Sunteți pe pagina 1din 22

ARTICLE IN PRESS

Journal of the Mechanics and Physics of Solids 56 (2008) 1730–1751


www.elsevier.com/locate/jmps

Finite deformation thermo-mechanical behavior of thermally


induced shape memory polymers
H. Jerry Qia,, Thao D. Nguyenb, Francisco Castroa, Christopher M. Yakackia,
Robin Shandasa
a
Department of Mechanical Engineering, University of Colorado, 427 UCB, ECME 124, Boulder, CO 80309, USA
b
Department of Mechanical Engineering, The Johns Hopkins University, Baltimore, MD 21218, USA

Received 12 June 2007; received in revised form 10 December 2007; accepted 13 December 2007

Abstract

Shape memory polymers (SMPs) are polymers that can demonstrate programmable shape memory effects. Typically, an
SMP is pre-deformed from an initial shape to a deformed shape by applying a mechanical load at the temperature TH4Tg.
It will maintain this deformed shape after subsequently lowering the temperature to TLoTg and removing the externally
mechanical load. The shape memory effect is activated by increasing the temperature to TD4Tg, where the initial shape is
recovered. In this paper, the finite deformation thermo-mechanical behaviors of amorphous SMPs are experimentally
investigated. Based on the experimental observations and an understanding of the underlying physical mechanism of the
shape memory behavior, a three-dimensional (3D) constitutive model is developed to describe the finite deformation
thermo-mechanical response of SMPs. The model in this paper has been implemented into an ABAQUS user material
subroutine (UMAT) for finite element analysis, and numerical simulations of the thermo-mechanical experiments verify
the efficiency of the model. This model will serve as a modeling tool for the design of more complicated SMP-based
structures and devices.
r 2007 Elsevier Ltd. All rights reserved.

Keywords: Shape memory polymers; Constitutive modeling; Finite deformation behavior; Stress–strain behavior; Thermo-mechanical
behavior

1. Introduction

Shape memory polymers (SMPs) have been investigated intensively because of their capability to recover a
predetermined shape in response to environmental changes, such as temperatures and light irradiations
(Lendlein et al., 2005a, b; Monkman, 2000; Otsuka and Wayman, 1998; Scott et al., 2005). Compared to other
shape memory materials, such as NiTi alloy, where the reported maximum deformation due to shape change is
8%, SMPs can exhibit large deformations exceeding 400% (Lendlein et al., 2005b). This advantage permits
applications such as microsystem actuation components, biomedical devices, aerospace deployable structures,
and morphing structures (Tobushi et al., 1996; Liu et al., 2004; Yakacki et al., 2007).

Corresponding author. Tel.: +1 303 492 1270; fax: +1 303 492 3498.
E-mail address: qih@colorado.edu (H.J. Qi).

0022-5096/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jmps.2007.12.002
ARTICLE IN PRESS
H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751 1731

Predeform (Program)
Permanent Shape

TH >Tg TH >Tg TL < Tg

Deploy (Recover)
TD >Tg TL < Tg
Temporary Shape

Fig. 1. A typical thermo-mechanical loading/unloading cycle in a SMP application.

