118Þ
or
-1
LeI = - FeI DpI FeI ð7:119Þ
-1
LeM = - FeM DpM FeM ð7:120Þ
Accordingly, the principle of virtual work can be rewritten for this special case
using Eqs. (7.106) and (7.197) as
Z -1

_ int = ΓeI : - FeI DpI FeI - 1 ΓpI :Dp pI þp
þJ
W þΓeM : - FeM DpM FeM
-1
þJ - 1 ΓM :DM
dV ð7:121Þ
V
or
_ int
W
Z h i
-T -T
J - 1 ΓpI - FeI ΓeI FeI : DpI þ J - 1 ΓpM - FeM ΓeM FeM
T T
= : DpM dV
V
ð7:122Þ
Since the velocity gradient is assumed to be zero for this special case, the velocity
field
will be also equal to zero. Therefore, external work will be equal to zero
_ ext = 0 . As a result, individual terms inside the parenthesis in Eq. (7.122) should
W
be equal to zero for arbitrary selection of V, DpI , and DpM :
-T
J - 1 ΓpI - FeI ΓeI FeI
T
=0 ð7:123Þ
-T
J - 1 ΓpM - FeM ΓeM FeM = 0
T
ð7:124Þ
Using definitions of stress measures in Eqs. (7.116) and (7.117), it can be shown
that
T -T
ΓpI = JFeI TI FeI ð7:125Þ
T -T
ΓpM = JFeM TM FeM ð7:126Þ
which form the definition of elastic symmetric Mandel stress in intermolecular

structure and molecular network structure as
T -T
MeI = JFeI TI FeI ð7:127Þ
T -T
MeM = JFeM TM FeM ð7:128Þ
whereas since the trace of plastic stretch rate is equal to zero due to incompressible
plastic flow, assumption stress conjugate to plastic stretch rate should be a deviatoric
tensor. Therefore, we can write

ΓpI = dev MeI ð7:129Þ

ΓpM = dev MeM ð7:130Þ
Finally, elastic second Piola-Kirchhoff stress tensor in intermolecular structure

and molecular network structure can be defined as follows:
-1 -T
SeI = JFeI TI FeI ð7:131Þ
-1 -T
SeM = JFeM TM FeM ð7:132Þ
Using definitions of elastic Mandel stress given by Eqs. (7.127) and (7.128) and
symmetric second Piola-Kirchhoff stress given by Eq. (7.131) and Eq. (7.132), we
can write the following relation:
MeI = CeI SeI ð7:133Þ

MeM = CeM SeM ð7:134Þ

Therefore,
it has been shown that stress measures ΓeI , ΓeM , ΓpI , ΓpM
[or ΤI , ΤM , MeI , MeM with deformation rate conjugates of DeI , DeM , DpI , DpM
form a frame indifferent framework for dual-mechanism elastic-viscoplastic consti-
tutive model. The thermodynamic restrictions on constitutive relations can be
obtained by substituting Eqs. (7.86), (7.89), and (7.93) into Eq. (7.84) as follows:
Jq
divðJs Þ - div =0 ð7:135Þ
θ
Jq
Js = ð7:136Þ
θ
1 ρr
γ ther = - div Jq ∇x ðθÞ þ > 0 ð7:137Þ
θ 2 θ
e e
∂ΨI EI , θ ∂ΨM CM , θ ∂ΨD ðA, θÞ
J -1 þ þ þ ρs θ_ = 0 ð7:138Þ
∂θ ∂θ ∂θ
e e
e e - 1 ∂Ψ EI , CM , A, θ
ρs EI , CM , A, θ = - J ð7:139Þ
∂θ
e e
∂ψ EI , CM , A, θ
s EeI , CeM , A, θ = - ð7:140Þ
∂θ
e
1 ∂ΨI EI , θ
: C_ I =
e
ΓeI : LeI - J - 1 ð7:141Þ
θ ∂CI e
e
- 1 ∂ΨI EI , θ
: C_ I
e
ΓI : LI = TI : LI = TI : DI = J
e e e e
ð7:142Þ
∂CeI
e
1 e-1 _ - 1 ∂ΨI EI , θ
e-T
: C_ I
e e
TI : DI =
e
F TI FI : CI = J ð7:143Þ
2 I ∂CeI
-1
e-T ∂ΨI EeI , θ
e
J FI TI FI = SI = 2
e
ð7:144Þ
∂CeI
e
1 - 1 ∂ΨM CM , θ
: C_ M =
e
ΓM : LM - J
e e
ð7:145Þ
θ ∂CM e
e
- 1 ∂ΨM CM , θ
: C_ M
e
ΓM : LM = TM : LM = TM : DM = J
e e e e
ð7:146Þ
∂CeM
e
1 e
_ - 1 ∂ΨM CM , θ
: C_ M
e e
TM : DM = FM TM FM : CM = J
e e
ð7:147Þ
2 ∂CeM
-1
e-T ∂ΨM CeM , θ
J FM TM FM = SM = 2
e e
ð7:148Þ
∂CeM
1 ∂ΨD ðA, θÞ _
γ mech = J - 1 ΓpI : LpI þ J - 1 ΓpM : LpM - J - 1 : A >0 ð7:149Þ
θ ∂A
Equation (7.136) relates entropy flux to heat flux. Equation (7.137) defines
irreversible entropy production per unit volume in a deformed configuration associ-
ated with heat conduction which is always positive according to Fourier’s law
(Eq. (7.150)). Equation (7.140) provides the relation between specific entropy and
specific Helmholtz energy. Equations (7.144) and (7.148) describe stress measures
derived from Helmholtz free energy functions. Equation (7.149) defines irreversible
entropy production due to mechanical dissipation per unit volume in a deformed
configuration. The reason for the appearance of J-1 term in front of Eq. (7.149) is
that all stress measures and their conjugate rate measures refer to intermediate
configuration (relaxed configuration), while irreversible entropy production rate
density refers to deformed configuration. Heat flux is given by
Jq = - k∇x ðθÞ ð7:150Þ
where k is the temperature-dependent thermal conductivity of the material. Specific

heat can be expressed in terms of specific entropy (Eq. (7.151)) and terms of
Helmholtz free energy (Eq. (7.153)) by exploiting the relation given by
Eq. (7.139). Using linear decomposition assumption in Eqs. (7.85) and (7.86),
specific heat can be rewritten in terms of separate components of specific entropy
and Helmholtz free energy as shown in Eq. (7.154):

∂s EeI , CeM , A, θ
c=θ ð7:151Þ
∂θ
e e
∂sI EI , θ ∂sM CM , θ ∂s ðA, θÞ
c=θ þ þ D ð7:152Þ
∂θ ∂θ ∂θ
-1 e
∂ J ∂Ψ EI , CM , A, θ
e
c= -θ ð7:153Þ
∂θ ρ ∂θ
e
∂ J - 1 ∂ΨI EI , θ ∂ΨM CeM , θ ∂ΨD ðA, θÞ
c= -θ þ þ ð7:154Þ
∂θ ρ ∂θ ∂θ ∂θ
7.3.2 Constitutive Relations
In constitutive modeling of amorphous polymers for large deformations, dual

decomposition of material response into two parallel working mechanisms of
intermolecular structure and molecular network structure is widely used (Gunel
and Basaran 2009, 2011a, b). A triple-mechanism constitutive model has also been
proposed to include a secondary mechanism for the molecular network structure
(Srivastava and Anand 2010). Both the dual- and triple-mechanism models are
proven to be successful in describing the large deformation behavior of amorphous
polymers at different isothermal conditions. To extend the applicability of such
models to non-isothermal conditions, several refinements on material property
definitions and viscoplastic flow rule definitions are necessary. Polymers have a
very low melting temperature; hence, using time-independent plasticity is not an
option.
In dual-mechanism constitutive models, a material response is assumed to be
controlled by two mechanisms working in parallel (intermolecular structure and
molecular network structure), as depicted in Fig. 7.2. Intermolecular (I ) and molec-
ular network mechanisms (M) work in parallel; deformation in both mechanisms is
equal to each other and equal to total deformation (Eq. (7.155)), while total stress is
the summation of stresses due to intermolecular interactions (I) and molecular
network interactions (M ) (Eq. (7.156)). Subscripts “I” and “M” represent
intermolecular and molecular network components of associated quantity,
respectively:
F = FI = FM ð7:155Þ
T = TI þ TM ð7:156Þ
Fig. 7.2 Schematic Molecular Network Resistance, M

representation of a material
model
Intermolecular Resistance, I
Intermolecular Resistance (I)

In most polymers, the initial elastic response due to intermolecular resistance is
governed by van der Waals bonds with surrounding molecules. A Helmholtz free
energy per unit volume in reference configuration is considered for constitutive
relation describing intermolecular resistance which was developed by Anand and
On (1979) and Anand (1986):
n 1

2 2
o
ΨI EeI , θ = G devðEIe Þj2 þ K - G tr EeI - - 3Kαðθ - θo ÞtrðEeI Þ ð7:157Þ
2 3
where θo is the initial temperature, θo = θ(r, to), and G, K, α are temperature-

dependent shear modulus, bulk modulus, e and coefficient of thermal expansion,
respectively.
Elastic logarithmic strain EI is related to the right elastic stretch
tensor CeI and elastic deformation gradient FeI through the following relations:
1
EeI =
ln CeI ð7:158Þ
2
T
1
EeI = ln FeI FeI ð7:159Þ
2
Utilizing Eq. (7.144) symmetric second Piola-Kirchhoff stress tensor SeI and Cauchy
stress tensor TI can be obtained from Helmholtz free energy density function
corresponding to intermolecular resistance as follows:

∂ΨI EeI , θ
SeI =2 ð7:160Þ
∂CeI
TI = J - 1 FeI SeI FeI

T
ð7:161Þ
Since Helmholtz free energy density corresponds to macroscopic elastic energy

stored, (ΨI) is an isotropic function of CeI elastic right Cauchy tensor; hence, CeI and
∂ΨI =∂CeI are coaxial and their product is a symmetric tensor, called elastic Mandel
stress tensor MeI which is given by
MeI = CeI SeI ð7:162Þ
Relation between elastic Mandel stress MeI and elastic logarithmic strain EeI can
be obtained from Helmholtz free energy function (Eq. (7.157)) and second Piola-
Kirchhoff stress tensor SeI , and the definition is given in Eq. (7.160); hence, we can
write

MeI = 2Gdev EeI þ K tr EeI - 3αðθ - θo Þ I ð7:163Þ
Kinematic hardening in intermolecular structure is modeled by a defect energy

function per unit volume in intermediate (relaxed) configuration (Anand et al. 2009):
h i
1
ΨD ðA, θÞ = B ln ða1 Þ2 þ ln ða2 Þ2 þ ln ða3 Þ2 ð7:164Þ
4
where ai represents eigenvalues of a stretch-like internal variable A which is a

symmetric unimodular tensor, detA(x, t) = 1. Defect energy (ΨD) is an isotropic
function of symmetric unimodular stretch-like tensor A. As a result, A and ∂ΨD/∂A
are coaxial and their product is a symmetric deviatoric back-stress tensor Mback
which is given by Eq. (7.165), and the evolution equation for A is given by
Eq. (7.166) with an initial condition defined by Eq. (7.167) as follows:
∂ΨðA, θÞ
Mback = 2dev A = B ln ðAÞ ð7:165Þ
∂A
A_ = DpI A þ ADpI - γA ln ðAÞνpI ð7:166Þ
AðX, 0Þ = I ð7:167Þ
where γ represents the dynamic recovery, B is the temperature-dependent back-stress

modulus, and νpI is the equivalent plastic stretch rate in intermolecular structure.
Driving stress for plastic flow in intermolecular structure is defined as

Meff = dev MeI - Mback ð7:168Þ
The equivalent plastic stretch rate is given by Eq. (7.169), effective equivalent
shear stress is given by Eq. (7.170), and mean normal pressure is defined in
Eq. (7.171), and they are all defined in terms of tensorial variables as follows:
pffiffiffi
νpI = 2 DpI ð7:169Þ
1
τI = pffiffiffi jMeff j ð7:170Þ
2
1
pI = - tr MeI ð7:171Þ
3
The evolution of plastic deformation gradient in the intermolecular mechanism

can be rewritten from Eq. (7.31) as follows:
F_ I = DpI FpI
p
ð7:172Þ
FpI ðr, 0Þ = I ð7:173Þ
Only the effective equivalent shear stress drives the plastic flow, and it is the
source of plastic dissipation. Some part of plastic work is stored as energy associated
with back-stress. In polymers, once the effective shear stress level reaches a critical
level so that the energy barrier to molecular chain segment rotation is exceeded,
plastic flow takes place. This is in contrast to the dislocation slip mechanism in metal
plasticity. It is important to point out that the concept of dislocation and dislocation
motion-induced plasticity model does not apply to amorphous polymers because the
concept of dislocation cannot be justified in amorphous polymers.
According to the cooperative model, viscous flow in a solid amorphous polymer
may take place only when several polymer segments move cooperatively which also
accounts for the significance of activation volume during the yield process. The flow
rule for amorphous polymers is essentially based on the energy distribution statistics
of individual segments (Fotheringham and Cherry 1978). In simple terms, the
cooperative model flow rule is based on the average probability of simultaneous
occurrence of n thermally activated transitions across an energy barrier (activation
energy, Q) inducing a macroscopic strain increment of νo (Fotheringham et al. 1976;
Fotheringham and Cherry 1978). Plastic flow characteristics of amorphous polymers
are strongly temperature- and rate-dependent. According to the strain rate-
temperature superposition principle, an increase in temperature will have the same
effect on the yield stress as a decrease in strain rate (Francisco et al. 1996).
Equivalence of time and temperature describes that the yielding of amorphous
polymers at low temperatures is comparable to that at high strain rates. Therefore,
Eyring plots (yield stress-temperature ratio vs. plastic strain rate curves) for various
temperatures can be shifted vertically and horizontally with respect to a reference
temperature (θref) to obtain a master curve describing yield stress behavior over a
wide range of temperatures and strain rates.
Richeton et al. (2006) proposed that both horizontal shift (ΔHh) and vertical shift
(ΔHv) should follow Arrhenius-type temperature dependence. The resulting yield
stress definition relates yield behavior of polymer with β mechanical loss peak at
temperatures below the glass transition temperature θg through introducing activa-
tion energy at β-transition temperature, i.e., yield behavior in polymers is controlled
by segmental motions of polymer chains, and the reference state for yielding is
chosen as β-transition. An increase in yield stress due to an increase in strain rate is
attributed to a decrease in molecular mobility of molecular chains, while a slow
deformation rate allows polymer chains to slip past each other, resulting in lower
resistance to flow. At low temperatures near secondary transition temperature (θβ),
secondary molecular motions are restricted, and chains become stiffer which also
increases the yield stress, while the increase in temperature provides more energy to
polymer chains facilitating relative motion between polymer molecular chains. For
temperatures above the glass transition temperature θg, characteristic plastic strain
rate equation was modified by Williams-Landel-Ferry (WLF) parameters (c1, c2).
Although characteristic plastic strain rate definitions at temperatures below and
above the glass transition temperature, θg, are continuous functions of temperature
in separate domains, piece-wise definition with respect to the glass transition tem-
perature, θg, results in an unrealistic change in plastic flow behavior around glass
transition temperature, i.e., the derivative of plastic strain rate equation is
discontinuous at θg. Srivastava and Anand (2010) proposed a modified version of the
flow rule in intermolecular structure which incorporates different values of activation
energy for the glassy region and rubbery region, but still, the constitutive model
makes an abrupt change in the activation energies at the glass transition temperature,
θg, which creates a discontinuity problem in computational mechanics simulations.
To provide a smoother transition in flow characteristics around glass transition
temperature, θg, characteristic plastic strain rate given by Eq. (7.174) and equivalent
shear plastic stretch rate given by Eq. (7.175) can be modified in the following
forms:
" !#
Q ln ð10Þc1 θ - θg
ν = νoI exp - I 1 þ exp ð7:174Þ
kB θ c2 þ θ - θ g
nI
p τI V
νI = ν sinh ð7:175Þ
2kB θ
where νoI is the pre-exponential factor, QI is the activation energy for plastic flow in
intermolecular structure, kB is Boltzmann’s constant, c1 and c2 are WLF parameters,
nI is the number of thermally activated transitions necessary for plastic flow, V is the
activation volume, and τI is the net effective stress which is defined as
τI = τI - SI - αP pI ð7:176Þ
where αp is hydrostatic pressure sensitivity parameter and SI is plastic flow resistance

in intermolecular structure. The evolution of intermolecular resistance to plastic flow
can be given by

S_ I = hI SI - SI νpI ð7:177Þ
with the initial condition given by
SoI = SI ðr, 0Þ ð7:178Þ
where hI is a parameter characterizing hardening-softening and SI is saturation value

for plastic flow resistance in intermolecular structure which can be defined as
follows:
SI = bðϕ - ϕÞ ð7:179Þ
where b is a temperature- and rate-dependent parameter that relates the saturation

value of plastic flow resistance to order function (ϕ - ϕ). Resistance
to plastic flow
(SI) increases with the order in the material and becomes constant SI when order
parameter (ϕ) reaches a critical value (ϕ) which is also a temperature- and rate-
dependent variable. When intermolecular resistance reaches saturation value, steady-
state plastic flow occurs and the plastic flow rate becomes equal to the applied strain
rate. The evolution equation for order parameter is defined as
ϕ_ = gðϕ - ϕÞνpI ð7:180Þ

ϕo = ϕðr, 0Þ ð7:181Þ
where g is a temperature-dependent parameter. The evolution equation for plastic

deformation gradient is given by Eqs. (7.172) and (7.173) which completes the
definition of material behavior in intermolecular structure. Strain hardening becomes
insignificant as temperatures approach glass transition temperature, θg, and
completely vanishes above the glass transition temperature θg (Richeton 2006).
Definition of a vanishing internal resistance right at the glass transition temperature
θg causes also discontinuity in yield behavior of polymers. Since annealing at high
temperatures well above the glass transition temperature, θg, clears past
thermomechanical history of a polymer by providing an alternative stationary
molecular configuration at a higher energy level, internal resistance is bound to
vanish above or around the glass transition temperature θg. Therefore, the underlying
problem is essentially the assumption that glass transition temperature takes place at
a single temperature point and internal resistance becomes zero abruptly exactly at
the glass transition temperature
θg. In the constitutive model presented in this
chapter, similarly, variables b, g, SI , ϕ , hI that characterize hardening-softening
behavior in the post-yield region shall also provide a smooth transition from
temperatures below the glass transition temperature θg to temperatures above θg.
It should be noted that viscoplastic models as of the writing of this book available
in the literature are all-phenomenological and serve as a mathematical tool to fit
experimentally observed behavior into a curve. These empirical models can provide
reasonably accurate predictions for viscoplastic characteristics of amorphous poly-
mers for only isothermal cases. In the case of non-isothermal loading, which
includes temperature change in the material concurrently with loading, most material
models available in the literature predict unrealistic results. A comparison of
viscoplastic models for amorphous polymers from literature and the improved
version of the dual-mechanism model is presented in Fig. 7.3.
Temperature variations of characteristic viscoplastic shear strain rates in different
models are presented by normalizing with respect to characteristic viscoplastic strain
rate at the reference glass transition temperature for PMMA (387 K). Material
properties are taken from Srivastava and Anand (2010), while WLF parameters in
the model discussed in this chapter and Richeton et al. (2006) are taken as their
original values. Viscoplastic models presented in this chapter are all applicable for
temperatures both above and below the glass transition temperature. In the model
derived earlier in this chapter, the temperature dependence of viscoplastic stretch rate
is directly employed by utilizing physically motivated Williams-Landel-Ferry
parameters in a new format as presented in Eq. (7.174). Temperature-dependent
activation energy approach used in Anand’s model does not provide a smooth
change in behavior of intermolecular mechanism but an abrupt increase in
viscoplastic strain rate over a relatively narrow temperature window. Therefore,
Fig. 7.3 Comparison of different viscoplastic models in literature in terms of the temperature
dependence of normalized characteristic viscoplastic shear strain rate νp =νpðθ = θg Þ , Gunel (2010)
according to Anand’s model, there is no rubbery region, but the material response is
liquid-like as the viscoplastic strain rate increases by six orders of magnitude in a
narrow temperature interval (2 °C) around the glass transition temperature. On the
other hand, in Richeton’s model, there is also a remarkable change in viscoplastic
strain rate due to the piece-wise definition with respect to the glass transition
temperature, but it also causes a discontinuous derivative of viscoplastic rate func-
tion at the glass transition temperature. Accordingly, the material response predicted
by Anand or Richeton viscoplastic models has a significant (abrupt) change in stress
value at the glass transition temperature. An improved version of the dual-
mechanism constitutive model presented in this chapter provides a gradual transition
in material response with respect to the temperature around the glass transition.
Though Anand’s model has a continuous definition of activation energy in the
temperature domain, a remarkable difference between activation energies in glassy
and rubbery regions still produces an abrupt change in the polymer’s response.
According to Fig. 7.3, Anand’s viscoplastic model is invariant of temperature above
glass transition temperature, while Richeton’s model provides a relatively gentler
transition in response. As a result of this rapid change in viscoplastic response in
Anand’s model, it cannot predict material behavior accurately under non-isothermal
conditions. These models were further studied by Gunel and Basaran (2010) in terms
of predictions for creep strain rate in a creep test with a stress level of 0.6 MPa which
is conducted at different temperatures as depicted in Fig. 7.4.
Figure 7.4 supports previous observations just discussed. All models predict the
same creep strain rates at temperatures below the glass transition. For assurance of
accurate and realistic modeling of material response around the glass transition
temperature, every aspect of material property definitions that describe hardening-
softening behavior and flow function must be continuous in the temperature domain
and must have continuous derivatives with respect to temperature.
Fig. 7.4 Comparison of different viscoplastic models in literature in terms of creep strain rates at
different temperatures in response to applied stress of 0.6 MPa
Molecular Network Resistance (M)

Resistance in the molecular network to deformation varies according to molecular
orientation and relaxation process. If there is enough stretch in polymer chains, the
network resists relaxation, and resistance increases with increasing stretch. In liter-
ature, there is a consensus on the modeling of the plastic flow behavior of
intermolecular structure based on the Eyring cooperative model, but there is still
debate on the modeling of molecular network resistance. Arruda and Boyce (1993)
modeled molecular network resistance with a rubber elasticity model based on the
eight-chain network of non-Gaussian chains similar to the transient response of
elastomers which represents nonlinear rate-dependent deviation from the equilib-
rium state (Bergstrom and Boyce 1998). Unfortunately, the network resistance
predicted by the eight-chain model is not accurate in describing the orientational
hardening behavior of amorphous polymers and simulations do not match experi-
mental observations. This is because of temperature dependence of rubbery modulus
and the number of rigid links between polymer chain segments, which control the
response. Therefore, molecular network description based on the eight-chain model
(Richeton et al. 2007; Arruda et al. 1995; Palm et al. 2006; Boyce et al. 2000) is
merely a numerical tool to match experimentally observed stress-strain response.
Instead, a simpler two-constant constitutive model for rubber networks developed by
Gent (1996) was shown to describe strain hardening due to polymer chain stretching
better than the statistical-mechanical entropic rubber elasticity model (eight-chain
model) while resulting in a similar stress-strain response according to Ames et al.
(2009) and Srivastava and Anand (2010). Gent’s (1996) model-free energy per unit
volume in reference configuration describes elastic energy stored in molecular
network structure in terms of the first invariant of stretch in polymer chains as
1 I -3
ΨM CeM , θ = - μM I M ln 1 - 1 ð7:182Þ
2 Im
where μM and IM are temperature-dependent rubbery shear modulus and limit the
extensibility of polymer chains, respectively. Since volume change in a polymer is
considered in the elastic deformation gradient associated with intermolecular struc-
ture, it is essential to define Gent free energy
in terms of distortional elastic
deformation gradient in network structure, FeM d , as given by Eq. (7.183) which
produces no change in volume. I1 is the first invariant of elastic distortional Cauchy
tensor in network structure, CeM d :
-1
FeM d = J =3 FeM ð7:183Þ

det FeM d = 1 ð7:184Þ
e T
CM d = FeM d FeM d ð7:185Þ

I 1 = tr CeM d ð7:186Þ
Second Piola-Kirchhoff stress tensor SeM and Cauchy stress tensor TM can be
derived from Gent free energy as follows:

∂ΨM CeM , θ
SeM =2 ð7:187Þ
∂CeM
TM = J - 1 FeM SeM FeM

T
ð7:188Þ
Using Gent free energy in stress definitions results in
- 1h i
I1 - 3 1 e e - 1
SeM = J - 2=3 μM 1 - I- tr CM d CM d ð7:189Þ
IM 3
-1
I1 - 3
TM = J - 1 μ M 1 - dev BeM d ð7:190Þ
IM

Elastic distortional Almansi tensor, BeM d , is defined in terms of distortional

elastic deformation gradient FeM d Þ as
T
BeM d
= FeM d FeM d ð7:191Þ
For elastic Mandel stress tensor MeM and equivalent shear stress τM and equivalent
plastic shear strain rate νpM for a molecular relaxation in-network, the structure can be
given by the following relations:
MeM = CeM SeM ð7:192Þ

-1
I1 - 3
MeM = μM 1 - dev CeM d ð7:193Þ
IM
1
τM = pffiffiffi MeM ð7:194Þ
2
p ffiffiffi
νpM = 2 DpM ð7:195Þ
The evolution of plastic deformation gradient in molecular network mechanism

F_ M = DpM FpM
p
ð7:196Þ
FpM ðr, 0Þ = I ð7:197Þ
A molecular network is responsible for resistance to chain alignment which

resists relaxation as stretch in-network increases. A similar observation for elasto-
mers was also found that plastic molecular chain stretch is inversely proportional to
the effective creep rate (Bergstrom and Boyce 1998). In laboratory experiments, it
was observed that in the post-yield region (at large deformations), the controlling
mechanism is the molecular network. The molecular network mechanism has a
dominant contribution to stress change in the post-yield region, while the amount
of elastic recovery upon unloading is associated with plastic strain in the molecular
network. Temperature dependence of molecular relaxation in network structure is
characterized by a classical Arrhenius function, and the plastic flow rule is described
with a simple power law as follows:
nM
QM τM
νpM = νoM exp - ð7:198Þ
kB θ SM
where νoM is the pre-exponential factor, QM is the activation energy for molecular
relaxation in network structure, kB is Boltzmann’s constant, nM is a strain-rate
sensitivity parameter, and SM is a stress measure describing the resistance of network
structure to relaxation which increases with increasing plastic stretch rate as defined
as follows:

S_ M = hM ðλpM - 1Þ SM - SM νpM ð7:199Þ
with initial condition

SoM = SM ðr, 0Þ ð7:200Þ
where hM is a parameter characterizing molecular relaxation in material, SM is

temperature- and rate-dependent saturation value of the network resistance, and λpM
is a plastic stretch that is related to plastic Almansi tensor in network structure BpM as
follows:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffi
trðBpM Þ
λM = ð7:201Þ
3
T
BpM = FpM FpM ð7:202Þ
Plastic deformation gradient evolution given by Eqs. (7.196) and (7.197) com-
pletes the constitutive modeling of material behavior in molecular network structure.
According to Eq. (7.199), network resistance will increase continuously as plastic
stretch ðλpM Þ in polymer chains increases and reaches a constant value SM which
depends on temperature and stretch rate. Plastic stretch-dependent evolution of
resistance to plastic flow also ensures the correct prediction of elastic recovery in
the unloading path. In Anand’s model, net driving stress for plastic flow in
intermolecular mechanism includes an additional resistance term accounting for
dissipative resistance to plastic flow ðSb or S2 Þ taking place at large deformations.
According to Anand’s model, this dissipative resistance evolves with plastic stretch
in the intermolecular mechanism or total stretch in the intermolecular mechanism. In
the model derived in this chapter, dissipative resistance (SM) at large deformations in
molecular network mechanism evolves with plastic stretch in molecular network
branch which controls material response at large deformations (post-yield region). In
Anand’s model, a third mechanism must be added for the true prediction of elastic
recovery during the unloading and cooling of a pre-heated sample. It was argued that
the third mechanism for molecular network structure was introduced due to a
necessity driven by the experimentally observed complex response. However, the
constitutive model presented in this chapter involves non-isothermal conditions
without introducing the additional third mechanism.
Finally, utilizing the relation given by Eq. (7.196) and specific Helmholtz free
energy definitions (Eqs. (7.157), (7.164), (7.182)), the temperature-dependent
governing constitutive equation can be given as
ρcθ_ = ∇x ðk∇x ðθÞÞ þ rþ

1
þJ - 1 τI þ γBj ln ðAÞj2 νpI þ τM νpM þ
2
! ð7:203Þ
-1
-1 1 eI _ e 1 ∂SeM _ e
∂S 1 ∂ Mback A _
þJ θ : CI þ : CM þ :A
2 ∂θ 2 ∂θ 2 ∂θ
7.4 Thermodynamics State Index 403
where the first two terms are heat conduction representing heat transfer within the
material during transient state and heat source due to passive heating (external
heating) or active heating (internally generated heat by internal friction or scattering,
etc.). The last two terms represent heat induced due to intrinsic dissipation and
thermoelastic effect representing conversion between mechanical and thermal
energy in the elastic range. In the case of coupled thermomechanical loading,
temperature increase due to mechanical work and temperature change due to heat
transfer between material and surroundings are all included in Eq. (7.203). Based on
descriptions of stress components and strain rate measures, irreversible entropy
production due to mechanical dissipation per unit volume in deformed configuration
2 3
∂ΨD ðA, θÞ
devð Þ - 2dev
J-1 6 A :DI þ 7p
MeI
γ mech = 6 ∂A 7>0 ð7:204Þ
θ 4 ∂Ψ D ð A, θ Þ 5
þγ A : ln ðAÞνpI þdevðMeM Þ:DpM
∂A
h i
J-1 1
γ mech = τI þ γBj ln ðAÞj2 νpI þ τM νpM > 0 ð7:205Þ
θ 2
Due to the associative plasticity rule, irreversible mechanical entropy production

