International Journal of Plasticity 25 (2009) 1495–1539

Contents lists available at ScienceDirect

International Journal of Plasticity

journal homepage: www.elsevier.com/locate/ijplas

A thermo-mechanically coupled theory for large

deformations of amorphous polymers. Part II: Applications
Nicoli M. Ames, Vikas Srivastava, Shawn A. Chester, Lallit Anand *
Department of Mechanical Engineering Massachusetts Institute of Technology, Cambridge, MA 02139, USA

a r t i c l e i n f o a b s t r a c t

Article history: We have conducted large-strain compression experiments on three

Received 22 July 2008 representative amorphous polymeric materials: poly(methyl
Received in final revised form 9 November methacrylate) (PMMA), polycarbonate (PC), and a cyclo-olefin
2008
polymer (Zeonex-690R), in a temperature range spanning room
Available online 27 November 2008
temperature to slightly below the glass transition temperature of
each material, in a strain rate range of  104 s1 to 101 s1 , and
compressive true strains exceeding 100%.
Keywords:
Amorphous polymers
The constitutive theory developed in Part I [Anand, L., Ames,
Viscoplasticity N.M., Srivastava, V., Chester, S., 2009. A thermo-mechanically
Large deformations coupled theory for large deformations of amorphous polymers.
Thermo-mechanically coupled Part 1: Formulation. International Journal of Plasticity] is special-
Finite elements ized to capture the salient features of the thermo-mechanically
coupled strain rate and temperature dependent large deforma-
tion mechanical response of amorphous polymers. For the three
amorphous polymers studied experimentally, the specialized
constitutive model is shown to perform well in reproducing
the following major intrinsic features of the macroscopic
stress–strain response of these materials: (a) the strain rate
and temperature dependent yield strength; (b) the transient
yield-peak and strain-softening which occurs due to deforma-
tion-induced disordering; (c) the subsequent rapid strain-hard-
ening due to alignment of the polymer chains at large strains;
(d) the unloading response at large strains; and (e) the temper-
ature rise due to plastic-dissipation and the limited time for
heat-conduction for the compression experiments performed at
strain rates ’0:01 s1 .
We have implemented our thermo-mechanically coupled consti-
tutive model by writing a user material subroutine for the finite
element program [Abaqus/Explicit, 2007. SIMULIA, Providence, RI].

* Corresponding author. Tel.: +1 617 253 1635.

E-mail address: [email protected] (L. Anand).

0749-6419/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijplas.2008.11.005
1496 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

In order to validate the predictive capabilities of our constitutive

theory and its numerical implementation, we have performed the
following validation experiments: (i) isothermal fixed-end large-
strain reversed-torsion tests on PC; (ii) macro-scale isothermal
plane-strain cold- and hot-forming operations on PC; (iii) macro-
scale isothermal, axi-symmetric hot-forming operations on Zeo-
nex; (iv) micro-scale hot-embossing of Zeonex; and (v) high-speed
normal-impact of a circular plate of PC with a spherical-tipped
cylindrical projectile. By comparing the results from this suite of
validation experiments of some key macroscopic features, such
as the experimentally-measured deformed shapes and the load-
displacement curves, against corresponding results from numerical
simulations, we show that our theory is capable of reasonably
accurately reproducing the experimental results obtained in the
validation experiments.
Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction

This, Part II of our two-part paper, is devoted to applications of the theory developed in Part I
(Anand et al., 2009).
The plan of the paper is as follows. In Section 2 we briefly describe our experiments on three rep-
resentative amorphous polymeric materials: poly(methyl methacrylate) (PMMA), polycarbonate
(PC), and a cyclo-olefin polymer (Zeonex-690R),1 in a temperature range spanning room temperature
to slightly below the glass transition temperature of each material, in a strain rate range of  104 s1
to 101 s1 , and compressive true strains exceeding 100%. In Section 3 we summarize the major aspects
of the general constitutive theory outlined in Part I (Anand et al., 2009). Then in Section 4 we special-
ize this constitutive theory to capture the salient features of the experimentally-measured mechanical
response of PMMA, PC and Zeonex. In an Appendix, Section A, we describe in reasonable detail our
heuristic method to calibrate the material parameters/functions appearing in our constitutive theory.
The quality of the fit of the specialized model to the experimentally-measured stress–strain curves
is discussed in Section 5, where we show that for the three amorphous polymers studied experimen-
tally, the constitutive model performs well in reproducing the following major intrinsic features of the
macroscopic stress–strain response of these materials: (a) the strain rate and temperature dependent
yield strength; (b) the transient yield-peak and strain-softening which occurs due to deformation-in-
duced disordering; (c) the subsequent rapid strain-hardening due to alignment of the polymer chains
at large strains; (d) the unloading response at large strains; and (e) the temperature rise due to plastic-
dissipation and the limited time for heat-conduction for the compression experiments performed at
strain rates ’0:01 s1 .
We have implemented our thermo-mechanically-coupled constitutive model by writing a user
material subroutine for the finite element program Abaqus/Explicit (2007). In Section 6, we present
results of a suite of experiments that we have conducted in order to validate the predictive capabil-
ities of our constitutive theory and its numerical implementation. By comparing the results of some
key macroscopic features from this set of validation experiments, such as the experimentally-mea-
sured deformed shapes and the load-displacement curves, against corresponding results from
numerical simulations, we show that our theory is capable of reasonably accurately reproducing
the experimental results obtained in the validation experiments. We close in Section 7 with some
final remarks.

1
From Zeon Chemicals; henceforth, simply called Zeonex in this paper. Relative to PMMA and PC, Zeonex is biocompatible, has
lower moisture uptake, has better light-transmittance characteristics, and it is also chemically-resistant to a wider variety of
solvents.
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1497

200 40
0.0003 1/s 0.0003 1/s
35

150 30
True Stress (MPa)

True Stress (MPa)

25 90C
25C
100 20

15 100C
50C
50 70C 10
110C
90C 5

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
True Strain True Strain

Fig. 1. Stress–strain curves in simple compression for PMMA at various temperatures ranging from 25 to 110 °C, at a strain rate
of 3  104 s1 . Note the change in scale for the stress axis between the two figures.

2. Simple compression experiments on three different amorphous polymers

We have conducted a suite of simple compression experiments on three technologically important

amorphous polymers: poly(methyl methacrylate) (PMMA), polycarbonate (PC), and a cyclo-olefin
polymer (Zeonex-690R). The nominal glass transition temperatures, #g , of these three materials are:
PMMA : #g  115  C; PC : #g  145  C; and Zeonex-690R : #g  135  C:
The cylindrical compression test specimens were 12.7 mm diameter and 12.7 mm tall, and were
annealed before and after machining to final shape by heating in a furnace at a temperature about
10 °C above the glass transition temperature of each material, and holding at that temperature for
two hours, before cooling to room temperature.2 The experiments were conducted using a servo-
hydraulic Instron testing machine fitted with a high-temperature furnace. Amorphous polymers are poor
thermal conductors; accordingly, in order to heat the compression specimens uniformly, we also used
heated steel compression platens in addition to the furnace. The platens were heated with cartridge heat-
ers, and thermocouples inserted into each platen were used to control the temperature. The top com-
pression platen also had an integrated spherical seat to help minimize any effects of misalignment
during compression testing. To reduce friction at the platen/specimen interface, the platens were pol-
ished, and thin Teflon (PTFE) films were used as lubricating layers between the specimen and the plat-
ens.3 Before a given experiment, each specimen was allowed to anneal at the testing temperature for one
hour prior to testing. Experiments on PMMA were conducted for the temperature range 25–110 °C at
four strain rates: 3  104 , 103 , 102 , and 101 s1 . On PC the experiments were conducted for the tem-
perature range 25–130 °C at three strain rates: 103 , 102 , and 101 s1 . While on Zeonex the experi-
ments were conducted for the temperature range 25–130 °C at four strain rates: 3  104 , 3  103 ,
3  102 , and 3  101 s1 . The compression tests were carried out at constant true strain rates to com-
pressive true strains exceeding ’100%; all strain measurements were made using an extensometer.
Fig. 1 shows representative stress–strain curves for PMMA at a strain rate of 3  104 s1 at tem-
peratures ranging from 25 °C through 110 °C, while Fig. 2 shows a more extensive set of stress–strain
curves for strain rates of 3  104 , 103 , 102 , and 101 s1 and temperatures of 25 °C through 110 °C.4
Referring to Fig. 1, we see that as the temperature increases in the glassy region, from 25 to 110 °C, the

2
After annealing, the PMMA and PC specimens were slowly cooled in the furnace to room temperature over a period of several
hours, while the Zeonex was quenched in water.
3
For true strains up to 100% our compression specimens showed very little or no bulging; however, for larger strain levels, in
spite of our precautions to minimize friction, some bulging did occur.
4
As is customary, in order to calculate the deformed cross-sectional area (and thence the true stress), we have assumed plastic
incompressibility to estimate the stretch in the lateral direction of the compression specimens.
1498 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

200 150
25C 0.1 1/s 50C 0.1 1/s
0.01 1/s 0.01 1/s
0.001 1/s 0.001 1/s
0.0003 1/s 0.0003 1/s
150

True Stress (MPa)

True Stress (MPa)

100

100

50
50

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
True Strain True Strain

70
70C 0.1 1/s 90C 0.1 1/s
100 0.01 1/s 0.01 1/s
0.001 1/s 60 0.001 1/s
0.0003 1/s 0.0003 1/s
80 50
True Stress (MPa)

True Stress (MPa)

60 40

30
40
20
20
10

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
True Strain True Strain
50 40
100C 0.1 1/s 110C 0.1 1/s
0.01 1/s 0.01 1/s
35
0.001 1/s 0.001 1/s
40 0.0003 1/s 0.0003 1/s
30
True Stress (MPa)

True Stress (MPa)

30 25

20
20 15

10
10
5

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
True Strain True Strain

Fig. 2. Stress–strain curves in simple compression for PMMA at strain rates of 3  104 , 103 , 102 , 101 s1 , and temperatures
of 25, 50, 70, 90, 100, and 110 °C. Note change in scale for the stress axis between the various figures.

yield strength decreases by an order of magnitude from  100 to  10 MPa, and the strain-hardening
observed at large strains due to limited extensibility of the polymer chains diminishes. Referring to
Fig. 2, which shows stress–strain curves at various fixed temperatures and the four different strain rates,
we see obvious strain-rate dependent features of the material response. In the low-temperature glassy
region, the yield strength of the material increases by about 10% for a one-decade increase in strain rate
at any given temperature. Another important strain rate dependent feature is the softening observed at
large strains at the two higher strain rates of 102 and 101 s1 at temperatures of 25, 50, and 70 °C.
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1499

The softening is attributable to (near) ‘‘adiabatic” heating at the higher strain rates. While we did not
measure the actual temperature rise in our specimens, Arruda et al. (1995) have shown that the surface
temperature of a compression specimen, for a test carried out at 20 °C, could increase by as much  20 C
after a 100% compressive strain at a strain rate of 101 s1 .
Stress–strain curves for PC and Zeonex will be presented in Section 5, where we compare experi-
mental stress–strain curves against those resulting from our constitutive model. Next in Section 3, we
summarize the theory developed in Part I; in Section 4, we specialize the theory for applications; and
then in Section 5 we show results of the fit of the specialized constitutive theory to the data from our
experiments on PMMA, PC, and Zeonex.

3. Theory

We begin by recalling the theory for isotropic elastic viscoplastic materials developed in Part I of
this paper (Anand et al., 2009). The theory relates the following basic fields:5

x ¼ vðX; tÞ motion
F ¼ rv; J ¼ det F > 0 deformation gradient
F ¼ Fe Fp elastic–plastic decomposition of F
Fe ; J e ¼ det Fe ¼ J > 0 elastic distortion
Fp ; J p ¼ det Fp ¼ 1 inelastic distortion
F ¼ RU ¼ VR polar decomposition of F
Fe ¼ Re Ue ¼ Ve Re polar decomposition of Fe
Fp ¼ Rp Up ¼ Vp Rp polar decomposition of Fp
C ¼ F> F; B ¼ FF> right and left Cauchy-Green tensors
Ce ¼ Fe > Fe ; Be ¼ Fe Fe > elastic right and left Cauchy-Green tensors
Cp ¼ Fp > Fp ; Bp ¼ Fp Fp > plastic right and left Cauchy-Green tensors
T ¼ T> Cauchy stress
TR ¼ JTF> Piola stress
wR free energy density per unit reference volume
n ¼ ðn1 ; n2 ; . . . ; nm Þ m scalar internal variables
A; A ¼ A> ; det A ¼ 1 tensorial internal variable
#>0 absolute temperature
r# referential temperature gradient
qR referential heat flux vector

3.1. Constitutive equations

1. Free energy
The free energy is given by
~ ð1Þ ðICe ; #Þ þ w
wR ¼ w ~ ð2Þ ðIC ; #Þ þ w
~ ðpÞ ðIA ; #Þ; ð3:1Þ
where ICe , IC , and IA are the lists of the principal invariants of Ce , C, and A, respectively. In (3.1),
~ ð1Þ is an energy associated with intermolecular interactions and modeled using the elastic Cauchy-
w
Green tensor Ce ; w~ ð2Þ is an energy associated with the stretching of the polymer chains and modeled

5
Notation: We use standard notation of modern continuum mechanics. Specifically: r and Div denote the gradient and
divergence with respect to the material point X in the reference configuration; grad and div denote these operators with respect to
the point x ¼ vðX; tÞ in the deformed body; a superposed dot denotes the material time-derivative. Throughout, we write
Fe1 ¼ ðFe Þ1 , Fp> ¼ ðFp Þ> , etc. We write trA, symA, skwA, A0 , and sym0 A respectively, for the trace, symmetric, skew, deviatoric,
and symmetric-deviatoric
pffiffiffiffiffiffiffiffiffiffiffi parts of a tensor A. Also, the inner product of tensors A and B is denoted by A : B, and the magnitude of A
by jAj ¼ A : A.
1500 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

using the total Cauchy-Green tensor C; and w ~ ðpÞ is an energy associated with plastic deformation,
and assumed to depend on the internal variable A. This last ‘‘defect energy” leads to the develop-
ment of a back-stress, and allows one to phenomenologically account for Bauschinger-like phenom-
ena; in addition, it contributes in an important manner to the plastic source term in the balance of
energy.