For a device design using thermally induced SMPs, the SMP will undergo a thermo-mechanical loading-
unloading cycle illustrated in Fig. 1. The SMP is isothermally predeformed (or programmed) from an initial
shape to a deformed shape by applying a mechanical load at the temperature TH. The material will maintain
its deformed shape after subsequently lowering the temperature to TL and removing the external mechanical
load. The unloaded deformed shape at TL is commonly referred to as the temporary or programmed shape.
The SMP can largely maintain this shape as long as the temperature does not change. The shape memory
effect is activated by raising the temperature to TD where the initial shape is recovered. In general, TH and TD
are above the glass transition temperature Tg, and TL is below Tg. Recent advances in polymer science has
made it possible to vary the Tg by controlling the chemistry and/or the structure of SMPs for a variety of
applications (Yakacki et al., 2007).
In principle, most polymers demonstrate a certain degree of shape memory behavior. However, in order to
achieve a highly recoverable and programmable shape change, crosslinking polymers are typically used.1 The
crosslinking can be chemical crosslinking, physical crosslinking, or macromolecular chain entanglements.
Lendlein et al. (2005b) and Liu et al. (2006) gave in-depth discussions of the underlying physical mechanism of
thermally induced shape memory effects. The shape memory effect is caused by the transition of a crosslinking
polymer from a state dominated by entropic energy (rubbery state) to a state dominated by internal energy
(glassy state) as the temperature decreases. At temperatures above the glassy transition temperature Tg,
individual macromolecular chains undergo large random conformational changes, which are constrained by
the crosslinking sites formed during material processing. Deforming the material reduces the possible
configuration and hence the configurational entropy of the macromolecular chains, leading to the well-known
entropic behavior of elastomers. After the removal of the external load at a temperature above Tg, the
tendency of the material to increase its entropy will recover the undeformed (processed) shape defined by
the spatial arrangement of crosslinking sites. However, this shape recovery can be interrupted by lowering the
temperature below Tg. There, the mobility of macromolecular chains is significantly reduced by the reduction
in free volume, and the conformational change of individual macromolecules becomes increasingly difficult.
Instead, cooperative conformational change of neighboring chains becomes dominant, and deformation thus
requires much higher energy. Therefore, the removal of the mechanical load at temperatures below Tg only
1
There are some other polymers that can be termed as ‘‘thermally induced shape memory polymers’’ but not necessarily be cross-linking
polymers. For example, the shape change of some liquid crystal elastomers can be activated by temperatures. However, the shape memory
effects in liquid crystal elastomers are due to a molecular transition between trans to cis state under proper temperature conditions. The
mechanism of shape memory effects is therefore distinct from SMPs studied in this paper. Significant research efforts have been carried on
liquid crystal elastomers, including constitutive modeling and computational implementation. The readers are referred to Wanner and
Terentjev (2003) for liquid crystal elastomers.
ARTICLE IN PRESS
1732 H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751

induces a small amount of shape recovery and most of the deformation incurred at the temperature above Tg
is retained (stored, or frozen). The shape memory effect is invoked as the temperature increases above Tg,
where the individual macromolecular chains become active again and the shape recovery mechanism described
above is permitted. In this sense, shape memory effect is simply a temperature-delayed recovery.
In the applications of SMPs, because of large and complicated deformation involved, it is highly desirable
that the deformation history of SMPs can be predicted and the recovery properties can be optimized. This
requires a finite deformation constitutive model that is based on the fundamental understanding of
structure–function relationships and can capture the thermo-mechanical response of SMPs. Most of the
existing constitutive models of SMPs have been limited to one-dimensional (1D) small deformations (Tobushi
et al., 1996; Liu et al., 2006). For example, considering the SMP as a mixture of two phases (active phase and
frozen phase), Liu et al. (2006) developed a 1D small deformation model. There, a ‘‘stored strain’’ was used to
memorize the predeformation. Note that the intensive research on shape memory alloys (SMAs) in the past
has resulted in the developments of sophisticated constitutive models for SMAs (such as Thamburaja and
Anand, 2002; Lagoudas et al., 2006, etc.). However, these models cannot be applied to SMPs because of the
fundamental differences in the underlying mechanism for shape memory effects. In this paper, thermo-
mechanical experiments were conducted to identify key features of finite deformation behaviors of SMPs.
Based on the experimental observations and the concept of phase transitions, a three-dimensional (3D) finite
deformation constitutive model that describes the thermo-mechanical response of SMPs is developed. This
model is implemented into a user material subroutine (UMAT) in the finite element software package
ABAQUS. The paper was arranged as the following. Section 2 presents finite deformation thermo-mechanical
experiments performed to explore the properties and shape memory behavior of SMPs. Based on observations
from these experiments, a 3D finite deformation constitutive model is proposed in Section 3. Comparisons
between model prediction and experimental results are presented in Section 4, and future work is discussed in
the concluding section.

2. Thermo-mechanical behavior of SMPs

2.1. Material and sample preparations

Sheets of SMPs were synthesized following the procedure in Yakacki et al. (2007). Briefly, tert-butyl acrylate
(tBA) monomer and crosslinker poly(ethylene glycol) dimethacrylate (PEGDMA) in liquid forms, and photo
polymerization initiator (2, 2-dimethoxy-2-phenylacetophenone) in powder form were mixed in a beaker
according to a pre-calculated ratio. The beaker was shaken for about 10 s to ensure a good mixture. The
solution was injected onto the surface of a specially designed glass slide or a glass tube, which then was placed
under the UV lamp for polymerization. After 10 min, the SMP material was removed from the glass slide or
tube and was put into an oven for 1 h at 80 1C. The SMP materials were machined into the proper shapes for
different experiments. In order to eliminate variations in the material properties caused by processing, the
samples from the same prepared batch were used for a series of thermo-mechanical experiments, including
dynamic mechanical analysis (DMA), isothermal uniaxial compression, and cooling/heating experiments.
Each experiment is described below, together with the corresponding result. For each type of experiment, at
least three tests were conducted to confirm the repeatability, but only result from one test was shown for the
sake of clarity.