(γ mech) is always positive:
DpI Meff
NpI = p = ð7:206Þ
DI j M eff j
DpM MeM
NpM = p = ð7:207Þ
DM MeM
7.4 Thermodynamics State Index
There are two major irreversible entropy production mechanisms, thermal dissipa-
tion (Eq. (7.137)) due to heat exchange between the system and surroundings and
mechanical work dissipation formulated by Eq. (7.205). Critical entropy (Scr) [also
referred to as fracture entropy in the literature] is a characteristic value of a material.
The link between irreversible material degradation and the amount of heat generated
due to non-conservative dissipation mechanisms (i.e., plastic work, internal friction,
scattering mechanisms, chemical reactions, etc.) or entropy production due to
mechanical work is well established. Heat conduction within metals is usually
very fast leading to negligible thermal gradients [at low-frequency loads, also
referred to as pseudo-static in some books] and leading to quick dissipation of heat
due to the high thermal diffusivity of metals. In the case of materials with low
thermal diffusivity such as polymers, heat transfer may take place over a prolonged
period with significant thermal gradients. However, thermal gradients cannot be

responsible for the failure of chemical bonds since material degradation (or damage)
is a consequence of the formation of small voids or cracks at the microscale by the
breakage of chemical bonds between molecules. Thermal dissipation may result in
deterioration of material properties which is usually insignificant compared to
degradation by mechanical dissipation.
Thermodynamics state index (TSI) is given by
h ΔS
i
Φ = Φcr 1 - e - ms R ð7:208Þ
When TSI reaches a critical value Φcr, corresponding to a critical entropy level
(Scr) that can be defined as failure or Φ = 1 can be defined as failure, depending on
the application. [Because TSI is an exponential function, it never reaches 1; com-
monly 0.999 is considered as the final value, depending on the computing
resources.] The critical TSI value depends on the critical entropy level (Scr), and it
is a characteristic property of material [like toughness] and can be calculated or
measured. Hence, we can calculate Φcr from the critical entropy value:
h ½Scr - So
i
Φcr = 1 - e - ms R ð7:209Þ
where So is initial internal entropy value, which may be taken as zero. Any irrevers-
ible process induces degradation in microstructure, and internal entropy increases
according to the second law of thermodynamics. Total entropy production due to
mechanical and thermal dissipation can be calculated at any time step as follows:
Zt
Smech = Smech jt = to þ γ mech dt ð7:210Þ
to
Zt h i
J-1 1 2 p p
Smech = Smech jt = to þ τI þ γBj ln ðAÞj νI þ τ M νM dt ð7:211Þ
θ 2
to
Zt
Sther = Sther jt = to þ γ ther dt ð7:212Þ
to
Zt
1 ρr
Sther = Sther jt = to þ - div J ∇ ð θ Þ þ dt ð7:213Þ
θ2
q x
θ
to
It is common practice to assume that there is no initial damage to the material. Of

course, this assumption is unrealistic, but it is for expediency:
7.5 Definition of Material Properties 405
Smech jt = to = Sther jt = to = 0 ð7:214Þ
At failure, internal entropy production reaches a critical value (Scr) which is of

course temperature-dependent. Unlike metals, the temperature dependence of critical
entropy is essential for the case of stretching of polymers due to changes in the
failure mode of amorphous polymer chains. Amorphous polymers display a brittle
failure at very low temperatures (θ < < θg) without any significant plastic defor-
mation, while at high temperatures (θ > θg), ductile failure occurs after a significant
amount of plastic work. Critical TSI parameter is defined in such a way that at
constant temperatures as S → Scr(θ), Φ → 1. Nonnegative entropy production
assures that Φ ≥ 0, while for an intact material we assume (S = 0), the damage is
assumed to be zero:
h i
m
ΦðθÞ = 1 - exp - s Scr ðθÞ ð7:215Þ
R
According to the incremental form of TSI evolution, degradation will increase at a

much faster rate at low temperatures (sudden brittle failure), while degradation rate
will be relatively slower at high temperatures (prolonged ductile failure). TSI
evolution merely depends on the thermodynamic fundamental equation which
incorporates all micro-mechanisms responsible for entropy generation due to
thermomechanical loading. Derivative of TSI with respect to entropy can be given
by

ΔΦ m m
= Φcr s exp - s Smech ð7:216Þ
ΔS R R
7.5 Definition of Material Properties
Material properties are vital for constitutive models. Material properties must be
defined over a large temperature and rate of loading range and should be continuous
and smooth over the transition region. Functions for material properties presented
herein are mathematical tools to describe the influence of temperature and loading
rate in a continuous form. Formulations for these material properties representing
temperature dependence are in a form that provides smooth continuity in the
temperature domain and has a continuous first derivative with respect to temperature.
Some material properties can be obtained by conducting isothermal tests at
different temperatures (both above and below the glass transition temperature)
and different loading rates such as E, ν, I m , νoI , QI , nI , Bg , X B , V, αp , γ . Some
material parameters are difficult to measure directly such as

hI , b, g, νoM , QM , hM , nM , μM , ϕ , SM , yet these properties/parameters can be
obtained by statistical methods.
According to the free-volume theory of William et al. (1955), the plastic flow rule
(yield function) can be constructed for equivalent plastic shear strain rate at temper-
atures above the glass transition temperature θg using Williams-Landel-Ferry (WLF)
equations. Similarly, rate dependence of glass transition temperature can be consid-
ered in terms of temperature-time equivalence of glass transition as follows:
8 g
> ν
< θref þ c2 log =νref ν > νref
θg =
g
cg1 - log ðν=νref Þ ð7:217Þ
>
:
θref
g ν≤ν ref
where cg1 and cg2 are WLF parameters associated with θg, νref is the reference stretch
rate, and ν is the equivalent stretch rate which is defined by
pffiffiffi
ν= 2 j D j ð7:218Þ
Temperature and rate dependence of elastic modulus (E) is given by

2 3
1 1 θ - θ g þ θE
E = 4 2 Eg þ Er - 2 Eg - E r tanh ΔE
5
þX E ½θ - ðθg þθE Þ
h i
ν
× 1 þ sE log ref ð7:219Þ
ν
( E
Xg θ ≤ θg þ θE
XE = ð7:220Þ
XrE
θ > θg þ θE
where Eg and Er are for glassy and rubbery elastic modulus corresponding to
temperatures confining the glass-rubber transition region. X gE and X rE represent the
rate of change of elastic modulus with respect to temperature in glassy and rubbery
domains, respectively. sE is the rate of loading sensitivity of elastic modulus, while
θE and ΔE define origin temperature and width of glass-rubber transition window,
respectively. Experimental studies on temperature and rate dependence of storage
modulus and elastic modulus of poly-methyl methacrylate (PMMA) indicate that
PMMA is highly sensitive to the rate of loading and temperature at constant
humidity. Modulus of PMMA continuously decreases with increasing temperature
with a remarkable drop around the glass transition temperature θg over a 10 C–20 C
temperature window, depending on the frequency of loading (Gunel 2010).
Poisson’s ratio ðνÞ is assumed to be only temperature-dependent and defined as
7.5 Definition of Material Properties 407

1 1 θ - θg þ θ E
ν = νg þ νr - νg - νr tanh ð7:221Þ
2 2 ΔE
where νg and νr are for glassy region and rubbery region Poisson’s ratios, respec-
tively. Shear modulus (G) and bulk modulus (K ) are defined as follows:
E
G= ð7:222Þ
2ð1 þ νÞ
E
K= ð7:223Þ
3ð1 - 2νÞ
Temperature and rate dependence of rubbery modulus (μM) is modeled like the
elastic modulus function as follows:
2 3
1 g 1 g θ - θg þ θμ h i
ν
μM = 4 2 μM þ μM - 2 μM - μM tanh
r r
Δμ 5 1 þ sμ log
þX μ ½θ - ðθg þθμ Þ
νref
ð7:224Þ
(
X gμ θ ≤ θg þ θμ
Xμ = ð7:225Þ
X rμ θ > θg þ θμ
Definitions of rubbery shear modulus parameters in Eq. (7.225) are identical to

those for elastic modulus. Temperature dependence of the critical value of order
parameter (ϕ), the limit of polymer chain extensibility
(IM), and saturation value of
plastic flow resistance of molecular network SM are defined below:

1 1 θ - θg þ θϕ
ϕ = ϕg þ ϕr - ϕg - ϕr tanh þ X ϕ θ - θg þ θϕ
2 2 Δϕ
ð7:226Þ
(
X gϕ
θ ≤ θg þ θϕ
Xϕ = ð7:227Þ
θ > θg þ θϕ
X rϕ

1 g 1 g θ - θg þ θM
I M = I M þ I M - I M - I M tanh
r r
þ X M θ - θg þ θM
2 2 ΔM
ð7:228Þ
(
X gM θ ≤ θg þ θμ
XM = ð7:229Þ
X rM θ > θg þ θμ

1 g 1 g θ - θg þ θS
SM= SM þ SM - SM - SM tanh
r r
2 2 ΔS

þ X S θ - θg þ θS ð7:230Þ
g
X S θ ≤ θg þ θS
Xs = ð7:231Þ
X rS θ > θg þ θS
Definitions of parameters above equations are identical

to those for elastic
modulus. Saturation value for network resistance SM and critical value of order
parameter (ϕ) were assumed to decrease with increasing temperature, while limited
chain extensibility (IM) increases with increasing temperature based on observations
in experiments. Similar to models of Richeton et al. (2005a, b) and Anand et al.
(2009), back-stress is assumed to vanish above the glass transition temperature, θg.
However, a decrease in back-stress modulus (B) with increasing temperature asymp-
totically approaches zero at a temperature around the glass transition temperature θg,
which can be defined as follows:
θ - θg
B = Bg 1 - tanh þ X B θg - θ ð7:232Þ
ΔB
(
X gB θ ≤ θg
XB = ð7:233Þ
0 θ > θg
Parameters b and g characterizing hardening-softening behavior in intermolecular

structure can be defined as follows:

1 1 θ - θg þ θg
g= g þ gr - g - gr tanh
2 g 2 g Δg

þ X g θ - θg þ θg ð7:234Þ
g
Xg θ ≤ θg þ θg
Xg = ð7:235Þ
0 θ > θg þ θg
!b3
νpI
b = b1 exp ðb2 θÞ p ð7:236Þ
νref
oOther parameters required for the constitutive model

νI , QI , V, αp , nI , hI , γ, νoM , QM , hM , nM are all basic constants.
7.6 Applications of Finite Deformation Models 409
7.6 Applications of Finite Deformation Models
For verification of the constitutive model, laboratory experiments of isothermal and

non-isothermal stretching of PMMA were performed. Simulation results were com-
pared with test data in terms of stress-strain curves for isothermal tests and
temperature-displacement-force histories for non-isothermal tests.
7.6.1 Material Properties
To simulate material response under isothermal and non-isothermal conditions, it is

essential to determine appropriate material parameters. Isothermal tests on PMMA
are required to obtain material parameters for the dual-mechanism viscoplastic
constitutive model presented in earlier sections. PMMA response is highly sensitive
to loading rate and temperature. Therefore, experiments must be conducted at
different loading rates and different temperatures at constant humidity. In addition,
the thermal properties of PMMA are also necessary for a fully coupled temperature-
displacement
finite element analysis. Most ofthe mechanical material properties of
PMMA E, ν, I m , νoI , QI , nI , Bg , X B , V, αp , γ can be obtained by conducting iso-
thermal tests at different temperatures and different strain (displacement) rates. Since
some material parameters such as (hI ,b,g,νoM ,QM ,hM ,nM ,μM ,ϕ ,SM ) cannot be
directly observed in macroscale experiments or their influence on material response
cannot be isolated from others, these parameters can be obtained by statistical
methods only.
Rate dependence of glass transition temperature of PMMA is based on the free-
volume theory of Williams et al. (1955), while WLF parameters cg1 , cg2 needed in
Eq. (7.224) are provided by Richeton et al. (2005a, b) and reference glass transition
temperature is provided by Nie (2005). Variation of glass transition temperature with
frequency (rate) can be obtained as shown in Fig. 7.5.
Fig. 7.5 Rate-dependent 384

glass transition temperature
of PMMA
382
380
θ.g(ν)
(°K)
378
376
374
10 20
10−5 ν (S-1) 25
1800
1600 H series
1400
Elasc Modulus (MPa)
M series
1200
1000 L series
800
600
400
200
0
0 20 40 60 80 100 120 140 160
Temperature (°C)
Fig. 7.6 Temperature- and rate-dependent elastic modulus of PMMA, Nie (2005)
5.0
4.8
Limited Chain Extensibility
4.6
4.4
4.2
M90
H series
4.0
3.8 M series
3.6
L series
3.4 H90
3.2
3.0
0 20 40 60 80 100 120 140 160
Temperature (°C)
Fig. 7.7 Temperature- and rate-dependent limited chain extensibility of PMMA
Figure 7.6 depicts the elastic modulus as a function of temperature for different
strain rates, H series (loading rate of 0.9 mm/s), M series (0.09 mm/s), and L
(0.009 mm/s).
Figure 7.6 indicates that the temperature sensitivity of elastic modulus becomes
smaller at slower loading rates. Parameters E g , E r , θE , ΔE , X gE , X rE , sE as a func-
tion of temperature and rate can be obtained from test data by statistical methods.
Temperature dependence of limited chain extensibility (IM) can be obtained from
fracture strain values in isothermal tests. Figure 7.7 shows the limited chain exten-
sibility as a function of temperature and strain rate.
Two data points (H90 and M90
values) are outliers. Using test data, parameters I gM , I rM , X gM , X rM , θM , ΔM can be
obtained as a function of temperature and loading rate.
Fig. 7.8 The ratio of yield stress to temperature vs. plastic strain rate plot of PMMA, Nie (2005)
The temperature variation of Poisson’s ratio in the glass-rubber transition region

can be assumedto be identical to that of elastic modulus (θE, ΔE), while glassy and
rubbery values νg , νr can be assumed to be constant. Poisson’s ratio in the glassy
regime is given by Nie (2005), and Poisson’s ratio in the rubbery regime is assumed
to be a value close to “0.5” to impose nearly incompressible conditions at high
temperatures. Material parameters characterizing viscoplastic features of deforma-
tion associated with an intermolecular structure νoI , QI , nI , Bg , X B , V, αp , γ can be
determined by using the ratio of yield stress to temperature (σ y/θ) versus plastic
strain rate (νp) plots (Eyring plots) obtained experimentally as presented in Fig. 7.8.
It is normally assumed that, when plastic flow occurs, the plastic rate is equal to
the total strain rate. Since PMMA displays brittle failure under tension at very low
temperatures without pronounced yielding, the data set for Eyring plots are limited.
A master curve can be constructed by shifting the data set horizontally and vertically
to a reference temperature (θref) as defined as
1 1
Δð log ðνpI ÞÞ = H h - ð7:237Þ
θ θref
σ
y 1 1
Δ = Hv - ð7:238Þ
θ θ θref
According to Richeton et al. (2005a, b), if θref is selected as θg, horizontal shift
(Hh) and vertical shift (Hv) factors can be related to material properties through the
following relation:
QI
Hh = ð7:239Þ
kB ln ð10Þ
γ
H v = Bg - X gB θg = ð7:240Þ
3
where QI is the activation energy for plastic flow in intermolecular structure, kB is

Boltzmann’s constant, Bg and X gB are parameters for bulk modulus, and γ is the
parameter characterizing dynamic recovery in hardening characteristics of the
intermolecular structure. The equation for the master curve can be derived from
the plastic flow rule in intermolecular structure (Eq. (7.174)) as follows:
!
σ y 2k B αp - 1
1
νpI nI
= 1- sinh - 1 ð7:241Þ
θg V 3 ν
where nI is the number of thermally activated transitions necessary for plastic flow,
V is the activation volume, and αp is the pressure sensitivity parameter. Temperature-
dependent characteristic viscoplastic flow rate (ν) is defined in Eq. (7.175), which is
derived from the flow rule of the theory of plasticity defined in Eq. (7.174). Effective
stress ðτI Þ at the glass transition temperature, θg, can be approximated from
Eq. (7.176) for a one-dimensional case. In this example, it is assumed that back-
stress (Mback) and plastic flow resistance in intermolecular structure (SI) vanish
around glass transition temperature, whereas applied stress at yielding is equal to
yield stress and normal pressure is one-third of applied stress:
σ1 = σy ð7:242Þ
Mback ffi 0 ð7:243Þ
SI ffi 0 ð7:244Þ
τI ffi σ 1 = σ y ð7:245Þ
1 1
pI = σ = σ ð7:246Þ
3 1 3 y
αp
τI = 1 - σ ð7:247Þ
3 y
Using regression analysis for fitting a master curve to experimental data with shift
factors Hh = 4900 K and Hv = - 40MPa material parameters, (νoI, QI, nI, Bg, XB, V,
αp and γ) can be calculated. Activation volume (V ) and activation energy (QI) were
assumed to be constant. Back-stress modulus asymptotically approaches zero around
the glass transition temperature. The remaining parameter is the back-stress modulus
definition ΔB which controls transition temperature range which was selected as 5 C
to ensure a smooth change in hardening characteristics of the material in
non-isothermal simulations (Fig. 7.9).
Implementing a 1-D version of the constitutive model in software, like
MATLAB, with isothermal conditions is an expedient way to determine the
remaining parameters (hI, b, g) in intermolecular
structure and the
parameters asso-
ciated with molecular network resistance νoM , QM , hM , nM , μM . These parameters
cannot be directly observed in macroscale experiments. Hence, it is necessary to run
simple 1-D simulations to estimate values for these parameters. Influence of param-
eters (hI, b, g) which control hardening-softening characteristics of response on
200
180
160
Back Stress Modulus (MPa)
140
120
100
80
60
40
20
0
0 20 40 60 80 100 120 140 160
Temperature (°C)
Fig. 7.9 Temperature-dependent back-stress modulus of PMMA

stress-strain curves and parameters νoM , QM , hM , nM , μM which control post-yield
responses is presented in Fig. 7.10a–h.
Arrows in Fig. 7.10a–h indicate an increasing trend of the corresponding param-
eter, while other parameters are held at a constant value. Parameters from
intermolecular network mechanism (b, g, hI) control strain hardening-softening
behavior in yield region, whereas parameters from molecular network mechanism
ðhM , μM , νpM , QM , nM Þ control post-yield behavior at large deformations. 1-D
MATLAB® simulations allow us to observe the influence of parameters on different
aspects of stress-strain curves and achieve an overall acceptable curve fitting to
stress-strain curves from isothermal test data. When parameters (b, g, hI) are deter-
mined, b is assumed to be both viscoplastic strain rate and temperature-dependent,
g is assumed to be temperature-dependent, and hI is assumed as a constant. Param-
eters for molecular network ðhM , νpM , QM , nM Þ are assumed to be constant except for
rubbery modulus which is taken as both temperature- and rate-dependent. The

critical

value for the order parameter (ϕ ) and saturation value for network resistance
SM are obtained from test data of Ames et al. (2009). Initial values for a parameter
(ϕ) and intermolecular resistance to plastic flow (SI) are usually assumed to be zero
(while molecularnetwork resistance to plastic flow (SM) is assumed to be at 10% of
saturation value SM Þ. Complete list of material parameters included in the consti-
tutive model is presented in Table 7.1:
ϕðr, 0Þ = 0,SI ðr, 0Þ = 0,SM ðr, 0Þ = 0:1SM ðθo Þ,θo = θðr, 0Þ

40 40
30 b 30
stress
stress
20
g
20
10 10
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
strain (b)
(a) strain
40 50
hM
40
30
30
hI
stress
20 stress
20
10
10
0 (d) 00
(c) 0 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1
strain strain
40 40
m no
M
30 M
30
stress
stress
20 20
10 10
0 0
(e) 0 0.2 0.4 0.6 0.8 1 (f) 0 0.2 0.4 0.6 0.8 1
strain strain
100 40
n
M
80
Q 30
M
60
stress
stress
20
40
10
20
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
(g) strain (h) strain

Fig. 7.10 Influence of material parameters b, g, hI , hM , μM , νoM , QM , nM on stress-strain curves
of PMMA. Gunel (2010)
7.7 Numerical Implementation of Dual-Mechanism Model 415
Table 7.1 Material parameters for PMMA constitutive model, Gunel (2010)
Parameters Parameters Parameters
θref
g ðK Þ
373 ϕg 0.001 gr 9.55
cg1 ðK Þ 32.58 ϕg 0 Δθ (K )
g
5
cg2 87.5 θϕ(K ) 0 θg (K ) 5

c1(K ) 9 Δϕ(K ) 10 X gg K - 1 0.07
c2 116 X gϕ ð1=K Þ -4.5 × 10-6 b1(MPa) 2.72 × 1010
-1 -2
ν (s )
ref
1.73 × 10 X rϕ ð1=K Þ 0 b2(1/K ) -4.58 × 10-2
Eg(MPa) 500 I gM 7.04 b3 4.04 × 10-2
Er (MPa) 1.5 I rM 4.6 νpref ðs - 1 Þ 0.001
θE(K ) -8 θM(K ) -20 νoI ðs - 1 Þ 2.43 × 1012
ΔE(K ) 15 ΔM(K ) 12 QI(J/K) 1.56 × 10-19
sE 0.06 X gM ð1=K Þ 0 V(m3) 1.39x10-27
X gE ðMPa=KÞ -15.05 X rM ð1=K Þ 0.001 αp 0.21
X rE ðMPa=KÞ 0 SgM ðMPaÞ 35 nI 2.17
νg 0.31 SrM ðMPaÞ 0.2 hI 40.42
νr 0.49 θS(K ) 5 γ(MPa) 60
μgM ðMPaÞ 9 ΔS(K ) 5 νoM ðs - 1 Þ 4.33 × 109
μrM ðMPaÞ 0.5 X gS ðMPa=KÞ -0.01 QM(J/K) 1.30 × 10-19
θμ(K ) 5 X rS ðMPa=KÞ 0 hM 14.43
Δμ(K ) 14 Bg (MPa) 10 nM 5
sμ 0.03 X gB ðMPa=KÞ 2.15 ms(g/mol) 100.13
X gμ ðMPa=KÞ -0.4 ΔθB (K ) 5
X rμ ðMPa=KÞ 0 gg 7.97
7.7 Numerical Implementation of Dual-Mechanism Model
The dual-mechanism constitutive model is implemented numerically based on the

staggered method with an isothermal split. In each loading step, the temperature
value at the end of the time increment is taken as constant, and after mechanical
equilibrium is satisfied, thermal equations are solved under a fixed configuration to
update the temperature increment in the following time increment. Since this scheme
is only conditionally stable, different time increments are used for isothermal cases at
different temperatures, and a suitable time increment is chosen based on conver-
gence study results. Objectivity (frame indifference) was ensured by using stress
components and their conjugate rate components in reference configuration
(in material coordinates). Final forms of the constitutive equations are obtained
after a transformation of evolution equations to reference configuration. The evolu-
tion of elastic and plastic deformation gradient tensors is approximated by the
exponential operator (Weber and Anand 1990). When an exponential operator is
used in combination with the backward Euler scheme, plastic incompressibility, and
symmetry of state variable tensors are conserved with sufficient accuracy by only
including the first two terms of series representation of the exponential function.
Rate dependency of some material properties (E, μM, θg) is implemented in a numer-
ical algorithm such that property values are calculated based on strain rates in the
previous time increment and strain rates are updated based on initial and final
deformation gradients at the end of the time increment. Therefore, equilibrium is
satisfied at discrete time increments assuring thermodynamic consistency.
Numerical integration of evolution equations of plastic deformation gradients
FpI ,FpM and stretch-like internal variable A based on exponential mapping is obtained
from Eqs. (7.172), (7.196), and (7.166) as presented below:

ðFpI Þnþ1 = exp Δt ðDpI Þnþ1 ðFpI Þn ð7:248Þ

ðFpM Þnþ1 = exp Δt ðDpM Þnþ1 ðFpM Þn ð7:249Þ

Anþ1 = An exp Δt ðDpI Þnþ1 þ exp Δt ðDpI Þnþ1 An - γAn ln ðAn Þ

exp Δt ðDpI Þnþ1
× pffiffiffi ð7:250Þ
2
where Δt = tn+1 - tn is the time increment, γ is a constant representing dynamic

recovery associated with kinematic hardening, νpI and νpM are equivalent plastic
stretch rates in intermolecular and molecular network structure, and DpI and DpM are
plastic stretch rate tensors in intermolecular structure and molecular network struc-
ture, respectively. Owing to unconditional stability of implicit time integration
scheme, the classical backward Euler method is preferred for integration of evalu-
ation equations of internal state variables, ς_ = f ðς, θÞ.
For a generalized internal state variable (ς), numerical integration of evolution
equation can be performed as
ςnþ1 = Δtf ðςnþ1 , θnþ1 Þ þ ςn ð7:251Þ
Further in-depth details of the implementation of the dual-mechanism

viscoplastic model are given by Gunel (2010).
7.7.1 Simulating Isothermal Stretching of PMMA
A fully coupled temperature-displacement analysis is performed to investigate the

adiabatic effect in stretching of PMMA. Eight-node linear brick elements (C3D8T)
are used as element types in Abaqus finite element code. Convergence studies using
different mesh sizes and time steps were conducted. In Fig. 7.11, the influence of
time increments on the convergence of different aspects of material response is
presented. In simulations, a rectangular prism model was uniaxially stretched at a
displacement rate of 1 mm/s for 50 s, while the temperature was kept constant at
90 C. In convergence studies for time increment, maximum axial stress (σ),
12
10
True Stress (MPa)
Δt=0.0001
6
Δt=0.001
4
Δt=0.01
2 Δt=0.1
(a) 0 0.2 0.4 0.6 0.8 1

True Strain
0.0445
0.0440
0.0435
Strain Rate (s-1 )
0.0430 νpI νpM d

0.0425
0.0420
0.0415
0.0410
0.0405
1.E-05 1.E-04 1.E-03 1.E-02 1.E-01 1.E+00
(b) me increment (Δt, sec)
Fig. 7.11 Convergence study results for time increment in the viscoplastic model in terms of (a)
true stress-strain curves and (b) strain rates
equivalent viscoplastic strain rate ðνpI , νpM Þ, and equivalent stretch rate (d) must be
monitored.
True stress-strain curves presented in Fig. 7.11a indicate a fast convergence even
for the largest time increment of Δt = 0.1 s. Convergence of equivalent plastic strain
rates ðνpI , νpM Þ requires a small time increment, which also indicates a slow conver-
gence of internal state variables (SI, SM, ϕ. . .). Another interesting point that can be
observed from Fig. 7.11b is that equivalent viscoplastic strain rate in an
intermolecular network ðνpI Þ is larger than the equivalent stretch rate (d ) which
might seem like an error. However, this is due to the numerical method that is used to
12 0.050
0.045
10
0.040
0.035
True Stress (MPa)
Strain Rate (s-1)