2. Cauchy stress
The Cauchy stress is given by

T ¼ Tð1Þ þ Tð2Þ ; ð3:2Þ

where

def
~ ð1Þ ðICe ; #Þ
@w
Tð1Þ ¼ J 1 ðFe Se Fe> Þ; with Se ¼ 2 ; ð3:3Þ
@Ce
and

def
~ ð2Þ ðIC ; #Þ
@w
Tð2Þ ¼ J 1 ðFSF> Þ; with S¼2 : ð3:4Þ
@C
3. Driving stresses for plastic flow
With

Me ¼ Ce Se ð3:5Þ

denoting a symmetric Mandel stress,

!
~ ðpÞ ðIA ; #Þ
@w
Mback ¼ 2 A ð3:6Þ
@A
0

a symmetric and deviatoric back-stress, and

Meeff ¼ Me  Mback ð3:7Þ

an effective Mandel stress, the driving stress for plastic flow is taken as the stress difference

ðMeeff Þ0 ¼ Me0  Mback : ð3:8Þ

4. Flow rule
The evolution equation for Fp is

F_ p ¼ Dp Fp ; ð3:9Þ

with Dp given by

p ðMeeff Þ0
Dp ¼ d Np ; Np ¼ ; ð3:10Þ
jðMeeff Þ0 j
where with

def
K ¼ fCe ; Bp ; A; n; #g ð3:11Þ
p
denoting a list of constitutive variables, the scalar flow rate d is obtained by solving the scalar
strength relation
p
jðMeeff Þ0 j ¼ Yðd ; KÞ; ð3:12Þ
p
for given ðMeeff Þ0 and K, where Yðd ; KÞ is the strength function, which is an isotropic function of its
arguments.
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1501

5. Evolution equations for internal variables

The internal variables n and A are presumed to evolve according to the differential equations
p
)
n_ i ¼ hi ðd ; KÞ;
p
ð3:13Þ
A_ ¼ Dp A þ ADp  GðKÞd ;
with the functions hi and G isotropic functions of their arguments, and since A is unimodular, the func-
tion G must satisfy trðGA1 Þ ¼ 0.
6. Fourier’s Law
The heat flux is taken to be given by Fourier’s law
qR ¼ jr#; ð3:14Þ
with jð#Þ > 0 the thermal conductivity.

The evolution equations for Fp , n and A need to be accompanied by initial conditions. Typical initial
conditions presume that the body is initially (at time t ¼ 0, say) in a virgin state in the sense that
FðX; 0Þ ¼ Fp ðX; 0Þ ¼ AðX; 0Þ ¼ 1; ni ðX; 0Þ ¼ ni0 ð¼ constantÞ; ð3:15Þ
e p e
so that by F ¼ F F we also have F ðX; 0Þ ¼ 1.

3.2. Partial differential equations for the deformation and temperature fields

The partial differential equation for the deformation is obtained from the local force balance
DivTR þ b0R ¼ qR €v; ð3:16Þ
>
where TR ¼ JTF is the Piola stress, b0R is the non-inertial body force, qR is the mass density in the
reference body, and €
v is the acceleration.
Balance of energy gives the following partial differential equation for the temperature
!!
~ ðpÞ ðIA ; #Þ
@w p
c#_ ¼ DivqR þ qR þ jðMeeff Þ0 j þ : GðKÞ d
@A
1 @Se _ e 1 @S _ @2w
~ ðpÞ ðIA ; #Þ
_
þ # :C þ # :Cþ# : A; ð3:17Þ
2 @# 2 @# @#@A
where the specific heat in the theory is given by
!
def @2w
~ ð1Þ ðICe ; #Þ @ 2 w
~ ð2Þ ðIC ; #Þ @ 2 w
~ ðpÞ ðIA ; #Þ
c ¼ # þ þ : ð3:18Þ
@#2 @#2 @#2

4. Specialization of the constitutive equations

In this section, based on experience with existing recent theories of isotropic viscoplasticity of
polymeric materials, we specialize the constitutive theory by imposing additional constitutive
assumptions.

4.1. Free energy wð1Þ

The spectral representation of Ce is

X
3
Ce ¼ xei rei  rei ; with xei ¼ ke2
i ; ð4:1Þ
i¼1

where ðre1 ; re2 ; re2 Þ are the orthonormal eigenvectors of Ce and Ue , and ðke1 ; ke2 ; ke3 Þ are the positive eigen-
values of Ue . Instead of using the invariants ICe , the free energy wð1Þ for isotropic materials may be
alternatively expressed in terms of the principal stretches and temperature as
1502 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

 e ; ke ; ke ; #Þ:
wð1Þ ¼ wðk ð4:2Þ
1 2 3

Then, by the chain-rule and (3.3))2, the stress Se is given by

 ð1Þ ðke ; ke ; ke ; #Þ
@w X3  ð1Þ ðke ; ke ; ke ; #Þ @ke X
@w 3  ð1Þ ðke ; ke ; ke ; #Þ @ xi
1 @w
Se ¼ 2 1
e
2 3
¼2 1
e
2 3 i
e ¼ e
1 2 3
: ð4:3Þ
@C i¼1
@ki @C i¼1
ki @kei @Ce

Assume that the squared principal stretches xei are distinct, so that the xei and the principal direc-
tions rei may be considered as functions of Ce ; then

@ xei
¼ rei  rei ; ð4:4Þ
@Ce
and, granted this, (4.4) and (4.3) imply that

X3  ð1Þ ðke ; ke ; ke ; #Þ
1 @w
Se ¼ e
1 2 3
rei  rei : ð4:5Þ
i¼1
ki @kei

Further, from (3.3)1,

!
1 e e e> e e e e e> e
X
3  ð1Þ ðke ; ke ; ke ; #Þ
@w
Tð1Þ
¼J FS F 1
¼J R U S U R 1
¼J R kei 1 2 3
ri  ri Re> :
e e
ð4:6Þ
i¼1
@kei

Next, since Me ¼ Ce Se (cf. (3.5)), use of (4.1) and (4.5) gives the Mandel stress as

X
3  ð1Þ ðke ; ke ; ke ; #Þ
@w
Me ¼ kei 1 2 3
rei  rei : ð4:7Þ
i¼1
@kei

Let

def
X
3
Ee ¼ ln Ue ¼ Eei rei  rei ; ð4:8Þ
i¼1

denote the logarithmic elastic strain with principal values

def
Eei ¼ ln kei ; ð4:9Þ
and consider an elastic free energy function of the form

w  ð1Þ ðEe ; Ee ; Ee ; #Þ;

 ð1Þ ðke ; ke ; ke ; #Þ ¼ w ð4:10Þ
1 2 3 1 2 3

so that, using (4.7),

X3  ð1Þ ðEe ; Ee ; Ee ; #Þ
@w
Me ¼ 1 2 3
rei  rei : ð4:11Þ
i¼1
@Eei

We consider the following simple generalization of the classical strain energy function of infinites-
imal isotropic elasticity which uses a logarithmic measure of finite strain6
 
 ð1Þ ðEe ; #Þ ¼ GjEe j2 þ 1 K  2 G ðtrEe Þ2  ð#  #0 Þð3K aÞðtrEe Þ þ ~f ð#Þ;
w ð4:12Þ
2 3
where ~f ð#Þ is an entropic contribution to the free energy related to the temperature-dependent spe-
cific heat of the material. The temperature-dependent parameters
Gð#Þ > 0; Kð#Þ > 0; að#Þ > 0; ð4:13Þ

6
This is a useful free energy function for moderately large elastic stretches, Anand (1979, 1986).
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1503

are the shear modulus, bulk modulus, and coefficient of thermal expansion, respectively, and #0 is a
reference temperature. Then, (4.11) gives
Me ¼ 2GEe0 þ KðtrEe Þ1  3K að#  #0 Þ1; ð4:14Þ
and on account of (4.6), (4.7) and (4.14),

Tð1Þ ¼ J 1 Re Me Re> : ð4:15Þ

4.2. Free energy wð2Þ

We denote the distortional part of F by7

def
Fdis ¼ J 1=3 F; detFdis ¼ 1: ð4:16Þ
Correspondingly, let
def
Cdis ¼ ðFdis Þ> Fdis ¼ J 2=3 C; ð4:17Þ
denote the distortional (or volume preserving) right Cauchy-Green tensor, and consider a free energy
function in the special form
 ð2Þ ðIC ; #Þ;
wð2Þ ¼ w ð4:18Þ
dis

where ICdis are the principal invariants of Cdis . Then using (3.4)2 the stress S with a free energy of the
form above is
 ð2Þ ðIC ; #Þ  >  ð2Þ
@w @Cdis @ w ðICdis ; #Þ
S¼2 dis
¼2 : ð4:19Þ
@C @C @Cdis
pffiffiffiffiffiffiffiffiffiffiffi
Next, since J ¼ det C, and since
@ det C
¼ ðdet CÞC1 ¼ J 2 C1 ;
@C
we have

@J 1 1 @J 2=3 1
¼ JC ; and ¼  J 2=3 C1 : ð4:20Þ
@C 2 @C 3
Also,
!
@Cdis @ðJ 2=3 CÞ 2=3 2=3 @J 2=3
¼ ¼J IþJ C ;
@C @C @C

or using (4.20)2,
   
@Cdis 1 1
¼ J 2=3 I  C  C1 ¼ J 2=3 I  Cdis  C1
dis ; ð4:21Þ
@C 3 3
where I is the fourth-order identity tensor. Thus, using (4.20)1 and (4.21) in (4.19), the stress S has the
form

1  ð2Þ ðIC ; #Þ
@w
S ¼ 2J 2=3 ðI  C1dis  Cdis Þ
dis

3 @Cdis
  ð2Þ   ð2Þ ðIC ; #Þ 1 
@ w ðICdis ; #Þ 1 @w
¼ 2J 2=3  Cdis : dis
C dis : ð4:22Þ
@Cdis 3 @Cdis

7
Since Je ¼ J, and since we have already accounted for a volumetric elastic energy for wð1Þ , we do not allow for a volumetric
elastic energy for wð2Þ .
1504 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

In order to model the stress increase due to the stretching and locking of polymer chains at
large strains, most previous theories for amorphous polymers (e.g., Parks et al., 1985; Boyce
et al., 1988; Arruda and Boyce, 1993; Wu and Van der Giessen, 1993a; Anand and Gurtin, 2003)
presume that polymer glasses behave like crosslinked rubber and use a free energy based on entro-
pic-network models. There is a conceptual difficulty with using statistical–mechanical ideas of the
theory of entropic rubber elasticity to describe the strain-hardening due to chain-alignment at
temperatures below the glass transition temperature, because at these temperatures the chains
do not have sufficient mobility in the amorphous state to sample all possible molecular conforma-
tions as visualized in the statistical–mechanical models of rubber elasticity (Anand and Ames,
2006). Here, we employ a simple phenomenological form for the free energy function wð2Þ due to
Gent (1996).8 With

def
I1 ¼ trCdis ð4:23Þ

denoting the first principal invariant of Cdis , the Gent free energy has the form
 
 ð2Þ ¼  1 l Im ln 1  I1  3 ;
w ð4:24Þ
2 R Im

which involves two temperature-dependent material parameters

lR ð#Þ > 0; Im ð#Þ > 3: ð4:25Þ

In particular, lR represents the ground-state rubbery shear modulus of the material, and Im represents
the upper limit of ðI1  3Þ associated with limited chain extensibility of polymeric molecules. For this
simple free energy, we note

 ð2Þ ðIC ; #Þ 1 
@w I1  3
1
@I1 1

I1  3
1
dis
¼ lR 1  ¼ lR 1  1: ð4:26Þ
@Cdis 2 Im @Cdis 2 Im

Using (4.26) in (4.22) gives

 1  
I1  3 1
S ¼ J 2=3 lR 1  1  ðtrCdis ÞC1
dis : ð4:27Þ
Im 3

Next, from (3.4)1, the contribution Tð2Þ to the Cauchy stress is

Tð2Þ ¼ J 1 FSF> ¼ J 1=3 Fdis SF>dis ; ð4:28Þ

and hence, using (4.27),
 1  
I1  3 1
Tð2Þ ¼ J 1 lR 1  Fdis F>dis  ðtrCdis ÞFdis C1 F >
dis dis :
Im 3
Then with
def
Bdis ¼ Fdis F>dis ; ð4:29Þ
 1  
I1  3 1
Tð2Þ ¼ J 1 lR 1  Bdis  ðtrBdis Þ1 ;
Im 3
or
 1
I1  3
Tð2Þ ¼ J 1 lR 1  ðBdis Þ0 : ð4:30Þ
Im

8
This model has been shown by Boyce (1996) to yield predictions for the stress–strain response similar to the entropic-network
model of Arruda and Boyce (1993).
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1505

4.3. Free energy wðpÞ

The spectral representation of A is

X
3
A¼ ai li  li ; ð4:31Þ
i¼1

where ða1 ; a2 ; a3 Þ are the positive eigenvalues, and ðl1 ; l2 ; l3 Þ are the orthonormal eigenvectors of A. The
principal invariants of A are:
9
I1 ðAÞ ¼ a1 þ a2 þ a3 ; >
=
I2 ðAÞ ¼ a1 a2 þ a2 a3 þ a3 a1 ; : ð4:32Þ
>
;
I3 ðAÞ ¼ a1 a2 a3 ¼ 1 ðsince detA ¼ 1Þ

Using (4.32), we express the defect energy as

~ ðpÞ ðIA ; #Þ ¼ w
wðpÞ ¼ w  ðpÞ ða1 ; a2 ; a3 ; #Þ: ð4:33Þ

Then, by the chain-rule

 ðpÞ ða1 ; a2 ; a3 ; #Þ X
@w 3  ðpÞ ða1 ; a2 ; a3 ; #Þ @ai
@w
¼ : ð4:34Þ
@A i¼1
@ai @A

Assume that ai are distinct, so that the ai and the principal directions li may be considered as func-
tions of A. Then,

@ai
¼ li  li ; ð4:35Þ
@A
and, granted this, (4.34) implies that

 ðpÞ ða1 ; a2 ; a3 ; #Þ X
@w 3  ðpÞ ða1 ; a2 ; a3 ; #Þ
@w
¼ li  li : ð4:36Þ
@A i¼1
@ai

Also, use of (4.31) and (4.36) in (3.6) gives the deviatoric back-stress as
!
X
3  ðpÞ ða1 ; a2 ; a3 ; #Þ
@w
Mback ¼ 2 ai li  li : ð4:37Þ
i¼1
@ai
0

Next, we consider the following simple defect energy:

h i
 ðpÞ ða1 ; a2 ; a3 ; #Þ ¼ 1 B ðln a1 Þ2 þ ðln a2 Þ2 þ ðln a3 Þ2 ;
w ð4:38Þ
4
with Bð#Þ > 0. Then
 ðpÞ ða1 ; a2 ; a3 ; #Þ 1 X
@w 3
ln ai 1
¼ B li  li ¼ Bðln AÞA1 ; ð4:39Þ
@A 2 i¼1 ai 2

where
def
X
3
ln A ¼ ln ai li  li ; ð4:40Þ
i¼1

and
X
3
A1 ¼ a1
i li  li : ð4:41Þ
i¼1
1506 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

Then, using (4.39) in (4.37) gives

Mback ¼ Bðln AÞ0 : ð4:42Þ
Note that since a1 a2 a3 ¼ 1 (cf. (4.32)3),
trðln AÞ ¼ ln a1 þ ln a2 þ ln a3 ¼ lnða1 a2 a3 Þ ¼ 0:
Hence the defect strain tensor ðln AÞ is traceless, and therefore
Mback ¼ B ln A; ð4:43Þ
we call the positive-valued constitutive parameter Bð#Þ P 0 the back-stress modulus.