2.2. DMA tests

DMA tests were conducted using a Perkin-Elmer DMA Tester (Model 7 Series) with Perkin-Elmer
Intracooler 2 cooling system to determine the glass transition temperature Tg of the SMP specimens. The test
procedure follows the one applied in Liu et al. (2006). A rectangular bar with dimensions of 10  2  1 mm3
was placed into a DMA three point bending device. A small dynamic load at 1 Hz was applied to the platen
and the temperature was lowered at a rate of 1 1C/min. Fig. 2 shows the storage modulus and tan d vs
temperature curves from one DMA test. It was determined that the glass transition temperature Tg of the SMP
samples was 49 1C.
ARTICLE IN PRESS
H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751 1733

103 2
Storage Modulus
tan 

1.6

Storage Modulus (MPa)


102
1.2

tan 
0.8
101

0.4

100 0
20 40 60 80 100
Temperature (°C)

Fig. 2. DMA test of the SMP dynamic mechanical properties.

0.2

0 1

-0.2 2.53×10−4

-0.4
(L/L0-1) (%)

-0.6

-0.8 1.48×10−4

1
-1

-1.2

-1.4

0 10 20 30 40 50 60 70
Temperature (°C)

Fig. 3. CTE measurement of the SMP during cooling.

2.3. CTE measurement

The coefficients of thermal expansion (CTE) were measured using the Perkin-Elmer DMA Tester.
Cylindrical samples of 10 mm in diameter and 10 mm in height were cut from the synthesized SMPs. The
sample was placed between two plates inside the DMA tester. A small compressive force of 1 mN was applied
and maintained to ensure the contact between the sample and the plates during the CTE measurement. As the
temperature was varied, the height of the sample changed due to thermal expansion. The displacement of the
top plate was recorded by the DMA tester. To measure the CTE, a sample was first heated from room
temperature (25 1C) to 70 1C at 5 1C/min and allowed to equilibrate at 70 1C for 20 min. The sample then was
cooled from 70 to 0 1C at 1 1C/min. The sample was kept at 0 1C for 10 min. It was then reheated to 70 1C at
1 1C/min. The height change of the sample during cooling and reheating was recorded.
Fig. 3 shows the measurement of CTEs during the cooling from 70 to 0 1C. The curve for heating from 0 to
70 1C shows similar result and is not presented here for the sake of clarity. In Fig. 3, the thermal expansion
strain (L/L01) is plotted versus the temperature. The reference height (L0) is the sample height at 70 1C.
ARTICLE IN PRESS
1734 H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751

The portions below and above Tg are used to obtain the slopes, which are the CTEs. In this paper, we used
linear regressions of the curve between 10 and 20 1C to obtain the CTE below Tg, and the curve between 50
and 60 1C to obtain the CTE above Tg. From Fig. 3, the CTEs were measured as a1 ¼ 2:53  104 for
temperatures above Tg and a2 ¼ 1:48  104 for temperatures below Tg.

2.4. Isothermal uniaxial compression tests

In order to investigate the large deformation behavior of SMPs under different temperature conditions,
isothermal uniaxial compression tests were conducted by using a universal materials testing machine (Instron
Model 5565 with load capacity of 5 kN). The machine is equipped with a temperature-controlled chamber
(Instron Model 3119-405-21 with a Euro 2408 controller). A thermocouple was placed close to the sample

70

60

50
-True Stress (MPa)

40
T = 0°C
30 T = 10°C
T = 20°C
20
T = 30°C

10
T = 60°C
0
0 0.25 0.5 0.75 1
-True Strain

70

rate = 0.01s-1
60 rate = 0.1s-1

50
-True Stress (MPa)