8
0.030
6 0.025
0.020
4
σ11 0.015
νpI
0.010
2 νpM
d 0.005
0 0.000
0 10 20 30 40 50 60
me (sec)
Fig. 7.12 Comparison of true stress and strain rate histories
capture strain-softening characteristics and is controlled by the parameters b and g in

the constitutive model.
In Fig. 7.12, true stress, equivalent viscoplastic strain rates ðνpI , νpM Þ, and equiv-
alent stretch rate (d ) histories are presented. At the beginning of a simulation, the
equivalent stretch rate (d ) is larger than the equivalent plastic strain rate in both
p p
e e ðνI , νM Þ, and continuous increase in elastic strain in both mechanisms
mechanisms
F I , F M causes an increase in stress level up to the onset of yielding at which
equivalent viscoplastic strain rate in intermolecular mechanism ðνpI Þ becomes larger
than the equivalent stretch rate (d ) and remains larger over some period which
corresponds to the yielding and strain-softening region. During this period, the
negative elastic strain rate results in a decrease in elastic strain in the intermolecular
mechanism FeI and hence decrease in stress level TI. After strain-softening, equiv-
alent viscoplastic strain rate in intermolecular mechanism ðνpI Þ remains equal to the
equivalent stretch rate (d). In the meantime, the difference between equivalent
viscoplastic strain rates in the molecular network ðνpM Þ and equivalent stretch rate
(d) starts to increase which leads to an increase in elastic strain in the molecular
network FeM . As a result, stress TM level in the molecular network also increases
which describes strain-hardening characteristics of material due to chain locking in
the molecular network. However, further stretch in the molecular network continu-
ously increases stress TM level in the molecular network causing the increase in
driving stress for plastic flow (τM) which in return starts to increase equivalent
viscoplastic strain rate in the molecular network ðνpM Þ once again. Therefore, the
change in stress level at large deformations gradually becomes smaller.
Mesh sensitivity of results is also very important in finite element analysis. In
Fig. 7.13, the influence of the size of mesh seeds (in Abaqus) on the convergence of
different aspects of material response is presented. A rectangular prism sample was

uniaxially stretched at a displacement rate of 1 mm/s for 50 s, while the temperature
was kept constant at 100 C. In mesh sensitivity studies, axial stress (σ) history,
equivalent viscoplastic strain rate ðνpI , νpM Þ, and equivalent stretch rate (d ) are all
monitored.
In Fig. 7.13a, results for different mesh sizes are presented. The seed size
encircled in Fig. 7.13a is used in the final simulations for PMMA since sufficient
convergence is achieved at this mesh size for true stress value (σ) and equivalent
plastic strain rate values ðνpI , νpM Þ as shown in Fig. 7.13b, c, respectively. A coarser
mesh than the selected one (seed size = 3) would also be reasonable for convergence
of (true) stress values except for the coarsest one (seed size = 10). However, a finer
mesh is always desirable for convergence of equivalent strain rate values and
associated state variables (Fig. 7.13c).
Results shown in Fig. 7.14 indicate that simulation results and experimental
measurements are reasonably in good agreement. The constitutive model is espe-
cially successful in describing temperature and rate dependence of material response
as well as strain-softening and strain-hardening characteristics. Due to the limited
size of the data set used for the determination of viscoplastic flow rule for
intermolecular mechanism, it is really difficult to determine material parameters
involved in the viscoplastic model which resulted in some minor differences in
response.
Dual-mechanism viscoplastic model for PMMA is verified and validated for more
challenging non-isothermal conditions as presented in the following section.
Entropy production values are calculated as an average of cumulative entropy
production of all Gauss integration points at the cross-section of the center of the
narrow section of the test sample. The difference between mean entropy production
and maximum entropy production in the cross-section is less than 5% for all
isothermal test simulations which indicates that no damage localization is predicted
at the cross-section which is similar to the case of brittle failure of materials. As
brittle failure takes place in the form of sudden rupture of the specimen with
concurrent crack formation and propagation at many locations, damage localization
is not observed contrary to ductile failure of materials for which damage localization
is an essential part of the crack initiation stage around microscopic defect regions,
while crack propagation takes place over a longer time compared to brittle failure.
Therefore, damage quantification based on average entropy production in the cross-
section is a proper method for brittle failure, while local thermodynamic state index
is essential for damage quantification in ductile failure. It is also important to note
that damage evolution and material degradation due to mechanical load are assumed
to result from only plastic energy dissipation excluding all other entropy generation
mechanisms for the sake of simplicity. Therefore, in order to fully account for the
evolution of microscopic defects that lead to stiffness and strength degradation in
material, it is necessary to include all entropy generation mechanisms associated
with energy loss associated with new surface creation at microscopic defect regions,
heat generation, chemical reactions, aging, etc. The latter approach would be more
accurate. It is possible to account for the difference between failure in tension and
1.E+05
number of elements (in log scale)
1.E+04
1.E+03
1.E+02
1.E+01
0 2 4 6 8 10 12
(a) seed size
7
True Stress (MPa)
5 seed size=10
4
seed size=5
seed size=2.5
3
seed size=1.25
2 seed size=0.625
1
seed size=3
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
(b) True Strain
0.070
0.065
0.060 νp1 νp2 d
0.055
Strain Rate (s-1)
0.050
0.045
0.040
0.035
0.030
0 2 4 6 8 10 12
(c) seed size
Fig. 7.13 Mesh sensitivity of the ABAQUS model in terms of (b) true stress-strain curves and (c)
strain rates (a) for different mesh densities
(a) 40
(b)
50 ISO-M30 (exp.)
ISO-H30 35
45 (exp.) ISO-M60 (exp.)
40 ISO-H60 30
(exp.) ISO-M90 (exp.)
35
Stress (MPa)
Stress (MPa)
ISO-H90 25
30 (exp.) ISO-M30 (sim.)
ISO-H30 20 ISO-M60 (sim.)
25 (sim.)
20 15 ISO-M90 (sim.)
15
10
10
5 5
0
0 0.1 0.2 0.3 0.4 0.5 0
0 0.2 0.4 0.6 0.8 1 1.2
Strain Strain
35
(c)
30 ISO-L30 (exp.)
ISO-L60 (exp.)
25
Stress (MPa)
ISO-L90 (exp.)
20
ISO-L30 (sim.)
15
ISO-L60 (sim.)
10
ISO-L90 (sim.)
5
0
0 0.5 1 1.5 2
Strain
Fig. 7.14 Comparison of Abaqus fea simulation and testing results for isothermal testing of
PMMA in (a) H series, (b) M series, and (c) L series tests
failure in compression or predict damage evolution in the case of ductile failure by

deriving a more comprehensive thermodynamic fundamental equation.
Total irreversible thermal entropy production in ISO-TEST simulations (Sirr, ther,
test is completely created by the transfer of heat generated due to plastic dissipation
)
during isothermal stretching. Temperature rise due to plastic dissipation in isother-
mal stretching is very small; hence, thermal gradients within specimen during testing
are also small, as a result producing a negligible amount of irreversible entropy
production in comparison with Sirr, ther, pre-test and total irreversible mechanical
entropy production during isothermal tests (Sirr, mech, test).
Mechanical damage in materials can be described as the formation of micro-
cracks and voids at the micro-level with a corresponding degradation of stiffness at
the macro-level. On the other hand, thermal damage in materials is the deterioration
of material properties which is the weakening of chemical bonds between molecules
or change in length of the molecular bond. The evolution of thermal damage is much
slower than the evolution of mechanical damage. Remarkable thermal gradients over
long periods are necessary to induce permanent cracks and voids in materials, e.g.,
freezing-thawing cycles, thermal shock, etc. Therefore, mechanical and thermal
damage induced in the material has different entropy generation sources, and both
must be defined separately in the case of thermomechanical loading. As failure in
isothermal stretching of PMMA was completely induced by mechanical entropy
generation mechanisms, TSI must be based on the evolution of irreversible mechan-
ical entropy production. In metal fatigue tests, it is found that cumulative-specific
entropy generation is constant at the time of failure, which is independent of the

structural geometry of a sample, load intensity, and loading rate for small frequen-
cies. The critical-specific entropy value is usually referred to as the fatigue fracture
entropy (FFE). It is observed by many researchers that the rate dependence of critical
entropy density is negligible (at small frequencies), while critical entropy density
increases with increasing temperature because of increasing plastic dissipation
before failure. Mechanical damage is completely based on irreversible mechanical
entropy production; however, due to intrinsic heat dissipation, the temperature
variation of plastic dissipation is also observed in critical entropy production values.
Critical entropy is also referred to as the fatigue fracture entropy (FFE) in some
publications. The logarithm of critical entropy value Sirr, cr for PMMA as a function
of temperature is given by Gunel (2010):
log ðSirr,cr Þ = 0:0405θ - 2:5653 ð7:252Þ
The selection of critical entropy is based on an assumed failure definition. If an

engineer defines a 50% reduction in mechanical stiffness or a 10% increase in
electrical resistance as a failure, that will be a different critical entropy value than
if the complete fracture is defined as the failure. In Eq. (7.252) complete fracture of
the sample is the failure definition. It is much easier to define a thermodynamic state
index value as a failure point than using the critical entropy, which must be
measured, as a failure.
7.7.2 Simulating Non-isothermal Stretching of PMMA
Dual-mechanism constitutive model (viscoplastic) is also applied to non-isothermal

stretching of PMMA to large deformations with specific emphasis on the transition
of material response around the glass transition temperature. In non-isothermal
mechanical tests on PMMA, the temperature of samples spanned over θg ± 50o C
range during which a complete glass transition is bound to take place and any change
in material response due to this large temperature difference could be monitored. It is
well-known that for some polymers viscoplastic behavior right around the glass
transition temperature exhibits significant change. This change is manifested as a
significant increase in stiffness during the cooling of a specimen. Contrary to other
polymer constitutive models, here it is shown that this PMMA model does not
experience any abrupt changes in response during stretching under non-isothermal
conditions, and the response is continuous and smooth.
It is important to point out that material constitutive models for finite-strain
behavior of amorphous polymers can only be verified in non-isothermal loading
conditions.
Concurrent heat transfer with axial stretching of PMMA requires a fully coupled
temperature-displacement analysis.
Fig. 7.15 Axial force histories in non-isothermal experiments and simulations of PMMA for H
series (0.9 mm/s forming rate)
Axial force histories of H, M, and L series with respect to normalized time (tn) are
presented in Figs. 7.15, 7.16, and 7.17. Simulation and experiment results in terms of
axial force histories are reasonably in good agreement. A unified mechanics formu-
lation of plastic flow rule for intermolecular structure and appropriate material
property definitions ensures smooth transition of response around the glass transition
temperature.
Here we conclude our discussion of the large deformation analysis of polymers.
Fig. 7.16 Axial force histories in non-isothermal experiments and simulations of PMMA for M
References 425
Fig. 7.17 Axial force histories in non-isothermal experiments and simulations of PMMA for L
References
Ames, N. M., et al. (2009). A thermo-mechanically coupled theory for large deformations of
amorphous polymers. Part II: Applications. International Journal of Plasticity, 25(8),
1495–1539.
Anand, L. (1986). Moderate deformations in extension-torsion of incompressible isotropic elastic
materials. Journal of the Mechanics and Physics of Solids, 34, 293–304.
Anand, L., & On, H. (1979). Hencky’s approximate strain-energy function for moderate deforma-
tions. Journal of Applied Mechanics, 46, 78–82.
Anand, L., et al. (2009). A thermo-mechanically coupled theory for large deformations of amor-
phous polymers. Part I: Formulation. International Journal of Plasticity, 25(8), 1474–1494.
Arruda, E. M., & Boyce, M. C. (1993). Evolution of plastic anisotropy in amorphous polymers
during finite straining. International Journal of Plasticity, 9(6), 697–720.
Arruda, E. M., Boyce, M. C., & Jayachandran, R. (1995). Effects of strain rate, temperature and
thermomechanical coupling on the finite strain deformation of glassy polymers. Mechanics of
Materials, 19(2–3), 193–212.
Bergström, J. S., & Boyce, M. C. (1998). Constitutive modeling of the large strain time-dependent
behavior of elastomers. Journal of the Mechanics and Physics of Solids, 46(5), 931–954.
Boyce, M. C., Socrate, S., & Llana, P. G. (2000). Constitutive model for the finite deformation
stress-strain behavior of poly(ethylene terephthalate) above the glass transition. Polymer, 41(6),
2183–2201.
Fotheringham, D. G., & Cherry, B. W. (1978). The role of recovery forces in the deformation of
linear polyethylene. Journal of Materials Science, 13(5), 951–964.
Fotheringham, D., Cherry, B. W., & Bauwens-Crowet, C. (1976). Comment on “the compression
yield behavior of polymethyl methacrylate over a wide range of temperatures and strain rates”.
Journal of Materials Science, 11(7), 1368–1371.
Francisco, P., Gustavo, S., & Élida, B. H. (1996). Temperature and strain rate dependence of the
tensile yield stress of PVC. Journal of Applied Polymer Science, 61(1), 109–117.
Gent, A. N. (1996). A new constitutive relation for rubber. Rubber Chemistry and Technology,
69(1), 59–61.
Gunel, E. M., & Basaran, C. (2009). Micro-deformation mechanisms in thermoformed alumina
trihydrate reinforced poly(methyl methacrylate). Materials Science and Engineering: A,
523(1–2), 160–172.
Gunel, E.M., & Basaran, C. (2010). Stress whitening quantification of thermoformed mineral filled
acrylics. Journal of Engineering Materials and Technology, 132(3), 031002–031011.
Gunel, E. M., & Basaran, C. (2011a). Damage characterization in non-isothermal stretching of
acrylics: Part II experimental validation. Mechanics of Materials, 43(12), 992–1012.
Gunel, E. M., & Basaran, C. (2011b). Damage characterization in non-isothermal stretching of
acrylics: Part I theory. Mechanics of Materials, 43(12), 979–991.
Hashin, Z., & Shtrikman, S. (1963). A variational approach to the theory of the elastic behavior of
multiphase materials. Journal of the Mechanics and Physics of Solids, 11(2), 127–140.
Mustafa Eray Gunel (2010), Large Deformation Micromechanics of Particle Filled Acrylics at
Elevated Temperatures, Dissertation submitted to Department of Civil, Structural and Environ-
mental Engineering, University at Buffalo, SUNY
Mandel, J. (1972). Plasticite classique et viscoplasticite (Lecture Notes). International Center for
Mechanical Sciences.
Nie, S. (2005). A micromechanical study of the damage mechanics of acrylic particulate composites
under thermomechanical loading. (Ph.D. Dissertation). In Civil, structural, and environmental
engineering. State University of New York at Buffalo.
Palm, G., Dupaix, R. B., & Castro, J. (2006). Large strain mechanical behavior of poly(methyl
methacrylate) (PMMA) near the glass transition temperature. Journal of Engineering Materials
and Technology, 128(4), 559–563.
Richeton, J., et al. (2005a). A formulation of the cooperative model for the yield stress of
amorphous polymers for a wide range of strain rates and temperatures. Polymer, 46(16),
6035–6043.
Richeton, J., et al. (2005b). A unified model for stiffness modulus of amorphous polymers across
transition temperatures and strain rates. Polymer, 46(19), 8194–8201.
Richeton, J., et al. (2006). Influence of temperature and strain rate on the mechanical behavior of
three amorphous polymers: Characterization and modeling of the compressive yield stress.
International Journal of Solids and Structures, 43(7–8), 2318–2335.
Richeton, J., et al. (2007). Modeling and validation of the large deformation inelastic response of
amorphous polymers over a wide range of temperatures and strain rates. International Journal
of Solids and Structures, 44(24), 7938–7954.
Srivastava, V., & Anand, L. (2010). A thermo-mechanically-coupled large-deformation theory for
amorphous polymers in a temperature range which spans their glass transition. International
Journal of Plasticity, 26(8), 1138–1182.
Weber, G. & Anand, L. (1990). Finite Deformation Constitutive Equation and a Time Integration
Procedure for Isotropic, Hyperelastic-Viscoplastic Solids. Computer Methods in Applied
Mechanics and Engineering, 79, 173–202.
Williams, M. L., Landel, R. F., & Ferry, J. D. (1955). The temperature dependence of relaxation
mechanisms in amorphous polymers and other glass-forming liquids. Journal of the American
Chemical Society, 77(14), 3701–3707.
Chapter 8
Unified Mechanics of Metals Under High
Electrical Current Density:
Electromigration and Thermomigration
8.1 Introduction
Electromigration and thermomigration are irreversible mass diffusion mechanisms

under high current density and high-temperature gradient, respectively. However,
thermomigration can take place alone in the absence of electrical current, while
electromigration cannot happen without thermomigration due to Joule heating,
except for special circumstances where boundary conditions prevent a thermal
gradient from happening.
8.2 Physics of Electromigration Process
When a metal is subjected to an electrical potential gradient, the current enters from
the anode side and travels to the cathode side, and the electrons travel from the
cathode to the anode side. Electromigration is a mass diffusion-controlled phenom-
enon. When a conductor is subject to a high current density, the so-called electron
wind transfer part of its momentum to the atoms (or ions) to make the atoms (or ions)
move in the direction of the current. As a result, the degradation of the metal
conductor occurs mainly in two forms: in the anode side, the atoms will accumulate
and finally form hillocks and the vacancy concentration on the cathode side will
form voids. Both hillocks and voids will cause the degradation of the material and
eventual failure. The damage evolution due to electromigration can be modeled as an
irreversible mass transport process. The purpose of this chapter is to present the
formulation of modeling the electromigration and thermomigration-induced material
degradation process using the unified mechanics theory
https://doi.org/10.1007/978-3-031-18621-9_8
428 8 Unified Mechanics of Metals Under High Electrical Current Density. . .
The physics of electromigration has been extensively investigated. It is believed

that in 1861 French physicist M. Gerardin was the first to observe electromigration in
liquid metal alloys. Skaupy (1914) introduced the concept of “electron wind,” which
facilitated the modeling of electromigration using the diffusion theory. Black (1969)
established an empirical relationship between the mean time to failure and current
density for confined thin films on a substrate. The experiments by Blech (1976)
revealed that the stress gradient could act as a counterforce to electromigration under
high current density. In addition, for thin films, Blech (1976) proposed a length scale
called “Blech’s critical length” below which mass diffusion due to electron wind
force is counterbalanced by stress gradient driving force in the opposite direction.
To solve an arbitrary boundary value problem with irregular boundaries or an
initial value problem with arbitrary conditions and composite material properties
involving electromigration, there must be a rational mechanics-based current
density-strain constitutive relation and a thermodynamic fundamental equation of
the material. Then this formulation can be implemented in a finite element method
(or any other computational mechanics procedure) to solve any boundary value
problem.
The classical definition of electromigration refers to the structural damage caused
by ion transport in metals because of high current density. Electromigration is
usually insignificant or nonexistent at low current density levels. Quantification of
what “high current density” is studied extensively by Ye et al. (2003a, b, c, d, e, f,
2004a, b, c, d, 2006), and Basaran et al. (2003). Electromigration is mass transport in
a diffusion-controlled process under certain driving forces. The driving force here is
more complicated than what is involved in a pure diffusion process, in which the
concentration gradient of the moving species is the only driving force component.
The electrical driving forces for electromigration consist of the electron wind force
and the direct field force. The electron wind force refers to the effect of momentum
exchange between the moving electrons and the ionic atoms when an electrical
charge—a direct field force—is applied to a conductor with metallic bonds. When
current density, which is proportional to the electron flux density, is high enough,
this momentum exchange effect becomes significant, resulting in a noticeable mass
transport referred to as electromigration. Low melting point alloys when used at
elevated homologous temperature (>0.4Tmelt Kelvin) are prone to have considerable
atomic diffusivity. For example, solder joints in microelectronics packaging are
prime examples of this category
8.2.1 Driving Forces of Electromigration Process
The Electric Field

The electrical current generates two driving mechanisms; one is attributed to the
electron wind forces, as originally suggested by Skaupy (1914), which refers to the
effect of the momentum exchange between the moving electrons and the ionic
atoms. This momentum exchange happens because of the scattering of free valence
8.2 Physics of Electromigration Process 429
electrons. Scattered electrons collide with metal atoms and push them in the direc-
tion of electron flow (Seith 1955). On the other hand, atoms move in the opposite
direction of the applied electric field when they are ionized, the latter mechanism, the
static force of the electric field, is considered the direct field force. The net effect of
these two forces is the so-called electrical field driving force of electromigration.
Stress Gradient
When atoms diffuse from the cathode side to the anode side, they leave behind
vacancies on the cathode side and mass accumulation on the anode side. As a result,
there is tensile stress on the cathode side and compressive stress on the anode side.
This stress differential is responsible for the stress gradient.
Of course, the mechanical stress gradient influences the electromigration process.
There is an interaction between the stress gradient and the electromigration driving
forces. Stress gradient can counteract or enhance the electromigration process,
depending upon the interaction among all diffusion driving forces. Essentially, stress
gradient is another driving force of mass transport.
However, in thermomigration in the absence of external stress gradients, the mass
moves from the hot side to the cold side, and compression on the cold side and
tension on the hot side develop, which counters the thermal gradient-induced force.
Thermal Gradient
The Joule heat generated under high current density is highly localized. Hence, there
is a thermal gradient in the medium, which leads to thermomigration. The thermal
gradient is one of the strongest driving forces of diffusion. Thermomigration cannot
be ignored especially when the thermal gradient is large, in which case the
thermomigration can be the dominant diffusion driving force.
The physical explanation behind thermomigration is not well understood. How-
ever, recent literature shows that thermomigration could play a significant role in the
electromigration-induced failure.
Atomic Vacancy Concentration Gradient
It has been estimated that the mass flux due to vacancy concentration gradient is
small compared to those induced by electrical field forces, stress gradient, and
thermal gradient. The atomic vacancy concentration gradient is usually small at
the beginning compared to the other diffusion driving forces. However, as the mass
migration progresses, vacancy concentration gradient increases significantly. At a
certain point in time, the vacancy concentration gradient becomes large enough to
slow the diffusion.
8.2.2 Laws Governing Electromigration

and Thermomigration
Vacancy diffusion is governed by the vacancy conservation equation; mechanical

deformation is governed by Newtonian mechanics; heat transfer is governed by
Fourier’s law; an electric field is governed by Maxwell’s equation of conservation of

charge; and degradation of the material is governed by the laws of thermodynamics.
Figure 8.1 shows a microelectronics solder joint before and after electromigration/
thermomigration failure.
In the following section, all these laws are used to model the unified mechanics of
electromigration and thermomigration processes.
When the only acting load is an electrical potential field, the primary deformation
mechanism is due to atomic diffusion, which is a mass transport by atomic motion
process. Diffusion in solids occurs only at the atomic or molecular level. However,
diffusion in solids may be observed on a macroscopic scale given enough time.
Diffusion that results in the net transport of matter over macroscopic distances is
considered a nonequilibrium process. It does not stop until the phase eventually
achieves fully thermodynamic equilibrium or diffusion driving force is removed or
counterbalanced.
The laws of diffusion were first developed by Adolf Fick when studying the
diffusion of the saltwater system (Fick 1855). Fick’s diffusion laws establish the
mathematical relationship between the rate of diffusion and the concentration gra-
dient, which is the driving force of mass transport.
Fick’s First Law
Fick discovered by direct observation that the magnitude of the mass flux is
proportional to the magnitude of the concentration gradient in the isotropic contin-
uum given by
AI Silicon Die
Ni UBM
100 μm
140 μm
Underfill
Solder
Cu
PCB
Ni UBM
Underfill
Solder
Cu Plate
40µm700X
Fig. 8.1 Electronics solder joint profile after electromigration and thermomigration failure
J = - D∇C ð8:1Þ
where:
J is the mass flux, which is defined as a vector quantity, with units of (molm-2s-1).
Q
D is diffusivity, with units of m2/s; it is a function of temperature D = D0 exp - RT d
where D0 is temperature-independent material constant, Qd is the activation

energy of diffusion, R is the gas constant (8.31 J/mole K), and T is the temperature
in Kelvin.
C is the concentration of the matter in question. Combining the mass conservation
equation with Fick’s first law yields Fick’s second law.
Fick’s Second Law: Nonsteady State Diffusion

Fick’s second law can be given in its most elementary form as
∂C
= D∇2 C ð8:2Þ
∂t
Until the twentieth century, diffusion in materials was addressed at a microscopic

scale only. Einstein showed the consistency of the random motion of microscopic
particles in the presence of molecules. Einstein derived a relationship for the chaotic
motion of small particles as
DF
vD = ð8:3Þ
kB T
where vD is the drift velocity, unit m/s; D is the diffusion coefficient; F is the driving
force of the diffusion, unit Newton; kB is Boltzmann’s constant; and T is absolute
temperature, Kelvin.
From Eq. (8.3) we can easily deduce that for self-diffusion, the driving force due
to concentration gradient has the following form:
F = - k T∇C ð8:4Þ
In addition, Fick’s second law has a more general form as follows:
∂C D∇F
= ð8:5Þ
∂t kT
where F is the driving force of the diffusion. In the presence of electrical current
density, the driving force has four primary components:
!
X
F= Fi = F em þ F tm þ F σ þ F s ð8:6Þ
i
where Fem is the electron wind force, Ftm is the driving force due to temperature
gradient, Fσ is the stress gradient driving force, and Fs is the driving force due to
chemical potential gradient. Each driving force will be introduced separately in the
following sections.
These forces also lead to grain coarsening and phase changes in the metals.
However, we are primarily interested in the mechanical implications in here.
8.2.3 Electromigration Electron Wind Force
The first-ever observation of electromigration occurred long before it became an

engineering problem. In 1861, a French scientist, M. Gerardin, first noticed the
phenomenon of electromigration. Nevertheless, at that time, no real scientific expla-
nation was provided for this phenomenon. Although Michael Faraday’s discovery of
electrochemistry a few decades earlier may have suggested an explanation, it would
have been incorrect because electrons had not been discovered yet. Without elec-
trons, a proper scientific explanation would not have been possible.
The first reasonable scientific explanation of electromigration driving force came
from Skaupy (1914), who recognized that in metals the moving free conduction band
electrons could drag atoms along through a frictional force (scattering), which he
called “electron wind.” After that, a series of publications by Huntington and Grone
(1961) Bosvieux and Friedel (1962), and Genoni and Huntington (1977) proposed
that at high current densities, conduction band electrons collide (scatter) with metal
atoms/ions and push them in the direction of the electron flow. In other words, the
“electron wind” force drives positive ions (or atoms) in the direction of the electron
flow, which is opposite to the direction of the electrostatic force of the electric field
(Fig. 8.2).
In electromigration studies, it has been customary to distinguish two types of
contributions: one from the scattering of the electrons with the ions in question (the
electron wind term) and that from the direct interaction of an atomic charge with the
applied electric potential field (the direct force) (Landauer and Woo 1974). There-
→
fore, the electromigration driving force, F em, exerted on an ion can be expressed as a
summation of these two forces acting on ions:
→ → →
F em = F wind þ F direct
→
= ð - Z wind þ Z el Þe E
→
ð8:7Þ
= Ze E
= Z eρj
Fig. 8.2 Kinematics of

electromigration
→
where e is the electron charge, E is the electric field potential (V/cm), ρ is the
resistivity of the metal (ohmmeter (Ωm)), j is the current density (Amp/cm2), and Z*
is a dimensionless quantity called the effective charge (also called an effective
nuclear charge or the affective valence), which is the combined net attractive positive
charge of nuclear protons acting on valence electrons. When a charged particle
penetrates through condensed matter, it polarizes the medium, which in turn reacts,
back slowing down the penetrating particle. To introduce a measure of the inertia of
ions, Brandt and Kitagawa (1982) introduced the concept of effective charge. The
effective charge, Z*, relates to how a diffusing atom interacts with the conduction
band electrons. The interaction with the conduction band electrons is a complex
function of the electronic structure, a quantum mechanical effect. The effective
charge is defined in terms of the stopping power of the charge Z(S) and the power
of the proton moving in the same medium at the same velocity (Sp) (Barberan and
Echenique 1986):
1
Z = S=Sp 2 ð8:8Þ
Z* is commonly negative in metal conductors due to the influence of the “electron

wind” Ye et al. (2004a, b, c, d). For high conductivity metals, i.e., Al or Cu, the value
of Z* is about [-1.0] (Lloyd et al. 2004). For pure tin, the Z* value was reported to
range from [-80] to [-160] (Kuz’menko and Osirovskii 1962; Lodding 1965; Sun
and Ohring 1976). For single-crystalline Sn, the Z* value was reported to range from
[-10] to [-16], while for polycrystalline, Sn is [-12] at about 200 °C. These values
seem to be favored by Sorbello’s electromigration theory, which predicted the
effective charge number for Sn at 185 °C as [-10] (Sorbello 1973). His theory is
based on the pseudopotential calculation of the driving force for atomic migration in
metals in the presence of electric current. The fact that Z* is temperature-dependent is
also reported by Singh and Ohring (1984) (see Fig. 8.3).
Fig. 8.3 Effective charge

number versus temperature
relation for Sn (After Singh
and Ohring (1984))
Fig. 8.4 Energy-

momentum relationship that
determines the value of
effective mass. (After Lloyd Ef
(2004)) Energy,E
1
m* ∝
d 2E
dk 2
Momentum,k
The effective charge, Z*, is determined by the effective mass, m*, of electrons
available for interaction (scattering) in which the energy produced is not exceeding
the Fermi energy, Ef, which is the total energy difference between the lowest
occupied conduction band and the highest conduction band, of uncharged metal at
absolute zero temperature. The effective mass m* is proportional to the second
derivative in the energy-versus-momentum, E - k relationship, a quantum mechan-
ical effect near Fermi level energy E = Ef (Lloyd et al. 2004). Above the inflection
point, m* is negative; below it m* is positive (Fig. 8.4).
8.2.4 Temperature Gradient Diffusion Driving Force
Thermomigration has been known since Ludwig (1856). As Chen et al. (2012)
stated, when an inhomogeneous binary alloy is annealed at an elevated isotropic
temperature field with no thermal gradient, it will become homogeneous. However,
when a homogeneous binary alloy is annealed under a temperature gradient, the
alloy will become inhomogeneous. This de-alloying phenomenon is called Soret
effect after Soret (1879). The phenomenon is caused by mass transport due to
temperature gradient.
The role of thermomigration in the presence of high current density in micro-
electronics solder joints has been ignored until recently. It has been shown by Ye
et al. (2003a, b, c, d, e, f), Abdulhamid et al. (2008), and Basaran et al. (2008) that
thermomigration plays a significant role in the presence of high current density.
Diffusion driving force due to thermal gradient can be bigger than electron wind
forces. If they are in the opposite directions, the thermal gradient can negate the
electron wind force and if they are in the same direction can hasten the failure
process significantly.
The temperature gradient-induced diffusion driving force is usually represented
by Huntington (1972):
Q
F tm = ∇T ð8:9Þ
T
where Q* is the heat of transport, which is the isothermal heat transmitted by the
moving atom in the act of jumping a lattice site less the intrinsic enthalpy, and T is
the temperature in Kelvin.
The most accepted thermomigration theory is very much based on the same
concepts as electromigration. In electromigration, the force acting on the diffusing
atom is the momentum exchange due to the collisions (scattering) of diffusing atoms
and scattered electrons in the conduction band. However, in thermomigration, in the
absence of electrical current, the driving force in the direction of the temperature
gradient results from the fact that the energy, therefore the momentum of the atoms
(phonons) at higher temperatures is greater than that of the lower temperature ones.
This gradient in the momentum produces a driving force for mass transport from the
hot side to the cold side.
Based on this argument, Q* for metal with a high Z* value should also be high.
Similar to Z*, the sign of Q* can either be positive or negative depending on the
effective mass of the charge carriers.
According to Huntington (1972), thermomigration driving force can be divided
into an intrinsic part and a part coupled to carrier electrons and phonons:
Q = Qin þ Qel þ Qph ð8:10Þ