4.4. Strength function. Internal variables

Here we are concerned with specializing the strength relation

p
jðMeeff Þ0 j ¼ Yðd ; KÞ with K ¼ Ce ; Bp ; A; n; # ; ð4:44Þ
appearing in (3.12).
First, we define an equivalent shear stress by
1
s def
¼ pffiffiffi jðMe eff Þ0 j; ð4:45Þ
2
and an equivalent shear-strain rate by
pffiffiffi pffiffiffi
mp def p
¼ 2d ¼ 2jDp j; ð4:46Þ
respectively.
Next, recalling (4.14) for the Mandel stress Me , we define the mean normal pressure by
 
1 1
 def
p ¼  trMe ¼ K trEe  3að#  #0 Þ ¼ K trðln Ce Þ  3að#  #0 Þ : ð4:47Þ
3 2
Also let
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kp def
¼ trBp =3 ð4:48Þ
define an effective plastic stretch. Then, as an (enormous) simplification of the theory, we assume that
the strength function Y is independent of A, and depends on Ce and Bp only through p  and 
kp .
Further, we restrict the list n of internal variables to three scalars
u; S1 ; and S2 :
The two parameters u and S1 are introduced to model the ‘‘yield-peak” of glassy polymers. A key
microstructural feature controlling the strain-softening associated with the ‘‘yield-peak” is the defor-
mation-induced disordering of glassy polymers.9 The variable u, a positive-valued dimensionless
‘‘order”-parameter, is introduced to represent such deformation-induced disordering; and S1 , a stress-
dimensioned internal variable, represents the corresponding transient resistance to plastic flow accompa-
nying the microstructural disordering. The parameter S2 , another positive-valued stress-dimensioned
internal variable, is introduced to model additional ‘‘isotropic”-hardening aspects of the stress–strain re-
sponse of these material as the chains are pulled taut between entanglements at large strains.
With these simplifications and internal variables, and using the definitions (4.45)-(4.48), we re-
write the strength relation (4.44)1 as
s ¼ gðmp ; #; p; kp ; u; S1 ; S2 Þ; ð4:49Þ

9
The deformation-induced disordering is often associated with the change in ‘‘free-volume” of glassy polymers. The ‘‘free-
volume” terminology was introduced by Cohen and Grest (1979) for simple atomic glasses (amorphous metals), and there is a
corresponding way to define it in glassy polymers (Shah et al., 1989). The deformation-induced disordering and its role in the yield
drop in amorphous materials has also been discussed recently by Argon and Demkowicz (2008) (in the context of amorphous
silicon).
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1507

; 
and assume further that at a fixed state ð#; p kp ; u; S1 ; S2 Þ the strength relation (4.49) is invertible, with
inverse
mp ¼ f ðs; #; p; kp ; u; S1 ; S2 Þ P 0: ð4:50Þ
Finally, guided by the literature (cf., e.g., Fotheringham and Cherry, 1976, 1978; Povolo and Hermi-
da, 1995, 1996; Richeton et al., 2005, 2006, 2007), for the flow function f in (4.50) we choose a ther-
mally-activated relation in the specific form
8
<0
> if se 6 0;
   1=m
mp ¼ Q sV ð4:51Þ
>
: m0 exp  k # sinh 2ke # if se > 0;
B B

where

se def
¼s   ðS1 þ S2 þ ap p
Þ ð4:52Þ
denotes a net shear stress for thermally-activated flow, and where ap P 0 is a parameter introduced to
account for the pressure sensitivity of plastic flow. Additionally, m0 is a pre-exponential factor with
dimensions of s1 , Q is an activation energy, kB is Boltzmann’s constant, V is an activation volume,
and m is a strain rate sensitivity parameter.
Some remarks: There are many models for the rate and temperature-dependent yield strength of
polymers in the literature which consider plastic flow as a thermally-activated process (cf., e.g., Eyring,
1936; Robertson, 1966; Argon, 1973). Most of these models give a reasonably acceptable representa-
tion of the variation of the yield strength with temperature and strain rate, but over limited ranges of
these variables. The flow function (4.51) used here is motivated by the recent work of Richeton et al.
(2005, 2006, 2007), who in turn base their model on the so-called ‘‘cooperative”-model of Fothering-
ham and Cherry (1976, 1978) and Povolo and Hermida (1995, 1996). Richeton et al. have shown that a
flow function of the form (4.51) may be used to satisfactorily represent the variation of the yield
strength of amorphous polymers over a wide-range of strain rates and temperatures.10 The major dif-
ference between the flow function proposed by Richeton et al. and the one considered here, is that in-
stead of a tensorial back-stress Mback (cf., (4.43)) to define an effective stress which drives plastic flow
(cf. (3.8)), they consider a temperature-dependent scalar internal stress in their theory. This results in
a profound difference between their model and the one considered here, specially in the ability of the
two models to capture unloading and cyclic loading phenomena, as well as in a proper accounting of
the energy dissipated during plastic flow. Also, the three-dimensional theory that they present in §3
of their 2007 paper is substantially different in its mathematical structure from that considered here.
Mulliken and Boyce (2006) have recently proposed an alternate model to describe the variation of
the yield strength of amorphous polymers over a wide-range of strain rates and temperatures, albeit
still for temperatures below the glass transition. Their model is a generalization of the model(s) pro-
posed by Bauwens, Bauwens-Crowet et al., and co-workers (cf., e.g., Bauwens et al., 1969; Bauwens,
1972; Bauwens-Crowet et al., 1969, 1972; Bauwens-Crowet, 1973), in which they introduce two rhe-
ological micro-mechanisms — designated as primary (or a) and secondary (or b) — which contribute to
the yield strength of the material. The primary a-mechanism represents the rotations of the main-
chain segments of the polymer, and the secondary b-mechanism represents the rotations of the ester
side groups in PMMA, and the rotations of the phenyl groups in the main chains of PC. These two
mechanisms are rate-limiting in different regimes of strain rates and/or temperatures; the a-mecha-
nism is the dominant rate-limiting mechanism at low rates (or high temperatures), and the b-mech-
anism is the dominant rate-limiting mechanism at high strain rates (or low temperatures). Mulliken
and Boyce assume that the a and b molecular processes are sufficiently decoupled, so that the overall
material response may be described by a simple superposition of the two mechanisms. Accordingly, in
the three-dimensional version of their theory, instead of using the standard Kröner (1960) decompo-
sition F ¼ Fe Fp as done here, they develop a theory which employs a decomposition of the form

10
Richeton et al. extend the flow rule (4.51) through the glass transition temperature, but in this paper we fix our attention in the
regime of temperatures below #g .
1508 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

F ¼ Fea Fpa ¼ Feb Fpb to account for the a and b mechanisms. This results in a substantially different and
more complex constitutive theory than that considered here, and also results in a doubling of the
material parameters concerning pre-exponential factors, activation energies, deformation resistances,
pressure-sensitivity parameters, and so on.
Whatever the physical merits of the multi-mechanism ða; bÞ-based model of Mulliken and Boyce,
for operational economy our preference here is not to follow their approach, but instead to adopt a
variant of the ‘‘cooperative”-model of Richeton et al. (2007), which from a macroscopic point of view
appears to achieve the same goal of being able to represent the variation of the yield strength of amor-
phous polymers over a wide-range of strain rates and temperatures.

4.5. Evolution equations for the internal variables

4.5.1. Evolution of u and S1

We assume that the material disorders, and is accompanied by a micro-scale dilatation as plastic
deformation occurs, resulting in an increase of the order-parameter u,11 and this increase in disorder
leads to a change in the resistance S1 , causing a transient change in the flow stress of the material as plas-
tic deformation proceeds. Accordingly, the evolution of the resistance S1 is coupled to the evolution of the
order-parameter u. Specifically, we take the evolution of S1 to be governed by12
)
S_ 1 ¼ H1 mp ; with initial value S1 ðX; 0Þ ¼ S1i ;
ð4:53Þ
H1 ¼ h1 S1  S1 ; and S1 ¼ b
S 1 ðmp ; #; uÞ;
and we assume that

u_ ¼ bmp with initial value uðX; 0Þ ¼ ui ;
ð4:54Þ
b ¼ gðu  uÞ; with ^  ðmp ; #Þ P 0;
u ¼ u
here b is a shear-induced disordering function.13
In these coupled evolution equations for S1 and u, the parameters h1 , g, S1i and ui are constants (pos-
sibly temperature-dependent). The function H1 represents the strain-hardening/softening function for
the resistance S1 during plastic flow: the material hardens (H1 > 0) if S1 < S1 , and softens (H1 < 0) if
S1 > S1 . The critical value S1 of S1 controlling such hardening/softening transitions is assumed to depend
on the current values of the plastic strain rate, temperature, and the order-parameter u. In the disorder-
ing function b, the parameter u represents a strain rate and temperature-dependent critical value for
the order-parameter: the material disorders (b > 0) when u < u , and becomes less disordered (b < 0)
when u > u . In a monotonic experiment at a given strain rate and temperature, the shear-induced
disordering vanishes (b ¼ 0) when u ¼ u . However, in an experiment in which the strain rate and tem-
perature are varying (e.g. strain rate or temperature jump experiments), the material will in general in-
crease or decrease in disorder, depending on the strain rate and temperature history, and because of the
coupling between the evolution equations for S1 and u, the resistance S1 will also vary.
Particular forms for the function u ^  ðmp ; #Þ and b
S 1 ðmp ; #; uÞ need to be specified. The function u con-
trols the amount of disordering the material undergoes during deformation and is both strain rate and
temperature dependent. The strain rate and temperature dependence of u is quite nonlinear; u is
expected to decrease with increasing temperature at a fixed strain rate, and increase with strain rate
at a fixed temperature. We model this temperature and strain rate dependence of u using the follow-
ing phenomenological form
8 h i s
< u 1 þ #c  # r mp for # 6 #c ;
^ p
u ðm ; #Þ ¼ r k mr ð4:55Þ
:
0 for # > #c ;

11
The micro-scale dilatation is extremely small, and at the macroscopic level we presume the plastic flow to be incompressible.
12
Coupled differential evolution equations of this type have previously been used to model yield-peaks in granular materials
(Anand and Gu, 2000), as well as amorphous polymeric materials (Anand and Gurtin, 2003) and amorphous metallic glasses
(Henann and Anand, 2008).
13
We concentrate only on deformation-induced disordering, and neglect any decrease in the degree of disorder due to
temperature-dependent recovery effects in the absence of macroscopic plastic deformation.
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1509

where #c is as strain rate dependent function given by14

( p

#g þ n ln vvr for vp > vr
#c ¼ ð4:56Þ
#g for vp 6 vr
with fur ; k; r; s; mr ; ng as constants.
Further, the function S1 , which controls the magnitude of the stress-overshoot, is taken as
S1 ¼ bðu  uÞ; ð4:57Þ
so that the value of S1
depends linearly on the difference between the current value of u and the
parameter u .
Thus, gathering the number of material parameters introduced to phenomenologically model the
yield-peak, we have the following rather large list
fh1 ; b; S1i ; g; ui ; ur ; k; r; s; mr ; ng;
with some of these parameters possessing additional temperature dependence. We note that modeling
the temperature and rate-sensitivity of the yield-peak over a wide-range of temperatures and strain rates is
notoriously difficult. If a simpler theory with fewer material parameters is desired, and if it is deemed
that modeling the yield-peak is not of interest, then there is no need to introduce the internal variables
u and S1 , and thereby also the attendant constants in their evolution equations.

4.5.2. Evolution of S2
The evolution of S2 is taken to be governed by

S_ 2 ¼ h2 ðkp  1ÞðS2  S2 Þmp with initial value S2 ðX; 0Þ ¼ S2i P 0; ð4:58Þ

where h2 is assumed to be constant and S2 ð#Þ
is a temperature-dependent material parameter. The
resistance S2 increases and the material hardens as long as S2 < S2 .

4.5.3. Evolution of A
Finally, the evolution equation for A is taken as

A_ ¼ Dp A þ ADp  cA ln Amp ; AðX; 0Þ ¼ 1; ð4:59Þ

where c P 0 is a constitutive parameter which governs the dynamic recovery of A. This evolution
equation is a generalization of the non-linear kinematic-hardening rule (Armstrong and Frederick,
1966) of the small deformation theory of classical metal viscoplasticity,15 but here applied to poly-
mer-viscoplasticity.

4.6. Summary of the specialized constitutive model. Partial differential equations for the deformation and
temperature fields

In this section, we summarize the specialized form of our theory, which should be useful in appli-
cations. We also list the partial differential equations for the deformation and temperature fields.

4.6.1. Constitutive equations

1. Free energy
We consider a separable free energy

wR ¼ wð1Þ þ wð2Þ þ wðpÞ : ð4:60Þ

14
This approximately models the rate dependence of the glass transition temperature of the material.
15
Cf., e.g., Chaboche (2008) for a recent review of the large variety of kinematic-hardening rules in classical small deformation
metal-viscoplasticity.
1510 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

With
X
3
Ue ¼ kei rei  rei ; ð4:61Þ
i¼1

denoting the spectral representation of Ue , and with

X
3
Ee ¼ Eei rei  rei ; Eei ¼ ln kei ; ð4:62Þ
i¼1

denoting an elastic logarithmic strain measure, we adopt the following special form for the free energy
wð1Þ :
 
1 2
wð1Þ ¼ GjEe j2 þ K  G ðtrEe Þ2  ð#  #0 Þð3K aÞðtrEe Þ þ ~f ð#Þ; ð4:63Þ
2 3

where ~f ð#Þ is an entropic contribution to the free energy related to the temperature-dependent spe-
cific heat of the material. The temperature-dependent parameters

Gð#Þ > 0 Kð#Þ > 0; að#Þ > 0; ð4:64Þ

are the shear modulus, bulk modulus, and coefficient of thermal expansion, respectively, and #0 is a
reference temperature.Next, with
def
I1 ¼ trCdis ð4:65Þ

denoting the first principal invariant of Cdis . We adopt the following special form for free energy wð2Þ
 
1 I 3
wð2Þ ¼  lR Im ln 1  1 ; ð4:66Þ
2 Im

where
lR ð#Þ > 0; Im ð#Þ > 3 ð4:67Þ

are two temperature-dependent material constants. In particular, lR represents the ground-state rub-
bery shear modulus of the material, and Im represents the upper limit of ðI1  3Þ, associated with lim-
ited chain extensibility.Further, with
X
3
A¼ ai li  li ; ð4:68Þ
i¼1

denoting the spectral representation of A, and with

X
3
ln A ¼ ln ai li  li ; ð4:69Þ
i¼1

denoting a defect logarithmic strain measure, we adopt a free energy wðpÞ of the form
1 h i
wðpÞ ¼ B ðln a1 Þ2 þ ðln a2 Þ2 þ ðln a3 Þ2 ; ð4:70Þ
4
where the positive-valued temperature-dependent parameter
Bð#Þ P 0; ð4:71Þ
is a back-stress modulus.
2. Cauchy stress. Mandel stress. Back-stress. Effective stress
Corresponding to the special free energy functions considered above, the Cauchy stress is given by

T ¼ Tð1Þ þ Tð2Þ ; ð4:72Þ

 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1511

with

Tð1Þ ¼ J 1 Re Me Re> ; ð4:73Þ

where

Me ¼ 2GEe0 þ KðtrEe Þ1  3K að#  #0 Þ1 ð4:74Þ

is the Mandel stress, and
 1
I1  3
Tð2Þ ¼ J 1 lR 1  ðBdis Þ0 : ð4:75Þ
Im

The symmetric and deviatoric back-stress is defined by

Mback ¼ B ln A; ð4:76Þ

and the driving stress for plastic flow is the effective stress given by
ðMeeff Þ0 ¼ Me0  Mback : ð4:77Þ
The corresponding equivalent shear stress and mean normal pressure are given by
1 1
s def  def
¼ pffiffiffi jðMeeff Þ0 j; and p ¼  trMe ; ð4:78Þ
2 3
respectively.