40 T = 0°C

30

20
T = 20°C

10
T = 60°C
0
0 0.25 0.5 0.75 1
-True Strain

Fig. 4. (A) Stress–strain behaviors of the SMP from isothermal uniaxial compression experiments at different temperatures. The strain
rate is 0.01/s. (B) Stress–strain behaviors of the SMP at different strain rates and at different temperatures.
ARTICLE IN PRESS
H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751 1735

surface to maintain the chamber temperature. Cylindrical samples of the same dimensions as those in the CTE
measurements were used. Previous studies showed that the effects of barreling and buckling could be reduced
as long as the ratio of sample height and diameter was within the range of 0.5–2.0 (Qi and Boyce, 2005;
Bergstrom and Boyce, 1998). In order to further reduce the effects of friction, Teflon sheets were placed
between the sample and the platens. In an isothermal test, the sample was first placed on the bottom platen;
the chamber temperature then was changed to the desired temperature; finally, the sample was given 20 min
to reach a thermal equilibrium before the uniaxial compression started. Uniaxial compression experiments
were conducted at two different strain rates of 0.01 and 0.1/s. Whilst most of the samples were tested to break,
a few samples were unloaded before their broken to observe the unloading behaviors. Only the results in
loading were presented in this paper.
Fig. 4A shows the stress–strain plots from uniaxial compression tests at T ¼ 0, 10, 20, 30 and 60 1C at the
strain rate of 0.01/s. Fig. 4B shows the stress–strain behaviors at T ¼ 0, 20, and 60 1C with two strain rates of
0.01 and 0.1/s at each temperature. The SMP demonstrates distinct behaviors at temperatures below and
above its glassy transition temperature Tg. Above Tg, it exhibits a typical hyperelastic behavior, with little
viscous effects for tested strain rates and almost no permanent deformation after unloading (unloading curve
not shown); at temperatures below Tg, it exhibits a typical glassy polymer behavior with a yield point, followed
by a softening, and then a slight hardening with significant permanent deformation observed after unloading
(unloading curve not shown). In addition, strong dependency on the mechanical loading rates can be observed
at temperatures below Tg (Fig. 4B).

2.5. Cooling/heating experiments

Cooling/heating experiments were conducted on two groups of samples for constrained recovery and free
recovery, respectively. The inset in Fig. 5 shows the mechanical loading and temperature histories for the
constrained recovery. The sample was first compressed by 20% at T ¼ 100 1C, then allowed to relax for
10 min. After that, the sample was cooled down to T ¼ 0 1C at the temperature rate of T_ ¼ 1 o C=min, then
reheated back to T ¼ 100 1C at T_ ¼ 1 o C=min. During the cooling and heating processes, the position of the
platen was maintained in its lowered position by using an extensometer, such that the sample was constrained
from recovering to its original shape during reheating. The free recovery experiment followed the same

2
Temperature

1.6
Strain
Normalized True Stress

1.2
Time
Heating
0.8

0.4 Cooling

0
0 20 40 60 80 100
Temperature (°C)

Fig. 5. Stress–temperature plot during cooling-heating for the constrained recovery case. The stress is normalized by the maximum
compressive stress immediately after loading. The inset shows the mechanical and thermal loading history for the constrained recovery.
ARTICLE IN PRESS
1736 H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751

Fig. 6. Snapshot images from the free recovery case: (A) before compression, T ¼ 100 1C; (B) immediately after compression, T ¼ 100 1C;
(C) after cooling, T ¼ 0 1C; (D) after the constraint was removed, T ¼ 0 1C; and (E) after reheating, T ¼ 100 1C.

procedure as the constrained recovery, except that the top platen was raised above its starting position after
cooling to permit a full recovery.
Fig. 5 shows the stress vs temperature plot for the case of constrained recovery. Fig. 6 shows the images at
different points of thermo-mechanical history for the free recovery case. During cooling, the constrained stress
gradually decreases to zero. This is due to two reasons: (1) the deformation at high temperatures being frozen
as the temperature passes the glassy transition temperature and (2) thermal contraction during cooling. Whilst
these two reasons cause the stress to decrease, heat transfer may delay the stress decrease. The heat transfer
problem is currently investigated by the authors. When reheating, the recovery stress does not follow the same
path as the one during cooling: It first overshoots to a much higher stress than the stress during cooling at the
same temperature, then decreases to the stress value close to cooling, follows the same path as cooling, and
finally reaches the stress value before the thermal cycle. The distinct force paths during cooling and reheating
are conjectured to be due to the structural relaxation and viscoelastic stress relaxation of the polymer
undergoing a glass transition, which depends on the temperature rate and the idling time between cooling and
reheating (McKenna, 1989). The effects of structural relaxation on the shape memory effects are currently
under investigation by the authors and are not in the scope of the model proposed in this paper.