The intrinsic considerations include the energy transported by the moving atom,
the energy required to prepare a place [a lattice site] to receive it, and, for the vacancy
diffusion mechanism, the formation energy for the counter-moving vacancy, which
is given by
Q = βEm - E f ð8:11Þ
where Em and Ef are the energies of motion and formation for the moving atoms,
respectively, and β is a dimensionless fraction less than 1 which gives the fraction of
the kinetic energy carried by the moving atom. Q* can be either positive or negative
depending on the magnitude of the kinetic energy and the formation energy.
According to Campbell and Huntington et al. (1969), β was reported to be about
0.81; hence, Q* is estimated to be about 1ev, which is in accordance with the
experimental measurement of 0.23 ev.
8.2.5 Stress Gradient Diffusion Driving Force
Due to mass transport from cathode to anode, and from the high-temperature side to
low-temperature side, there are more vacancies on one side and more mass accu-
mulation on the other side. Consequently, on the mass accumulation side, atoms are
subjected to compression, and on the vacancy accumulation side, atoms are
subjected to tension. Therefore, there is a stress gradient between two sides of the
continuum. This stress gradient is also another diffusion driving force. Tensile stress
makes diffusion easier and compressive stress makes diffusion more difficult to
happen.
The role of mechanical stress gradients as a diffusion driving force in conjunction
with electromigration was first explored in a series of experiments by Blech (1976)
and Blech and Herring (1976). It was shown that a critical current density threshold
exists below which mass transport is stopped in a thin-film aluminum sample. This
threshold was found to be inversely proportional to the thin-film Al stripe length. It
was suggested that electromigration in a short segment tends to induce back-stress
when they did an electromigration test on a set of short Al thin-film strips deposited
on a substrate as depicted in Fig. 8.5.
Extrusion Extrusion
Length
Void Void
AI AI AI
e ~ TIN e~
Fig. 8.5 Schematic of Al thin-film strips of different lengths showing stress gradient influence due
to electromigration
Figure 8.5 shows the effect of stress gradient on Al thin film on a substrate. The
Al thin-film strips are deposited on a substrate of TiN. It can be observed that the
longer the length, the larger the depletion, and extrusion on the cathode and anode
sides, respectively, due to electromigration. When the thin film is below a “critical
length,” there is no observable depletion or extrusion. Stress gradient diffusion
driving force is neutralizing the electron wind force-driven diffusion.
The dependence of mass transport on the strip length was explained by the effect
of the stress gradient. When the electromigration process transports Al atoms from
cathode to anode, the anode will be in compression, while the cathode will be in
tension. Based on the Nabarro-Herring model of equilibrium vacancy concentration
in a stressed solid, the tensile region has more, and the compressive region has fewer,
vacancies than the unstressed region. As a result, there exists a vacancy concentra-
tion gradient between the cathode side and the anode side. The stress gradient
induces an atomic flux of mass diffusing from anode to cathode, and it opposes
the mass flux driven by electromigration from the cathode side to the anode side.
Blech’s critical length comes into effect when a thin film on a substrate is
subjected to current density with blocking boundary conditions. A stress gradient
provides a mass transport force in opposition to electromigration. If the current
density is sufficiently low, a stress gradient will be generated that can stop electron
wind force-driven diffusion.
The stress gradient depends on the length of the strip; the shorter the thin film
strip, the greater the gradient. At a certain length defined as the “critical length,” the
stress gradient is large enough to counterbalance electromigration force so no
depletion at the cathode or extrusion at the anode occurs. The critical length can
be calculated when the net force due to electromigration and stress gradient is zero,
as shown in Eq. (8.12). The force equilibrium equation in the x-direction for an atom
can be rewritten as
∂σ
F = F e þ F σ = - Z eρj þ f Ω =0 ð8:12Þ
∂x
where f is the atomic vacancy relaxation ratio, the ratio of atomic volume to the
volume of an atomic vacancy, Ω is the atomic volume, and σ is the spherical stress.
Solving Eq. (8.12),
∂σ Z eρj
=
∂x fΩ
ð8:13Þ
fΩ
j∂x = ∂σ
Z eρ
The integration of Eq. (8.13) produces one of the most important characteristics
of electromigration. Given maximum stress that can be sustained by the conductor,
the product of current density and the length of the conductor determines whether
electromigration diffusion can occur or not. This product was discovered by I. A.
Blech and is referred to as the “Blech product” (Blech 1976; Blech and Herring
1976):
fΩ
jlB = ðσ - σ 0 Þ ð8:14Þ
Z eρ m
where lB is known as Blech’s critical length, σ m is the maximum spherical stress that
can be supported by the conductor, while σ 0 is the initial spherical stress in the
conductor. If the product of the current density and the length is exceeded,
electromigration mass transport can occur; if it is not, no electromigration can
happen.
If all thin-film conductors could be designed so that jlB stays below the critical
value, electromigration can be avoided since mass transport comes to a
complete halt.
However, this scenario is only true for a pure metal thin film on a substrate,
because it assumes that metal is not an alloy and there is no temperature gradient
acting on the system.
8.3 Laws of Conservation
Electromigration is an electron flow-assisted diffusion process. The process can be

assumed to be controlled by a vacancy diffusion mechanism, in which the diffusion
takes place by vacancies switching lattice sites with adjacent atoms. In isothermal
conditions, the process is driven by electrical current caused by mass diffusion, stress
gradient-induced diffusion, and diffusion due to atomic vacancy concentration. In
the presence of electrical current, due to electrical resistance, there is always Joule
heat production that leads to thermal gradient, which interacts with other diffusion
driving forces. Under the presence of these four forces, the atomic vacancy flux
equation can be given by combining Huntington’s (1972) and Kirchheim’s (1992)
flux definitions, and then adding the influence of temperature gradient and vacancy
concentration yields the following equations:
∂C v → →
= -∇ q þG ð8:15Þ
∂t
" → #

→ → Cv Z e → Cv → C v Q ∇T
q = - Dv ∇C v þ -ρ j - ð - f ΩÞ∇σ þ ð8:16Þ
kT kT kT T
Combining these two equations yields

→ →
∂C v Z eρ → fΩ → Q →
= D v ∇2 C v - ∇ Cv j þ ∇ C v ∇σ þ 2 ∇ðC v ∇T Þ þ G
∂t kT kT kT
ð8:17Þ
8.3 Laws of Conservation 439
8.3.1 Vacancy Conservation
Based on mass/vacancy conservation law, the following derivation can be given

(Fig. 8.6):

qx - qxþΔx ΔyΔz þ GΔxΔyΔz Δt = ðc þ ΔcÞðV þ ΔV Þ - cV
ð - ∇q þ GÞVΔt = ΔðcV Þ
set Δt → 0
∂ðCV Þ ∂c ∂V
) ð - ∇q þ GÞV = =V þc ð8:18Þ
∂t ∂t ∂t
∂c ∂V=V ∂c ∂εV
) - ∇q þ G = þc = þc
∂t ∂t ∂t ∂t
∂c ∂εV
) = - ∇q þ G - c
∂t ∂t
where CV0 is the thermodynamic equilibrium vacancy concentration in the absence

of stress field, c is the normalized vacancy concentration given by c = CCV0V , CV is the
instantaneous vacancy concentration, and εV is the volumetric strain. It is assumed
that a diffusing atom leaves behind a vacancy or takes an interstitial position (both
cases lead to volumetric strain only at the lattice site), and t is the time.
Substituting Eqs. (8.6), (8.8), and (8.11) into Eq. (8.5), we obtain the vacancy
flux, q, in the following form:

Ze cf Ω cQ
q = - Dv C v0 ∇c þ ð∇ϕÞc þ ∇σ þ 2 ∇T
sp
ð8:19Þ
kT kT kT
where:
Initial State
Δx, Δy, Δz, V= ΔxΔyΔz, c
After Δt
qx qx+Δ
x
Δx+ έxΔtΔx
Δy+ έyΔtΔy
Δ
Δz+ έzΔtΔz
z
V+ ΔV
Δ c+Δc
y
Δ
x
έxΔtΔ
x
Fig. 8.6 Illustration of vacancy conservation law

Q
Dv = vacancy diffusivity, Dv = Do e - kT
Z = effective charge number
e = electron charge
ϕ = electric potential
j = current density (vector)
f = vacancy relaxation ratio, ratio of atomic volume to the volume of a vacancy
Ω = atomic volume
k = Boltzmann’s constant
T = absolute temperature in Kelvin
σ sp = spherical part of the stress tensor, σ sp = trace(σ ij)/3
Q = heat of transport, the isothermal heat transmitted by the moving atom in the
process of jumping a lattice site less the intrinsic enthalpy
G = vacancy generation/annihilation rate can be expressed by (Sarychev et al. 1999)
Cv - C ve
G= - ð8:20aÞ
τs
Cev = thermodynamic equilibrium vacancy concentration given by
ð1 - f ÞΩσ sp
Cve = Cv0 e kT ð8:20bÞ
Cv0 = equilibrium vacancy concentration in the absence of stress

τs = characteristic vacancy generation/annihilation time. Hence,

ð1 - f ÞΩσ sp
c - exp kT
G = - C v0 ð8:20cÞ
τs
To solve Eq. (8.15) using the finite element method (method of weighted
residuals), we can write the following relationship:
Z
∂c
δc C V0 þ ∇ q - G dV = 0 ð8:21Þ
V ∂t
By expanding Eq. (8.21), we can write
ð1 - f ÞΩσ sp
Z Z exp -c
∂c kT
δc CV0 dV þ C V0 δc dVþ
∂t τs
V
Z V

Ze cf Ω cQ
δc∇ DV C V0 ∇c þ ∇ϕ c þ ∇σ sp þ þ 2 ∇T dV = 0
V kT kT kT
ð8:22Þ
8.3 Laws of Conservation 441
The concentration vector can be represented in terms of nodal values, that is,
δc = N T δci ð8:23Þ
Using Eq. (8.23) into Eq. (8.22), after simple organization, we can derive the
stiffness matrices. However, first, we have to define the terms in Eq. (8.22).
If we define C = Cv/Cv0 as the normalized concentration, then the vacancy
diffusion equation could be rewritten as
∂C v
∂t
→ → →
Z eρ → fΩ → Q →
= Dv ∇2 C þ ∇ C j þ ∇ C∇σ þ 2 ∇ C∇T
kT kT kT
G
þ ð8:24Þ
C vo
where initially, C = 1 (or Cv = Cv0).

A vacancy can be considered as a substitutional species at the lattice site with a
smaller relaxed volume than the volume of an atom. When a vacancy switches lattice
site with an atom or when another vacancy is generated/annihilated, local volumetric
strain occurs. As proposed by Sarychev et al. (1999), the vacancy causes volumetric
strain at the lattice site, because the volume of the vacancy is different from the
volume of the atom. This volumetric strain is composed of two parts, εm ij , the
g
volumetric strain due to vacancy flux divergence, and εij , the volumetric strain due
to vacancy generation:
1 → →
ε_ m
ij = f Ω∇ q δij ð8:25Þ
3
1
ε_ gij = ð1 - f Þ ΩGδij ð8:26Þ
3
where δij is the Kronecker delta. Thus, the total volumetric strain rate due to electrical
current is
h i
Ω → →
ε_ elec _m
ij = ε _ gij =
ij þ ε f ∇ q þ ð1 - f ÞG δij ð8:27Þ
3
The total volumetric strain rate due to current stressing is then

h → i
→
ε_ elec = Ω f ∇ q þ ð1 - f ÞG ð8:28Þ
The volumetric strain caused by the current stressing is superimposed onto the
total strain tensor with strains due to other loadings; thus, total strain [for small strain
case] can be given by
ε_ total
ij = ε_ mech
ij þ ε_ therm
ij þ ε_ elec
ij ð8:29Þ
where ε_ total
ij is the total strain rate tensor, ε_ mech
ij is the strain rate due to mechanical
loading, ε_ ij
therm
is the strain rate due to thermal load, and ε_ elec
ij is the volumetric strain
rate due to electromigration.
Following the standard procedure to obtain the finite element method stiffness
matrices leads to
Z
1 1
K cc = NT NdV ð8:30Þ
V Δt
2 3
Z ∂ fΩ sp Q ∇T nþ1
2 ∂N T 6 ∂x þ kT nþ1 ∇σ nþ1 þ kT 2 þ 7
K cc = DV 6
4
nþ1 sp 7 NdV ð8:31Þ
V ∂x Z e
c f Ω ∂∇σ nþ1 5
þ ∇ϕnþ1 þ nþ1
kT nþ1 kT nþ1 ∂cnþ1
2ð1 - f ÞΩ sp 3
Z ð1 - f ÞΩσ nþ1 ∂σ sp
3 exp nþ1
∂cnþ1
- 1
K cc = - N T 4 5 NdV
kT nþ1 kT nþ1
ð8:32Þ
V τs
Eventually, we have
cc
K nþ1 = K 1cc þ K 2cc þ K 3cc ð8:33Þ
By taking derivative with respect to temperature, we can obtain the coupled

stiffness matrix of vacancy concentration and temperature as follows:
Z
1 ð1 - f ÞΩσ sp sp
ð1 - f ÞΩ ∂σ nþ1 σ sp
½K cT = - NT exp nþ1
- nþ1
NdV
V τs kT nþ1 kT nþ1 ∂T nþ1 T nþ1
Z "
∂N T cf Ω ∂∇σ sp ∇σ sp cQ ∂ ∇T nþ1
þ DV nþ1
- nþ1
þ 2 -2
V ∂x kT nþ1 ∂T nþ1 T nþ1 kT nþ1 ∂x T nþ1

Z e 1
- ∇ϕnþ1 cnþ1 NdV
kT nþ1 T nþ1
ð8:34Þ
By taking derivative with respect to displacement u, we can write the following

relation:
Z
sp
∂N T cnþ1 f Ω ∂ ∂σ nþ1
K 1cu = DV ΒdV ð8:35Þ
V ∂x kT ∂x ∂εnþ1
8.4 Newtonian Mechanics Force Equilibrium 443
sp
Z ð1 - f ÞΩσ
2
nþ1
ð1 - f ÞΩ e kT nþ1 ∂σ sp
K cu = - NT nþ1
ΒdV ð8:36Þ
V kT nþ1 τs ∂εnþ1
where the strain displacement relation is given by εM

nþ1 = Β u .
M
And we can obtain the coupled stiffness matrix of concentration and mechanical
displacement by the following:
cu
K nþ1 = K 1cu þ K 2cu ð8:37Þ
The proper approach would be also to take a derivative of the potential with
respect to entropy and find the related stiffness matrices. However, that will add
entropy as a nodal unknown to our equations. For sake of computational simplicity,
we take the entropy generation rate values from the previous step and calculate them
at the Gauss integration point. Assuming the increments are small, the error should
be small. However, this simplification is not mathematically true and does contribute
to error.
8.4 Newtonian Mechanics Force Equilibrium
The force equilibrium equation must be written according to the laws of unified
mechanics theory. However, here we try to provide a simple implementation, which
ignores the derivative of displacement with respect to entropy. The thermodynamic
state index is introduced into the formulation later on. In the absence of body forces,
the Newtonian mechanics force equilibrium equation has the following form:
σ ij,j = 0 ð8:38Þ
For small strains, the elastic strain-stress constitutive model can be established as
σ = C ðε - εvp - εD - εTE Þ ð8:39Þ
where:
εvp is the viscoplastic strain vector
εD is strain vector due to diffusion
εTE is strain vector due to thermal expansion
C is the tangential constitutive tensor
Using the principle of virtual work, we can write
Y Z Z
= δε σ dV = δε C ðε - εvp - εD - εTE ÞdV ð8:40Þ
u V V
After the standard derivation process, we end up with the mechanical stress-
induced deformation-related stiffness matrices as follows:
Z

nþ1 =
K uu BT C B dV ð8:41Þ
V
The stiffness matrix for unit vacancy concentration change-induced deformation

can be obtained by taking the derivative of Eq. (8.40) with respect to concentration c:
Z
∂σ sp
nþ1 = ΒT NdV ð8:42Þ
nþ1
K uc
V ∂cnþ1
With a similar operation, we can obtain the stiffness matrix components

concerning the deformation induced by unit temperature change as follows:
Z
uT ∂σ sp
K nþ1 = ΒT nþ1
NdV ð8:43Þ
V ∂T nþ1
8.5 Heat Transfer
The transient heat transfer equation has the following form:
∂T
ρC p - ∇ðkh ∇T Þ - ρQ = 0 ð8:44Þ
∂t
where:
ρ = mass density of the material
Cp = specific heat
kh = coefficient of heat transfer
Q = heat generated within the body, which can be expressed as
Q = QJ þ QP þ QV ð8:45Þ
where:
QJ = Joule heating, which can be written as
Z
1
QJ = QJ ðt Þdt ð8:46Þ
Δt Δt
Using linear assumption, Eq. (8.46) can be expanded as

8.5 Heat Transfer 445
1 1 1 1
QJ = E nþ1 E - E nþ1 ΔE nþ1 þ ΔE nþ1 ΔEnþ1 ð8:47Þ
R nþ1 R 3 R
where E is the electrical field intensity defined as
∂ϕ
E= - ð8:48Þ
∂x
Based on Ohm’s law, the flow of electrical current description is given by
E 1 ∂ϕ
j= =- ð8:49Þ
R R ∂x
QP is heat generated due to plastic deformation at step n + 1 which can be

described as
QP = σ nþ1 : ε_ pl
nþ1 ð8:50Þ
where ε_ pl is the plastic strain rate. The entire plastic work is not dissipated as heat.
Some part is stored in dislocations which contribute to hardening. Equation (8.50) is
an acceptable simplification because plastic work in electromigration is negligible.
QV is the heat due to vacancy flux which can be expressed by
QV = q : F k ð8:51Þ
where:
q is vacancy flux
Fk is the effective driving force which has the following form:
Q kT
F k = - Z e∇ϕ þ f Ω∇σ sp þ ∇T þ ∇c ð8:52Þ
T c
Equation (8.51) can be expanded at step n + 1 as
QV
Q ∇T nþ1
2
cnþ1 kT nþ1
= Dv C V ∇cnþ1 þ Z e∇ϕnþ1 þ f Ω∇σ sp
nþ1 þ
kT nþ1 cnþ1 T nþ1
ð8:53Þ
Using the method of weighted residuals, Eq. (8.44) can be rewritten as

Z
∂T
δT ρC p - ∇ðkh ∇T Þ - ρQ dV = 0 ð8:54Þ
V ∂t
Using integration by parts and divergence theorem, we obtain

Z Z Z
∂T k 1
δT dV - h ∇δT ∇TdV - δT QdV = 0 ð8:55Þ
V ∂t ρC p V Cp V
Substituting δT = NT δT into Eq. (8.54) after simple finite element method

operations, we arrive at the following equation:
Z Z Z
∂T k ∂N
N T Cp dV þ h ∇TdV - N T QdV δT = 0 ð8:56Þ
V ∂t ρ V ∂x V
By taking the derivative of Eq. (8.56) with respect to temperature, T, we can find
the heat capacity matrix and heat transfer matrix which can be separately expressed
as
Z
CP
nþ1 =
CTT NT
Δt
NdV ð8:57Þ
V
Z
∂N T kh ∂N
K 1TT = dV ð8:58Þ
V ∂x ρ ∂x
Z
∂Qnþ1
J
K 2TT = NT NdV
∂T nþ1
Z V
1 ∂R ∂ϕnþ1 ∂ϕnþ1 ∂ϕnþ1 ∂Δϕnþ1
= N
T
- 2 - ð8:59Þ
R ∂T nþ1 ∂x ∂x ∂x ∂x
V
1 ∂Δϕnþ1 ∂Δϕnþ1
þ NdV
3 ∂x ∂x
Z Z
3 ∂Qnþ1
V
DV C V cnþ1
K TT = NT NdV = NT
∂T nþ1 k
" V V #
k∇cnþ1 ∂σ sp Q ∂N Q ∇T nþ1
k
F nþ1 N - 2F nþ1
k
N þ f Ω nþ1 N þ - N
cnþ1 ∂T nþ1 T nþ1 ∂x T 2nþ1
dV
T 2nþ1
ð8:60Þ
Eventually, we can get the temperature “stiffness” matrix by adding them all
together:
8.6 Electrical Conduction Equations 447
TT 1 2 3
nþ1 = Cnþ1 þ K TT þ K TT þ K TT
K TT ð8:61Þ
The term stiffness is normally used for the mechanical force-displacement rela-
tionship. Here we are using the term “stiffness” as an analogy. The coupled term
concerning concentration field-induced temperature change can be obtained by
taking the derivative of the total potential equation with respect to c which yields
Z Z
∂Qnþ1
V
DV C V k 2 ∂σ sp ∂N
Tc
K nþ1 =- NT NdV = - F nþ1 N - 2F nþ1
k
cnþ1 f Ω nþ1
∂cnþ1 kT nþ1 ∂cnþ1 ∂x
V V
kT nþ1 ∂N
þ cnþ1 - ∇c N
c2nþ1 ∂x
ð8:62Þ
Similarly, we can get the change of temperature induced by plastic deformation

by taking the derivative of Eq. (8.56) with respect to u:
Z Z
∂QPnþ1 ∂σ dev 1 ∂ΔεPl
nþ1 = - NT NdV = - NT þ BdV
nþ1 nþ1
K Tu
V ∂unþ1 V ∂εnþ1 Δt ∂εnþ1
ð8:63Þ
where σ dev represents deviatoric stress tensor.

And Joule heating can be considered by adding the following term into the
general stiffness matrix:
h i Z Z
T ∂Qnþ1
J
1 ∂N
Tϕ
K nþ1 = - N NdV = = N T J - ΔJ dV ð8:64Þ
V ∂ϕ nþ1 V 3 ∂x
8.6 Electrical Conduction Equations
The electrical field in metal is governed by Maxwell’s equation of conservation of

charge. Assuming steady-state direct current and no internal volumetric current
source, Maxwell’s equation reduces to
Z
j n dA = 0 ð8:65Þ
A
where:
A is the surface of a control volume
n is the outward normal to surface
j is the electrical current density defined in Eq. (8.49)
The divergence theorem can be used to convert the surface integral into a volume
integral which yields to
Z
∂
jdV = 0 ð8:66Þ
V ∂x
The equivalent weak form is obtained by introducing an arbitrary variation in the

electrical potential field, δ ϕ, and integrating it over the volume:
Z
∂
δϕ jdV = 0 ð8:67Þ
V ∂x
Using first the chain rule and then the divergence theorem, this statement can be
rewritten as
Z
∂δϕ
jdV = 0 ð8:68Þ
V ∂x
By introducing the definition of (8.48) and (8.49), the governing conservation of

charge equation becomes
Z
∂δϕ 1 ∂ϕ
dV = 0 ð8:69Þ
V ∂x R ∂x
Introducing δϕ = NT δϕN into the above equation yields

Z
∂N T 1 ∂ϕ
dV δϕN = 0 ð8:70Þ
V ∂x R ∂x
Utilizing Eq. (8.70), we can obtain the stiffness terms:

h i Z
∂N T 1 ∂N
K ϕϕ
nþ1 = dV ð8:75Þ
V ∂x R ∂x
The influence of temperature change on the current field can be taken into account
by adding the following term:
h i Z
∂N T 1 ∂R
K ϕT = E NdV ð8:76Þ
nþ1
V ∂x R2 ∂T
nþ1
Eventually, we obtain the total stiffness matrix for an element:

8.7 Thermodynamic Fundamental Equation for Electromigration and Thermomigration 449
2 38 → 9 8 9
½K uu
nþ1 ½Kucnþ1 ½K uT
nþ1 0 < u = < fu =
½
½K nþ1 fug = 4 ½K nþ1
K cu
½Kccnþ1 ½K cT
nþ1 0 5 cv
= qqc ð8:73Þ
nþ1 ½KTc
nþ1 ½K TT
nþ1 ½ nþ1
K Tϕ : T ; : T;
Tu
0 0 ½K ϕT
nþ1 ½Kϕϕnþ1 ϕ j
It is easy to notice that the finite element nodal point has displacement, vacancy
concentration, temperature, and electrical potential as nodal unknowns. An accurate
analysis, according to unified mechanics theory, would also include entropy gener-
ation rate as a nodal unknown. Of course, it does complicate the computation
process.
∂σ sp
nþ1
∂σ sp ∂σ sp nþ1 ∂σ nþ1
dev
∂ΔεPnþ1
Terms ∂cnþ1 , ∂Tnþ1
nþ1
, ∂ε nþ1
, ∂ε nþ1
, and ∂εnþ1
in Eq. (8.73) can be derived individually
once we introduce the viscoplastic model.

for Electromigration and Thermomigration
In this section, the thermodynamic fundamental equation accounting for most but
not all entropy production mechanisms is derived in a way appropriate for imple-
mentation in the finite element method. Ignored mechanisms included but are not
limited to grain coarsening and phase change and are assumed to account for a small
portion of the total entropy generation. Of course, including their contribution is
more accurate. In the derivation below, for the sake of completeness, and clarity,
some obvious thermodynamics relations are restated.
According to the second law of thermodynamics for any macroscopic system, the
entropy of the system is a state function.
The differential of the entropy, dS, may be written as the sum of two terms:
dS = dSe þ dSi ð8:74Þ
where dSe is the entropy supplied to the system and dSi is the entropy produced
inside the system. The second law of thermodynamics states that dSi must be equal to
zero for reversible and positive for irreversible processes:
dSi ≥ 0 ð8:75Þ
We establish the local form of entropy as

Z
S= ρsdV ð8:76Þ
V
Z
dSe
=- Js,tot dΩ ð8:77Þ
dt
Ω
Z
dSi
= γdV ð8:78Þ
dt
V
where s is the specific entropy per unit mass, Js, tot is the total entropy flow per unit
area and unit time, and γ is the entropy source strength or internal entropy production
per unit volume per unit time.
Using the local form of entropy, we can get the local form of the variation of
entropy as follows:
∂ρs
= - divJs,tot þ γ ð8:79Þ
∂t
γ≥0 ð8:80Þ
With the help of the material time derivative of the volume integral, which is
given by
d ∂
= þ v grad ð8:81Þ
dt ∂t
Eq. (8.79) can be rewritten in a slightly different form as follows:
ds
ρ = - divJs þ γ ð8:82Þ
dt
where the entropy flux, Js, is the difference between the total entropy flux and a
convective term:
Js = Js,tot - ρsV ð8:83Þ
In obtaining (8.80) and (8.82), it is assumed that the definition given by

Eqs. (8.74) and (8.75) also hold for infinitesimally small parts of the system. This
assumption agrees with the Boltzmann (1877) relation where the entropy state is
related to the disorder parameter via Boltzmann’s constant using statistical mechan-
ics, discussed earlier in Chap. 3.
8.7.1 Entropy Balance Equations
Based on the above definitions, the degradation of the system is directly related to
the rate of change of the entropy (thermodynamic state index). This enables us to
obtain more explicit expressions for the entropy flux and internal entropy
production rate.
With the assumption of a caloric equation of state, Helmholtz free energy, Ψ, in
differential form can be written as
dΨ = du - Tds - sdT ð8:84Þ
where u is internal energy and T is the temperature (here we use upper case, Ψ, to
distinguish from the electrical potential ϕ). Rearranging terms in Eq. (8.84) leads to
Tds = du - dΨ - sdT ð8:85Þ
To get the expression for free energy, we need to specify a function with a scalar
value concave with respect to temperature T and convex with respect to other
variables. Here we use
Ψ = Ψðε, εe , εp , T, V k Þ ð8:86Þ
where Vk can be any internal state variable determined by the specific application at
hand and physical processes that is leading to generating entropy for our specific
definition of failure.
In small strain theory, the strain can be written in the form of their additive
decomposition as follows:
εe = ε - εp ð8:87Þ
so that
Ψ = Ψðεe , T, V k Þ ð8:88Þ
The time derivative of the free energy is given by
dΨ ∂Ψ dεe ∂Ψ dT ∂Ψ dV k
= e: þ : þ : ð8:89Þ
dt ∂ε dt ∂T dt ∂V k dt
From Eq. (8.85), we obtain
ds du dΨ dT
T = - -s ð8:90Þ
dt dt dt dt
Using the first principle of thermodynamics, the conservation of energy equation

can be given by
du X
ρ = - divJq þ σ : gradð~vÞ þ Jk Fk ð8:91Þ
dt k
where u is the total internal energy, Jq is the heat flux, σ is the stress tensor, ~v is the
rate of deformation, Jk is the diffusional flux of component k, and Fk is the body
force acting on the mass of component k.
Using the conservation of energy equation in the form given by Eq. (8.91) and
Eqs. (8.82) and (8.83), we can write the rate of change of entropy density as follows:
ds 1 X ∂Ψ dεe ∂Ψ dT
ρ = - divJq þ σ : gradðvÞ þ Jk Fk - ρ : þ :
dt T ∂εe dt ∂T dt
k ð8:92Þ
∂Ψ dV k dT
þ : -s
∂V k dt dt
where
σ : gradð~vÞ = σ : ðD þ WÞ ð8:93Þ
and D (symmetric) and W (skew-symmetric) are the rate of deformation tensor and
spin tensor, respectively.
Due to the symmetry of σ, we obtain the following relation:
σ : ðD þ W Þ = σ : D ð8:94Þ
For a small strain case, we can make the following assumption:
dε dεe dεp
σ : D=σ : =σ : þ ð8:95Þ
dt dt dt
After rearranging Eq. (8.92) and comparing it with Eq. (8.82), we can obtain
Js = T1 Jq and the following entropy production rate term:
1 1X 1 dεp 1 dεe ∂Ψ dεe

γ= - J gradðT Þ þ
2 q
Jk Fk þ σ : þ σ: -ρ e : þ
T T k T dt T dt ∂ε dt
ρ ∂Ψ dT ρ ∂Ψ dV k
þ sþ - :
T ∂T t T ∂V k dt
ð8:96Þ
In materials with internal friction, all deformations cause a positive entropy

production rate γ ≥ 0, which is also referred to as the Clausius-Duhem inequality.
Using the following relations:
σ = ρð∂Ψ=∂εe Þ ð8:97Þ
∂Ψ
s= - ð8:98Þ
∂T
If we assume that entropy generation due to elastic strain rate is much smaller
compared to other mechanisms, we can simplify Eq. (8.96) as follows:
1 1X 1 ρ ∂Ψ dV k
γ= - J q grad ð T Þ þ Jk Fk þ σ : e_ p - : ð8:99Þ
T 2 T k
T T ∂V k dt
or if the heat flux term Jq is replaced by
1
Jq = CjGradðT Þj2 ð8:100Þ
T2
where C is the thermal conductivity tensor, the simplified entropy production rate
can be given by
1 1X 1 ρ ∂Ψ dV k
γ= - CjGradðT Þj2 þ J F þ σ : ε_ p - : ð8:101Þ
T 2 T k k k T T ∂V k dt
We identify Jk in Eq. (8.101) as the flux, and Fk as the effective driving force
term, as defined in the Onsager relation:

→ Q → kT →
Fk = Z ejρ þ ð - f ΩÞ∇σ - ∇T - ∇C ð8:102Þ
T C
In Eq. (8.101) the irreversible dissipation includes two parts: the first term is
called heat dissipation caused by conduction inside the system, while the second,
third, and fourth terms account for other irreversible processes in the system, and we
will call it intrinsic dissipation.
It is important to point out that the simplification we did in the entropy generation
rate term is not necessarily accurate. However, it simplifies computation. An accu-
rate derivation should include all entropy-generating micro-mechanisms in the
thermodynamic fundamental equation. Ignoring the entropy generation due to elastic
deformations as well, using Eqs. (8.101) and (8.102), we can write total entropy as
Z t1
Δs = γ dt ð8:103Þ
to
Δs
Zt → 2
1 CD Q∇T kT →
CjgradðT Þj2 þ v effective Z l eρj - f Ω∇σ þ þ ∇C
T2 kT 2 T C
=t0
1 ρ ∂Ψ dV k
þ σ : εp - : dt
T T ∂V k dt
ð8:104Þ
Thermodynamic state index (TSI) was given by
Fig. 8.7 Cross-section of test module and solder joints (Ye et al. 2003a)
Table 8.1 Time to failure