3. Internal variables
The internal variables of the theory
u P 0; S1 P 0; S2 P 0;
represent aspects of the intermolecular shear resistance to plastic flow. The parameter u is a dimen-
sionless order-parameter representing a local measure of disorder of the polymeric glass; S1 and S2
have dimensions of stress and, respectively, represent aspects of a transient shear resistance accom-
panying microstructural disordering, and aspects of increased shear resistance to plastic flow as the
chains are pulled taut between entanglements at large strains.

4. Flow rule
The evolution equation for Fp is
9
F_ p ¼ Dp Fp ; Fp ðX; 0Þ ¼ 1; >
>
 e  >
>
ðMeff Þ0
>
>
Dp ¼ mp ; >
>
2s
 >
>
>
=
se ¼s  ðS1 þ S2 þ ap p Þ; ð4:79Þ
8 >
<0
> if se 6 0; >
>
>
>
   1=m >
>
mp ¼ >
>
> Q
: m0 exp  k # sinh 2ke #
sV if se > 0; >
>
;
B B

where se denotes a net shear stress for thermally-activated flow; ap is a pressure-sensitivity parameter;
m0 is a pre-exponential factor with units of 1=time; Q is an activation energy; kB is Boltzmann’s constant;
V is an activation volume; and m is a strain rate sensitivity parameter.

5. Evolution equations for the internal variables S1 , u, S2 , and A

The internal variables S1 and u are taken to obey the coupled evolution equations:

S_ 1 ¼ h1 S1  S1 mp ; with S1 ¼ bðu  uÞ; and S1 ðX; 0Þ ¼ S1i ; ð4:80Þ

1512 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

9
u_ ¼ gðu  uÞvp ; with uðX; 0Þ ¼ ui ; >
>
8 h i  >
>
< u 1 þ #c  # r vp s for # 6 # ; > >
>
 p r m c > >
and u ðv ; #Þ ¼ k r =
: ð4:81Þ
0 for # > #c ;
(  >
>
p >
>
#g þ n ln mm for vp > mr ; >
>
>
where #c ¼ r >
>
p ;
#g for v 6 mr ;
with fh1 ; b; S1i ; g; ui ; ur ; k; r; s; mr ; ng as material parameters.
The evolution of S2 is taken to be governed by
S_ 2 ¼ h2 ðkp  1ÞðS2  S2 Þmp ; with S2 ðX; 0Þ ¼ S2i P 0; ð4:82Þ
where h2 and S2 are constants. Also, the evolution equation for A is taken as

A_ ¼ Dp A þ ADp  cA ln Amp ; with AðX; 0Þ ¼ 1; ð4:83Þ

where c P 0 is a constitutive parameter which governs the dynamic recovery of A.

6. Fourier’s Law:
The heat flux is taken to be governed by Fourier’s law
qR ¼ jr#; ð4:84Þ
where jð#Þ > 0 is the thermal conductivity.

4.6.2. Partial differential equations for the deformation and temperature fields
The partial differential equation for the deformation is obtained from the local force balance
DivTR þ b0R ¼ qR €v: ð4:85Þ
Also, balance of energy (3.17), when specialized, gives the following partial differential equation for
the temperature,
 
1
c#_ ¼ DivqR þ qR þ s þ Bcj ln Aj2 mp
2
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
rate of plastic dissipation
    
1 @Me 1 @S 1 @Bð#Þ
þ # Ce1 : C_ e þ # : C_ þ ðln AÞA1 : A_ ; ð4:86Þ
2 @# 2 @# 2 @#
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
“thermoelastic-coupling” terms
where
9
@Me ¼2
@Gð#Þ e @Kð#Þ
E0 þ ðtrEe Þ1  3 @ ðKð#Það#Þð#  #0 ÞÞ1; >
>
>
@# @# @# @# =
 1 ! ð4:87Þ
@S ¼ @ lR ð#Þ 1  I1  3 F1 ðBdis Þ0 F> ; >
>
@# @# Im ð#Þ >
;
and the specific heat in the theory is given by
" #
@2w
 ð1Þ ðICe ; #Þ @ 2 w
 ð2Þ ðIC ; #Þ @ 2 w
 pÞ ðIA ; #Þ
c ¼ ^cðIeC ; IC ; IA ; #Þ ¼ # þ þ : ð4:88Þ
@#2 @#2 @#2
At this stage of the development of the theory and the concomitant experimental database, the
‘‘thermoelastic-coupling” terms in (4.86) which give rise to a temperature change due to variations
of Ce , C and A are not well-characterized, nor is the dependence of the specific heat c on these quan-
tities. Much work needs to be done to characterize these dependencies. Here, as approximations, (i) we as-
sume that c  ^cð#Þ (independent of Ce , C and A), and may be obtained from experimental
measurements; and (ii) we neglect the thermoelastic-coupling terms, and assume instead that only
a fraction 0/x/1 of the rate of plastic-dissipation contributes to the temperature changes. Under
these approximative assumptions (4.86) reduces to
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1513

 
1
 þ Bcj ln Aj2
c#_ ¼ DivqR þ qR þ x s mp ; with c ¼ ^cð#Þ: ð4:89Þ
2

4.7. Temperature dependence of material parameters

Temperature dependence of the thermoelastic moduli G, K, and a:

For polymeric materials the magnitude of the elastic shear modulus G decreases as the temperature
increases, and then decreases drastically as the temperature increases through the glass transition
temperature #g of the material. For temperatures below #g we approximate the change of G with tem-
perature by

Gð#Þ ¼ G0  Mð#  #g Þ for # < #g ; ð4:90Þ

where G0 and M are constants, and #g is the glass transition temperature.

Below the glass transition temperature, the Poisson’s ratio of the material is approximated as a
constant

mpoi  constant;
and the temperature dependence of the bulk modulus K is then obtained by using the standard
relation
2ð1 þ mpoi Þ
Kð#Þ ¼ Gð#Þ  : ð4:91Þ
3ð1  2mpoi Þ
In the temperature range of interest, the coefficient of thermal expansion is also approximated to
be constant
a  constant:
Temperature dependence of the back-stress modulus B:
The back-stress modulus B is assumed to decrease linearly with temperature,
Bð#Þ ¼ Xð#g  #Þ for # < #g ; ð4:92Þ
where X > 0 is a constant.

Temperature dependence of the plastic flow parameters Q, V and m:

For temperatures below #g , the activation energy Q, the activation volume V, and the strain rate
sensitivity parameter m appearing in the thermally-activated model (4.51) are assumed to be
constants.

Temperature dependence of the material parameters in the evolution equations for u, S1 , S2 , and A
In the coupled evolution Eqs. 4.53 and 4.54 for S1 and u, the material parameters are
fh1 ; b; S1i ; g; ui ; ur ; k; r; s; mr ; ng:
We assume that all but g are independent of temperature. The parameter g in (4.54) controls the
width of the yield-peak where a higher value of g results in a narrower peak; this parameter is as-
sumed to increase linearly with temperature
gð#Þ ¼ g 1 þ g 2 # for # < #g : ð4:93Þ
In the evolution Eq. 4.58, the material parameters are h2 and S2 .
We take h2 to be a temperature
independent constant, while the saturation value S2 is taken to decrease linearly with temperature,
with S2 vanishing above #g :
S2 ð#Þ ¼ l1  l2 # for # < #g ; ð4:94Þ
with l1 and l2 constant.
1514 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

Table 1
Material parameters for PMMA, PC and Zeonex.

Parameter PMMA PC Zeonex 690R

#g ðKÞ 388 418 408
q ðkg m3 Þ 1200 1200 1010
a ðkg1 Þ 7  105 6:5  105 7  105
G0 (MPa) 296 638 482
M ðMPa K1 Þ 10 0.74 0.16
mpoi 0.35 0.37 0.40
X ðMPa K1 Þ 9.4 1.5 0.7
c 34.6 26.0 6.92
ap 0.2 0.116 0.116
m0 ðs1 Þ 2  1016 2:1  1016 3:2  1011
m 0.218 0.08 0.16
Q (J) 1:81  1019 1:46  1019 1:81  1019
V ðm3 Þ 3:655  1028 2:95  1028 1:97  1027
S1i (MPa) 0 0 0
h1 70 58 173
b (MPa) 5850 5850 5850
g1 4.92 5.66 16.17
g 2 ðK1 Þ 0.0318 0.0381 0.0693
ui 0 0 0
ur 1:0  104 6:6  104 7:2  104
k (K) 0.2 0.16 0.16
r 0.59 0.25 0.24
s 0.050 0.010 0.045
mr ðs1 Þ 5:2  104 5:2  104 5:2  104
n (K) 1.0 0.5 1.6
S2i (MPa) 0 0 0
h2 0 0.12 3.6
l1 (MPa) 0 300 75
l2 ðMPa K1 Þ 0 0.35 0.16
l0 (MPa) 0.2 4.0 3.0
N ðMPa K1 ) 20:0  102 11:4  102 6:2  102
Im 5.5 7.8 6.2
1
c0 ðJ kg K1 Þ 1710 1630 2120
1
c1 ðJ kg K2 Þ 4.1 3.6 8
j0 (Watt m1 K1 Þ 0.190 0.187 0.467
j1 0.22 0.22 0.46
x 0.65 0.8 0.8

In the evolution Eq. 4.83, the only material parameter is c; we take this to be temperature
independent.

Temperature dependence of lR and Im :

For the two material parameters lR and Im in (4.25), experimental results indicate that the rubbery
shear modulus lR decreases with increasing temperature, and the parameter Im , which is related to
limited chain extensibility, is approximately constant. The empirical function chosen to fit the exper-
imentally-observed temperature dependence of lR is
lR ð#Þ ¼ l0  Nð#  #g Þ for # < #g ; ð4:95Þ
where l0 and N are constants.

Temperature dependence of specific heat c and thermal conductivity j:

For temperatures below #g , the specific heat c and the thermal conductivity j are assumed to have
the following empirical temperature dependencies (cf., e.g. Van Krevelen, 1990; Bicerano, 1993):
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1515

200 200
0.0003 1/s Experiment 0.001 1/s Experiment
25C 25C
−− Model −− Model

150 150
True Stress (MPa)

True Stress (MPa)

50C 50C

100 100
70C 70C

50 50
90C 90C
110C 110C
0 0
0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1 1.2
True Strain True Strain
200 200
0.01 1/s Experiment 0.1 1/s Experiment
−− Model −− Model
25C
150 150
True Stress (MPa)
True Stress (MPa)

50C 25C

50C
100 100
70C 70C

50 90C 50 90C

100C 100C

0 0
0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1 1.2
True Strain True Strain

Fig. 3. Fit of the constitutive model to the experimental stress–strain curves for PMMA at various temperatures ranging from 25
to 110 °C, and strain rates ranging from 3  104 to 101 s1 . The experimental data is plotted as solid lines, while the fit is
shown as dashed lines.

cð#Þ ¼ c0  c1 ð#  #g Þ for # < #g ; ð4:96Þ

 j1
#
jð#Þ ¼ j0 for # < #g : ð4:97Þ
#g

5. Material parameters for PMMA, PC, and Zeonex

We have implemented our thermo-mechanically-coupled constitutive model by writing a user

material subroutine for the finite element program Abaqus/Explicit (2007).
The material parameters appearing in our model were calibrated by fitting the experimental
stress–strain data for PMMA, PC and Zeonex with the help of a MATLAB implementation of a one-
dimensional version of our model which is detailed in Appendix A, as well as three-dimensional finite
element simulations using a single element. Under certain circumstances, when it became necessary
to account for heat generation and thermal conduction in the simple compression experiments,16 fully
thermo-mechanically-coupled multi-element simulations were required. Our heuristic material param-
eter calibration procedure for our model is described in the Appendix. The material parameters for
PMMA, PC and Zeonex determined by using this procedure are listed in Table 1.

16
Typically to fit the experimental data at a strain rate of 0.01 s-1.
1516 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

200 200
0.001 1/s Experiment 0.01 1/s Experiment
25C 25C
−− Model −− Model

150 150
True Stress (MPa)

True Stress (MPa)

70C
70C

100 105C 100 105C

130C 130C

50 50

0 0
0 0.5 1 1.5 0 0.5 1 1.5
True Strain True Strain
200
0.1 1/s Experiment
−− Model
25C
150
True Stress (MPa)

70C

100
105C
130C

50

0
0 0.5 1 1.5
True Strain

Fig. 4. Fit of the constitutive model to the experimental stress–strain curves for PC at various temperatures ranging from 25 to
130 °C, and strain rates ranging from 103 to 101 s1 . The experimental data is plotted as solid lines, while the fit is shown as
dashed lines.