3. Constitutive model

3.1. General considerations

As discussed previously, it is proposed that the shape memory effect of an SMP is due to two concurrent
processes: (1) the transition from rubbery behavior dominated by entropic energy at high temperatures to
glassy behavior dominated by internal energy at low temperatures and (2) the storage of the deformation
incurred at high temperatures during cooling. In the first process, the transition of energy can be described as
the change in Helmholtz energy:
H total ¼ f r ðTÞH r þ f g ðTÞH g , (1)

where Htotal is the total Helmholtz energy at a given temperature T; Hr is the Helmholtz energy of the rubber
material at TbTg; Hg is the Helmholtz energy of the glassy material at T5Tg; and fr and fg are functions of
the temperature and fr+fg ¼ 1. At TbTg, frE1 and fgE0 so that HtotalEHr; at T5Tg, frE0 and fgE1 so that
HtotalEHg. Therefore, evolving fr and fg will capture the transition of Helmholtz energy.
Noting that this energy consideration resembles a description of a first order phase transition process, for
the sake of model description, we phenomenologically assume there exist a rubbery phase (RP) and a glassy
phase in the material and the phase transition between these two phases is realized through the change of
volume fraction of each phase. Here, we point out that such a phase transition description between RP and
glassy phase is phenomenological and in a real polymer systems, there may or may not exist distinct phases.
We also note that the phase transition concept has been used previously for the development of a 1D model of
SMPs (Liu et al., 2006), where an SMP was considered as a mixture of two phases.
ARTICLE IN PRESS
H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751 1737

The second process, or the deformation storage process, implies that the glassy phase formed during cooling
may have different deformation history. We therefore further divide the glassy phase into two phases: the
initial glassy phase (IGP) and frozen glassy phase (FGP). Therefore, there are three phases in the model2:

1. Rubbery phase (RP): The RP dominates at temperatures well above the glass transition temperature Tg.
The volume fraction of the RP is fr. As the temperature decreases, the volume fraction of this phase
decreases, and vice versa. In the limit, TbTg, fr ¼ 1.0, and when T5Tg, fr ¼ 0.0.
2. Frozen glassy phase (FGP): FGP refers to the newly formed glassy phase caused by a decrease in the
temperature. The volume fraction of the FGP is fT. As the temperature decreases, the volume fraction of
this phase increases, and vice versa.
3. Initial glassy phase (IGP): IGP refers to the glassy phase in the initial configuration of the material. This
phase of the material deforms when an external load is applied at the beginning of an analysis. The volume
fraction of the IGP is fg0. Since decreasing the temperature will induce new glassy phases, which is the FGP,
the volume fraction of the IGP will not change. On the other hand, as the temperature increases, the volume
fraction of the IGP decreases. Note that in most of SMP applications, the predeformation step is performed
at TbTg. In such cases, fg0E0. However, it is also possible that one start with loading the sample at a
temperature in the vicinity of Tg. For the completeness of the model, we consider a general case where an
analysis starts from a temperature in the vicinity of Tg and fg06¼0.

The volume fractions of the phases should satisfy f g0 þ f T ¼ f g , and f r þ f g0 þ f T ¼ 1:0. Here, although fg,
fg0, and fT are the volume fractions of the glassy phases, they refer to glassy phases with different thermo-
mechanical history: fg is the total volume fraction of the glassy phase in the material at a given temperature; fg0
is the glassy phase existing since the start of an analysis; fT is the glassy phase formed during cooling. As
discussed below, due to the different thermo-mechanical history, the FGP carries the deformation only after
its formation whilst the IGP carries the deformation since an analysis starts.
With three phases defined above, the total Helmholtz energy can be written as
H total ¼ f r H r þ f g0 H g0 þ f T H T . (2)
The total stress therefore is
T ¼ f r Tr þ f g0 Tg0 þ f T TT . (3)
The evolution rule for volume fractions and the determination of deformations and stresses for individual
phases are described below. It is noted that Eqs. (1)–(3) follow the simple Voigt mixing rule for composite.
Although more sophisticated composite theories can be used to capture the transition of the energy, this paper
focuses on developing a theoretical framework that captures the shape memory effects and we therefore use
this simplest model.