Finite element simulations
Current density Experiment time to failure Time to failure
(Amp/cm2) (hours) (hours)
0.6 × 104 1058.7 1098.2
0.8 × 104 446.6 435.33
1.0 × 104 228.7 222.41
h i
Φ = 1 - e - Δs R
ms
Table 8.2 Material properties for SAC405

Property Unit 95.5%Sn4%Ag0.5%Cu
Vacancy relaxation time (Ts) seconds 1.80 × 10-03
Effective charge number (Z*) N/A 10
Resistivity (R) μm3s-3A-2Kg RðT Þ = 1:52 × 1011
þ3:50 × 108 T ðK Þ
Atomic volume (Ω) μm-3 2.71 × 10-11
Vacancy concentration at stress-free state (Cv0) μm-3 1.107E+06
Young’s modulus (E) GPa E(T ) = 57.7 - 0.056T(K )
Poisson’s ratio N/A 0.33
pffiffiffi
Initial yield stress (σ y0) MPa σ y ðT Þ = 79:98 þ 95= d
- 0:2133T ðK Þ
Coefficient of thermal expansion (CTE) K-1 18.9 × 10-06
Linear kinematic hardening cons (c1) Kg.s.μm-1 9.63 × 103
Nonlinear kinematic hardening cons (c2) Kg.s.K-1.μm-1 7.25 × 102
Saturation isotropic hardening (R1) Kg.s.μm-1 0
Isotropic hardening rate (c) N/A 383.3
Dimensionless constant (A) N/A 7.60 × 109
Burger’s vector (B) μm 3.18 × 10-04
Initial average grain size μm 2.45
Grain size exponent N/A 3.34
Stress exponent N/A 6.65
Creep activation energy μm2.s-2.Kg. 7.95 × 1016
mol-1
Heat of transport(Q*) Kg.μm2.s-2 -3.68 × 10-08
Density Kg.μm-3 7.39 × 10-15
Specific heat μm2.s-2.K-1 2.19 × 1014
Heat conductivity Kg.μm.s-3.K-1 5.73 × 107
Fig. 8.8 TSI evolution (a) in three different solder joints at -20 °C (b) in solder #7 at different
temperatures
80
Experimental
70 Simulational
60
Time to Failure (hour)
50
40
30
20
10
0
-50 -45 -40 -35 -30 -25 -20 -15 -10 -5 0
Ambient Temperature (ć)
Fig. 8.9 Comparison of experiment and simulation result
8.8 Example
The solder joint in the electronic package shown in Fig. 8.7 was studied using the
formulation given in this chapter.
Ye et al. (2003a, b, c, d, e, f) published laboratory test data and finite element
simulation results of time to failure shown in Table 8.1 for solder joints under
different current density levels. Material properties used in the analysis are given
in Table 8.2, Figs. 8.8 and 8.9.
In these example simulations, critical TSI is set for Φcr = 0.094. Here we
conclude our discussion on electromigration and thermomigration. More examples
on this topic are provided in the publications listed in the references section.
References 457
References
Abdulhamid, M., Li, S., & Basaran, C. (2008). Thermomigration in lead-free solder joints.
International Journal of Materials and Structural Integrity, 2(1/2), 11–34.
Barberan, N., & Echenique, P. M. (1986). Effective charge of slow ions in solids. Journal of
Physics B: Atomic and Molecular Physics, 19(3), L81.
Basaran, C., Lin, M., & Ye, H. (2003). A thermodynamic model for electrical current induced
damage. International Journal of Solids and Structures, 40(26), 7315–7327.
Basaran, C., Li, S., & Abdulhamid, M. (2008). Thermomigration induced degradation in solder
alloys. Journal of Applied Physics, 103, 123520.
Black, J. R. (1969). Electromigration failure modes in aluminum metallization for semiconductor
devices. Proceedings of the IEEE, 57(9), 1587–1594.
Blech, I. A. (1976). Electromigration in thin aluminum films on titanium nitride. Journal of Applied
Physics, 47(4), 1203–1208.
Blech, I. A., & Herring, C. (1976). Stress generation by electromigration. Applied Physics Letters,
29(3), 131–133.
Boltzmann, L. (1877). Sitzungberichte der Kaiserlichen Akademie der Wissenschaften.
Mathematisch-Naturwissen Classe. Abt. II, LXXVI, 373–435.
Bosvieux, C., & Friedel, J. (1962). Electrolysis of metallic alloys. Physics Chemistry Solids, 23,
123–136.
Brandt, W., & Kitagawa, M. (1982). Effective stopping-power charges of swift ions in condensed
matter. Physical Review B, 25, 5631. Erratum Phys. Rev. B 26, 3968 (1982).
Campbell, D. R. & H.B. Huntington. (1969). Thermomigration and Electromigration in Zirconium.
Physical Review, 179(3):601–601.
Chen, C., Hsiao, H.-Y., Chang, Y.-W., Ouyang, F., & Tu, K. N. (2012). Thermomigration in solder
joints. Materials Science and Engineering: R: Reports, 73(9–10), 85–100.
Fick, A. (1855). On liquid diffusion. Philosophical Magazine and Journal of Science, 10, 31–39.
Genoni, T. C., & Huntington, H. B. (1977). Transport in nearly-free-electron metals.
IV. Electromigration in zinc. Physical Review B, 16, 1344.
Huntington, H. B., & Grone, A. R. (1961). Current-induced marker motion in gold wires. Journal of
Physics and Chemistry of Solids, 20(1–2), 76–87.
Huntington, H. B. (1972). Electro-and thermomigration in metals. Diffusion, papers presented at a
seminar of the American Society of Metals.
Kirchheim, R. (1992). Stress, and electromigration in Al-lines of integrated circuits. Acta
Metallurgica et Materialia, 40(2), 309–323.
Kuz’menko, P. F., & Osirovskii, L. F. (1962). Electric transfer of silver in copper. Izvestiya
Vysshikh Uchebnykh Zavedenii, Chernaya Metallurgiya, 11, 146–149.
Landauer, R., & Woo, J. W. F. (1974). Driving force in electromigration. Physical Review B,
10, 1266.
Lloyd, J. R., Tu, K. N., & Jasvir Jaspal, J. (2004). The physics and materials science of
electromigration and thermomigration in solders, chapter 20. In K. J. Puttlitz & K. A. Stalter
(Eds.), Handbook of lead-free solder technology for microelectronic assemblies. Taylor &
Francis.
Lodding, A. (1965). Current induced motion of lattice defects in indium metal. Journal of Physics
and Chemistry of Solids, 26(1), 143–151.
Ludwig, C. (1856). Akademie der Wissenschaften. Mathematisch-Naturwissenschaftliche Klasse,
20, 539.
Sarychev, M. E., et al. (1999). General model for mechanical stress evolution during
electromigration. Journal of Applied Physics, 86(6), 3068–3075.
Seith, W. (1955). Diffusion in Metallen, Platzwechselreaktionen. Springer.
Singh, P., & Ohring, M. (1984). Tracer study of diffusion and electromigration in thin tin films.
Journal of Applied Physics, 56(4), 899–907.
Skaupy, F. (1914). Electrical conduction in metals. Verband Deutscher Physikalischer

Gesellschaften, 16, 156–157.
Sorbello, R. S. (1973). A pseudopotential based theory of the driving forces for electromigration in
metals. Journal of Physics and Chemistry of Solids, 34(6), 937–950.
Soret, C. (1879). Archives des Sciences Physiques et Naturelles. Geneve, 3, 48–61.
Sun, P. H., & Ohring, M. (1976). Tracer self-diffusion and electromigration in thin tin films.
Journal of Applied Physics, 47, 478.
Wever, H., Seith, W. & Elektrochem, Z. (1955). Huntington, H.B. Electro- and thermomigration in
metals, proceedings of the seminar of the American Society of Metals. 59, 942.
Ye, H., Basaran, C., & Hopkins, D. (2003a). Thermomigration in Pb-Sn solder joints under joule
heating during electric current stressing. Applied Physics Letters, 82(8), 1045.
Ye, H., Hopkins, D., & Basaran, C. (2003b). Measuring joint reliability: applying the moire
interferometry technique. Advanced Packaging, 1, 17–20. Review article.
Ye, H., Basaran, C., & Hopkins, D. (2003c). Damage mechanics of microelectronics solder joints
under high current densities. International Journal of Solids and Structures, 40(15), 4021–4032.
#34.
Ye, H., Basaran, C., & Hopkins, D. (2003d). Numerical simulation of stress evolution during
electromigration in IC interconnect lines. IEEE Transactions on Components and Packaging
Technologies, 26(3), 673–681.
Ye, H., Basaran, C., & Hopkins, D. (2003e). Mechanical degradation of microelectronics solder
joints under current stressing. International Journal of Solids and Structures, 40(26),
7269–7284.
Ye, H., Basaran, C., & Hopkins, D. (2003f). Measurement of high electrical current density effects
in solder joints. Microelectronics Reliability, 43(12), 2021–2029.
Ye, H., Basaran, C., & Hopkins, D. (2004a). Pb Phase coarsening in eutectic Pb/Sn flip-chip solder
joint under electrical current stressing. International Journal of Solids and Structures, 41,
2743–2755.
Ye, H., Basaran, C., & Hopkins, D. (2004b). Deformation of microelectronic solder joints under
current stressing and numerical simulation I. International Journal of Solids and Structures, 41,
4939–4958.
Ye, H., Basaran, C., & Hopkins, D. (2004c). Deformation of microelectronic solder joints under
current stressing and numerical simulation II. International Journal of Solids and Structures, 41,
4959–4973.
Ye, H., Basaran, C., & Hopkins, D. (2004d). Mechanical implications of high current densities in
flip chip solder joints. International Journal of Damage Mechanics, 13(4), 335–346.
Ye, H., Basaran, C., & Hopkins, D. (2006). Experimental damage mechanics of micro/power
electronics solder joints under electrical current stresses. International Journal of Damage
Mechanics, 15, 41–68.
Chapter 9
Fatigue of Materials
Fatigue life prediction of materials has been widely studied. An extensive literature
survey of fatigue life prediction models is available in Lee and Basaran (2021);
hence, it is not necessary to repeat a literature survey herein. However, most of the
research is based on empirical curve fitting models under the framework of Newto-
nian mechanics. Fatigue test data is used for curve fitting a degradation evolution
function using stress, strain, or energy as a variable. Unified mechanics theory
(UMT), on the other hand, unifies the universal laws of motion of Newton by
incorporating the second law of thermodynamics directly into Newton’s laws at
the ab initio level. UMT introduces an additional linearly independent axis called
thermodynamic state index (TSI), which can have values between 0 and 1. Evolution
along the TSI axis follows Boltzmann’s entropy formulation and the thermodynamic
fundamental equation of the material. As a result, governing differential equations of
any system automatically include energy dissipation and degradation evolution.
When UMT is used, there is no need for an empirical degradation evolution function
obtained by curve fitting to fatigue test data. However, UMT does require deriving
the analytical thermodynamic fundamental equation of the material. In this chapter,
thermodynamic fundamental equations for several different fatigue problems are
presented.
The theory of elasticity assumes that there is no irreversible deformation during
elastic loading. However, it is well-known that under cyclic loading in a linear elastic
regime, any metal will still fatigue. Therefore, the second law of thermodynamics is
not violated. There is always irreversible entropy generation in any closed system,
which is ignored in the Newtonian mechanics-based theory of elasticity. If we apply
a cyclic load on metal within its elastic range [below the yield stress], the
corresponding strain would always be the same at every cycle, according to the
theory of elasticity, because the second law of thermodynamics is not included in the
universal laws of Newton directly. In summary, there is no TSI axis in the Newto-
nian space-time coordinate system. Moreover, in Newtonian mechanics, the laws of
thermodynamics are satisfied separately from than laws of Newton [not in the same
https://doi.org/10.1007/978-3-031-18621-9_9
460 9 Fatigue of Materials
differential equation] (such as satisfying the thermodynamic consistency using the

Clausius-Duhem inequality). As a result, in Newtonian continuum mechanics,
derivative of displacement with respect to entropy is zero.
9.1 Predicting High Cycle Fatigue Life of Metals
Fatigue life prediction with the unified mechanics theory eliminates the need for
curve fitting an empirical evolution function to fatigue test data. In unified mechanics
theory, the degradation evolution occurs along with the linearly independent ther-
modynamic state index (TSI) axis which is quantified by the thermodynamic fun-
damental equation of the material. The coordinate value of TSI starts at 0 and ends at
1. However, because TSI is an exponential function, computations are usually
stopped before it reaches 1.00, which is possible only if the power term is infinite.
The thermodynamic fundamental equation provides the entropy generation quantity
in the material under a given load. Based on this entropy value, a new TSI coordinate
is calculated, which is the normalized form of Boltzmann’s entropy formulation of
the second law of thermodynamics. In the formulation, the free energy of the system
is degraded by the entropy generated.
9.1.1 Thermodynamic Fundamental Equation
In this section, we derive the thermodynamic fundamental equation for high cycle
fatigue where the sample is subjected to a nominal stress level that is well below the
yield stress of the material and the loading frequency is quasi-static. The following
assumptions are made in the derivation:
1. Applied maximum nominal stress is below the yield stress of the metal; hence, no
macroscopic plastic deformation is expected to happen.
2. A mechanism called microplasticity is expected to happen at some micro-defect
sites at the micro-level (Lemaitre et al., 1999; Doudard et al., 2005;).
3. Point defects, including atomic vacancies, interstitials, and impurities, can be
built-in with the original crystal growth or created by energy input during the
fatigue process.
4. Input mechanical energy increases atomic vacancies and dislocation densities.
However, the increasing dislocation density only causes hardening in the micro-
level and never induces macroscopic plastic deformation as the maximum applied
nominal stress is below the metal’s yield stress value. The vacancy generation/
diffusion and dislocation motion (e.g., cross-slip) around inclusions induce irre-
versible deformations at a micro-level in elastic cyclic loads, (Marti et al., 2020; ;
Ho et al., 2017; Mughrabi, 2009).
9.1 Predicting High Cycle Fatigue Life of Metals 461
5. Vacancy concentration gradient in the specimen will result in vacancy diffusion,

and temperature gradient in the specimen will result in thermomigration.
6. Temperature evolution in the specimen is assumed to happen by atomic-friction-
generated heat, heat conduction, microplastic work, and thermoelastic damping.
9.1.2 Entropy Generation Mechanisms
Six mechanisms generate entropy during high cycle metal fatigue at quasi-static
frequencies. These are:
I. Configurational entropy, ΔSc
II. Vibrational entropy, ΔSvib
III. The entropy generation due to vacancy concentration gradient-driven diffusion,
ΔSd
IV. The entropy from heat conduction, ΔST
V. The entropy generation due to atomic-friction-generated heat, ΔSr
VI. The entropy generation due to microplasticity ΔSμp.
Because entropy is an additive property, we can write the thermodynamic
fundamental equation for the total entropy production as follows:
ΔS = ΔSc þ ΔSvib þ ΔSd þ ΔST þ ΔSr þ ΔSμp ð9:1Þ
I. Configurational Entropy Generation

The micro-mechanisms in the crystal include the rearrangement of atoms which
results in entropy production during cycling elastic loading. The configurational
entropy Sc is a concept from statistical thermodynamics that uses the binomial
distribution formula given by
N!
ð9:2Þ
ðN - nÞ!n!
where N is the number of lattice sites and n is the number of vacancies. Replacing the
disorder parameter W in Boltzmann-Planck entropy equation yields the following
form of configurational entropy (Kelly et al., 2012; Abbaschian & Reed-Hill, 2009):

N!
Sc = kB ln ð9:3Þ
ðN - nÞ!n!
Atomic defects are atoms missing from their proper lattice sites. These vacancies
are rearranged due to the thermally activated mass transport. The change in config-
urational entropy ΔSc from the temperature state of 1 to the temperature state of 2 is
given by

N! N!
ΔSc = kB ln - k B ln ð9:4Þ
ðN - n2 Þ!n2 ! ðN - n1 Þ!n1 !
where n1 and n2 are the number of vacancies at the temperature state of 1 and the
temperature state of 2, respectively. Equation (9.4) can also be written in the
following form:

1 - C v2 1 - C v1
ΔSc = kB N C v2 ln - Cv1 ln ð9:5Þ
C v2 C v1
where Cv1 and Cv2 are the thermodynamic equilibrium vacancy concentrations at the
temperature state of 1 and the temperature state of 2, respectively.
II. Vibrational Entropy Generation
The vibrational entropy is also a concept from statistical thermodynamics that
replaces disorder parameter W in the Boltzmann-Planck entropy equation with the
phase state of the atoms as they vibrate, which is defined by momentum and position
coordinates (Fultz, 2010; Atkins et al., 2018; Wollenberger, 1996; Laughlin &
Hono, 2014; Kelly et al., 2012; Abbaschian & Reed-Hill, 2009; Fultz, 2010).
The vibrational state of the atoms changes when vacancies are created. There are
various models proposed to precisely calculate vibrational entropies when a certain
number of vacancies are removed (Mishin et al., 2001).
We can assume that the change in vibrational entropy is the same when each
atomic vacancy is created (Abbaschian & Reed-Hill, 2009; Burton, 1972). There-
fore, the total vibrational entropy in the system can be given by
Svib = nΔSv ð9:6Þ
In which n is the number of atomic vacancies created and ΔSv is the change in
vibrational entropy when one atomic vacancy is created. Equation (9.6) is valid
because of the additive property of entropy.
The variation ΔSvib is given by
ΔSvib = ðn2 - n1 ÞΔSv ð9:7Þ
Comparing Configurational Entropy and Vibrational Entropy Magnitudes

The parameters necessary to calculate the configurational and vibrational entropy
magnitudes are listed in Table 9.1.
Entropy calculations for these two mechanisms are based on the number of
vacancies [or the vacancy concentration], which is related to the temperature at
that state. If we assume that after several cycles of loading, the temperature in the
metal specimen increases from T0 = 298 K to Ti = 398 K, then we can calculate the
evolution of ΔSc and ΔSvib. Figure 9.1 shows configurational entropy as a function
of temperature. Figure 9.2 shows vibrational entropy as a function of temperature.
We should point out that during high cycle fatigue at quasi-static frequencies, the
Table 9.1 Parameters to calculate configurational and vibrational entropy

Parameters Value Units
Avogadro’s number NA 6.023 × 1023 atoms=mol
Atomic weight AwFe 55.85 gr=mol
The mass density of steel ρsteel 7.894 gr=cm3
Vacancy formation energy for Fe Δhf 1.08 eV

Boltzmann’s constant kB 8.62 × 10-5 eV/K
1.38 × 10-23 J/K
Vibrational entropy for bcc crystals Δsv 2.4 kB/atom
Fig. 9.1 Configurational entropy (JK-1 m-3) versus temperature evolution
temperature increase in the sample is about 15 °C. The purpose of this example
where ΔT = 100 °C is used is to show the magnitude of entropy generation for a
hypothetical case.
III. Entropy Generation due to Vacancy Concentration Gradient-Driven
Diffusion
During the fatigue process concentration of vacancies in the metal are not uniform.
As a result, there is a vacancy concentration gradient. Vacancy concentration
gradient-driven diffusion and thermomigration are governed by the vacancy conser-
vation equation (or mass conservation) (Basaran & Lin, 2007; Ye et al., 2004; Ye
et al., 2006).
Fig. 9.2 Vibrational entropy (JK-1 m-3) versus temperature evolution
The concentration of atomic vacancies should be higher around the edge and
lower around the center for a specimen with a round cross-section (cracks usually
develop from outside to inward); hence, they cause the vacancy diffusion. The
entropy generation from vacancy gradient-driven diffusion ΔSd is extensively stud-
ied by Basaran and Lin (2008), Li et al. (2009), and Yao and Basaran (2013a, b, c). It
is given by
Z 2 !
t
C v Dv Q ∇T kB T
ΔSd = þ ∇c dt ð9:8Þ
t0 kB T 2 T c
where kB is Boltzmann’s constant, Cv0 is the thermodynamic equilibrium vacancy

concentration in the absence of a stress field, Cv is instantaneous atomic vacancy
concentration, c is normalized vacancy concentration c = Cv/Cv0, Dv is effective
vacancy diffusivity, T is absolute temperature, and Q is the heat of transport, which
is the isothermal heat transmitted by the moving atom in the process of jumping a
lattice site.
To compare the order of magnitude of Eq. (9.8) with other entropy generation
mechanisms, the following condition is assumed: the specimen is initially placed at
room temperature (298 K). After several cycles, the temperature at the specimen’s
gage section uniformly increases to 398 K. We assume the vacancy concentration at
the center is ten times smaller than at the edge (Fig. 9.3). The thermodynamic
equilibrium vacancy concentration Cv can be calculated by
Fig. 9.3 An illustration of a vacancy concentration gradient

TΔSv - Δhf
Cv = exp ð9:9Þ
kB T
The vacancy concentration at room temperature (298 K) and 398 K can be

obtained as follows:
C v,398 K = 2:35 × 10 - 13
ð9:10Þ
C v0 = C v,298 K = 6:1 × 10 - 18
The normalized vacancy concentration c and normalized vacancy gradient are

therefore given by
C 0:1C v,388 K
cedge = v,398 K , ccenter =

C v0

Cv0 ð9:11Þ
∇c = cedge - ccenter =1 cm = 3:5 × 106 m - 1
The effective vacancy diffusivity Dv is given by

-Q
Dv = D0 exp ð9:12Þ
RT
For BCC iron (α-Fe), D0 = 2:8 × 10 - 4 ms , activation energy Q = 251 kJ/mole,

2

and R = 8:3145 moleJ K [63]. At 398 K, Dv = D0 exp -RTQd ≈ 3 × 10 - 37 ms .
2
We can calculate the value of Eq. (9.8) as follows:
m2 2
2:35 × 10 - 13 × 3 × 10 - 37
1:38 × 10 - 23 J=K × 398 K
_
Sd = s 3:5 × 10 m
6 -1
1:38 × 10 - 23 J=K × ð398 KÞ2 3852
= 8 × 10 - 91 JK - 1 s - 1
ð9:13Þ
From this calculation, it is clear that the entropy generation due to mass transport
is negligible compared to other entropy generation mechanisms.
IV. Entropy Generation due to Heat Conduction

In metal high cycle fatigue, the temperature gradient causes an irreversible heat flow.
The entropy generation equation is given by
Z t
∇T ∇T
ΔST = - kh dt ð9:14Þ
t0 T2
where kh is coefficient of heat conduction and ∇T is the temperature gradient. Again,

to calculate the order of magnitude of this entropy generation mechanism, we
assume a simple one-dimensional temperature gradient ΔT Δx = 10 =cm from the gage
K
section to the grip section (Fig. 9.3). The thermal conductivity of steel is K h = 50 Km
w
.
After several cycles, the temperature increases from T0 = 298 K to Ti = 398 K and
then remains in a steady-state until near failure at around 1.4 × 105 cycles (operating
frequency at 30 Hz). The evolution of entropy generation due to heat conduction
ΔST for a specimen is shown in Fig. 9.4.
V. Entropy Generation due to Atomic-Friction (Scattering)-Generated Heat
The entropy generation due to atomic-friction-generated heat is given by
Z t
ρr
ΔSr = dt ð9:15Þ
t0 T
Fig. 9.4 Entropy production due to thermal conduction (JK-1 m-3) versus the number of cycles
Temperaure varation for uniaxial compression

600
550
temperature(K)
500
450
400
350
300
250
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
time(picosecond)
0.001/ps 0.005/ps 0.01/ps
Fig. 9.5 Temperature variation with time under uniaxial compression under different strain rates
At the atomic scale, the temperature is defined by the intensity of atomic

vibrations. The atomic-friction-generated heat per unit mass, r, is due to the increas-
ing intensity of atomic vibrations in the lattice from micro-mechanisms such as the
breaking of atomic bonds and phonon-phonon scattering, phonon-electron scatter-
ing, and electron scattering (called internal friction) (Gao, 1997; Ragab & Basaran,
2009; Chu et al., 2015). This term is essentially different than the heat generated
through the macroscopic observable plastic work or thermoelastic sources.
It is possible to estimate the internal heat generation due to internal frictions using
molecular dynamics simulations. Figures 9.5 and 9.6 show temperature variation
with time under uniaxial compression and tension under different strain rates
obtained by molecular-level simulations (Lee & Basaran, 2022).
It is shown that the temperature fluctuation during uniaxial compression and
tension are not completely reversible at higher strain rates (e.g., for 0.01/picosecond
at 20 ps). As a result, temperature increases by 250 K for compression but decreases
by 90 K for tension. However, at lower strain rates, the temperature fluctuation is
negligible (e.g., for strain rates below 0.001/ps). It is therefore concluded that at low
strain rates, the temperature fluctuation is solely the result of thermoelastic coupling.
Atomic-friction-generated heat is negligible and can be neglected below 0.001/
picosecond rate of loading.
VI. Entropy Generation due to Microplasticity
The term microplasticity is used in two-scale models, which was initially developed
by Lemaitre et al. (1999) and then reformulated by Doudard et al. (2005). These
researchers developed the microplasticity model to calculate the amount of
microplastic strain during high cycle fatigue and then used the microplastic strain
as a variable in an empirical function to curve fit the test data to predict the fatigue
life. In UMT-based model, microplastic strain is needed just to calculate the amount
of entropy generation.
Temperature variation for uniaxial tension

320
300
temperature (K)
280
260
240
220
200
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
time(picosecond)
0.001/ps 0.005/ps 0.01/ps
Fig. 9.6 Temperature variation with time under uniaxial tension for different strain rates
In two-scale models, the fatigue in the material is investigated at macroscopic and

microscopic scales independently. From a macroscopic point of view, the material
deforms elastically during the elastic cyclic fatigue loading, while from the micro-
scopic point of view, some microplasticity is activated, because of high-stress
concentration, at defects and localized dislocation slip planes (Fig. 9.7).
The two-scale model is based on an RVE (representative volume element) and
divided into two parts: the elastic matrix part and the elastic-plastic inclusion part.
The law of localization and homogenization is applied to deduce the relationship
between macroscopic stress tensors and the microscopic stress and microscopic
plastic strain tensors (Lemaitre et al., 1999; Doudard et al., 2005; Fan et al., 2018).
Microscopic stress tensor is given by
σ μ = Σ - 2μð1 - bÞεμp ð9:16aÞ

6ðK þ 2μÞ
b= ð9:16bÞ
5ð3K þ 4μÞ
where σ is microscopic stress tensor, Σ is macroscopic stress tensor, μ is Lame’s

constant, K is bulk modulus, and εp is the microscopic plastic strain tensor.
Charkaluk and Constantinescu (2009) proposed a modified version of Eq. (9.16a)
using a self-consistent scheme proposed by Kröner (1961). In the case of isotropic
elastic behavior with a classically defined deviatoric plasticity, one can write the
following relationship:
σ μ = Σ - 2μð1 - bÞð1 - f v Þεμp ð9:17Þ

Fig. 9.7 A representative

volume element of the
two-scale model, Σ is a
macroscopic elastic stress
tensor, E is a macroscopic
elastic strain tensor, σ μ is
microscopic stress tensor,
and εμp is the microscopic
plastic strain tensor
where fv is the volume fraction of micro-defects experiencing (slip) microplasticity.

Assuming that the material experiences the same elastic behavior at the mesoscopic
and the macroscopic scale, the previous relation implies the following relation:
εμe = E - ½ð1 - bÞð1 - f v Þεμp ð9:18Þ
where εμe is the microscopic elastic strain tensor.