The graphical fit of the constitutive model to the experimental stress–strain curves for PMMA at
various temperatures ranging from 25 to 100 °C and strain rates ranging from 3  104 to 101 s1
is shown in Fig. 3.
The fit of the constitutive model to our experimental stress–strain curves for PC at various temper-
atures ranging from 25 to 130 °C and strain rates ranging from 103 to 101 s1 is shown in Fig. 4. The
fit of the model to the high strain rate experimental stress–strain data from Garg et al. (2008) for PC at
rates of 0:5 s1 and 3400 s1 , at an initial temperature of 25 °C, is shown in Fig. 5a.17 The corresponding
rise in the surface temperature of the compression specimens, as measured by Garg et al., and that pre-
dicted by the model are shown in Fig. 5b.
Finally, Fig. 6 shows the fit of the constitutive model to the experimental stress–strain curves for
Zeonex at various temperatures ranging from 25 to 130 °C and strain rates ranging from 3  104 to
3  101 s1 .
For all three amorphous polymers (PMMA, PC and Zeonex) studied in this paper, our continuum-
mechanical, thermodynamically-consistent, large deformation constitutive model performs accept-
ably in reproducing the following major features of the macroscopic stress–strain response of these
materials: (a) the strain rate and temperature-dependent yield strength; (b) the transient yield-peak
and strain-softening which occurs due to deformation-induced disordering; (c) the subsequent rapid
strain-hardening due to alignment of the polymer chains at large strains; (d) the unloading response

17
Also see Bjerke et al. (2002) who report on temperature rise measurements in high rate experiments on PC.
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1517

a 200 __ Experiment b 50
Experiment
−− Model
Model
40

Temperature Rise (K)

150
True Stress (MPa)

3400 1/s 3400 1/s

30
100
20
0.5 1/s
50 0.5 1/s
10

0 0
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
True Strain True Strain

Fig. 5. (a) Fit of the constitutive model to the high strain rate experimental stress–strain curves for PC at rates of 0:5 s1 and
3400 s1 , at an initial temperature of 25 °C. (b) The corresponding rise in the surface temperature of the compression
specimens. The experimental data (from Garg et al., 2008) is plotted as solid lines, while the fit is shown as dashed lines.

250 0.0003 1/s Experiment 250 0.003 1/s Experiment

−− Model −− Model
25C 25C
200 200
True Stress (MPa)

True Stress (MPa)

150 70C 150 70C

120C
100 100
130C
120C
50 130C 50

0 0
0 0.5 1 1.5 0 0.5 1 1.5
True Strain True Strain

250 0.03 1/s Experiment 25C 250 0.3 1/s Experiment

−− Model −− Model

200 200 25C

True Stress (MPa)

True Stress (MPa)

70C 70C
150 150
120C
100 130C 100 120C
130C

50 50

0 0
0 0.5 1 1.5 0 0.5 1 1.5
True Strain True Strain

Fig. 6. Fit of the constitutive model to the experimental stress–strain curves for Zeonex-690R at various temperatures ranging
from 25 to 130 °C, and strain rates ranging from 3  104 to 3  101 s1 . The experimental data is plotted as solid lines, while
the fit is shown as dashed lines.

at large strains; and (e) the temperature rise due to plastic-dissipation and the limited time for heat-
conduction for the compression experiments performed at strain rates ’0:01 s1 . Of particular note is
1518 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

the feature of the constitutive model to acceptably capture the deformation response of PC over a large
range of strain rates: from 103 to 3:4  103 s1 .18

6. Validation experiments and simulations

In order to validate the predictive capabilities of our constitutive theory and its numerical imple-
mentation, in this section we show results of some non-homogeneous experiments (which were not
used to determine the material parameters in our theory), and compare the results of some key mac-
roscopic features of the experimental results against those from the corresponding numerical simula-
tions. Our validation experiments have been performed on either PC or Zeonex. The particular
validation experiments considered below are: (i) isothermal fixed-end large-strain reversed-torsion
on PC; (ii) macro-scale isothermal plane-strain cold- and hot-forming operations on PC; (iii) macro-
scale isothermal, axi-symmetric hot-forming operations on Zeonex; (iv) a micro-scale hot-embossing
of Zeonex; and (v) high-speed normal-impact of a circular plate of PC with a spherical-tipped cylindri-
cal projectile.

6.1. Fixed-end large-strain reversed-torsion

The torsion of a solid circular bar is a seemingly simple deformation mode. However, when the
large-strain torsion is conducted with axially traction-free ends, a measurable axial extension also
develops; this fascinating and complex nonlinear phenomenon is known as the Swift-effect.19 A com-
plementary phenomenon is the development of an axial force when the ends of the bar are axially fixed
during the large-strain torsion.
Free- or fixed-end large-strain torsion experiments provided simple yet effective means for assess-
ing the validity of large-strain constitutive models for elastic–plastic materials.20 Large-strain inelastic
torsion of amorphous polymeric materials has been previously numerically studied by Wu and Van der
Giessen (1993b). Here, for purposes of validating our constitutive theory, we study fixed-end large-strain
reversed-torsion of a solid cylindrical specimen. The torsion experiment was conducted at room tempera-
ture on a PC specimen with geometry shown in Fig. 7a. In the gage section, the torsion specimen has a
diameter of D0 ¼ 31:75 mm and a gage length of L0 ¼ 8:89 mm. With / denoting the angle of twist in
radians, the shear-strain at the outer surface of the gage-section of such a specimen is
/D0
C¼ : ð6:1Þ
2L0
The reversed-cycle torsion experiment was performed on an Instron tension-torsion servohydrau-
lic machine equipped with precision-aligned hydraulic grips. The machine was programmed to fix the
axial displacement, and twist the specimen by rotating the grips relative to each other at an angular
velocity of 0:25 deg=s, which corresponds to a surface shear-strain rate of C _ ¼ 7:8  103 s1 dur-
ing the reversed-torsion experiment. The maximum surface shear strains achieved during the exper-
iment, without initiating fracture, is C/  1:4.
For the corresponding finite element simulation, we have modeled only the gage-section and the
chamfered-section of the specimen leading into the gage-section. The finite element mesh, consisting
of 4,801 ABAQUS-C3D8R elements, is shown in Fig. 7b. The deformed geometry at a surface shear
strain of C ¼ 1:4 is shown in Fig. 7c. Note that for the specific geometry of the torsion specimen used
here, the deformation is essentially confined to the gage-section of the specimen.

18
High rate data for Zeonex is not currently available. Split-Hopkinson-pressure-bar high-rate compression experiments for
PMMA have been conducted by Mulliken and Boyce (2006), but the data is unreliable because the material crazes after relatively
small strains.
19
First studied by Swift (1947) for metals.
20
For metallic materials it has been firmly established in recent years that these axial effects in large-strain torsion arise due to
the development of crystallographic texture, and that the predictions of the axial effects during torsion are strongly dependent on
the constitutive model used to predict such effects (cf., e.g., Bronkhorst et al., 1992).
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1519

a 8.89mm 31.75mm

50mm

o
30

b d 400
Experiment
300 Simulation

200

Torque (N-m)
100

−100

−200

−300

−400
−1.5 −1 −0.5 0 0.5 1 1.5
Γ

c e 6
Experiment
Simulation
4
Axial Force (kN)

−2

−4

−6
−1.5 −1 −0.5 0 0.5 1 1.5
Γ

Fig. 7. (a) Geometry of torsion specimen. (b) Undeformed finite element mesh for the torsion simulation. (c) Deformed finite
element mesh at a surface shear strain of C ¼ 1:4. (d) Torque versus surface shear-strain response for PC under reversed fixed-
end torsion. (e) Axial-force versus surface shear-strain.

Fig. 8. Schematic of the plane-strain cruciform-forging experiment.

1520 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

Fig. 7d shows an excellent agreement between the numerically-predicted and the experimentally-
measured torque versus surface shear-strain response for both forward and reversed straining. Fur-
ther, Fig. 7e shows the ability of our constitutive theory to capture the major trends of the induced
axial-force versus the surface shear-strain response for both forward and reversed-torsional straining.
Although the precise magnitudes of the axial forces are not as well-predicted as the torque response,
the prediction of the actual trends for the variation of the axial forces as the shear strain is cycled, is
quite remarkable.

6.2. Macro- and micro-scale, isothermal forming operations

6.2.1. Macro-scale, plane-strain, cold- and hot-forging of PC

Channel-die, plane-strain, cold- and hot-forging experiments were performed on PC specimens.
The plane-strain forming operation under consideration converts a cylindrical specimen with a circu-
lar cross-section into a specimen with a cross-section which is in the shape of a ‘‘cruciform”. A sche-
matic of a forging experiment is shown in Fig. 8. The PC specimens had an original diameter of
12.7 mm, and were 12.7 mm deep in the plane-strain direction, which is into the plane of the paper.
The split-dies which impart the cruciform shape to the workpiece were made from hardened tool
steel, and the interfaces between the workpiece and the dies were lubricated to minimize frictional
effects. The forging experiments were carried out at 25 and 120 °C, at a constant die-closing velocity
of 0.02 mm/s. The forging experiments at 25 °C were carried to three different die-displacement levels
of 2.8, 4.6, and 5.4 mm, while the experiment at 120 °C was only carried out to a final die-displace-
ment of 5.4 mm.
For the finite element simulation of such a process, we make use of the symmetry of the geometry
and only mesh one-quarter of the geometry, as shown in Fig. 9a. The quarter-circle of the workpiece
cross-section is meshed with 976 ABAQUS-CPE4R elements, and the cruciform-die is modeled as a ri-
gid surface. Since the physical experiment was well-lubricated, the contact between the die and the
workpiece was modeled as frictionless.
Fig. 9b compares the numerically-predicted and the experimentally-measured, load–unload force
versus displacement curves for the cruciform-forging processes at 25 and 120 °C. The agreement be-
tween the predicted and the measured force–displacement responses at 25 °C for die-displacements of
2.8, 4.6, and 5.4 mm is very good, as is the agreement between the prediction and the experimental
result for the experiment at 120 °C for a die-displacement of 5.4 mm.
After unloading, each forged specimen was sectioned, polished, and then photographed. Fig. 10a
and b compare the numerically-predicted and the experimentally-measured deformed shapes after

a b 50
Experiment
Simulation
40
Rigid Surface
Force (kN)

30

120C
20
Symmetry

25C

10

0
0 1 2 3 4 5 6
Symmetry
Displacement (mm)

Fig. 9. (a) Quarter-symmetry finite element mesh for the workpiece and the rigid surface used in the plane-strain cruciform-
forging simulations for PC. (b) Comparison of numerically-predicted and experimentally-measured force–displacement curves
for forgings at 25 and 120 °C.
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1521

Fig. 10. Comparison of numerically-predicted and experimentally-measured unloaded deformed shapes for the cruciform
forging. (a) For a forging at 25 °C at a die-displacement of 5.4 mm. (b) For a forging at 120 °C at a die-displacement of 5.4 mm. (i)
experimental macrographs; (ii) deformed meshes; and (iii) outlines of simulated shapes (thick black lines) superimposed over
the experimentally-measured shapes.

Fig. 11. Schematic of the axi-symmetric forging experiment.

a b 10
Experiment
Simulation
Rigid Surface 8
Force (kN)

6
Axis of Symmetry

90C
4

Rigid Surface
2 120C

0
0 1 2 3 4 5
Displacement (mm)

Fig. 12. (a) Half-symmetry finite element mesh for the workpiece and the rigid surfaces used in the axi-symmetric cruciform-
forging simulations for Zeonex. (b) Comparison of numerically-predicted and experimentally-measured force–displacement
curves for forgings at 90 and 120 °C.
1522 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

Fig. 13. Comparison of numerically-predicted and experimentally-measured unloaded deformed shapes for the axi-symmetric
forgings at 120 °C after die-displacements of 2 mm (top) and 4.5 mm (bottom).

unloading the test specimens at 5.4 mm of die-displacement for the forgings at 25 and 120 °C. The
agreement between numerically-predicted and experimentally-measured deformed geometries is
also quite good.

6.2.2. Macro-scale, axi-symmetric, hot-forging of Zeonex

Axi-symmetric, hot-forging experiments were performed on Zeonex specimens. The axi-symmetric
forming operation under consideration converts a cylindrical specimen with a circular cross-section
into a specimen with a circular base, a bulged-middle, and a tapered neck. A schematic of a forging
experiment is shown in Fig. 11. The Zeonex specimens had an original diameter of 10.16 mm, and
were 10.16 mm tall. The split-dies which impart the particular shape to the workpiece were made
from hardened tool steel, and (in contrast to the lubricated plane-strain forging experiments for PC)
the interfaces between the Zeonex workpiece and the dies were not lubricated. The axi-symmetric
forging experiments were carried out at 90 and 120 °C, at a constant die-closing velocity of
0.02 mm/s. The forging experiments at 90 °C were carried out to a final die-displacement of
4.5 mm, while the experiments at 120 °C were carried to two different die-displacement levels of
2 mm and 4.5 mm.
For the finite element simulation of such a process we make use of the axial-symmetry of the
geometry, and mesh only a slice of the geometry, as shown in Fig. 12a. The workpiece was meshed
with 802 ABAQUS-CAX4R elements, and the top and bottom forging dies were modeled as rigid sur-
faces; the axis of symmetry is labelled in Fig. 12a. Since no lubricant was used in the physical exper-
iment, the contact between the die and the workpiece was modeled as ‘‘rough” with full-sticking.
Fig. 12b compares the numerically-predicted and the experimentally-measured, load–unload force
versus displacement curves for the axi-symmetric-forging processes at 90 and 120 °C. The fact that
numerically-predicted loads are slightly higher than the experimentally-measured loads is to be ex-
pected because the numerical simulation assumed perfect-sticking, while in the physical experiment
the frictional conditions are less severe. Given the uncertainty in the precise frictional conditions at
the interface between the dies and the workpiece in the physical experiment, the agreement between
the predicted and the measured load-displacement responses at both temperatures is quite
reasonable.
After unloading, the specimens that were forged at 120 °C to die-displacement levels of 2 mm and
4.5 mm were photographed. Fig. 13 compares the numerically-predicted and the experimentally-mea-
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1523

a b

Fig. 14. (a) Schematic of plane-strain tool (not to scale). (b) SEM image of the metallic glass tool.

sured deformed shapes after die-displacements of 2 and 4.5 mm. The numerically-predicted shapes
are quite similar to those which were experimentally-measured.