3.2. Evolution rules for volume fractions

As the temperature changes, the volume fractions of individual phases will change. In this model, we only
consider the equilibrium phase separation condition, i.e., the volume fractions of RP and glassy phase are sole
functions of the temperature. The volume fraction of each phase as a function of the temperature is defined as
1
fr ¼ , (4a)
1 þ exp½ðT  T r Þ=A

1
fg ¼ 1 , (4b)
1 þ exp½ðT  T r Þ=A

2
Following the argument of dividing the glassy phase into an initial glassy phase and a frozen glassy phase, one can also distinguish the
rubbery phase into an initial rubbery phase and a softening rubbery phase. Such a distinction for the rubbery phase is only necessary when
one increases the temperature after the initial mechanical loading. However, such a case is not a typical thermo-mechanical loading in
SMP applications. We therefore do not distinguish these two rubbery phases.
ARTICLE IN PRESS
1738 H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751

where A is a parameter that characterizes the width of the phase transition zone, Tr is a reference temperature
and is close to Tg. Note that these equations are similar to the VTF functions (Fulcher, 1925; Vogel, 1921;
Tammann and Hesse, 1926), which are typically used to characterize the viscosity of a liquid as a function of
the temperature at temperatures above Tg. As in VTF equations, Tr is typically a temperature some distance
below the Tg.
The volume fraction of RP is defined by Eq. (4a) during cooling and heating. For the glassy phase, however,
the IGP and the FGP evolve differently during cooling and reheating. Prior to a thermal loading, the initial
values of the IGP and the FGP are
f g0 jt¼0 ¼ f g jt¼0 , (5a)

f T jt¼0 ¼ 0:0. (5b)


During cooling, Dfg volume fraction of the RP transforms into the glassy phase. Here, we define that the entire
Dfg becomes the FGP, i.e.,
f g0 ¼ f g jt¼0 , (6a)

f T jt¼t2 ¼ f T jt¼t1 þ Df g , (6b)


where t24t1, t2 is the time immediately after t1.
During reheating, Dfg volume fraction of glassy phase transforms into the RP. Here, both the IGP and the
FGP transform into RP. We assume the IGP and the FGP transform into the RP in a similar way. Therefore,
the partition of these two phases depends on the relative volume fraction ratio, i.e.,
f g0
Df g0 ¼ Df , (7a)
f g0 þ f T g

fT
Df T ¼ Df , (7b)
f g0 þ f T g
where Dfg0 is the volume fraction from the IGP, DfT is the volume fraction from the FGP. Therefore,
f g0 jt¼t4 ¼ f g0 jt¼t3  Df g0 , (8a)

f T jt¼t4 ¼ f T jt¼t3  Df T , (8b)


where t44t3, t4 is the time immediately after t3.

3.3. Deformations and stresses

3.3.1. Rubbery phase


As demonstrated in Section 2, the material response at the temperature above Tg shows rubber-like
hyperelastic behavior. It is therefore possible to capture this behavior using a hyperelastic model, such as the
Mooney–Rivlin model, Arruda–Boyce eight chain model, and the Ogden model, etc. For comparisons of these
models, one is referred to Treloar (1958) and a review paper by Boyce and Arruda (2000). For an isotropic
homogeneous elastomer, the Langevin chain based Arruda–Boyce eight-chain model (Boyce and Arruda,
1993) captures the hyperelastic behavior of the material up to large stretches and is used here to represent the
material behavior of the RP. The Cauchy stress tensor is defined as
pffiffiffiffiffiffi  
mr N r 1 lchain 0
Tr ¼ L pffiffiffiffiffiffi B þ kr ½J  1  3a1 ðT  T 0 ÞI, (9)
3J lchain Nr
where mr is the initial shear modulus, and Nr is the number of ‘‘rigid links’’ between the two crosslink sites
(and/or strong physical entanglements), k is bulk modulus, a1 is the thermal expansion coefficient at
temperature above Tg, and T0 is the initial temperature in an analysis. Among these parameters, mr and Nr are
the fitting parameters that can be determined by the stress–strain behavior at high temperatures; a1 is
determined from experimental measurements; the bulk modulus can be chosen to be two or three order of
ARTICLE IN PRESS
H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751 1739

Hyperelastic
spring Viscoplastic
component

Fig. 7. One-dimensional rheologic representation of viscoplastic model of a glassy polymer.

magnitude higher than mr to represent the incompressibility of rubbery behavior. In this model, F is the overall
deformation gradient and contains a small volumetric strain due to both material compressibility and thermal
expansion. The volumetric strain is taken out through F ¼ ð1=J 1=3 ÞF, where J ¼ det[F]. B is the isochoric left
qffiffiffiffiffiffiffiffiffiffi
T 0
Cauchy–Green tensor, B ¼ F F , and B ¼ B  13 trðBÞI is the deviatoric part of B. lchain ¼ I 1 =3 is the stretch
on each chain in the eight-chain network, and I 1 ¼ trðBÞ is the first invariant of B. L is the Langevin function
defined as
1
LðbÞ ¼ coth b  . (10)
b