Charkaluk and Constantinescu (2009) report that, although it cannot exactly be
correlated to the volume ratio, it is assumed that the relative surface ratio covered by
the activated dislocation slip bands can be used to represent fv. For low carbon steel,
which has a fatigue limit [meaning no microplasticity happens] of 235 MPa, Cugy
and Galtier (2002) reported empirical values of fv = 3% for a nominal stress
amplitude of 180 MPa, a value of fv = 10% for a nominal stress amplitude of
250 MPa, and a value of fv = 20% for a nominal stress amplitude of 300 MPa. We
believe that fv may be calculated by dislocation dynamics simulations for a given
initial defect ratio or measured by SEM.
A computational scheme for the microplastic strain increment was proposed by
Charkaluk and Constantinescu (2009), as follows:
qffiffi
3
2 Anþ1 - σ μy0 Anþ1
Δεμp = qffiffi ð9:19Þ
3
2 2μð1 - bÞð1 - f v Þ þ 23 h Anþ1
in which h is the hardening modulus and Anþ1 is given by

2
Anþ1 = devðΣn Þ - 2μð1 - bÞð1 - f v Þ þ h εμp,n þ devðΔΣÞ ð9:20aÞ
3
6ðK þ 2μÞ
b= ð9:20bÞ
5ð3K þ 4μÞ
The yield criterion is given by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffi
2 μ
f = ðSμ - αμd Þ : ðSμ - αμd Þ - σ =0 ð9:21Þ
3 y0
and the hardening rule is given by

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 μ 2 μp μp 1 μ
α_ μ = h_eμp ðσ - αμ Þ = h E_ E_ ðσ - α μ Þ ð9:22Þ
σ μy0 3 ij ij σ μy0
where Sμ is the microscopic deviatoric stress, αμd is the deviatoric part of the
microscopic back-stress of kinematic hardening, eμp is the microscopic equivalent
plastic strain, σ μy0 is the microscopic yield stress, (σ μ - αμ) is the translational
direction of the microscopic yield surface under Ziegler’s rule, and h is the kinematic
linear hardening modulus defined by the slope of the stress-strain curve for a finite
plastic strain value as follows: h = σ μ - σ μy0 =εμp .
The entropy generation due to microplasticity is then given by
Z t σ μ : ε_ μp
ΔSμp = ϕf v dt ð9:23Þ
t0 T
in which σ μ and ε_ μP are micro-stress and strain rate tensors. During the microscopic
plastic deformation, some plastic work is stored as dislocation stored energy, which
is accounted for as the hardening of the material in Eq. (9.22). fv is a constant that
establishes a relation between the macro-stresses and micro-stresses utilizing the law
of localization and homogenization. It is the ratio of the activated volume of the
micro-defects VD to the elastic matrix Vmatrix (Charkaluk & Constantinescu, 2009):
VD
fv = ð9:24Þ
V matrix
Some researchers determined fv value by high-resolution microscopic inspection

of the surface area along the central axis of the specimen around the center of the
gage part during the test to identify activated slip bands in different stages of fatigue
(Cugy & Galtier, 2002). For high cycle fatigue of DP600 steel under conventional
low-frequency tension-compression loading (e.g., operating at 30 Hz), the material
deformation is expected to happen in an athermal regime, where trans-granular crack
initiation that leads to surface failure occurs due to local stress and strain concen-
tration at slip bands in ferrite grains (Torabian et al., 2017). The percentage of the
observed slip bands’ area in the microscope inspection surface area can be used as a
first approximation to calculate the volume fraction of micro-defects in the matrix,
VD/Vmatrix. This value can be determined by using microscopy such as TEM or SEM,
as shown in Fig. 9.9.
Fig. 9.8 Visualization of microplasticity entropy calculation
Table 9.2 Material parameters for DP600 steel

Young’s modulus 210,000 MPa Macroscopic yield stress 440 MPa
Hardening 1000 MPa Microscopic yield stress 260 MPa
coefficients
Poisson’s ratio 0.3 Thermal expansion coefficient 10 × 10-
6 o
coefficientcoefficients /C
Density 7894 kg m-3 Heat capacity 460 J kg-
1 -1
K
Equivalent thermal exchange duration, τeq, 80 s
Figure 9.8 depicts the visualization of microplasticity, inside the hexagon region,
microscopic stress and microplastic strain. In Table 9.2 material constants used for
simulation are given.
The microscopic yield stress in this microplasticity model is the mean fatigue
limit of the specimen. We should clarify that a fatigue limit does not mean that there
is no fatigue failure if the stress is below the fatigue limit. Not resulting in fatigue at
any stress level would violate the second law of thermodynamics. It just indicates
that no microplasticity occurs below the macroscopic fatigue limit stress. However,
all other entropy generation mechanisms are still active, regardless of stress level and
all materials’ fatigue at stress levels below the fatigue limit.
In Eq. (9.23), TSI, (ϕ), is a metric of entropy evolution. It is assumed that the
increase in the percentage of activated dislocation slip planes follows the second law
of thermodynamics, which is represented by TSI. However, each microplasticity site
is considered to be independent and has a constant fv. Microplasticity sites are not
expected to coalesce before the onset of final failure (Fig. 9.8).
Fig. 9.9 Dislocation slip bands in ferrite grains on the surface of the specimen activated at 250 MPa
stress amplitude. N = 107 cycles, ΔT < 15 ° C. (After Torabian et al. (2017))
There is no fv data available in the literature obtained by TEM or SEM for DP600
steel at hand. However, Torabian et al. (2017) provided fatigue test data on DP600
steel and the results of fractography studies. The appearance of slip bands in ferrite
grains on the surface of the specimen around the gage part under 30 Hz fatigue
loading is shown in Fig. 9.9. The arrows indicate the slip bands, which are estimated
to be around 10% of the observed surface. We use fv=10% as the initial value for the
calculations. Of course, fractography studies for individual cases are necessary, since
the defect ratio in materials is the highly variable depending quality of the material
and manufacturing methods.
The microplasticity calculation in this section is based on the following
assumptions:
1. The hardening behavior of the material is approximated by a bilinear kinematic
hardening curve. Hardening modulus is defined as the slope between (lower-
yielding strength) LYS and UTS (ultimate tensile strength) from the stress-strain
curve, as a first approximation.
2. The microscopic hysteresis loops are stabilized, and they are the same at each
defect site. The evolution of TSI, ϕ, from 0 to 1 only induces microplasticity at
more defect sites, without any macroplasticity.
3. The energy dissipation due to kinematic hardening at microplasticity sites is small
enough to be ignored (Naderi et al., 2010)
Fig. 9.10 Micro-stress-strain curve for each inclusion (defect) site under 300 MPa nominal stress
amplitude
A microstress-microstrain hysteresis loop at a defect location, calculated by

Eqs. (9.17)–(9.22), is shown in Fig. 9.10. The entropy production due to
microplasticity for a specimen undergoing 1.4 × 105 cycles (at a frequency of
30 Hz) is shown in Fig. 9.11.
9.1.3 Comparison Between Different Entropy Generation

Mechanisms
The magnitude of each entropy generation mechanism is summarized in Table 9.3.

In Table 9.3, it is observed that the configurational entropy and vibrational
entropy generation mechanisms are around the same order of magnitude. They are
very small compared to heat conduction and microplasticity. The diffusion mecha-
nism and atomic-friction-generated heat are also very small. Furthermore, between
the microplasticity and heat conduction, entropy generation due to microplasticity is
Fig. 9.11 Entropy generation (MJK-1 m-3) due to microplasticity versus the number of cycles
under 300 MPa nominal stress amplitude
Table 9.3 Order of magnitude of different entropy generation mechanisms

Mechanisms The magnitude of entropy generation
1 Configurational entropy ≈10-7 JK-1 m-3
2 Vibrational entropy ≈10-8 JK-1 m-3
3 Vacancy diffusion Negligible due to low diffusivity
4 Thermal conduction ≈104 JK-1 m-3
5 Atomic-friction-generated heat Negligible based on molecular dynamics simulations
6 Microplasticity ≈106 JK-1 m-3
two orders of magnitude bigger. Hence, we conclude that the total entropy produc-
tion, the thermodynamic fundamental equation, can be simplified as
ΔS = ΔSconf þ ΔSvib þ ΔSd þ ΔST þ ΔSr þ ΔSμp ≈ ΔSμp ð9:25Þ
Therefore, the thermodynamic state index can be given by

0 0 Z t 11
σ μ :ε_μP
- ms ϕf v dt
B B ρT CC
ϕ = ϕcr B B CC ð9:26Þ
t0
@1 - exp @ R AA
Incorporating the TSI into microplastic work accounts for the increased proba-
bility of activated micro-defect sites that emerge in the specimen during the fatigue
process.
TSI is an exponential function, hence never reaches unity. In practice, we
determine a critical value ϕcr as a threshold. In this example, the specimen is
assumed to fail when ϕ reaches ϕcr = 0.995, since the probability of reaching
maximum entropy at this state is 99.5%.
9.1.4 Temperature Evolution due to Mechanical Work
The temperature of the specimen plays an important role in high cycle fatigue life.
From the entropy point of view, the cumulative entropy production and TSI calcu-
lation are related to the temperature of the specimen. Inspecting Eq. (9.26), if the
temperature is controlled (i.e., the specimen is air-cooled, e.g., near room tempera-
ture), the resulting entropy production will be larger, hence resulting in shorter
fatigue life. However, self-heating during loads at low frequencies (i.e., 30 Hz)
does not exceed ΔT = 15 K. This amount of temperature difference will not change
the fatigue life significantly. However, as the frequency of the loading increases, ΔT
increases. In ultrasonic frequencies, ΔT reach several hundred degrees.
The temperature of the specimen is governed by the fully coupled
thermomechanical equation derived from classical continuum mechanics as follows:

∂σ e ∂Ak _
k h ∇2 T = ρC T_ - σ : ε_ p - ρr - T : ε_ þ Ak - T Vk ð9:27Þ
∂T ∂T
∂s
where ρ is the mass density; C = T ∂T is specific heat; εe and εP are elastic and plastic
strain vectors, respectively; σ is the stress tensor; r is the strength per unit mass of the
internal distributed heat source; and Ak is a thermodynamic force associated with the
internal thermodynamic variables, Vk.
Using Eq. (9.27), we can calculate the evolution of temperature in the specimen
due to mechanical work. To simplify this equation, we can ignore plastic strain ε_ p ,
internally distributed heat source r, and other thermodynamic variables, Ak V_ k. These
assumptions are justified, because of the following reasons:
1. The sample is subjected to elastic loads only, as such, there is no uniform
macroplastic strain. However, the contribution of microplasticity is not neglected.
2. Based on molecular dynamics simulations, the contribution of r is negligible at
30 Hz loading.

3. Ak - T ∂Ak V_ k represents the non-recoverable energy corresponding to the
∂T
internal coupling source (such as grain coarsening, phase transformation, etc.).
However, for metals, this non-recoverable energy only represents 5–10% of the
mechanical dissipation (plastic work) and is often negligible for high cycle
fatigue.
Therefore, Eq. (9.27) can be simplified to the following expression:

∂Σ _e
ρC T_ - ϕf v σ μ : ε_μP - k h ∇2 T - T : E =0 ð9:28Þ
∂T
If we ignore the thermal fluctuation due to thermoelastic damping and simplify

the conduction term as - k h ∇2 T ffi ρC τθeq, Eq. (9.28) can be written in the following
time integration form:

T nþ1 - T n εμP,nþ1 - εμP,n T - T0
ρC - ϕf v σ μn : þ ρC n =0 ð9:29Þ
Δt Δt τeq

σ μn : εμP,nþ1 - εμP,n T - T0
T nþ1 = T n þ ϕf v - n Δt ð9:30Þ
ρC τeq
The temperature in the sample is calculated at each cycle using Eq. (9.30).
Figure 9.12 shows the temperature evolution for DP600 steel operating at 30 Hz
for various stress amplitudes obtained using Eq. (9.30). The entropy generated due to
microplasticity is strongly affected by the stress amplitude. A rapid climbing stage
followed by a steady state is observed. However, Eq. (9.29) cannot capture the
abruptly raising temperature when a specimen is near failure when the temperature
spikes.
Fig. 9.12 Temperature evolution versus the number of cycles at various stress amplitudes
9.1.5 Entropy and TSI Calculations
Figure 9.13 shows accumulated entropy at failure for various stress amplitudes.
Accumulated total entropy at failure is a constant value and is sometimes also
referred to as the fatigue fracture entropy (FFE), which is known to be independent
of stress amplitude.
Figure 9.14 shows the thermodynamic state index as a function of the number of
cycles. TSI, ϕ, reaches 1 around 8.5 × 104 cycles under 325 MPa applied stress. For
stress amplitude of 300 MPa, TSI, ϕ, reaches 1 at 1.4 × 105 cycles. For stress
amplitude of 275 MPa, TSI, ϕ, reaches 1 at around 3.6 × 105 cycles.
If we relate the stress amplitude with the number of cycles to failure, a bilinear
curve can be plotted, in Fig. 9.15.
In Fig. 9.14, it is easy to observe that the TSI increases slowly in the early stage of
cycling and then suddenly begins to evolve rapidly after a certain critical threshold.
This trend can be explained by Eq. (9.28), in which TSI is included in the
microplasticity entropy production equation. In the early stages, there are too few
microplasticity sites. Because microplasticity is the largest contributor to entropy
generation, the corresponding TSI calculated by the entropy production is small at
the early stages. After a certain number of cycles, enough micro-defects are acti-
vated, and more entropy is produced.
Fig. 9.13 Accumulated total entropy production versus the number of cycles for various stress
amplitudes [this is also referred to as the fatigue fracture entropy]
Fig. 9.14 TSI evolution versus the number of cycles for various stress amplitudes
Fig. 9.15 Fatigue life prediction with UMT for stress amplitude versus the number of cycles for
various stress amplitudes [when TSI reaches ϕcr = 0.995, the specimen is considered as failed]
Fig. 9.16 Comparison of fatigue data and unified mechanics theory predictions for tension-
compression fatigue tests on dual-phase steel (DP600) without any pre-strain [the experimental
SN curve is obtained from Torabian et al., 2017]
In Fig. 9.16, simulation results are compared with test data. It shows that the
predicted cycles to failure obtained from unified mechanics theory and the test data
match reasonably well. Test data in Fig. 9.16 shows expected scatter, especially at
lower stress levels. It is very well-known that fatigue test data is always stochastic.
Unified mechanics theory, which is based on Boltzmann’s entropy formulation, also
yields stochastic expected life.
In this section, unified mechanics theory is used for predicting the high cycle
fatigue life of DP600 steel. The thermodynamic fundamental equation of metals
under high cycle fatigue is derived. In the fatigue life prediction, no empirical curve
fitting function is needed. The entropy generation due to atomic vacancy configura-
tion, atomic vibration, and mass transport is extremely small compared to entropy
generation due to microplasticity. It is also assumed that entropy generation due to
grain coarsening during high cycle fatigue is negligible.
The entropy generation due to microplasticity is the most dominant entropy
generation mechanism. This mechanism can be visualized as some microplastic
inclusions at localized defect sites inside an elastic matrix which has its micro-
stress and micro-strain based on laws of localization and homogenization. Energy is
dissipated through the microplastic work at the locations of these inclusions. The
number of inclusions increases as TSI ϕ increases from 0 to ϕcr. Metal high cycle
fatigue life prediction results based on the theory of unified mechanics are compared
with experimental test data. It is shown that the prediction and test data match
very well.
9.2 Predicting Ultrasonic Vibration Fatigue Life
Low cycle, high cycle, very high cycle, and ultrahigh cycle fatigue occur on different
entropy generation mechanisms. In each one, the magnitude of entropy generation
for each mechanism is different. For example, during high cycle fatigue, internal
heat generation is usually less than ≈15 ° C temperature change in a test sample,
while in very high cycle, fatigue temperature change can go to ≈150 ° C or much
higher. On the other hand, while microplasticity is the dominant entropy generation
mechanism in a high cycle and very high cycle fatigue, it is assumed that it does not
happen in ultrahigh cycle fatigue, which happens when the stress value is below the
nominal fatigue limit of a metal.
In this section, predicting the very high cycle fatigue life of metals with the
unified mechanics theory is discussed. Simulation results are compared with the
fatigue life test data. It is shown that the unified mechanics theory can predict very
high cycle fatigue life very well, without relying on traditional curves fitting an
empirical function to fatigue test data.
The objective is to predict the fatigue life of metals subjected to ultrasonic
vibration where the maximum nominal stress is below the yield stress of the material.
The nomenclature used in this section is given in Table 9.4.
Table 9.4 Nomenclature used for ultrasonic vibration fatigue life model
Nomenclature
ϕ Thermodynamic state index ρ Mass density
ϕcr Critical thermodynamic state index r Internal heat generation [IH]
Δs Change in total specific entropy rdrag IH due to drag mechanism
ms The molar mass of the material rdis IH due to dislocation motion
R Universal gas constant Bdrag Drag coefficient
ST Thermal conduction entropy v Dislocation velocity
Sr Internal friction entropy μ Shear modulus
Sμp Microplasticity entropy b The magnitude of the Burger’s
vector
kh Thermal conductivity ϱ Total dislocation density
σμ Microscopic stress tensor ϱm Mobile dislocation density
εμe The microscopic elastic strain tensor αH Taylor’s constant
εμP The microscopic plastic strain tensor γ_ Shear strain rate
Σ Macroscopic stress tensor a Lattice constant
E Macroscopic strain tensor E Young’s modulus
fv The volume fraction of activated micro- υ Poisson’s ratio
defects
Sμ Microscopic deviatoric stress α Thermal expansion coefficient
αμd Deviatoric of microscopic back-stress cV Specific heat capacity
eμp Microscopic equivalent plastic strain θ Total change in temperature
h Hardening modulus τeq Equivalent conduction time
σy Microscopic yield stress (fatigue limit) t eq Fraction of heat conducted
Ψf Frequency coefficient
9.2 Predicting Ultrasonic Vibration Fatigue Life 481
9.2.1 Thermodynamic Fundamental Equation
During the very high cycle fatigue of metals, six prominent mechanisms generate
entropy. These are:
I. Configurational entropy Sc.
II. Vibrational entropy, Svib
III. Entropy generation due to diffusion Sd
IV. Entropy generation due to heat conduction ST
V. Entropy generation due to internal heat generation Sr
VI. Entropy generation due to microplasticity, Sμp
Among these six mechanisms, the first three are too small compared to other
mechanisms (Lee & Basaran, 2021). Therefore, we discuss the entropy generation
due to the last three mechanisms. Entropy is an additive property; hence, we can
write the following equation for the total entropy to obtain the thermodynamic
fundamental equation:
ΔS = ΔST þ ΔSr þ ΔSμp ð9:31Þ
The following derivation is based on these assumptions:

1. Maximum nominal stress during fatigue testing is below the yield stress of the
material; hence, no macroscopic plastic deformation is expected. However,
microplasticity is expected to happen at defect sites at the micro-level.
2. Input mechanical energy increases atomic vacancies and dislocation density.
However, the increasing dislocation density only causes hardening at a micro-
level and never induces macroscopic plastic deformation, as the maximum
nominal stress is below the metal’s yield stress. The vacancy generation/diffusion
and dislocation motions (e.g., cross-slip) around defects induce localized
microplasticity at a micro-level.
3. Temperature evolution in the specimen is due to several independent mecha-
nisms, such as internal atomic-friction-generated heat, heat conduction,
microplastic work, and thermoelastic damping.
I. Entropy Generation due to Heat Conduction

Temperature gradient causes an irreversible heat flow across the specimen during
fatigue loading. The entropy change is given by
Z t
∇T ∙ ∇T
t0 T2
where kh is coefficient of heat conduction and ∇T is the temperature gradient from

the gage section to the grip section of the tested sample, respectively. T are the
absolute temperature at the gage center.
In Eq. (9.32), the amount of entropy generation due to heat conduction is strongly
dependent on the test duration. To calculate the order of magnitude of this entropy
generation mechanism for ultrasonic vibration fatigue, the following boundary
conditions are used:
According to measurements during ultrasonic vibration testing, the temperature at
the gage section of the steel [EN 1.0577] test specimen raises from 300 K to about
450 K, while the nominal normal stress amplitude is 400 MPa. The thermal
conductivity of the material is kh = 50WK-1 m-1. Assume the temperature varies
parabolically along the centerline of the sample from the gage section to the
grip section, as T = C1x2 + C2x + C3. If the boundary conditions are set to be
Tgage(t) = T(x = 0, t), Tgrip(t) = T(x = ± L/2, t) where L2 = 30 mm is the length
between gage and grip sections. The one-dimensional thermal gradient
that changes
as the number of cycles increases can be obtained as ∇T = L8x2 T grip ðt Þ - T gage ðt Þ.
II. Entropy Generation due to Microplasticity
During very high cycle fatigue, microplasticity happens at defect sites. The concept
of microplasticity was introduced in the previous section of this chapter. Hence, we
just present the final equation. The entropy generation equation due to
microplasticity is given by
Z !
t
σ μ : ε_μP
ΔSμp = Ψf ϕf v dt ð9:33Þ
t0 T
Compared to the microplasticity-generated entropy formulation in the previous

section, Eq. (9.23), the new Eq. (9.33), has an extra term, the frequency coefficient
Ψf, which is necessary to relate the forced vibration frequency with the entropy
generation rate. It is well understood that the fatigue life measured for a specimen is
affected by the loading frequency. For example, under the same stress amplitude,
fatigue life is shorter when loading is at a 10–50 Hz frequency range, compared to
the fatigue life when the loading frequency range is 15–30 kHz. For FCC metals, it is
well-known that the difference in fatigue life is due to the occurrence of time-
dependent cross-slip of dislocations and vacancy production/diffusion involved in
the persistent slip band formation.
Marti et al. (2020) established a correlation between low- and high-frequency
fatigue life of pure polycrystalline copper with mechanisms of slip band formation.
The failure and early slip marking S-N curves they reported from experiments show
that the fatigue life is longer at higher frequencies, and the number of cycles for early
slip markings to emerge is also higher at “high” than “low” frequencies, for a given
stress amplitude. The similarity of these tendencies suggests a correlation between
both failure and early slip marking S-N curves. It is assumed that such a correlation is
a reasonable hypothesis because the formation and emergence of slip bands play a
key role in crack initiation and fatigue failure in ductile single-phase metals. The
initiation of persistent slip bands at the surface of fatigued specimens requires cross-
slip of screw dislocations to promote slip localization and irreversibility. In addition,
the generation of vacancies in persistent slip bands and their diffusion toward the
matrix favor their emergence. Marti et al. (2020) concluded that cross-slip activation
is the main mechanism that leads to early slip markings.
For the ultrasonic vibration of BCC metals, Torabian et al. (2017) reported that
there is a transition between material deformation mode from thermally activated
regime at stress amplitudes below the fatigue limit to athermal mode at stress
amplitudes above the fatigue limit. Since in this example the macroscopic stress
amplitude is above the fatigue limit, the athermal regime where the mobilities of
screw and edge dislocations are equivalent and screw dislocations can cross-slip is of
interest. The cross-slip probability P is given by Marti et al. (2020):

l δt τ - τcsR
P=β exp css V ð9:34Þ
l0 t 0 kB T
where β is a normalization coefficient ensuring that 0 < P < 1, l/l0 is the ratio of the
length of a screw dislocation segment to a reference length l0 = 1 mm, δt/t0 is the
ratio of the considered time to a reference time, V is the activation volume associated
with cross-slip, (τcss - τcsR) compares the resolved shear stress τcss on the cross-slip
system to a threshold critical stress τcsR required to activate cross-slip, kB is
Boltzmann’s constant, and T is the absolute temperature. The duration in which
τcss is greater than τcsR in one cycle is given by Marti et al. (2020):
0 1
τcsR
2 arcsin τcss
Δt 1cycle = @1 - A1 ð9:35Þ
π f
where f is the frequency of cyclic loading. Assuming all the parameters in Eq. (9.34)
are frequency insensitive, Eq. (9.35) shows that the time to activate cross-slip is
inversely related to the frequency. For two different loading frequencies, 30 Hz and
20 KHz at the same stress amplitude, the following equation must be satisfied to
obtain the same cumulated probability of cross-slip activation:
T A = Δt 1cycle,30 Hz × N 30 Hz = Δt 1cycle,20k Hz × N 20 kHz ð9:36Þ
where TA is the duration when τcss is greater than τcsR and N is the number of cycles.
According to Eqs. (9.35) and (9.36), at a given stress amplitude, the material
subjected to f1 loading frequency reaches the same cumulative probability of acti-
vating initial micro-defects (early slip marking) as the material under f2 loading
frequency when
N1 f 1
= ð9:37Þ
N2 f 2
where N1 is the number of cycles experienced under frequency f1 and N2 is the

number of cycles experienced under frequency f2. Test data indicates that the fatigue
life in a frequency range of 10–50 Hz is not strongly frequency sensitive. If we
set the operating frequency fr = 10~50 Hz as the reference frequency, and
fp = 15~30 KHz as the frequency during ultrasonic vibration testing, then Ψf can
be defined as Ψf = fr/fp, and Eq. (9.33) can be rewritten in the following
conditional form:
Z !
t
σ μ : ε_μP
t0 T
8
< 1, if f p < 100 Hz ðtypical servo - hydraulic machineÞ
Ψf = 100 ð9:39Þ
: , if f p is in the order of KHz ðultrasonic vibrationÞ
fp
The frequency coefficient Ψf = fr/fp is used to account for the fact that at higher
frequencies, more cycles are needed to activate the same volume of dislocation slip
bands. It indicates that at a given stress amplitude, under the same number of cycles,
the specimen produces lower cumulative microplasticity-generated entropy at high
frequency than at low frequency, because under the same number of cycles, the
probability of generating microplasticity is smaller at high frequency than at low
frequency. The high and low frequency mentioned here refer to the ultrasonic
vibration frequency and traditional servo-hydraulic machine frequency, respectively.
Equation (9.39) is only applicable to up to the frequencies in the order of kHz,
because at higher frequencies (e.g., MHz or GHz [ultrahigh cycle fatigue]), different
micro-mechanisms and fatigue failure occur (Marti et al., 2020; Torabian et al.,
2017). We should also point out that Eq. (9.39) is also applicable only for the class of
materials where dislocation slip band formation dominates the crack initiation
process (Mughrabi, 2002).
In the literature, many micromechanics models for very high cycle fatigue life
prediction utilize cyclic irreversible slip, which is equivalent to fv in the formulation
presented above.
III. Entropy Generation due to Atomic-Friction (Scattering)-Generated Heat
The entropy change due to internal atomic-friction (scattering)-generated heat is
given by
Z t
ρr
ΔSr = dt ð9:40Þ
t0 T
At the atomic level, the temperature is defined as the intensity of atomic vibra-
tions. The internally generated heat per unit mass, r, is the increasing intensity of
atomic vibrations in the lattice due to events such as the breaking of atomic bonds,
phonon-phonon scattering, phonon-electron scattering, electron scattering (referred
to as internal friction here), and dislocation motions (Gao, 1997; Ragab & Basaran,
2009; Chu et al., 2015). We should emphasize that this term is different than the heat
generated through the macroscopic observable plastic work or thermoelastic sources.
In the following section, the equation governing internal heat generation is

derived analytically. It has been shown that the rate of internal heat generation is
composed of two mechanisms:
1. The drag (friction due to internal scattering) process involves phonon drag,
electron drag, and radiation drag (Blaschke et al., 2020; De Hosson et al., 2001;
Galligan et al., 2000). The internal heat generation equation for this mechanism is
given by
ρr drag = ϱBdrag v v ð9:41Þ
2. The dislocation motion during the plastic deformation (Huang et al., 2008; Parvin
& Kazeminezhad, 2016). The internal heat generation equation for this mecha-
nism is given by
1 2 pffiffiffi
ρrdis = μb ϱ_ - αH μb ϱγ_ ð9:42Þ
2
where ρ is the mass density and Bdrag and v represent the effective drag coefficient
and velocity of dislocation, respectively. The terms μ, b, ϱ, and αH are the shear
modulus, the magnitude of Burger’s vector, total dislocation density, and Taylor’s
hardening parameter, respectively. In Eq. (9.41), the velocity of dislocation is related
to the applied shear strain rate, γ_ , as follows:
γ_
v= ð9:43Þ
ϱm b
where ϱm is the density of mobile dislocations. Hence, the entropy change due to
internal friction can be given by
Z Z t drag pffiffiffi
t
ρr ϱB v v þ 12 μb2 ϱ_ - αH μb ϱγ_
ΔSr = dt = dt ð9:44Þ
t0 T t0 T
Using Eq. (9.44), the evolution of ΔSr for the specimen under ultrasonic vibration
with 400 MPa stress amplitude is shown in Fig. 9.20.
Thermodynamic Fundamental Equation
Finally, the thermodynamic fundamental equation can be given in the
following form:
Z t Z t !
∇T ∙ ∇T σ μ : ε_μP
Δs = - kh dt þ Ψf ϕf v dt
t0 ρT 2 t0 ρT
Z t drag pffiffiffi
ϱB v v þ 12 μb2 ϱ_ - αH μb ϱγ_
þ dt ð9:45Þ
t0 ρT
To calculate the TSI evolution versus the number of cycles, material constants
need to be obtained beforehand. The temperature evolution in the sample is com-
puted using a fully coupled thermal-mechanical equation. To verify the computed
temperature from the fully coupled thermal-mechanical equation, a thermographic
camera is used to record the temperature profile of the specimen focusing on the gage
center during fatigue testing.
Modal Analysis
An ultrasonic vibration fatigue test is usually carried out in the resonance frequency
of the sample in the axial direction. Therefore, modal analysis is necessary to
determine the exact natural frequencies of the tested specimen. To compute the
natural frequencies of the test specimen, usually finite element analysis is utilized.
The tension-compression mode [in the axial direction] resonance frequency of the
sample was determined to be 19,864 Hz.
Then, for the axial direction resonance frequency, the stress distribution in the
sample must be determined for selected levels of the vibration amplitude applied at
the support end of the sample. A linear relationship between the vibration amplitude
and the average stress in the center of the sample is obtained:
σ = 20,510 A ð9:46Þ
where σ is the average normal stress in the center of the sample in MPa and A is the
vibration amplitude in mm. Finite element analysis results indicate that stress is
greater on the surface of the specimen and smaller in the center. These stresses are in
the range of about ±1% of the average stress in the narrow gage section.
The average stress is σ = 380 MPa for vibration amplitude of A = 0.0185 mm
(Fig. 9.17).
It is shown that a temperature gradient between the specimen’s gage section and
grip section increases slowly in phase 2 and then rises rapidly when a macro crack
forms in phase 3.
Thermodynamic State Index (TSI) Computation
Material properties used in the example are given in Table 9.5.
The microscopic yield stress is the mean fatigue limit of the specimen; it is
assumed that no microplasticity occurs below the macroscopic fatigue limit.
It should be noted that the microplasticity model in this subsection is based on the
following assumptions:
1. The hardening behavior of the material is approximated by a bilinear kinematic
hardening. Hardening modulus is defined as the slope between (lower yield
strength) LYS and UTS (ultimate tensile strength) from the test stress-strain
curve, as a first approximation.
2. The microscopic hysteresis loops are stabilized and they are the same at each
defect site. The evolution of TSI, ϕ, from 0 to 1 only induces microplasticity at
more defect sites; however, they never coalesce until the final rupture.
Fig. 9.17 (a) Sample dimensions along the axis, (b) normal stress distribution along with the
sample, (c) displacement distribution along the specimen axis for average stress of σ = 380 MPa.
(After Lee et al. (2022))
Table 9.5 Material parameters for EN 1.0577 steel

Young’s modulus 197,265 MPa Macroscopic yield stress 400 MPa
Hardening coefficient 879 MPa Microscopic yield stress 368 ± 3.3 MPa
[fatigue limit]
Poisson’s ratio 0.27 Thermal expansion 12.610-6 K-1
coefficient
Mass density 7820 kg m-3 Specific heat capacity cV 470 J kg-1 K-1
The volume fraction of active 20% Frequency coefficient Ψf 510-3
defects fv
Critical thermodynamic state ϕcr = 0.97
index
After Lee et al. (2022)
Computing Internal Heat Generation

It has been shown that during the deformation of iron and iron alloys, the proportion
of the main types of activated dislocations is approximately 60% of 12 ah1 1 1i, 20%
of ah1 0 0i, and 20% of ah1 1 0i, which are unaffected by the temperature of
deformation, alloy content, amount of strain, dislocation density, or rate of straining
(Dingley & Hale, 1966). The magnitude of Burger’s vectors can be calculated by

a pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi
kbk = h2 þ k 2 þ l2 For BCC and FCC lattice
2 ð9:47Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
k k
b = ð a Þ h2 þ k 2 þ l2 For simple cubic lattice
where kbk is the magnitude of Burger’s vector, a is the lattice constants in units of an
angstrom (Å), and h, k, l are the components of Burger’s vector.
The parameters in Eqs. (9.41)–(9.43) can be obtained by discrete dislocation
dynamics. In the literature, these values are available for the material used in this
example, as follows:
1. Burger’s vector values are provided in Dingley and Hale (1966). The terms
related to Burger’s vector are summed according to their proportion.
2. Taylor’s hardening parameter αH is a dimensionless parameter ranging from 0.05
to 2.6. It is used to characterize the relation between the plastic flow stress of a
pffiffiffi
material and its dislocation density, τ = τ0 þ αμb ρ . Taylor’s parameter from
Davoudi and Vlassak (2018) is used, since it has a similar stress-strain curve as
the material of interest in this example.
3. The initial dislocation density value for mild steel is given by Bin Jamal
et al. (2021).
4. Moving dislocations experience a drag due to their interaction with the crystal
structure, and this drag coefficient Bdrag determines the dislocation glide time
between obstacles (Blaschke, 2019; Blaschke et al., 2019). In terms of the
dominant drag mechanism, it is assumed that the phonon scattering with dislo-
cations [phonon-defect scatter] constitutes the dominant mechanism. Bdrag value
is provided by Blaschke et al. (2019).
5. The evolution of dislocation density during plastic deformation has been exten-
sively studied in the literature. It is widely accepted that when nominal macro-
stress is well below the yield stress, the microplastic deformation that results in
stabilized hysteresis micro-stress-strain loops happens only at micro-defect sites.
Therefore, a constant dislocation generation rate Δϱ at each defect site is
assumed.
The material parameters needed in Eq. (9.45) are summarized in Table 9.6.
Figure 9.19 shows entropy generation, Eq. (9.44), due to internal heat generation.
Table 9.6 Parameters for mild steel EN 1.0577

Parameter Symbol Value Units
Lattice constants a 2.856 10-10 m
The magnitude of Burger’s vector b 60% of 0.87a Angstrom
20% of 1.00a
20% of 1.41a
Shear modulus μ 7.76 × 104 MPa
Taylor’s hardening parameter αH 1 –
Initial dislocation density ϱ 0.88 × 1014 m/m3
Drag coefficient due to transverse phonon drag Bdrag 0.02 mPa-s
Dislocation density increment in each cycle Δϱ 1.523 × 1014 m/m3
9.2.2 Temperature Evolution due to Mechanical Work
The temperature increase due to mechanical work is governed by the fully coupled
thermomechanical equilibrium equation given below:

_ ∂σ e ∂Ak _
ρcV T - k∇ T = σ : ε_ þ T
2 p
: ε_ þ ρr þ T Ak - Vk ð9:48Þ
∂T ∂T
where the first term is the rate of heat storage, the second term is the heat conduction,
the third term is the energy dissipation due to macroplastic work, the fourth term is
the thermoelastic coupling source, and the fifth term is the heat generation due to
internal frictions, and the last term is non-recoverable energy corresponding to the
∂s
internal source. cV = T ∂T is defined as specific heat capacity; εe and εP are elastic
and plastic strain tensors, respectively; σ is the stress tensor; r is the strength per unit
mass of the internal distributed heat generation; and Ak is a thermodynamic force
associated with the internal thermodynamic variable, Vk.
Equation (9.48) yields the evolution of temperature due to mechanical work with
properly imposed boundary conditions. To simplify this equation, we can ignore
plastic strain ε_ p and other thermodynamic variables Ak V_ k . These assumptions are
valid, because
1. We are investigating mechanical response under elastic loads only, as such, there
is no uniform plastic strain at the macro-level. However, the contribution from
microplasticity is not neglected.