6.2.3. Micro-scale hot-embossing of Zeonex

As a simple example of a micro-hot-embossing process, we consider the embossing of a series of
long channels into a Zeonex substrate. The pattern consists of channels which are 55 lm wide,
43:5 lm deep, and are spaced 92 lm apart. To carry out the micro-hot-embossing in the polymer, a
Zr-based metallic-glass tool with a negative of the desired channel pattern was manufactured by mi-
cro-scale thermoplastic forming (Henann and Anand, 2008). Fig. 14a shows a schematic of the pattern
of the tool, and Fig. 14b shows a SEM photomicrograph of a portion of the metallic glass tool.
The hot-embossing experiment was carried out on a servo-hydraulic Instron testing machine
equipped with heated compression platens. A 25.4 mm square and 2 mm thick sheet of Zeonex, and
a 11.7 mm square patterned metallic glass tool were aligned and placed between the heated compres-
sion platens. The embossing experiment was conducted under nominally isothermal conditions at a
temperature of 130 °C in air. The load was ramped up to 13 kN to produce a nominal pressure of
95 MPa in 10 s, and held for 2 min before unloading, after which the tool was quickly removed from
the substrate. The force-cycle for the micro-hot-embossing process is schematically shown in Fig. 15a.
Since the channels are long relative to their width, and there are a large number of them aligned in
parallel, we employ a plane-strain idealization in our numerical simulation, and consider only a single
half-segment, with suitable boundary conditions. Fig. 15b shows the finite element mesh. The Zeonex
substrate is modeled using a mesh consisting of 849 ABAQUS-CPE4R plane-strain elements, and the
metallic glass tool is modeled using an appropriately shaped rigid surface. Contact between the sub-
strate and tool was approximated as frictionless. The displacement boundary conditions on the por-
tions AD and BC of the mesh boundary are u1 ¼ 0, while on the portion CD of the mesh, u1 ¼ u2 ¼ 0
are prescribed. The predicted embossed pattern in the Zeonex after hot-embossing is shown in
Fig. 15c. The numerically-predicted pattern shown in Fig. 15c has been mirrored and repeated during
post-processing to ease comparison with the corresponding experimental result, which is shown in
Fig. 15d. The final geometry of the embossed channels predicted by the simulation agrees well with
the result from the micro-hot-embossing experiment. The simulation, Fig. 15c, predicts that at the
embossing temperature of 130 °C and nominal pressure of 95 MPa, the micro-hot-embossing should
result in channel heights which are the same as the depths in the embossing tool, but there is incom-
plete die-filling and the edges of the channels are rounded; this is also the result seen in the physical
experiment, Fig. 15d.21

21
In order to get complete die-filling, it would be ideal to conduct the hot-embossing at temperatures above the glass transition
temperature of the polymer, a regime that is of considerable practical interest for the manufacture of microfluidic devices by
micro-hot-embossing.
1524 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

Embossing Tool

a b A B

14

12

10
Force (kN)

2
2
0
0 20 40 60 80 100 120 140
1 D C
Time (s)

c d

e 100
Experiment
Simulation
Feature Height (µm)

50

0 50 100 150 200 250 300

Feature Width (µm)
Fig. 15. (a) The micro-hot-embossing was carried out at 130 °C under load control; the process force history is shown. (b) Finite
element mesh for a plane-strain simulation showing the meshed substrate and the tool modeled as a rigid surface. The
displacement boundary conditions on the portions AD and BC of the mesh boundary are u1 ¼ 0, while on the portion CD of the
mesh, u1 ¼ u2 ¼ 0 are prescribed. (c) Predicted deformed shape. (d) SEM image of the micro-channels embossed in Zeonex. (e)
Comparison of numerically-predicted channel profile (dashed line) with corresponding profilometer measurements.

We further investigated the quality of the embossed features by using optical profilometry
methods. Fig. 15e compares representative cross-sections of the embossed features in the Zeonex
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1525

Rigid Projectile
Rigid Upper Clamp

Axis of Symmetry Rigid Lower Clamp

Fig. 16. Finite element mesh used in the thermo-mechanically-coupled analysis of the plate impact experiment.

a
c

Fig. 17. (a) Final shape of the impacted plate from the experiment. (b) Corresponding numerical prediction. (c) Comparison of
traced surface profile of the specimen after impact with the numerically-calculated profile.

(circles), against the numerically-predicted channel profile (dashed line). The depth of the embossed
features closely match with the numerical prediction; note that the optical profilometry method
that we used to measure the channel profile is not capable of providing data for the sharp vertical
features.

6.3. Normal impact of a clamped circular plate of PC by a spherical-tipped cylindrical projectile

As a final validation experiment — one which is not quasi-static, conducted at high strain rates, and
is not isothermal — we consider the normal impact of a circular plate of PC with a spherical-tipped
cylindrical projectile. Experiments of this type are of substantial practical interest in the design and
testing of transparent lightweight armor.
The circular plate specimen of PC, 203.2 mm in diameter and 5.334 mm thick (with bolt-holes for
clamping), was fabricated using a water-jet machine. The PC plate was clamped (using steel clamping
plates and bolts) in an Instron-Dynatup testing machine, and subjected to normal impact by a spher-
ical-tipped cylindrical steel projectile with a mass of 80 kg at an impact velocity of 3.6 m/s. The impact
conditions were specially chosen such that the plate only deforms plastically at the high rates, and
does not fracture. The force versus time was recorded during the impact, and the impacted plate spec-
imen was recovered.
For the finite element simulation we make use of the axial-symmetry of the geometry, and mesh
only a slice of the geometry, as shown in Fig. 16. The PC plate is modeled using 304 ABAQUS-CAX4RT
reduced-integration, thermo-mechanically-coupled, axi-symmetric elements. The actual clamping
boundary conditions are modeled by rigid surfaces representing the clamping plates, but instead of
individual clamping bolts, the surface interaction between the rigid surfaces representing the
1526 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

a 35
Experiment
b
Simulation Temperature (K)
30 343
336
25 329
322
Force (kN)

20 314
307
300
15 298
10

0
0 5 10 15 20 25
Time (ms)

Fig. 18. (a) Comparison of numerically-predicted and experimentally-measured reaction force versus time response for the
projectile. (b) Contours of temperature in the deformed plate immediately after the impact.

clamping plates and the PC plate is modeled using a high Coulomb friction coefficient of 0.75; thus the
polymer is not completely constrained to remain in contact with the clamping surfaces. The spherical-
tipped cylindrical steel projectile is modeled as a rigid body with a mass of 80 kg, and given an initial
velocity of 3.6 m/s towards the plate specimen. The projectile/polymer interface is modeled as
frictionless.
Fig. 17a shows an image of a sectioned one-half of the specimen after the experiment, while
Fig. 17b shows the corresponding numerically-predicted result. The predicted deformed profile of
the polycarbonate plate is qualitatively very similar to that in the experiment. More quantitatively,
Fig. 17c shows a comparison of the traced surface profile of the specimen after impact with the numer-
ically calculated profile — the two compare very favorably. Fig. 18a shows the excellent agreement be-
tween the experimentally measured, and the simulated force-time response on the projectile — up to
the time for which the experimental data was available. Lastly, Fig. 18b shows the temperature distri-
bution in the plate 25 ms after the impact, when the projectile has rebounded and lost contact with
the plate. As expected, the temperature rise is largest under the tip of the projectile, where it increases
by approximately 45 K, from 298 to 343 K.

7. Concluding remarks

We have developed a thermo-mechanically-coupled theory for large deformations of amorphous

polymers. A specialized version of the theory has been shown to perform well in reproducing the ma-
jor intrinsic features of the macroscopic stress–strain response of three representative materials:
PMMA, PC, and Zeonex. The thermo-mechanically-coupled theory has been implemented in the finite
element program Abaqus/Explicit (2007). The predictive capabilities of the constitutive theory and its
numerical implementation have been validated by comparing the results from a suite of validation
experiments of some key macroscopic features, such as the experimentally-measured deformed
shapes and the load-displacement curves, against corresponding results from the numerical
simulations.
Some important extensions of the theory that remain to be carried out are:

1. To extend the theory to model the large-deformation response of amorphous polymeric materials
in a temperature range spanning their glass transition temperatures. Such theories are still in their
infancy (e.g., Buckley and Jones, 1995; Dooling et al., 2002; Boyce et al., 2000; Dupaix, 2003; Dupaix
and Boyce, 2007). What is needed in this temperature range is a unified constitutive framework to
model the transition from a viscoelastic–plastic solid-like response below the glass transition tem-
perature, to a rubbery-viscoelastic response above the glass transition temperature of the material.
We will present such an extension in a forthcoming paper.
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1527

2. To extend the theory to account for crazing and cavitation and to include suitable damage and fail-
ure criteria (cf., e.g., Gearing and Anand, 2004a,b).

When suitably extended, the theory should be useful not only for modeling of fracture initiation
from cracks and notches under high-rate loading, but also for modeling and simulation of a variety
of polymer processing operations, and for predicting the relationship between processing methods
and the subsequent mechanical properties of amorphous polymeric products.

Acknowledgements

This work was supported by the National Science Foundation under Grant No. DMI-0517966, and
the MST program of the Singapore-MIT Alliance. We are grateful to Mr. David Henann of our Labora-
tory for providing the metallic-glass micro-hot-embossing die. Access to the Instron-Dynatup impact
tester in the Impact and Crashworthiness Laboratory at MIT was provided by Professor Tomasz Wie-
rzbicki; the help of Mr. Carey Walters and Miss Emily Houston in conducting the impact experiments
is gratefully acknowledged. Discussions with Professor A.S. Argon concerning the free-volume concept
and its role in strain-softening of amorphous materials are gratefully acknowledged.

Appendix A. Calibration of material parameters in the constitutive model

In this appendix we briefly outline a heuristic procedure for estimating values of the material
parameters in the constitutive model. As an example, the procedure is applied to determine the mate-
rial parameters for the amorphous polymer Zeonex. For an isotropic theory such as the one presented
in this paper, it is most convenient to use an implementation of a one-dimensional version of our mod-
el (described below) in the computer program MATLAB to conduct appropriate simulations to esti-
mate the material parameters.

A.1. One-dimensional version of the constitutive theory

In this section we present an approximate one-dimensional version of the model, which substan-
tially aids in the calibration of material properties from experimental data. The approximation is pri-
marily in that we cannot account for Poisson’s type lateral contractions, and attendant volume
changes, in a one-dimensional setting. The underlying constitutive equations relate the following ba-
sic fields:

U>0 stretch
Up plastic stretch
U e ¼ UU p1 elastic part of the stretch
 ¼ ln U logarithmic strain
e ¼ ln U e logarithmic elastic strain
n ¼ ðu; S1 ; S2 Þ scalar internal variables
A>0 squared stretch-like internal variable
#>0 absolute temperature
w ¼ wð1Þ U e ; # þ wð2Þ U; # þ wðpÞ A; # free energy density
r Cauchy stress

A.1.1. Free energy. Stress. Back-stress

For wð1Þ we use a simple linear elastic form for the free energy
1
wð1Þ ¼ Eðe Þ2  Eað#  #0 Þe þ ~f ð#Þ; ð8:1Þ
2
1528 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

where Eð#Þ > 0 is Young’s modulus, a is the coefficient of thermal expansion, #0 is a reference temper-
ature, and ~f ð#Þ is an entropic contribution to the free energy related to the specific heat of the material.
This free energy contributes a component

rð1Þ ¼ Ee  Eað#  #0 Þ; ð8:2Þ

to the total Cauchy stress r.

For wð2Þ , consider first a symmetric positive definite stretch tensor U which satisfies det U ¼ 1. Let
ðU 1 ; U 2 ; U 3 Þ denote the set of principal stretches, with U 1 U 2 U 3 ¼ 1. The first invariant I1 of the squared-
stretch tensor U2 in three-dimensions is defined by
def
I1 ¼ U 21 þ U 22 þ U 23 : ð8:3Þ

In terms of I1 , the Gent (1996) free energy is

 
1 I 3
wð2Þ ¼  lR Im ln 1  1 ; ð8:4Þ
2 Im

where lR ð#Þ > 0 and Im > 3 are two material parameters, with lR representing the ground-state rub-
bery shear modulus of the material, and Im representing the upper limit of ðI1  3Þ, associated with
limited chain extensibility. With rð2Þ denoting the contribution to the Cauchy stress from this free en-
ergy, standard relations of finite deformation incompressible elasticity give the principal values of the
corresponding stress as

@wð2Þ
rð2Þ
i ¼ Ui  P; ð8:5Þ
@U i
ð2Þ
with P an arbitrary ‘‘pressure.” In simple tension/compression, with r1 rð2Þ and rð2Þ ð2Þ
2 ¼ r3 ¼ 0, we
get
  
@wð2Þ @wð2Þ @wð2Þ @I1 @I1 @wð2Þ
rð2Þ ¼ U 1  U2 ¼ U1  U2 ¼2 U 21  U 22 ; ð8:6Þ
@U 1 @U 2 @I1 @U 1 @U 2 @I1

or equivalently, with U 1 U and U 2 ¼ U 3 ¼ U 1=2 ,

@wð2Þ 2
rð2Þ ¼ 2 ðU  U 1 Þ; ð8:7Þ
@I1
and hence, for the Gent free energy (8.4),
 1
I1  3
rð2Þ ¼ lR 1  ðU 2  U 1 Þ: ð8:8Þ
Im
The total Cauchy stress in simple tension/compression is

r ¼ rð1Þ þ rð2Þ : ð8:9Þ

ðpÞ
Next, for w , consider a symmetric positive definite squared-stretch-like tensor A which satisfies
det A ¼ 1. Let ða1 ; a2 ; a3 Þ denote the set of principal values of A, with a1 a2 a3 ¼ 1. We assume a plastic
energy of the form
1 h i
wðpÞ ¼ B ðln a1 Þ2 þ ðln a2 Þ2 þ ðln a3 Þ2 ; ð8:10Þ
4
where Bð#Þ P 0 is a back-stress modulus. With rðbackÞ denoting a stress from this free energy, standard
relations of finite deformation incompressible elasticity give the corresponding principal values of the
back-stress as

@wðpÞ
rðbackÞ
i ¼ 2ai  P; ð8:11Þ
@ai
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1529

with P an arbitrary ‘‘pressure,” so that in a simple tension/compression, with rðbackÞ

1 rback ,
rðbackÞ
2
ðbackÞ
¼ r3 ¼ 0,

@wðpÞ @wðpÞ
rback ¼ 2a1  2a2 : ð8:12Þ
@a1 @a2

When the free energy is given by (8.10), (8.12) reduces to

rback ¼ Bðln a1  ln a2 Þ; ð8:13Þ

or equivalently, with a1 ¼ A, and a2 ¼ a2 ¼ A1=2 ,

3
rback ¼ B ln A: ð8:14Þ
2
In a one-dimensional setting, the driving stress for plastic flow is the effective stress given by

rð1Þ
eff ¼ r
ð1Þ
 rback ; ð8:15Þ

and the equivalent tensile stress and the mean normal pressure are

1 ð1Þ
r def ð1Þ
¼ jreff j and ¼
p r ; ð8:16Þ
3

respectively.