3.3.2. Initial glassy phase


The deformation of the IGP is represented by the total deformation gradient, F. The viscoplastic behavior
of SMPs at low temperatures can be captured by many models (Simo, 1987; Govindjee and Simo, 1991;
Govindjee and Reese, 1997; Reese and Govindjee, 1998; Miehe and Keck, 2000; Lion, 2000; Boyce et al.,
1988a–c, 1989, 2001; Bergstrom and Boyce, 1998; Qi and Boyce 2005; Weber and Anand, 1990). The method
by Boyce and co-workers is used here, though the constitutive framework can accommodate other internal
variable based viscoplastic models. The viscoplastic behaviors of a polymer can be modeled by decomposing
the stress response into an equilibrium time-independent behavior and a non-equilibrium time-dependent
behavior. Fig. 7 shows a 1D rheological representation. The total stress is
Tg0 ¼ Trg0 þ Tve
g0 . (11)
The hyperelastic spring (equilibrium response) can be modeled using the Arruda–Boyce eight-chain model, as
outlined in the previous section,3 but with different material parameters, i.e.,
pffiffiffiffiffiffi !
mg N g l chain 0
Trg0 ¼ L1
pffiffiffiffiffiffi B̄ þ kg ½J  1  3a2 ðT  T 0 ÞI. (12)
3J lchain Ng
Note here, since the Eq. (12) is for the material behaviors below Tg, the thermal expansion coefficient at
temperatures below Tg a2 is used here. For the viscoplastic deformation, the elastic deformation gradient is
determined from the multiplicative decomposition of F into elastic and viscoplastic contributions,
1
Fe ¼ FFv , (13)

3
Since hyperelastic models (with different material parameters) are used for both RP and IGP, one can combine these two into one
model, which will reduce the total number of material parameters. However, as discussed in Appendix B, the hyperelastic model used in
the glassy polymer model and the hyperelastic model for rubbery behavior captures different features of stress-strain behaviors of the
material at different temperatures, using two separate models offers the flexibility to capture the more general behavior of SMPs. In
addition, the short range deformation mechanisms of the glass and long range deformation mechanisms of the rubber are fundamentally
different and warrant different descriptions.
ARTICLE IN PRESS
1740 H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751

where Fv is a relaxed configuration obtained by elastically unloading Fe. The stress due to viscoplastic
deformation can be calculated using Fe, i.e.,
1 e
Tve
g0 ¼ ½L : Ee  a2 ð3lg þ 2G g ÞðT  T 0 ÞI, (14a)
Je
where J e ¼ detðFe Þ, Ee ¼ ln Ve , Ve ¼ Fe Re , and Le is the fourth order isotropic elasticity tensor and
Le ¼ 2G g I þ lg I  I, (14b)
where Gg and lg are Lamé constants, I is the fourth order identity tensor and I is the second order identity
tensor.
To evolve Fe in Eq. (14a), we apply the decomposition of the spatial velocity gradient
_ 1 ¼ F_ e Fe1 þ Fe l v Fe1 ,
l ¼ FF (15)
v v1
_ v _
where F is the material velocity gradient, l ¼ F F v
is the spatial velocity gradient. The l can be further
decomposed into:
v
l v ¼ F_ Fv1 ¼ Dv þ Wv , (16)
v v v
where D and W are the rate of stretching and the spin, respectively. We take W ¼ 0 without losing
generality in the isotropic case as shown in Boyce et al. (1988b). The viscoplastic stretch rate Dv is
constitutively prescribed to be
g_ v 0
Dv ¼ pffiffiffi Tg0 , (17)
2t̄
where Tg0 is the stress acting on the viscoplastic component convected to its relaxed configuration
Tg0 ¼ Re Tg0 Re , the prime denotes the deviator; t̄ is the equivalent shear stress defined as
h i
0 0 1=2
t̄ ¼ 12Tg0  Tg0 . (18)

g_ v Denotes the viscoplastic shear strain rate and is constitutively prescribed to take the form
    
v DG t
g_ ¼ g_ 0 exp  1 , (19)
kT s
where g_ 0 is the pre-exponential factor, DG is the zero stress level activation energy, and s is the athermal shear
strength, which represents the resistance to the viscoplastic shear deformation in the material. To further
consider the softening effects observed in the experiments, the evolution rule for s as defined below can be
used,
s_ ¼ h0 ð1  s=ss Þ_gv , (20a)
and the initial condition
s ¼ s0 when gv ¼ 0, (20b)
where s0 is the initial value of athermal shear strength, and ss is the saturation value. h0 is a parameter, and g_ v
is defined in Eq. (19). When s04ss, Eqs. (20) represent an evolution rule that characterizes a softening
behavior of a material.