2. Ak - T ∂Ak V_ k represents the non-recoverable energy corresponding to the
∂T
internal sources (e.g., sound waves, grain coarsening, phase transformation,
etc.). However, for metals, this non-recoverable energy only represents 5% or
less of the total dissipation.
Hence, Eq. (9.48) can be rewritten into the following form:

h i pffiffiffi
ρcV T_ - k∇2 T = Ψf ϕf v σ μ : ε_μP þ ϱBdrag v v þ μb2 ϱ_ - αH μb ϱγ_
1
ð9:49Þ
2
where ρ is the mass density and Bdrag and v represent the effective dislocation-drag
coefficient and velocity of dislocation, respectively. The terms kh, ϕ, and cV are
coefficient of heat conduction, the thermodynamic state index, and the specific heat
capacity, respectively. The relative surface ratio covered by activated micro slip
bands is represented by fv. The terms μ, b, ϱ, and αH are the shear modulus of the
bulk material, the magnitude of Burger’s vector, total dislocation density, and
Taylor’s hardening coefficient, respectively.
If we ignore the thermal fluctuation due to thermoelastic dissipation and assume
that the temperature field is uniform, we can simplify the conduction term
as - k h ∇2 T ffi ρcV τθeq (Torabian et al., 2017).
Where the terms, θ, and τeq are the total change in temperature (Tn - T0) and an
equivalent conduction time used to characterize the heat transfer between the
specimen and the surroundings (Charkaluk & Constantinescu, 2009; Zhang et al.,
2013; Chrysochoos et al., 2012).
Then, the following equation is obtained:
h i
ðT - T 0 Þ
= Ψf ϕf v σ μ : ε_μP þ ϱBdrag v ∙ v þ μb2 ϱ_
1
ρcV T_ þ n
τeq 2
pffiffiffi
- αH μb ϱγ_ ð9:50Þ
The time discretization of the above equation is given by
μ
Δt Δt Δt γ_ p nþ1 2
ΔT nþ1 = - ðT n - T 0 Þ þ Ψf ½ϕf v ðσ μ : ˙εμP Þ þ ϱBdrag
τeq ρcV ρcV ϱm b
Δt 1 2 Δϱnþ1 Δt pffiffiffi
þ μb - α μb ϱγ_ = 0:
ρcV 2 Δt ρcV H
ð9:51Þ
After rearranging the terms in Eq. (9.51), we get the following equation for
calculating the change in temperature:
h i μ
Ψf γ_ p nþ1 2
ϕf v σ μ nþ1 ε_μP Δt þ
1
ΔT nþ1 = - t eq ðT n - T 0 Þ þ ϱBdrag Δt
ρcV ρcV ϱm b
1 1 2 1 pffiffiffi
þ μb Δϱnþ1 - α μb ϱγ_ μp Δt
ρcV 2 ρcV H
ð9:52Þ
Fig. 9.18 Measured surface temperature at the gage center of the 11 monitored samples versus the
number of cycles (in millions) during fatigue tests. (After Lee et al. (2022))
Δt
In Eq. (9.52) dimensionless time factor, t eq = τeq , represents the fraction of heat
conducted. Equivalent conduction time, τeq, the value measured and reported by
Charkaluk and Constantinescu (2009), is used in this example.
Temperature evolution in the material is calculated at each cycle using Eq. (9.52).
It should be emphasized that Eq. (9.52) is used to simulate the phase 1 and phase
2 temperature evolution. Equation (9.52) cannot capture the abrupt temperature rise
in phase 3 when a macro crack initiates, because the mechanisms of macroplasticity,
macro crack formation, and crack propagation are not included in the formulation.
Figure 9.18 shows the measured surface temperature at the gage center versus the
number of cycles, and Fig. 9.19 shows the simulations for the same based on
Eq. (9.52).
Fig. 9.19 Surface temperature predictions at gage center versus the number of cycles under various
stress amplitudes based on Eq. (9.52). Compared to test data (Fig. 9.18). (After Lee et al. (2022))
9.2.3 Comparison Between Different Entropy Generation

Mechanisms
The entropy generation due to heat conduction, microplasticity, and internal frictions
during phase 1 and phase 2 is obtained using Eq. (9.45). The measured temperature
rise during phase 1 (shown in Fig. 9.18, up to 2.5 × 105 cycles) matches the
predictions shown in Fig. 9.19. Some researchers reported a change in temperature
in phase 1, ΔTphase1, to be a constant value (Guo et al., 2015), analogous to the elastic
modulus of a material. The temperature rise during phase 2 follows a parabolic
function. When the simulation results shown in Fig. 9.19 are compared with test data
shown in Fig. 9.18 [before the onset of phase 3], we observe a small difference at the
end of phase 2. This discrepancy may be due to the uncertainty of the formation of
cross-slip mechanisms [which leads to microplasticity], introduced by the frequency
effects. However, the difference between the prediction and measured temperature is
very small for all practical purposes.
Figure 9.20 shows that the entropy generation due to microplasticity is small at
the beginning and starts to increase after a critical threshold. This is because the
evolution of activated micro-defects follows the TSI. As more micro-defects are
activated, more entropy is produced.
Fig. 9.20 Total entropy generation (MJK-1 m-3) versus the number of cycles under 400 MPa
nominal normal stress, for different mechanisms. (After Lee et al. (2022))
The order of magnitude of each entropy generation mechanism is summarized in

Table 9.7
From Table 9.7, it is obvious that during ultrasonic vibration, very high cycle
fatigue, the entropy generation due to thermal conduction, atomic-friction-generated
heat, and microplasticity are about the same order of magnitude. Still,
microplasticity is the main contributor. Therefore, none of these three entropy
generation mechanisms should be ignored. The thermodynamic state index can
then be given by
2 0 Z t Z t drag 1 2 pffiffi Z t 13
ðB v vþ2μb ϱ_ - αH μb ϱγ_ Þ σ μ :ε_μ
ms k h ∇TT∙2∇T dt - dt - Ψf ϕf v T P dt
6 B T C7
ϕ = ϕcr 6 B
41 - exp @
t0 t0 t0 C7
A5
ρR
ð9:53Þ
Table 9.7 Order of magnitude of different entropy generation mechanisms

Mechanism The magnitude of entropy generation
1 Thermal conduction ≈0.25 MJK-1 m-3
2 Internal friction ≈0.85 MJK-1 m-3
3 Microplasticity ≈3 MJK-1 m-3
Equation (9.53) never reaches one, because it is an exponential function. In

practice, we determine a critical value of TSI ϕcr as a threshold that we consider
the specimen is failed taking into account computational time. In this example,
ϕcr = 0.97 was chosen as a critical value.
9.2.4 Computing Thermodynamic State Index
The computation of TSI requires a summation of the entropy generation for all
mechanisms. The material properties necessary for calculating TSI are presented in
Tables 9.5 and 9.6. The material properties can be regarded as constant between
25 and 150 °C based on test data for EN 1.0577 steel, which shows a small change in
material properties in this temperature window. The temperature in the specimen
during the ultrasonic vibration test does not exceed 130 °C (around 400 K) before
rapid temperature rise at phase 3 when a macro crack initiates. Therefore, material
constants are used without a temperature correction.
Using Eq. (9.53), we can compute the TSI evolution for EN 1.0577 steel
operating at 20 kHz. The entropy generation and TSI evolution for different stress
amplitudes are shown in Figs. 9.21 and 9.22, respectively. The cumulative specific
entropy at failure [fatigue fracture entropy] is validated to be a material constant
independent of stress amplitude, specimen’s dimensions and geometry, and loading
frequency. It is commonly used as a damage indicator to predict the number of cycles
to failure. Figure 9.21 shows that fatigue fracture entropy (FFE) is the same
(a material constant) regardless of the stress level during testing. This is similar to
test results reported widely by Naderi et al. (2010), Yun and Modarres (2019), and
Ribeiro et al. (2020). Figure 9.22 shows the TSI evolution between 0 and 1. The
evolution of TSI is based on the cumulative specific entropy of the system; therefore,
for the same critical TSI (i.e., ϕcr = 0.999), the cumulative specific entropy will be
the same regardless of the stress amplitude, frequency of loading, or geometry of the
sample (Fig. 9.21, Fig. 9.23).
The comparison of the unified mechanics theory-based model simulation S-N
curve and the experimental S-N curve is presented in Fig. 9.24.
From Fig. 9.21, it is observed that the cumulative entropy generation for 1.0577
steel under various stress amplitudes is the same (4.1 MJK-1 m-3) at failure. This
value is a material constant regardless of the test frequency, loading path, loading
type, geometry of the specimen, or stress state. However, the time it takes to reach
the fatigue fracture entropy will be different in each case.
Fig. 9.21 Entropy generation versus cycles for 1.0577 steel under various stress amplitudes. (After
Lee et al. (2022))
A unified mechanics theory-based model is formulated for predicting the very

high cycle fatigue life of EN 1.0577 steel under ultrasonic vibration operating at
20 kHz. The thermodynamic fundamental equation under ultrasonic fatigue is
derived. Contributions of different mechanisms to the total entropy generation are
compared. The entropy generations due to microplasticity, thermal conduction, and
atomic-friction-generated heat are the dominant entropy generation mechanisms.
Microplasticity, which makes the biggest contribution to entropy generation, can
be visualized as having some microplastic inclusions in an elastic matrix. Energy is
dissipated through the microplastic work at these inclusions. The number of inclu-
sions increases as TSI, ϕ, increases from zero to the predefined ϕcr critical vale of
TSI. The evolution trend of numbers of inclusions leads to very little entropy
generation in the beginning, because in the beginning, very few inclusion sites are
activated. However, after cycles accumulate, the number of inclusions experiencing
microplasticity gradually increases and entropy generation due to microplasticity
becomes more evident.
To obtain the absolute temperature of the specimen needed in the thermodynamic
fundamental equation without test data, a fully coupled thermomechanical equation
that includes a heat storage term, a heat conduction term, a microplasticity term, and
Fig. 9.22 TSI evolution versus cycles for 1.0577 steel under various stress amplitudes. (After Lee
et al. (2022))
a self-heating term due to internal frictions is derived. However, the derived thermal-
mechanical equation cannot capture the abrupt temperature rise in phase 3 when
macro cracks initiate because the mechanisms of macroplasticity, crack formation,
and crack propagation are not included in the model.
Very high cycle fatigue life prediction results based on the unified mechanics
theory are compared with experimental test data. The comparison shows that the
prediction and test data match very well.
9.3 Predicting Low Cycle Fatigue Life
In low cycle fatigue, where the nominal stress is above the yield limit, the plastic
work is the dominant dissipation mechanism. Entropy generation in the plastic
dissipation process is given by
Z t2
1
Δs = σ : dεp ð9:54Þ
ρT t1
9.3 Predicting Low Cycle Fatigue Life 497
Fig. 9.23 Unified mechanics theory simulated S-N data points for 1.0577 steel under various stress
amplitudes. (After Lee et al. (2022))
where ρ is the mass density of the material and σ and ε p are the stress and plastic
strain, respectively. T represents the temperature. Integral limits t1 and t2 represent
the time bounds of the mechanical loading. For one-dimensional case, the total
plastic strain, ε p(t), is calculated as follows:
σ ðt Þ
εp ðt Þ = εtotal ðt Þ - ð9:55Þ
E
where εtotal(t) is the total strain at time t and σ(t) and E are to the stress at time t and
Young’s modulus, respectively.
The true stress-strain graph published by Carrion et al. (2017) for Ti-6Al-4V
alloy, is used for the comparison between experimental data and the numerical
predictions by Bin Jamal et al. (2020). Material constants used are given in
Table 9.8. Assuming that the gage section of a dog bone AST standard sample
experiences uniform strain, 5 mm section was considered in the computational
model to reduce the computational cost. The diameter of the specimen is 6.35 mm.
In ABAQUS, linear brick elements, C3D8R are used. One end of the sample is
defined with a zero displacement (fixed) boundary condition and the other end is
Fig. 9.24 UMT simulated S-N data points for 1.0577 steel compared with experimental data. The
blue line and squares are simulated results from UMT. (After Lee et al. (2022))
Table 9.8 Material parame- Material parameter Value Unit

ters used in the numerical
Young’s modulus, E 106 GPa
model for tensile loading in
Ti-6Al-4V alloy Poisson’s ratio, ν 0.31
Mass density, ρ 4540 kg/m3
Critical TSI, Φc 1
Hardening parameter, K 968.00 MPa
Hardening exponent, r 0.64
Yield strength, σ y0 992.00 MPa
Molar mass, ms 0.047867 kg/mol
The reference temperature, T 298 K
Bin Jamal et al. (2020)
subjected to controlled displacement loading in the axial direction. After a mesh

convergence analysis, an optimum seed size of 0.9 mm is fixed for all the simula-
tions. Figures 9.25 and 9.26 show a comparison between the test data and numerical
simulation for monotonic tension and compression experiments, respectively. Fig-
ure 9.27 depicts simulation results for engineering stress-strain hysteresis loops for
1.2% strain amplitude. Figure 9.28 presents TSI evolution for different strain
amplitudes during cyclic loading. Figure 9.29 shows a comparison between the
test data and computational predictions for low cycle fatigue life at different strain
amplitudes.
9.3 Predicting Low Cycle Fatigue Life 499
1.2x109
A
9
1.0x10
True Stress (Pa)

1.1x109
1.0x109
True Stress (Pa)
1.0x109
8.0x108 1.0x109
9.8x108
9.6x108
9.4x108
9.2x108
6.0x108 9.0x108
0.007 0.008 0.009 0.010 0.011 0.012
Strain
4.0x108 Enlarged view of A
Test data
2.0x108
Numerical simulation
0.0
0.00 0.01 0.02 0.03
Strain
Fig. 9.25 Comparison between monotonic tensile stress-strain graphs obtained from the test data
(Carrion et al., 2017) and numerical simulation (Bin Jamal et al., 2020)
1.5x109
1.2x109 B
1.12x109
True Stress (Pa)
1.10x109
True Stress (Pa)
8 1.08x109
9.0x10 1.06x109
1.04x109
1.02x109
1.00x109
0.010
0.015
0.020
0.025
8
6.0x10
Strain
Enlarged view of B
3.0x108
Test data
Numerical (Seed size - 0.64 mm)
0.0 Numerical (Seed size - 0.85 mm)
Numerical (Seed size - 1.0 mm)
0.0 0.1 0.2 0.3

Strain
Fig. 9.26 Comparison between test data (Carrion et al., 2017) and numerical simulation (Bin Jamal
et al., 2020), for monotonic compressive stress-strain
1.5x109 1.5x109
Hysteresis loops at 1.2% strain amplitude Hysteresis loops at 1.2% strain amplitude
1.0x109 1.0x109
Load cycle-1 Load cycle-1

5.0x108 5.0x108
Stress (Pa)
Stress (Pa)
0.0 0.0
-5.0x108 -5.0x108
Load cycle-50 Load cycle-50

-1.0x109 -1.0x109
-1.5x109 -1.5x109
-0.015 -0.010 -0.005 0.000 0.005 0.010 0.015 -0.015 -0.010 -0.005 0.000 0.005 0.010 0.015
Strain Strain
a b
Fig. 9.27 Predictions for engineering stress-strain hysteresis loops for 1.2% strain amplitude. (a)
Hysteresis loops at 1.2% strain amplitude for 50 cycles; (b) comparative hysteresis plot for the first
cycle and 50th cycle (Bin Jamal et al., 2020)
1.0
Damage (TSI)
0.5
1.2% Strain
0.0 1% Strain
0.8% Strain
0
0
0
00
00
00
00
00
00
00
00
00
00
00
00
00
-50
50
10
15
20
25
30
35
40
45
50
55
60
65
70
Number of Cycles
Fig. 9.28 Predicted damage for different strain amplitudes during cyclic loading (Bin Jamal et al.,
2020)
Here we conclude our discussion on fatigue life prediction with the unified
mechanics theory. Further details may be found in the references cited below.
References 501
0.024
0.022 Test data

Total strain amplitude Analytical approach: 1-D
0.020 Numerical approach: 3-D
0.018
0.016
0.014
0.012
0.010
0.008
0 2000 4000 6000 8000

Nf
Fig. 9.29 Comparison between the test data (Carrion et al., 2017) and predictions (Bin Jamal et al.,
2020) for low cycle fatigue life (Nf) at different strain amplitudes
References
Abbaschian, R., & Reed-Hill, R. E. (2009). Physical metallurgy principles. Van Nostrand, ISBN
9780495438519.
Atkins, P., Paula, J., et al. (2018). Physical chemistry (11th ed.). Oxford University Press, ISBN
9780198769866.
Basaran, C., & Lin, M. (2007). Electromigration induced strain field simulations for nanoelectronics
lead-free solder joints. International Journal of Solids and Structures, 44(14–15), 4909–4924.
Basaran, C., & Lin, M. (2008). Damage mechanics of electromigration induced failure. Mechanics
of Materials, 40(1–2), 66–79.
Bin Jamal, M. N., Kumar, A., Lakshmana Rao, C., & Basaran, C. (2020). Low cycle fatigue life
prediction using unified mechanics theory in Ti-6Al-4V alloys. Entropy, 2020(22), 24.
Bin Jamal, M. N., Rao, C., & Basaran, C. (2021). A unified mechanics theory-based model for
temperature and strain rate dependent proportionality limit the stress of mild steel. Mechanics of
Materials, 155, 103762.
Blaschke, D. (2019). Properties of dislocation drag from phonon wind at ambient conditions.
Materials, 12(6), 948.
Blaschke, D., Mottola, E., & Preston, D. (2019). Dislocation drag from phonon wind in an isotropic
crystal at large velocities. Philosophical Magazine, 100(5), 571–600. https://doi.org/10.1080/
14786435.2019.1696484
Blaschke, D. N., Mottola, E., & Preston, D. L. (2020). Dislocation drag from phonon wind in an
isotropic crystal at large velocities. Philosophical Magazine, 100, 571–600.
Burton, J. J. (1972). Vacancy-formation entropy in cubic metals. Physical Review B, 5, 2948.
Charkaluk, E., & Constantinescu, A. (2009). Dissipative aspects in high cycle fatigue. Mechanics of
Materials, 41(5), 483–494, ISSN 0167-6636.
Chrysochoos, A., Boulanger, T., & Morabito, A. (2012). Dissipation and thermoelastic coupling
associated with fatigue of materials. Mechanics, Models, and Methods in Civil Engineering,
147–156. https://doi.org/10.1007/978-3-642-24638-8_7
Chu, Y., Gautreau, P., Ragab, T., & Basaran, C. (2015). Temperature dependence of Joule heating
in zigzag graphene nanoribbon. Carbon, 89, 169–175.
Cugy, P., & Galtier, A. (2002). Microplasticity and temperature increase in low carbon steels. In
A. F. Blom (Ed.), Proceedings of the 8th international fatigue congress, 3–7 June 2002
(pp. 549–556). EMAS, Barnsley.
Davoudi, K., & Vlassak, J. (2018). Dislocation evolution during plastic deformation:
Equations vs. discrete dislocation dynamics study. Journal of Applied Physics, 123(8),
085302. https://doi.org/10.1063/1.5013213
De Hosson, T. M., Roos, A., & Metselaar, E. D. (2001). Temperature rise due to fast-moving
dislocations. Philosophical Magazine A, 81(5), 1099–1120.
Dingley, D., & Hale, K. (1966). Burgers vectors of dislocations in deformed iron and iron alloys.
Proceedings of the Royal Society of London Series A. Mathematical and Physical Sciences,
295(1440), 55–71. https://doi.org/10.1098/rspa.1966.0225
Doudard, C., Calloch, S., Cugy, P., Galtier, A., & Hild, F. (2005). A probabilistic two-scale model
for high-cycle fatigue life predictions. Fatigue & Fracture of Engineering Materials & Struc-
tures, 28(3), 2005.
Fan, J., Zhao, Y., & Guo, X. (2018). A unifying energy approach for high cycle fatigue behavior
evaluation. Mechanics of Materials, 120, 15–25, ISSN 0167-6636.
Fultz, B. (2010). Vibrational thermodynamics of materials. Progress in Materials Science, 55(4),
247–352.
Galligan, J., McKrell, T., & Robson, M. (2000). Dislocation drag processes. Materials Science And
Engineering: A, 287(2), 259–264.
Gao, S. (1997). Quantum kinetic theory of vibrational heating and bond breaking by hot electrons.
Physical Review B, 55(3), 1876. https://doi.org/10.1103/PhysRevB.55.1876
Guo, Q., Guo, X., Fan, J., Syed, R., & Wu, C. (2015). An energy method for rapid evaluation of
high-cycle fatigue parameters based on intrinsic dissipation. International Journal of Fatigue,
80, 136–144. https://doi.org/10.1016/j.ijfatigue.2015.04.016
Ho, H. S., Risbet, M., & Feaugas, X. (2017). A cyclic slip irreversibility based model for fatigue
crack initiation of nickel base alloys. International Journal of Fatigue, 102, 1–8, ISSN
0142-1123.
Huang, M., Rivera-Díaz-Del-Castillo, P. E. J., Bouaziz, O., & Van Der Zwaag, S. (2008). Irrevers-
ible thermodynamics modeling of plastic deformation of metals. Materials Science and Tech-
nology, 24, 495–500. https://doi.org/10.1179/174328408X294125
Kelly, A., & Knowles, K. (2012). Crystallography and crystal defects. Wiley, ISBN
9780470750148.
Kröner, E. (1961). Zur plastischen verformung des vielkristalls. Acta Metallurgica, 9(2), 155–161.
Laughlin, D., & Hono, K. (2014). Physical metallurgy (5th ed.). Elsevier, ISBN 9780444537713.
Lee, H. W., & Basaran, C. (2021). A review of damage, void evolution, and fatigue life prediction
models. Metals, 11(4), 609.
Lee, H. W., & Basaran, C. (2022). Predicting high cycle fatigue life with unified mechanics theory.
Mechanics of Materials, 164, 104116.
Lee, H. W., Basaran, C., Egner, H., Lipski, A., Piotrowski, M., Mroziński, S., Noushad Bin Jamal,
M., & Rao, C. L. (2022). Modeling ultrasonic vibration fatigue with unified mechanics theory.
International Journal of Solids and Structures, 236–237, 111313.
Lemaitre, J., Sermage, J., & Desmorat, R. (1999). A two scale damage concept applied to fatigue.
International Journal of Fracture, 97, 67.
Li, S., Abdulhamid, M. F., & Basaran, C. (2009). Damage mechanics of low temperature
electromigration and thermomigration. IEEE Transactions on Advanced Packaging, 32(2),
478–485.
Marti, N., Favier, V., Gregori, F., & Saintier, N. (2020). Correlation of the low and high frequency
fatigue responses of pure polycrystalline copper with mechanisms of slip band formation.
Materials Science and Engineering: A, 772, 138619, ISSN 0921-5093.
Mughrabi, H. (2002). On ‘multi-stage’ fatigue life diagrams and the relevant life-controlling
mechanisms in ultrahigh-cycle fatigue. Fatigue & Fracture of Engineering Materials & Struc-
tures, 25(8–9), 755–764.
References 503
Mishin, Y., Sørensen, M. R., & Voter, A. F. (2001). Calculation of point-defect entropy in metals.
Philosophical Magazine A, 81(11), 2591–2612.
Mughrabi, H. (2009). Cyclic slip irreversibilities and the evolution of fatigue damage. Metallurgical
and Materials Transactions A: Physical Metallurgy and Materials Science, 40(6), 1257–1279.
fracture. Proceedings of the Royal Society A, 466, 423.
Parvin, H., & Kazeminezhad, M. (2016). Modeling the temperature rise effect through high-
pressure torsion. Materials Science and Technology (United Kingdom), 32, 1218–1222.
Patricio E. Carrion, Nima Shamsaei, Steven R. Daniewicz, Robert D. Moser. (2017). Fatigue
behavior of Ti-6Al-4V ELI including mean stress effects, International Journal of Fatigue, 99,
87–100.
Ragab, T., & Basaran, C. (2009). Joule heating in single-walled carbon nanotubes. Journal of
Applied Physics, 106, 063705.
Ribeiro, P., Petit, J., & Gallimard, L. (2020, July). Experimental determination of entropy and
exergy in low cycle fatigue. International Journal of Fatigue, 136, 105333.
Torabian, N., Favier, V., Dirrenberger, J., Adamski, F., Ziaei-Rad, S., & Ranc, N. (2017). Corre-
lation of the high and very high cycle fatigue response of ferrite based steels with strain rate-
temperature conditions. Acta Materialia, 134, 40–52, ISSN 1359-6454.
Wollenberger, H. J. (1996). Physical metallurgy (4th ed.). Elsevier, ISBN 9780444898753.
Yao, W., & Basaran, C. (2013a). Electromigration damage mechanics of lead-free solder joints
under pulsed DC: A computational model. Computational Materials Science., 71, 76–88.
Yao, W., & Basaran, C. (2013b). Computational damage mechanics of electromigration and
thermomigration. Journal of Applied Physics, 114, 103708.
Yao, W., & Basaran, C. (2013c). Damage mechanics of electromigration and thermomigration in
lead-free solder alloys under alternating current: An experimental study. International Journal
of Damage Mechanics, 23, 203–221.
Ye, H., Basaran, C., & Hopkins, D. C. (2004). Deformation of solder joint under current stressing
and numerical simulation––I. International Journal of Solids and Structures, 41(18–19),
4939–4958.
Ye, H., Basaran, C., & Hopkins, D. C. (2006). Experimental damage mechanics of micro/power
electronics solder joints under electric current stresses. International Journal of Damage
Mechanics, 15(1), 41–67.
Zhang, L., Liu, X., Wu, S., Ma, Z., & Fang, H. (2013). Rapid determination of fatigue life based on
temperature evolution. International Journal of Fatigue, 54, 1–6. https://doi.org/10.1016/j.
ijfatigue.2013.04.002
Chapter 10
Corrosion-Fatigue Interaction
10.1 Corrosion
In this section, first, we summarize the basics of the metal corrosion process. The
information presented below is a summary of the corrosion chapter in Callister and
Rethwisch (2014).
Metals experience chemical interaction with their environment. As a result of this
interaction, they experience material loss by dissolution and by the formation of
nonmetallic scale or film.
Corrosion is an electrochemical process, where there is a transfer of electrons
from one material to another. Metals have a metallic bond where positive ions float
in a sea of free negatively charged valence electrons. During the corrosion process,
metals give up electrons in what is called an oxidation reaction. For example, an iron
atom (Fe) has two valence electrons in its electronic configuration
[1s22s22p63s23p63d64s2]. Fe can experience oxidation by giving up these two
valence electrons; as a result, it becomes (2+) a positively charged ion. This can
be represented in chemistry notation as follows:
yields
Fe → Fe2þ þ 2e - ð10:1Þ
where e- symbolizes an electron. The site where oxidation happens is called the
anode. The term anodic reaction is also used to refer to the oxidation process. The
electrons given up by a metal must be physically transferred to other chemical
species, in what is termed a reduction reaction. Given up electrons become part of
the other chemical species. The site where reduction reaction occurs is called the
cathode.
An electrochemical reaction contains one oxidation reaction (anodic site) and a
reduction reaction (cathodic site). The total rate of oxidation equals the total rate of
reduction, or, in other words, all electrons released by the oxidation site must be
absorbed by the reduction site. As an example, we can formulate rusting of iron, Fe,
https://doi.org/10.1007/978-3-031-18621-9_10
506 10 Corrosion-Fatigue Interaction
in water with dissolved oxygen, which takes place in two separate steps. Electron
configurations of hydrogen and oxygen are given by [1s1], [1s22s22p4].
Step 1—Fe is oxidized Fe2+, as Fe(OH)2, in chemistry notation:
1 yields yields
Fe þ O2 þ H2 O → Fe2þ þ 2OH - → FeðOHÞ2 ð10:2Þ
2
Step 2—To Fe3+ as Fe(OH)3 according to
1 yields
2FeðOHÞ2 þ O2 þ H2 O → 2FeðOHÞ3 ð10:3Þ
2
The compound Fe(OH)3 is called rust. Readers are referred to Callister and
Rethwisch (2014) for a more in-depth study of corrosion.