A.1.2. Flow rule

The evolution equation for U p is
9
U_ p ¼ Dp U p U p ð0Þ ¼ 1; >
>
>
>
Dp ¼ _ signðr Þ;
p ð1Þ
_ p P 0; >
>
>
>
def
>
=
re ¼r  ðS1 þ S2 þ ap pÞ;
ð8:17Þ
8
>
< 0 if re 6 0; >
>
>
>
   1=m >
>
_ p ¼ >
> _ 0 exp  Q
: kB #
sinh re V
2kB #
if re > 0: >
>
;

Here _ p is the equivalent tensile plastic strain rate, and re denotes a net equivalent tensile stress for
thermally-activated flow; ap is a pressure-sensitivity parameter; _ 0 is a pre-exponential factor with units
of s1 ; Q is an activation energy; kB is Boltzmann’s constant; V is an activation volume; and m is a strain
rate sensitivity parameter.
When _ p > 0, (8.17), using (8.15) and (8.16), may be inverted to give
" m #
ð1Þ 1 2kb # 1 _ p
jr  rback j þ ap rð1Þ ¼ S1 þ S2 þ sinh ; ð8:18Þ
3 V _  ð#Þ
with
 
Q
_  ð#Þ def
¼ _ 0 exp  : ð8:19Þ
kB #

A.1.3. Evolution equations for the internal variables S1 , u, S2 , and A

The internal variables S1 and u are taken to obey the coupled evolution equations:

S_ 1 ¼ h1 S1  S1 _ p ; with S1 ¼ bðu  uÞ; and S1 ð0Þ ¼ S1i ; ð8:20Þ

1530 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

9
u_ ¼ gðu  uÞ_ p ; with uðX; 0Þ ¼ ui ; >
>
8 h  i s >
>
< u 1 þ #c  # r _ p >
>
for # 6 #c ; >
>
>
and u ð ; #Þ ¼
 _p r k _ r >
=
:
0 for # > #c ; ð8:21Þ
8   >
>
>
>
< # þ n ln _ p >
_ r for  > r ;
g _p _ >
>
where #c ¼ >
>
: >
;
#g for _ p 6 _ r ;
with fh1 ; b; S1i ; ui ; ur ; k; r; s; _ r ; ng constants, and g temperature-dependent.
The evolution of S2 is taken to be governed by

S_ 2 ¼ h2 ðkp  1ÞðS2  S2 Þ_ p ; with initial value S2 ð0Þ ¼ S2i P 0; ð8:22Þ

where
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kp def
¼ ðU p2 þ 2U p1 Þ=3 ð8:23Þ

is an effective plastic stretch, h2 a constant, and S2 temperature-dependent.

Also, the evolution equation for A is taken as

A_ ¼ 2ADp  cðA ln AÞ_ p ; Að0Þ ¼ 1; ð8:24Þ

where c P 0 is a constitutive parameter which governs the dynamic recovery of A.

A.1.4. Evolution equation for temperature

For one-dimensional tests at the highest strain rates, which may be approximated as adiabatic, the
temperature is taken to evolve according to
 
1
c#_ ¼ x re þ Bcj ln Aj2 _ p : ð8:25Þ
2

A.2. Material parameter calibration

With the full three-dimensional and simplified one-dimensional version of the theory in place, we
are in position to estimate the material parameters/functions appearing in the theory by fitting the
experimental data. We illustrate our heuristic material parameter calibration procedure for Zeonex;
the procedure for PMMA and PC is essentially identical.
We have implemented the one-dimensional model of Section A.1 in MATLAB using an explicit inte-
gration scheme, and we use it to calibrate the material parameters from the experiments described in
Section 2. The one-dimensional calibration process consists of four sequential steps which are outlined
in this section. The four steps cover calibration of the following aspects of the stress–strain response:
(1) elastic modulus; (2) initial yield stress; (3) large-strain behavior; and (4) yield-peak and back-
stress.

A.2.1. Elastic modulus

For polymeric materials the magnitude of the Young’s modulus E decreases as the temperature in-
creases. We assume that the temperature dependence of the Young’s modulus may be adequately
approximated by (cf. (4.90))22
Eð#Þ ¼ E0  ME ð#  #g Þ;
where E0 and M E are constants. Using the experimental data for the E versus #, we estimate

E0 ¼ 1350 MPa; ME ¼ 0:45 MPa K1 :

22
We ignore all rate-sensitivity of the initial stiffness.
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1531

100

80

True Stress (MPa)

60

40

Yield Stress
20

0
0 0.2 0.4 0.6 0.8
True Strain

Fig. 19. Schematic showing the ‘‘yield stress” defined as the intersection of the pre-peak stress–strain curve with the back-
extrapolated tangent to the stress–stain curve at a strain of 0.4.

A.2.2. Initial yield stress

Most previous models for amorphous polymers have identified the peak stress in a stress–strain
curve from a simple compression test as a ‘‘yield stress” for the material. Since the stress-peak is asso-
ciated with the transient disordering of the material, and the actual level of a peak is very dependent
on the initial thermal history of the material, here we follow a different approach. We identify a ‘‘yield
stress” in a compression experiment as a back-extrapolated value of the intersection of the initial elas-
tic slope with the tangent to the stress–strain curve at a strain of, say, 0:4, a strain level by which all
transients of the yield-peak have died out, and the chain-locking effects giving rise to the stress–strain
curve are minimal. Accordingly, at this point in the calibration procedure we ignore the effects of the
yield-peak and define the ‘‘yield stress” as the intersection of the pre-peak stress–strain curve with the
back-extrapolated tangent to the stress–strain curve at approximately 0.4 strain; this is shown sche-
matically in Fig. 19. 23
Since

jrð1Þ  rback j ¼ ðrð1Þ  rback Þ signðrð1Þ  rback Þ

and since in a monotonic compression test

signðrð1Þ  rback Þ ¼ signðrð1Þ Þ ¼ signðrback Þ;

we have

jrð1Þ  rback j ¼ jrð1Þ j  jrback j;

and hence, from (8.18),
" m #
ap  ð1Þ 2kb # 1 _ p
1 jr j ¼ S1 þ S2 þ jrback j þ sinh : ð8:26Þ
3 V _  ð#Þ
Thus, neglecting the contribution from the internal variables S1 (which is associated with the transient
yield-peak) and the contribution from S2 (since this only manifests itself at large stretches), for fully-
developed flows when _ p  _ (taken to be positive in compression) and with jrð1Þ j ¼ ry , (8.26) gives
the following approximate expression for yield stress ry as a function of temperature # and strain rate _ :
" m #
ap  2k # _
1 ry  rback ð#Þ þ B sinh1 ; ð8:27Þ
3 V _  ð#Þ
where we have introduced the notation

23
This is a non-standard definition of the yield stress for polymeric materials.
1532 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

a 0.2 b 0.1

0.09
298 K
0.08
0.15
0.07
|σ|/ (MPa/K)

|σ|/ (MPa/K)
0.06
0.1 343 K
0.05

0.04
393 K
0.05
0.03
403 K
0.02

0 0.01
−4 −3 −2 −1 0 −4 −2 0 2 4 6
log ˙(s − 1 ) log ˙(s − 1 )
Fig. 20. (a) Ratio of compressive yield stress to temperature as a function of logarithm of strain rate. The data plotted as bullets
( ) are the yield stress values estimated from the compression experiments, and the dashed lines are estimated isotherms. (b)
Master curve constructed at 408 K by shifting the yield stress data. The shifted experimental data is plotted as triangles (M), and
the solid line indicates a fit of flow function to the master curve.

def
rback ð#Þ ¼ jrback ð#Þj: ð8:28Þ
Here, rback ð#Þ represents a temperature-dependent saturation value of the back-stress in compres-
sion.24 Because of the assumed temperature dependence (4.92) of the back-stress modulus, rback de-
creases linearly with temperature,
rback ¼ Rð#g  #Þ for # 6 #g ; ð8:29Þ
where R is a material parameter. Finally, recalling (8.19),
 
def Q
_  ð#Þ ¼ _ 0 exp  : ð8:30Þ
kB #
To summarize, from (8.27), (8.29) and (8.30), there is a list of six material parameters
fap ; V; m; R; _ 0 ; Q g ð8:31Þ
that must be calibrated from the experimental data for ry as a function of strain rate _ and tempera-
ture #. The value of the pressure-sensitivity parameter ap is not determinable from simple compres-
sion experiments alone. As reviewed by Crist (1997), for amorphous polymers the pressure-sensitivity
parameter ap in simple tension/compression for PMMA is  0:35, that for PC is  0:2, and for amor-
phous polymers is generally in the range 0:1 to 0:4. We are not aware of any data for the pressure sen-
sitivity of yield for Zeonex in the literature. Here, we assume that
ap  0:2 ð8:32Þ
for Zeonex. This reduces the list (8.31) to
fV; m; R; _ 0 ; Q g; ð8:33Þ

24
For the purpose of obtaining material parameters associated with the ‘‘yield stress,” we ignore the evolution of the back-stress
and use the temperature-dependent saturation value for the back-stress as an internal stress in the one-dimensional theory. In
order to make connection with the work of Richeton et al. (2005, 2006, 2007), one may identify rback ð#Þ with their internal stress
ri ð#Þ. Note, however, that in the work of Richeton et al., ri ð#Þ is always a positive-valued scalar internal stress which leads to
isotropic hardening, whereas in our more general theory the back-stress may in general be positive or negative, and is not only
temperature dependent, but also evolves with strain to give rise to kinematic-hardening.
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1533

which need to be calibrated from the experimental data for ry as a function of strain rate _ and tem-
perature #.
Following the back-extrapolation method of Fig. 19, values of the yield stress ry as a function of
temperature # and strain rate _ have been estimated from the compression stress–strain curves for
Zeonex in the temperature range 25–130 °C at four strain rates. The ratio of these yield stresses to test
temperatures, ry =#, as a function of the logarithm of strain rate, log10 _ are shown in the Eyring-plot of
Fig. 20a. Estimated isotherms have been drawn to visually connect the yield points for a given test
temperature. For a given temperature we have only four data points spanning a relatively narrow
strain rate range, which makes fitting the flow function (8.27) difficult. However, by utilizing the shift-
ing and superposition ideas of Richeton et al. (2005, 2006), we can form a master curve of all 16 data
points at a single reference temperature that covers a much wider range of strain rates. To obtain the
master curve, the experimental data is shifted along both axes by temperature-dependent shift factors
defined below:
 9
Horizontal shift : Dðlog10 _ Þ ¼ Hh 1
#
 #1 ;>
=
shift
 ð8:34Þ
;>
ry 1 ;
Vertical shift : D #
¼ Hv #
 #1
shift

where # is the temperature of the experiment, #shift is the temperature that the data is shifted to, and
Hh and Hv are shift parameters. Richeton et al. (2005, 2006) have argued that these shift factors may be
equated with the material parameters appearing in the cooperative flow model such that
)
Q
Hh ¼ kB ln 10
;
ð8:35Þ
Hv ¼ rback ð# ¼ 0Þ ¼ Rhg :

The master curve constructed at #shift ¼ #g ¼ 408 K using the shift factors

Hh ¼ 5:7  103 K; Hv ¼ 70 MPa;

is shown in Fig. 20b, and the values of Q and R, calculated using (8.35), are

Q ¼ 1:81  1019 J; and R ¼ 0:172MPa K1 :

For a master curve constructed at #shift ¼ #g , the back-stress term from the flow function (8.27) van-
ishes, and (8.27) simplifies to
" m #
ry 2kB ap 1 1 _
¼ 1 sinh ; ð8:36Þ
#g V 3 _  ð#g Þ
with the list of unknown parameters reduced to {_ 0 , V, m}. A non-linear least-squares fitting method
was used in MATLAB to obtain these parameters from the shifted experimental data. This gives

_ 0 ¼ 1:8  1011 s1 ; V ¼ 1:14  1027 m3 ; and m ¼ 0:16;

and the resulting fit of (8.36) to the shifted data at 408K is shown in Fig. 20b as a solid line.

A.2.3. Stress–strain response at large strains

Here, we focus on estimating the material parameters: (1) lR and Im in the expression (8.8), to-
gether with the temperature dependence of lR given in (4.95); and (2) h2 and S2 , together with the
temperature dependence of S2 given in (4.94) — parameters which account for the stress increase
associated with chain-locking at large stretches.
To begin, we neglect the transient response associated with the yield-peak and set ui ¼ S1i ¼ 0, and
correspondingly ignore the evolution Eqs. 8.20 and 8.21 for u and S1 ; we return to determining the
material parameters appearing in these coupled evolution equations later. We also ignore the evolu-
tion of the back-stress, and set it constant, using the temperature-dependent saturation value, such
that
rback ð#Þ ¼ rback ð#Þ signðrback Þ ¼ Rð#g  #Þ for # < #g ; ð8:37Þ
1534 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

and determine material parameters associated with the evolution of the back-stress later.
Below the glass transition temperature, the parameter Im is presumed to be temperature indepen-
dent, as is the parameter h2 in the evolution Eq. 8.22, while the temperature dependence of lR ð#Þ and
S2 ð#Þ is presumed to follow
lR ð#Þ ¼ l0  Nð#  #g Þ for # < #g ; ð8:38Þ
and
S2 ð#Þ ¼ l1  l2 # for # < #g ; ð8:39Þ
(cf., (4.95) and (4.94)). Using the one-dimensional MATLAB implementation of the model, together
with the material parameters estimated to this point, estimates for the desired parameter list
fl0 ; N; Im ; S2i ; h2 ; l1 ; l2 g
are relatively easily obtained by curve-fitting both the loading as well as the unloading response at
large strains for the stress–strain data at the lowest strain rate.25 A few trials give the estimates as

l0 ¼ 3 MPa; N ¼ 6:2  102 MPa K1 ; Im ¼ 6:2;

h2 ¼ 6:24; l1 ¼ 130 MPa; l2 ¼ 0:27 MPa K1 :

A.2.4. Yield-peak and back-stress evolution

Finally we calibrate material parameters associated with the yield-peak and the back-stress evo-
lution. This last step in the calibration procedure is an iterative process, and requires fitting the tran-
sient stress-overshoot in the simple compression stress–strain response together with the creep
response, iteratively, several times in order to get a good fit. The steps in the iterative procedure
are listed below.
Step 1:
The parameters related to the change of back-stress rback with strain and temperature are c and X
(cf. (8.24), (4.92)). To begin, we note that for compression

Dp ¼ _ p ; ð8:40Þ

and we may then rewrite the evolution equation for A (8.24) as

A_ ¼ ð2 þ c ln AÞA _ p : ð8:41Þ

It follows then that the saturation value of A in compression is

 
2
A ¼ exp  : ð8:42Þ
c
Combining this result with the equation for the back-stress (8.14) gives the saturation value of the
back-stress as a function of the material parameters Bð#Þ and c
Bð#Þ
rback ð#Þ ¼ 3 : ð8:43Þ
c
Equating the saturation value for the back-stress using (4.92) and (8.29)1 we obtain
Xð#g  #Þ R
3 ¼ Rð#g  #Þ ) X¼ c; ð8:44Þ
c 3
and since R has already been determined, we obtain the fixed value for the ratio X=c.