of Corrosion-Fatigue
Entropy is an additive property. Hence, we need to identify all entropy-generating

mechanisms involved.
10.2.1 Entropy Generation Mechanisms During Corrosion
I. Entropy Generation due to Electrochemical Reaction: Activation

Overpotential, ΔSact
Overpotential is the additional potential (beyond the thermodynamic equilibrium)
needed to drive an electrochemical reaction at a certain rate. Accordingly, activation
overpotential is the additional electric potential needed for the activation of electrode
reaction at the desired rate, which includes the activation of oxidation reaction at
anode and cathode, and the activation of reduction reaction at anode and cathode.
The thermodynamic fundamental equation [entropy generation mechanisms] for
metal corrosion has been extensively studied based on the theory of chemical
kinetics and fundamentals of irreversible thermodynamics. Entropy generation can
be computed by multiplying the thermodynamic flux (corrosion current density) and
thermodynamic force (electrochemical affinity) based on the Onsager reciprocal
relations. Entropy generation for activation overpotential can be given in the form
of the sum of the product of fluxes, Ji, and forces, Pi (Gutman, 1998):
1 Xm
ΔSact = J ∙ Pi
i-1 i
ð10:4Þ
T
10.2 Thermodynamic Fundamental Equation of Corrosion-Fatigue 507
1
~ þ J c - αc A
~

ΔSact = J a αa A ð10:5Þ
T

~
α A
J a = j0 exp a -1 ð10:6aÞ
RT

~
- αc A
J c = - j0 exp -1 ð10:6bÞ
RT
where Ja and Jc are the fluxes (current densities) of anodic flux (direct) and cathodic
flux (reverse) half-reactions, respectively. Ja and Jc are defined by Butler-Volmer
equation, which formulates the relationship between the electrical current passing
through an electrode and the voltage difference between the electrode and the
electrolyte, j0 is the exchange current in a dilute solution in the equilibrium state
~ = 0 and Ja + Jb = 0), and αa and αc are the charge transfer coefficients of the
(A
~ is the electrochemical
electrochemical reactions, anodic and cathodic, respectively. A
Pn
~= -
affinity of the total reaction which is expressed as A vi μi, in which vi are the
i=1
stoichiometric coefficients, μi are the chemical potentials, and n is the total number
of components (initial substances and products). For an electrochemical reaction, the
electrochemical affinity A~ is usually determined by replacing μi with electrochemical
potential μ
~i :
μ
~i = μi þ zi Fη ð10:7Þ
where zi is the effective charge number of the ith chemical specie, F is Faraday’s
constant, η is the overpotential defined by η = (E - Ecorr) in which Ecorr is the
equilibrium electrode potential (open circuit potential, corrosion potential), and E is
the working electrode potential on the metal specimen.
For the simple electrochemical half-reaction
Fe $ Fe2þ þ 2e - ð10:8Þ
~ = zFη.
the electrochemical affinity is given by A
Butler-Volmer equation is given by
h i
α zFη α zFη
J = J a þ J c = j0 exp a - exp - c ð10:9Þ
RT RT
The exchange flux, j0, is a function of the concentrations of the reduced and
oxidized species:
j 0 = F ðk c C O Þα ðk a C R Þ1 - α ð10:10Þ
where ka and kc are the anodic and cathodic reaction rate constants, respectively. CO
and CR are the concentration of oxidized and reduced species, respectively. There-
fore, Eq. (10.5) can be rewritten as
1 ~ þ J c αc A
~

ΔSact = J αA ð10:11Þ
T a a
Equation (10.11) is the entropy generation equation for a simple oxidation

reaction considering the chemical affinity of A~ = z F ðE - E corr Þ. It should be noted
that for single-step reactions, αa + αc = 1. In this specific case, they are called
symmetry factors (Bockris & Nagy, 1973).
Imanian and Modarres (2015) extended Eq. (10.11) to a general oxidation-
reduction (redox) reaction, as follows [using their original notation]:
1 ~ M þ J M,c ð1 - αM ÞA ~ O þ J O,c ð1 - αO ÞA
~ M þ J O,a αO A

~ O ð10:12Þ
ΔSact = J M,a αM A
T
where JM, a and JM, c are the irreversible anodic and cathodic activation fluxes
(thermodynamic fluxes) for oxidation reaction and JO, a and JO, c are the anodic and
cathodic activation fluxes (thermodynamic fluxes) for reduction reaction, respec-
tively. αM and αO are the charge transport coefficients for oxidation reaction and
~ M and A
reduction reaction, respectively. A ~ O are the electrochemical affinity (ther-
modynamic force) for oxidation and reduction reactions induced by the electro-
chemical potential, respectively. Figure 10.1 explains the four different fluxes
formulated in Eq. (10.12).
The flux can be related to current density by using Faraday’s second law as
follows:
I M,a = zM FJ M,a ð10:13Þ
where IM, a is the anodic current density for the metal dissolution, zM is the number of
electrons involved in the metal dissolution reaction, JM, a is the irreversible anodic
activation flux, and F is Faraday’s constant.
Imanian and Modarres (2015) included three additional entropy production
mechanisms in addition to the activation overpotential. These three mechanisms are:
1. The entropy generation due to chemical reaction overpotential, ΔSreact.
2. The entropy generation due to diffusional overpotential, ΔSconc
3. The entropy generation due to ohmic loss, ΔSΩ
Entropy generations due to the diffusional overpotential and the ohmic loss are
negligible when the solution is well mixed (very low chemical concentration gradi-
ent) and when the electrolyte has a strong conductivity, respectively (Onsager,
1931). Therefore, these two mechanisms are trivial
II. Entropy Generation due to Chemical Reaction Overpotential, ΔSreact
Chemical reaction overpotential entropy generation is given by
Fig. 10.1 Schematic showing the anodic and the cathodic parts of the iron corrosion reaction in a
neutral electrolyte. (Adapted from Popov (2015) The red dashed line in the middle constructs the
Tafel plot)
1 Xr
ΔSreact = υA
j=1 j j
T
where r is the number of chemical reactions involved, T is the absolute temperature,

υj is the chemical reaction rate in units of mol/sec, and Aj is the chemical reaction
affinity in the units of KJ/mol.
III. Entropy Generation due to Diffusion Overpotential, ΔSconc
The diffusion overpotential happens when the initial concentration of the solution
fails to remain constant due to the consumption of reactant at the electrode. This
mechanism is predominant at a high rate of cathodic reactions, which leads to the
depletion of dissolved species in the adjacent solution.
IV. Entropy Generation due to Ohmic Loss, ΔSΩ
The ohmic loss is due to the resistance to the flow of electrons in the electrodes and
protons in the electrolyte. This loss becomes nontrivial at high current density or
when the electrolyte has low conductivity.
Hence, the thermodynamic fundamental equation is obtained by summation of all
entropy generation mechanisms as follows:
ΔScorr = ΔSact þ ΔSreact þ ΔSconc þ ΔSΩ ð10:13aÞ

1 ~ þ J M,c ð1 - αM ÞA ~ M þ J O,a αO A

~ O þ J O,c 1 - αO ÞA~O

ΔScorr = J α A
T M,a M M
1 Xr 1
þ j=1
υj A j þ J M,c E M conc,c
þ J O,c E O conc,c
þ ΔSΩ
T T
ð10:13bÞ
Usually, during a corrosion test conditions are imposed to limit entropy genera-
tion due to activation overpotential and chemical reaction overpotential only. There-
fore, the thermodynamic fundamental equation can then be simplified to include
only these two mechanisms:
1 ~ M þ J M,c ð1 - αM ÞA ~ O þ J O,c ð1 - αO ÞA
~ M þ J O,a αO A ~O

ΔScorr = J M,a αM A
T
1 X
r
þ v A
T j=1 j j
ð10:14Þ
Calculation of Entropy Production

To compute the corrosion entropy production, the irreversible anodic and cathodic
activation fluxes JM, a and JM, c for oxidation and the irreversible anodic and cathodic
activation fluxes JO, a and JO, c for reduction must be determined. They are given by

~M
αM,a A
J M,a = j0M exp -1 ð10:15Þ
RT

~M
- αM,c A
J M,c = - j0M exp -1 ð10:16Þ
RT

α A ~
J O,a = j0R exp O,a O - 1 ð10:17Þ
RT

~O
- αO,c A
J O,c = - j0R exp -1 ð10:18Þ
RT
where j0M and j0R are the exchange fluxes for the oxidation reaction and reduction
reaction, respectively. These exchange fluxes can also be expressed in the
following form: j0M = j0Fe , j0R = j0O2 . αM,a, αM,c, αO,a, αO,c are the charge transfer
coefficients. A ~ M and A
~ O are chemical affinities related to the overpotential, for
oxidation and reduction, respectively.
It is usually assumed that the redox reaction during the electrochemical corrosion
only involves the metal oxidation and the reduction of dissolved oxygen, and they
are both single-step reactions that follow the symmetry property. Therefore,
αM,a þ αM,c = 1; αO,a þ αO,c = 1 ð10:19Þ
The effective charge coefficients αM, a and αO, C can be calculated based on the
slope of the Tafel plot:
2:303RT
βa = ð10:20aÞ
αa zF
2:303RT
βc = ð10:20bÞ
αc zF
where coefficients βa and βc are the slopes of the two branches of the Tafel plot (also
known as Tafel slope, or Tafel constant).
Exchange current density is a function of the concentrations of the reduced and
oxidized species. They are usually in the following range depending on the test
conditions (Popov, 2015; Chen et al., 2018):
j0Fe = 10 - 4 10 - 5 A=cm2 , j0O2 = 10 - 6 3 × 10 - 7 A=cm2 ð10:21Þ
In summary, the calculation of corrosion entropy production requires the corro-

sion potential, corrosion currents, and charge transfer coefficients. The detailed
procedure for the determination of these variables is discussed later in the chapter.
10.2.2 Entropy Generation Mechanisms: Fatigue
Entropy is an additive property. Therefore, we can add the entropy generation due to
fatigue and entropy generation due to corrosion.
The entropy generation during ultrasonic vibration fatigue has six entropy gen-
eration mechanisms during very high cycle fatigue. However, only three are dom-
inant entropy generation mechanisms:
ΔSmec = ΔST þ ΔSr þ ΔSμp ð10:22Þ
where ΔST is the entropy generation due to thermal conduction, ΔSr is the entropy
generation due to internal heat, and ΔSμp is the entropy generation due to
microplastic work. They are given by the following equations:
Z t
∇T ∙ ∇T
t0 T2
Z t Z t drag pffiffiffi
ρr drag þ ρr dis ϱB v v þ 12 μb2 ϱ_ - αH μb ϱγ_
ΔSr = dt = dt ð10:24Þ
t0 T t0 T
Z t !
σ μ : ε_μP
t0 T
where kh is coefficient of heat conduction, ∇T is the temperature gradient, ρ is the

mass density, r is the internal heat generation due to drag and dislocation motion
mechanisms, and Bdrag and v represent the effective drag coefficient and velocity of
dislocations, respectively. The terms μ, b, ϱ, and αH are the shear modulus, the
magnitude of Burger’s vector, total dislocation density, and Taylor’s hardening
parameter, respectively. Ψf is an ultrasonic vibration frequency coefficient, ϕ is the
thermodynamic state index, fv is the activated volume fraction of micro-defects
(inclusions, vacancies, dislocations), σ μ is microscopic stress tensor, and εμP is the
microscopic plastic strain tensor Figure 10.2 depicts the comparison of fatigue life
for uncorroded and corroded samples.
Fig. 10.2 Stress vs. the number of cycles to failure test data for the uncorroded and saltwater-
corroded ASTM A656 structural steel samples subjected to 20 KHz vibrations and corrosion. (After
Lee et al. (2022))
10.4 Comparing Simulation Results and Test Data 513
10.3 Thermodynamic State Index (TSI)
The evolution of TSI is based on the cumulative-specific entropy production Δs

during corrosion fatigue. Δs is the summation of all entropy generation mechanisms
discussed above. The degradation of the material happens according to the second
law of thermodynamics which is the basis for the thermodynamic fundamental
equation of the material and progresses along the TSI axis:
1 1
Δs = ðΔScorr þ ΔSmec Þ = ΔSact þ ΔSreact þ ΔST þ ΔSr þ ΔSμp
ρ ρ
1 ~ M þ J M,c ð1 - αM ÞA

~ O þ J O,c 1 - αO ÞA
~ M þ J O,a αO A ~O

Δs = J M,a αM A
ρT
Z t Z t drag
1 Xr ∇T ∙ ∇T r þ r dis ð10:26Þ
þ υ A - k dt þ dt
ρT j = 1 j j
t0
h
ρT 2
t0 T
Z t !
μ
σ : εP _
μ
þ Ψf ϕf v dt
t0 ρT
10.4 Comparing Simulation Results and Test Data
Using the formulation presented above, the cumulative entropy generation calcu-
lated is shown in Fig. 10.3. Fatigue fracture entropy (FFE) is calculated and
measured to be 4.1 MJ K-1 m-3. The TSI evolution is presented in Fig. 10.4. TSI
value during mechanical loading starts from the endpoint of the TSI value due to
corrosion, instead of 0. Because entropy is an additive property, the contributions
from corrosion and fatigue are added.
Stress versus the number of cycles to failure curves for the uncorroded samples
and corroded samples are presented in Fig. 10.5. Corrosion degradation leads to a
decrease in fatigue life. At the stress amplitude of 400 MPa, the fatigue life decreases
from 2.79∙106 cycles to 1.58∙106 cycles due to the corrosion degradation, which is
around a 40% reduction in fatigue life.
It is important to point out that in Fig. 10.5, the test data fluctuates around the
unified mechanics theory solution, because Boltzmann’s formulation of the second
law is for all possible complexions. If we did an infinite number of experiments, the
data will always fluctuate around Boltzmann’s solution. A simple explanation can be
given by flipping a coin; we will get 50% head and 50% tail if we flip the coin an
infinite number of times.
Here we conclude our discussion of corrosion-fatigue interactions.
Fig. 10.3 Cumulative entropy production in the corroded samples subjected to ultrasonic vibration
3 because samples were first
MK
fatigue. The cumulative entropy production starts from 0.25 Km
corroded and then subjected to fatigue loading. (After Lee et al. (2022))
Fig. 10.4 TSI evolution of the corroded samples subjected to ultrasonic vibration fatigue. (After
Lee et al. (2022)). TSI due to corrosion is 0.2; hence, the curves start from 0.2
References 515
Fig. 10.5 Unified mechanics theory-based simulated stress vs. number of cycles to failure curves
for the corroded and uncorroded samples, compared with experimental data. (After Lee et al.
(2022))
References
Bockris, J., & Nagy, Z. (1973). Symmetry factor and transfer coefficient. A source of confusion in
electrode kinetics. Journal of Chemical Education, 50(12), 839. https://doi.org/10.1021/
ed050p839
Chen, L., Hu, J., Zhong, X., Zhang, Q., Zheng, Y., Zhang, Z., & Zeng, D. (2018). Corrosion
behaviors of Q345R steel at the initial stage in an oxygen-containing aqueous environment:
Experiment and modeling. Materials, 11(8), 1462. https://doi.org/10.3390/ma11081462
Gutman, E. (1998). Mechanochemistry of materials. Cambridge International Science Publishing.
Imanian, A., & Modarres, M. (2015). A thermodynamic entropy approach to reliability assessment
with applications to corrosion fatigue. Entropy, 17(12), 6995–7020. https://doi.org/10.3390/
e17106995
Lee, H. W., Basaran, C., Egner, H., Lipski, A., Piotrowski, M., Mroziński, S., Bin Jamal, N., & Rao,
L. (2022). Modeling ultrasonic vibration fatigue with unified mechanics theory. International
Journal of Solids and Structures, 236–237, 111313. https://doi.org/10.1016/j.ijsolstr.2021.
111313
Onsager, L. (1931). Reciprocal relations in irreversible processes. I. Physical Review, 37(4),
405–426. https://doi.org/10.1103/physrev.37.405
Popov, B. (2015). Electrochemical kinetics of corrosion. Corrosion Engineering, 93–142. https://
doi.org/10.1016/b978-0-444-62722-3.00003-3
William D. Callister, Jr. and David G. Rethwisch (2014), Materials Science and Engineering: An
Introduction, 9E, Wiley, 2014
Index
A Diffusion, 130, 134, 179, 224, 225, 241, 250,

Acrylic, 3, 210 427–431, 435–438, 441, 443, 460, 461,
Alumina trihydrate (ATH), 310, 311, 463, 464, 473, 474, 481–483, 509
342, 349, 361 Disorder, 111, 114–118, 135, 153, 200,
202, 203, 210–212, 249, 250, 257,
450, 461, 462
B Dissipation, 2, 86, 87, 91, 99, 100, 109, 113,
Boltzmann, 3, 116, 117, 148, 150, 153–156, 124–130, 133, 134, 142, 143, 149,
159, 161, 163, 164, 176, 180, 195, 199, 218–221, 223–226, 228, 257, 261,
200, 204, 210, 211, 241, 298, 396, 401, 265–267, 365, 384, 391, 395, 403, 404,
412, 431, 440, 450, 459, 460, 463, 464, 419, 421, 422, 453, 459, 472, 475, 489,
479, 483, 513 490, 496
Boltzmann entropy, 135, 151–153, 200
E
C Electromigration, 3, 427–456
Composite material, 3, 211, 263, 309, 352, 428 Electronics packaging, 428
Continuum mechanics, 1–3, 5, 13, 73–76, 81, Energy, 2, 3, 12, 86, 88, 91–94, 96–115,
89, 91, 93, 98–100, 112–114, 116, 118, 117–124, 126, 127, 133–138, 140–145,
119, 123, 125, 130, 133–145, 147–150, 148, 150, 153–156, 158–161, 165–167,
251, 258, 260, 263, 272–275, 295, 309, 170–177, 179–181, 183–188, 190–193,
460, 475 196–209, 211, 213, 214, 217, 218, 221,
Corrosion, 3, 130, 147, 505–507, 509–514 224–226, 228, 230, 238, 241, 243,
Corrosion fatigue, 3, 147, 506, 513 249–251, 254–257, 259–261, 263–266,
Couple stress, 20, 21, 96, 274–276, 278–290, 276–278, 280–282, 289, 298, 299, 367,
297–299 383–385, 391, 393–403, 412, 419, 431,
434–436, 451, 452, 459, 460, 463, 465,
470, 472, 475, 479, 481, 489, 495
D Entropy, 3, 12, 91, 98, 100–105, 107–119, 121,
Damage, 119, 142, 145–150, 242, 247, 263, 123, 125, 126, 128–130, 134–138,
346, 404, 405, 419, 421, 422, 427, 428, 141–157, 194–196, 199–203, 205, 207,
494, 500 208, 210–212, 214, 215, 217, 218,
Deformation gradient tensor, 113, 121, 254, 221–226, 229, 231, 246, 249–251, 254,
375–380, 415 255, 257–263, 266, 267, 272, 296,
© The Editor(s) (if applicable) and The Author(s), under exclusive license to 517
Springer Nature Switzerland AG 2022
https://doi.org/10.1007/978-3-031-18621-9
518 Index
298–299, 354, 355, 358, 365, 367, 368, Modeling, 2, 3, 28, 57, 117, 124, 127, 142, 309,
383–385, 391, 403–405, 419, 421, 422, 346–348, 371, 375, 384, 392, 398, 399,
443, 449–454, 459–482, 484, 485, 488, 402, 427, 428
492–496, 506, 508–511, 513, 514
Eshelby, 309, 312, 314–317, 332
Euler equation, 106, 107 N
Eulerian descriptions, 38, 49, 55–57, 59 Newtonian mechanics, 81, 83, 86, 91, 96, 98,
113, 119, 123, 124, 126, 133, 136, 138,
139, 142, 148, 150, 217, 219–223, 226,
F 229, 237, 238, 253–255, 263, 265, 275,
Failure, 11, 99, 108, 115, 117, 133, 135, 142, 276, 278, 279, 288, 313, 366, 429,
146, 147, 149, 150, 210, 211, 215, 217, 443–444, 459
312, 404, 405, 411, 419, 421, 422,
427–430, 435, 451, 454, 456, 466, 470,
471, 476, 477, 479, 482, 484, 494, 512, O
513, 515 Onsager reciprocal relations, 128, 129, 506
Fatigue, 3, 99, 108, 117, 125, 130, 133, 137,
141, 145–148, 150, 215, 216, 263,
265–272, 346, 367, 421, 422, 459–501, P
511, 513, 514 Particle filled, 3
Fatigue corrosion interaction, vii, 3, 505–515 Particle filled composite, 3, 311, 338, 371
Field theories, 11 Piola–Kirchoff stress, 76–78, 389, 390, 393,
Finite deformations, 375–425 400
Finite strain, 3, 34, 57–73, 78, 136, 383–403, Plasticity, 31, 81, 124, 126–128, 136, 137, 139,
422 142, 215, 237, 238, 240, 241, 263, 272,
Flow theory, 237–241, 288, 294 273, 281, 288–295, 297, 309, 392, 395,
Fluid mechanics, 18, 49 403, 412, 468
Fracture, 133, 142, 145, 147, 150, 263, 367, Polymer, 3, 148, 210, 321, 362, 365, 375, 384,
403, 410, 422, 494, 513 392, 393, 395, 397–400, 402, 403, 405,
407, 422, 423
Poly-methyl-meth-acrylate (PMMA), 310, 311,
G 349, 354, 355, 362, 366, 397, 406,
Gibbs–Duhem relation, 104–107 409–411, 413–425
H R
Homogenizations, 309, 468, 470, 479 Rational mechanics, 428
Reliability, 150
L
Lagrangian description, 38, 48, 54, 57 S
Lagrangian mechanics, 86, 87, 91, 133 Solder alloys, 273, 298
Large strain, 34, 55, 60, 273, 366 Solder joints, 3, 267–272, 428, 430, 435,
Laws of motion, 2, 5, 11, 12, 50, 81, 91, 119, 455, 456
133, 142, 152, 154, 459 Solid mechanics, 18, 39, 73, 251
Life prediction, 146, 150, 459, 460, 478, 479, Strain, 1–3, 33–37, 40, 41, 43, 44, 46, 52–57,
484, 496, 500 60, 62–76, 96, 97, 108, 111, 113, 117,
119, 123, 125–127, 130, 136, 137, 139,
140, 143, 145, 148, 214, 216, 226, 237,
M 239–246, 253, 254, 256, 262–267,
Manufacturing, 346, 472 272–278, 280–282, 284, 285, 288–291,
Metal corrosion, 505, 506 295–299, 301, 303, 309, 312–317,
Microplasticity, 469–475, 477, 479, 492–495 322–324, 327, 329, 330, 333, 336, 337,
Index 519
352–355, 358, 364, 367–370, 385, 393, Thermodynamics, 2, 3, 12, 86, 91–93, 97–102,
395–399, 401, 403, 406, 409–411, 413, 104, 105, 107–116, 118–128, 130,
416–420, 439, 441–443, 445, 451–453, 133–138, 140, 142, 144, 145, 148–153,
459, 467–471, 475, 480, 485, 488, 489, 155–157, 198, 201, 210, 211, 213, 226,
497, 498, 500, 501, 512 228, 229, 242, 243, 249–251, 255,
Stress, 1, 2, 12–15, 18–32, 35, 37, 41, 73, 257–264, 266, 267, 272, 296–298, 354,
76–79, 83–85, 95, 96, 99, 108, 111, 117, 358, 368, 383–392, 403–405, 416, 419,
119, 121, 123, 126, 130, 137–142, 146, 422, 430, 439, 443, 449–451, 454,
150, 217, 238–245, 253, 257, 261, 263, 459–462, 464, 471, 474, 475, 477, 480,
266, 273–282, 284, 288–290, 295–299, 485–487, 489, 490, 493–496, 506, 508,
301, 309, 311–314, 316–323, 327–338, 512, 513
350, 352, 353, 358, 366, 378, 379, Thermomechanical analysis, 272–288
386–396, 398–403, 408, 411–413, 415, Thermomigration, 3, 117, 266, 427–456,
416, 418, 419, 428, 429, 432, 436–440, 461, 463
447, 452, 455, 459, 460, 464, 468–489,
492–497, 512, 513, 515
V
Very high cycle fatigue, 480–482, 484, 493,
T 495, 496, 511
Thermal mechanical analysis, 237–306 Viscoplasticity, 126, 241, 263, 295, 350–353
Thermodynamic fundamental equation, 3, 102,
107, 113, 134, 139, 147, 153, 210, 212,
214, 220, 224–226, 228, 229, 231, Y
249–250, 260, 263, 354, 365, 368, 384, Yield surface, 31, 124, 127, 136–138, 140,
405, 421, 428, 449–456, 459–461, 474, 238–240, 245, 246, 272, 289, 290, 295,
479, 481–488, 495, 506, 509, 510, 513 297, 298, 352, 353, 358, 470

Unified Mechanics Theory

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

Unified Mechanics Theory

Uploaded by

Copyright:

Available Formats

Cemal Basaran

Thermodynamic State Index

100 years old

ISBN 978-3-031-18620-2 ISBN 978-3-031-18621-9 (eBook)

1918 Nobel Prize in Physics for his work on quantum theory.

Friedrich-Wilhelms-Universität Max Planck

Buffalo, NY, USA Cemal Basaran

Callen, H. B. (1985). Thermodynamics and an introduction to thermostatistics (2nd ed.). Wiley.

2.5 Rate of Deformation Tensor and Rate of Spin Tensor . . . . . . . . 53

3.3.9 Thermodynamic Potential . . . . . . . . . . . . . . . . . . . . . . 118

5.3 Return Mapping Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . 243

6.4 Pairwise Interacting Solution for Two-Phase Composites . . . . . . 323

7.7 Numerical Implementation of Dual-Mechanism Model . . . . . . . 415

10 Corrosion-Fatigue Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505

1.1 What Is the Mechanics of Continuous Medium?

Probably one of the best deﬁnitions of mechanics of a continuous medium is

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 1

what we mean by “incomplete.” For example, if a soccer ball is given an initial

Hill, R. (1965). Continuum micro-mechanics of elastoplastic polycrystals. Journal of the Mechan-

Newton, I. (1687). Philosophiæ Naturalis Principia Mathematica, .

2.1 Newton’s Universal Laws of Motion

2.1.1 First Universal Law of Motion

The original text in Latin reads:

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 5

Fig. 2.1 Original 1687

Translated to English, this reads:

(Modern Translation) Law I: In an inertial frame of reference, an object either

2.1.1.1 Formulation of the First Law

2.1.2 Second Universal Law of Motion

Isaac Newton’s original Latin text reads:

This was translated by Motte in 1729 as:

2.1.2.1 Formulation of Newton’s Second Law

The second law can also be stated in terms of an object’s acceleration. It is

An impulse is deﬁned by a force F acting on an object over an interval of time Δt

Since force is the time derivative of momentum, it follows that,

2.1.3 Third Universal Law of Motion

Newton’s original Latin text reads:

Direct translation to English, this reads:

2.1.4 Range of Validity of Newton’s Universal Laws

2.1.5 Relation to the Thermodynamics

2.2.1 Deﬁnitions of Stress and Traction

Fig. 2.3 Deﬁnition of stress

2.2.2 Stress Vector on an Arbitrary Plane

and σ (z) are deﬁned.

Fig. 2.4 Stresses acting at a point in Cartesian coordinates

n2x þ n2y þ n2z = 1

Fig. 2.6 Geometry of the tetrahedron at a point

The volume of the pyramid is given by

substituting h in the previous equation yields,

Similarly, we can write

Eliminating ΔA from each term leads to the general equilibrium equation.

σ ðnÞ = σ ðxÞ nx þ σ ðyÞ ny þ σ ðzÞ nz ð2:1Þ

In some textbooks, Eq. (2.1) is obtained by allowing the tetrahedron h to go to

σ ij is a second-order tensor (linear vector function); therefore the components of the

½σ = ½N T ½σ ½N ½σ = ½N ½σ ½N T

2.2.3 Symmetry of Stress Tensor

When there are no distributed couple-stresses (moments) (body or surface) or the

Writing equilibrium of moment about z axis leads to ∑Mz = 0

Similarly writing ∑My = 0 and ∑Mx = 0 equilibrium equations lead to

However, we should make it clear that we obtained this result because we

2.2.4 Couple Stresses

Therefore σ xy ≠ σ yx, similarly σ xz ≠ σ zx, σ yz ≠ σ zy.

It is important to point out that the ½ multiplier in front of the couple-stress

Fig. 2.8 Couple stresses y