25
The internal variable S2 , together with its evolution (8.22), is essential for a proper modeling of the unloading response of the
material after large strains. We assume that the material begins in a well-annealed, ‘‘ground” state and take S2i to be zero. For
PMMA the experimental data, to which the model is fit, to does not include data at very large strains, therefore we ignore the
material parameters associated with the isotropic hardening at large strains and set h2 ; l1 ; l2 to be zero for this material.
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1535

100 100

80 80
True Stress (MPa)

True Stress (MPa)

60 60

40 40

20 20

0 0
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
True Strain True Strain
100 100

80 True Stress (MPa) 80

True Stress (MPa)

60 60

40 40

20 20

0 0
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
True Strain True Strain
Fig. 21. Schematics of the effects of the material parameters {h1 ,b, g, u } on the shape of the yield-peak. Arrows indicate
changes as the values of the respective parameters are increased.

a 30 b 0.06

25 0.05
48 MPa
Back Stress (MPa)

20 0.04
True Strain

35 MPa
15 0.03
Increasing γ
10 0.02

5 0.01

0 0
0 0.2 0.4 0.6 0.8 1 1.2 0 1000 2000 3000
True Strain Creep Time (s)

Fig. 22. (a) The dependence of the evolution of back-stress on the material parameter c: effect of sequentially doubling the
material parameter c from 5 to 160 on the back-stress for a constant ratio of B=c. (b) Creep test results under simple
compression at two stress levels below the yield-peak (solid lines), together with one-dimensional MATLAB simulations
(dashed lines).
1536 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

Fig. 23. Temperature and rate dependance of u . The arrow indicates an increase in the strain rate.

Step 2:
In this step we estimate a value for c, and calculate the corresponding value for X from (4.53) to get
an estimate for the parameters involved in the evolution of the back-stress. This leaves one with a list
of parameters fS1i ; h1 ; b; ui ; g; u g in the evolution Eqs. 8.20 and 8.21 for u and S1 to calibrate the yield-
peak.
We assume the material begins in a well-annealed ‘‘ground-state,” so that we may take the initial
value of the order-parameter u and stress like internal resistance S1 to be zero,
ui ¼ 0 and S1i ¼ 0:
To find fh1 ; b; g; u g, several simulations are performed using different values of parameters to
approximately match the shape of the yield-peak at the various strain rates and temperatures. As
an aid to the iterative curve-fitting procedure, Fig. 21 shows how the parameters fh1 ; b; g; u g affect
the shape of the yield-peak. The parameter h1 controls the initial slope of the yield-peak, the param-
eters b and u control the height of the yield-peak, while the parameter g controls the width of the
yield-peak.
Step 3:
With the parameters for yield-peak estimated, one returns to refining the values of the material
parameters in the back-stress evolution. To get refined estimates for the recovery parameter c and
the temperature sensitivity parameter X for the back-stress modulus B, we first note that c controls
the rate of saturation of the back-stress. This is shown in Fig. 22a, where the back-stress versus axial
strain response is shown for varying values of c at a constant ratio of B=c: as c increases, the back-
stress approaches its saturation value more rapidly.
The parameters c and B significantly affect the creep response of the material.26 In order to get more
refined estimates for these parameters, we turn to a limited set of available data for room-temperature
creep of Zeonex shown in Fig. 22b as solid lines. The value of c is chosen such that the creep response is
adequately represented, as shown by the dashed lines in Fig. 22b.
Steps 2 and 3 are iteratively repeated until the yield-peaks in the total stress–strain response of the
material, as well as the creep response, are satisfactorily calibrated.
Once fh1 ; b; g; u g are determined for each stress–strain curve, we have found that, to a good
approximation, the parameters h1 and b may be taken as constants; g as temperature dependent,
and u as both temperature and strain rate dependent. The temperature dependence of g was then
fit to the functional form (4.93), while the temperature and strain rate dependence of u was fit to

26
Cyclic tension-compression stress–strain curves at different temperatures may also be used to fit the back-stress parameters,
but we have not conducted the necessary extensive set of such experiments.
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1537

the functional form (8.21); Fig. 23 shows a schematic of the variation of u with temperature and
strain rate.
The material parameters for Zeonex that give a reasonable fit for the yield-peak for the range of
temperatures and strain rates under consideration, and also adequately reproduce the limited creep
data, are

h1 ¼ 300; b ¼ 10:13  103 MPa; g 1 ¼ 28; g 2 ¼ 0:12;

4
ur ¼ 7:2  10 ; k ¼ 0:16 K; r ¼ 0:24; s ¼ 0:045;
mr ¼ 3  104 s1 n ¼ 1:6; c ¼ 12; X ¼ 0:7 MPa K1 ;

A.3. Parameters for the three-dimensional model

Except for the list of parameters fm0 ; ap ; V; S1i ; h1 ; b; g 1 ; g 2 ; S2i ; h2 ; l1 ; l2 ; mr ; cg, the values of the one-
dimensional material parameters are unchanged when used in the three-dimensional equations. Not-
ing that
pffiffiffi m
sm ¼ r_ ; r ¼ 3s; _ ¼ pffiffiffi ; ð8:45Þ
3
the list of parameters fm0 ; ap ; V; S1i ; h1 ; b; g 1 ; g 2 ; S2i ; h2 ; l1 ; l2 ; mr ; cg may be converted from the one-
dimensional compression form to the three-dimensional shear form using
pffiffiffi 9
_ 0 ¼ p1ffiffi3 m0 aðcompÞ
p
ðshearÞ
¼ 3ap ; V ðcompÞ ¼ p1ffiffi3 V ðshearÞ ; >
>
pffiffiffi ðshearÞ pffiffiffi ðshearÞ pffiffiffi ðshearÞ >>
>
;>
ðcompÞ ðcompÞ ðcompÞ
S1i ¼ 3S1i ; h1 ¼ 3h 1 ; b ¼ 3b >
>
ðcompÞ
pffiffiffi ðshearÞ ðcompÞ
pffiffiffi ðshearÞ ðcompÞ
pffiffiffi ðshearÞ =
g1 ¼ 3g 1 ; g2 ¼ 3g 2 S2i ¼ 3S2i ;> ð8:46Þ
pffiffiffi ðshearÞ pffiffiffi ðshearÞ pffiffiffi ðshearÞ >>
; >
ðcompÞ ðcompÞ ðcompÞ
h2 ¼ 3h2 ; l1 ¼ 3l 1 ; l2 ¼ 3 l2 >
>
pffiffiffi >
>
;
cðcompÞ ¼ 3c ðshearÞ
; _ r ¼ pffiffi mr :
1
3

Further, to convert the temperature dependence parameters for the Young’s modulus E to those for
the shear modulus G, we use the standard relations

E0 ME
G0 ¼ ; M¼ ; ð8:47Þ
2ð1 þ mpoi Þ 2ð1 þ mpoi Þ

with mpoi assumed to be temperature independent in the temperature range under consideration.
The material parameters for the three-dimensional theory that were determined by following the
procedure described in this Appendix are listed for PMMA, PC, and Zeonex in Table 1.

References

Abaqus/Explicit, 2007. SIMULIA, Providence, RI.

Anand, L., 1979. On H Hencky’s approximate strain–energy function for moderate deformations. ASME Journal of Applied
Mechanics 46, 78–82.
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., Gu, C., 2000. Granular materials: constitutive equations and strain localization. Journal of the Mechanics and Physics
of Solids 48, 1710–1733.
Anand, L., Gurtin, M.E., 2003. A theory of amorphous solids undergoing large deformations, with applications to polymeric
glasses. International Journal of Solids and Structures 40, 1465–1487.
Anand, L., Ames, N.M., 2006. On modeling the micro-indentation response of an amorphous polymer. International Journal of
Plasticity 22, 1123–1170.
Anand, L., Ames, N.M., Srivastava, V., Chester, S., 2009. A thermo-mechanically coupled theory for large deformations of
amorphous polymers. Part 1: Formulation. International Journal of Plasticity, accepted for publication.
Argon, A.S., 1973. A theory for the low temperature plastic deformation of glassy polymers. Philosophical Magazine 28, 839–
865.
Argon, A.S., Demkowicz, M.J., 2008. What can plasticity of amorphous silicon tell us about plasticity of metallic glasses?
Metallurgical and Materials Transactions A 39A, 1762–1778.
1538 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539

Armstrong, P.J., Frederick, C.O., 1966. A mathematical representation of the multiaxial Bauschinger effect. Report RD/B/N731,
CEGB, Central Electricity Generating Board, Berkeley, UK.
Arruda, E.M., Boyce, M.C., 1993. Evolution of plastic anisotropy in amorphous polymers during finite straining. International
Journal of Plasticity 9, 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, 193–212.
Bauwens, J.C., Bauwens-Crowet, C., Homeś, G., 1969. Tensile yield-stress behavior of poly(vinyl chloride) and polycarbonate in
the glass transition region. Journal of Polymer Science 7 (Part A-2), 1745–1754.
Bauwens, J.C., 1972. Relation between the compression yield stress and the mechanical loss peak of bisphenol-A-polycarbonate
in the b transition range. Journal of Materials Science 7, 577–584.
Bauwens-Crowet, C., Bauwens, J.C., Homeś, G., 1969. Tensile yield-stress behavior of glassy polymers. Journal of Polymer Science
7 (Part A-2), 735–742.
Bauwens-Crowet, C., Bauwens, J.C., Homeś, G., 1972. The temperature dependence of yield of polycarbonate in uniaxial
compression and tensile tests. Journal of Materials Science 7, 176183.
Bauwens-Crowet, C., 1973. The compression yield behaviour of polymethyl methacrylate over a wide range of temperatures and
strain-rates. Journal of Materials Science 8, 968–979.
Bjerke, T., Li, Z., Lambros, J., 2002. Role of plasticity in heat generation during high rate deformation and fracture of
polycarbonate. International Journal of Plasticity 18, 549–567.
Bicerano, J., 1993. Prediction of Polymer Properties. Marcel Dekker Inc., New York.
Boyce, M.C., Parks, D.M., Argon, A.S., 1988. Large inelastic deformation of glassy polymers Part 1: rate dependent constitutive
model. Mechanics of Materials 7, 15–33.
Boyce, M.C., 1996. Direct comparison of the Gent and the Arruda-Boyce constitutive models of rubber elasticity. Rubber
Chemistry and Technology 69, 781–785.
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, 2183–2201.
Bronkhorst, C.A., Kalidindi, S.R., Anand, L., 1992. Polycrystalline plasticity and the evolution of crystallographic texture in fcc
metals. Philosophical Transactions of The Royal Society of London A 341, 443–477.
Buckley, C.P., Jones, D.C., 1995. Glass-rubber constitutive model for amorphous polymers near the glass transition. Polymer 36,
3301–3312.
Chaboche, J.L., 2008. A review of some plasticity and viscoplasticity constitutive theories. International Journal of Plasticity.
doi:10.1016/j.ijplas.2008.03.009.
Cohen, M.H., Grest, G.S., 1979. Liquid–glass transition, a free volume approach. Physical Review B 20, 1077–1098.
Crist, B., 1997. Yield processes in glassy polymers. In: Haward, R.N., Young, R.J. (Eds.), The Physics of Glassy Polymers, second ed.
Chapman & Hall, London, pp. 155–212.
Dooling, P.J., Buckley, C.P., Rostami, S., Zahalan, N., 2002. Hot-drawing of poly(methyl methacrylate) and simulation using a
glass-rubber constitutive model. Polymer 43, 2451–2465.
Dupaix, R.B., 2003. Temperature and rate dependent finite strain behavior of poly(ethylene terephthalate) and poly(ethylene
terephthlate)-glycol above the glass transition temperature, Ph.D. Dissertation, Massachusetts Institute of Technology.
Dupaix, R.B., Boyce, M.C., 2007. Constitutive modeling of the finite strain behavior of amorphous polymers in and above the
glass transition. Mechanics of Materials 39, 39–52.
Eyring, H., 1936. Viscosity, plasticity, and diffusion as examples of absolute reaction rates. Journal of Chemical Physics 4, 283–
291.
Fotheringham, D.G., Cherry, B.W., 1976. Comment on the compression yield behaviour of polymethyl methacrylate over a wide
range of temperatures and strain-rates. Materials Science Letters 11, 1368–1371.
Fotheringham, D.G., Cherry, B.W., 1978. The role of recovery forces in the deformation of polyethylene. Journal of Materials
Science 13, 951–964.
Gearing, B.P., Anand, L., 2004a. Notch-sensitive fracture of polycarbonate. International Journal of Solids and Structures 41, 827–
845.
Gearing, B.P., Anand, L., 2004b. On modeling the deformation and fracture response of glassy polymers due to shear-yielding
and crazing. International Journal of Solids and Structures 41, 3125–3150.
Garg, M., Mulliken, A.D., Boyce, M., 2008. Temperature rise in polymeric materials during high rate deformation. Journal of
Applied Mechanics 75, 011009 (8 pages).
Gent, A.N., 1996. A new constitutive relation for rubber. Rubber Chemistry and Technology 69, 59–61.
Henann, D., Anand, L., 2008. A constitutive theory for the mechanical response of amorphous metals at high temperatures
spanning the glass transition temperature: application to microscale thermoplastic forming. Acta Materialia 56, 3290–3305.
Kröner, E., 1960. Allgemeine kontinuumstheorie der versetzungen und eigenspannungen. Archive for Rational Mechanics and
Analysis 4, 273–334.
Mulliken, A.D., Boyce, M.C., 2006. Mechanics of the rate-dependent elastic–plastic deformation of glassy polymers from low to
high strain rates. International Journal of Solids and Structures 43, 1331–1356.
Parks, D.M., Argon, A.S., Bagepalli, B., 1985. Large elastic–plastic deformation of glassy polymers, part 1: constitutive modelling.
Report, MIT, Program in Polymer Science and Technology.
Povolo, F., Hermida, E.B., 1995. Phenomenological description of the strain rate and temperature-dependent yield stress of
PMMA. Journal of Applied Polymer Science 58, 55–68.
Povolo, F., Hermida, E.B., 1996. Temperature and strain rate dependence of the tensile yield stress of PVC. Journal of Applied
Polymer Science 61, 109–117.
Richeton, J., Ahzi, S., Daridon, L., Rémond, Y., 2005. A formulation of the cooperative model for the yield stress of amorphous
polymers for a wide range of strain rates and temperatures. Polymer 46, 6035–6043.
Richeton, J., Ahzi, S., Vecchio, K.S., Jiang, F.C., Adharapurapu, R.R., 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, 2318–2335.
 N.M. Ames et al. / International Journal of Plasticity 25 (2009) 1495–1539 1539

Richeton, J., Ahzi, S., Vecchio, K.S., Jiang, F.C., Makardi, A., 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, 7938–7954.
Robertson, R.E., 1966. Theory for the plasticity of glassy polymers. Journal of Chemical Physics 44, 3950–3956.
Shah, V.M., Stern, S.A., Ludovice, P.J., 1989. Estimation of the free volume in polymers by means of a Monte Carlo technique.
Macromolecules 22, 4660–4662.
Swift, H.W., 1947. Length changes in metals under torsional overstrain. Engineering 163, 253–257.
Van Krevelen, D.W., 1990. Properties of Polymers, third ed. Elsevier, Amsterdam.
Wu, P.D., Van der Giessen, E., 1993a. On improved network models for rubber elasticity and their applications to orientation
hardening of glassy polymers. Journal of the Mechanics and Physics of Solids 41, 427.
Wu, P.D., Van der Giessen, E., 1993b. On large strain inelastic torsion of glassy polymers. International Journal of Mechanical
Sciences 35, 935–951.