3.3.3. Frozen glassy phase


During cooling, new glassy phase will be formed (‘‘frozen’’ from the RP). As discussed earlier, the
deformation in the RP is also ‘‘frozen’’, implying that the newly formed glassy phase does not inherit the
deformation of the rubber phase and will behave as an undeformed material, i.e., the newly formed glassy
phase takes the current configuration of the RP as its initial configuration. The initial deformation of the FGP
upon formation is zero. However, due to the 3D nature of deformation, the vanishing deformation of the
transforming the RP will cause a redistribution of deformation in the material, which in turn will cause a new
deformation in the FGP. Here, we emphasize that this new deformation is very different from the overall
Table 1
Summary of the model

H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751


Total stress T ¼ frTr+fg0Tg0+fTTT

Phases Rubbery phase fr Initial glassy phase: fg0 Frozen glassy phase: fT

ARTICLE IN PRESS
Volume fraction evolution rule 1 1
fr ¼ fg ¼ 1
1 þ exp½ðT  T r Þ=A 1 þ exp½ðT  T r Þ=A
8 8
<0
> cooling > Df
< g
cooling
Df g0 ¼ f g0 Df heating Df T ¼ fT
>
: fg g : f Df g
> heating
g

(
Deformation gradient F F Fnþ1 ðFn Þ1 if DTa0
DFnþ1
T ¼ and Fnþ1
T ¼ DFnþ1
T FT
n
I if DT ¼ 0

pffiffiffiffiffi
  pffiffiffiffiffi !
Stress mr N r 1 lchain 0 Ng lchain Same as IGP but using FT
Tr ¼ 3J lchain L
pffiffiffiffiffiffi B̄ þ kr ½J  1  3a1 ðT  T 0 ÞI m
Trg0 ¼ 3Jg
0
L1 pffiffiffiffiffiffiffi B̄ þ kg ½J  1  3a2 ðT  T 0 ÞI
Nr lchain
Ng
1 e
Tve
g0 ¼ ½L : Ee  a2 ð3lg þ 2G g ÞðT  T 0 ÞI
Je    
v DG t̄
g_ ¼ g_ 0 exp  1
kT s

1741
ARTICLE IN PRESS
1742 H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751

deformation of the material. Since this new deformation is due to the redistribution of overall deformation,
the incremental deformation gradient for the FGP DFnþ1
T is
(
nþ1
Fnþ1 ðFn Þ1 if DTa0;
DFT ¼ (21)
I if DT ¼ 0;

where Fn and Fn+1 are the overall deformation gradient at the increment n and n+1, respectively. Here, we
only consider a monotonic thermal loading, i.e., the cooling or reheating processes are not interrupted by
isothermal mechanical loading. This implies that during the cooling or reheating processes the only
deformation is due to the temperature change and mechanical constraints. Note that such a monotonic
thermal loading is a typical loading condition in most SMP applications.
Strictly speaking, the FGPs formed at different time during the cooling process have a different deformation
history. Computationally, tracking the phases formed at different time will be prohibitively expensive.
Therefore, we simplify this process by assuming a temporal and spatial average, i.e., we assume that all the
FGPs have the same deformation gradient and therefore will not distinguish the FGPs formed at different
time. The total deformation gradient acting on the FGP is

Fnþ1
T ¼ DFTnþ1 FnT , (22)
where FnT and Fnþ1T are the deformation gradient for the FGP at the increment n and n+1, respectively. It is
noted that in a typical SMP applications, the programming is achieved by holding the sample in a deformed
shape then lowering the temperature. In this process, the new deformation in the FGP should be small as
compared to the deformation imposed to the material at high temperatures. Therefore, Eq. (22) provides a
reasonable approximation of the deformation gradient in the FGP.
The stress in the FGP can then be calculated using Eqs. (9)–(20), with FT instead of F in Eqs. (9) and (13).
It is also noted that the glassy phases formed at different temperature and time history may have
slightly different mechanical properties. In this model, for the sake of simplicity, we assume all the glassy
phases have the same mechanical properties, and therefore, the material parameters for all the glassy phases
are the same.
The definition of the FGP plays a key role in capturing the shape memory effects. The advantage of this
definition is that it does not require the introduction of a 3D finite deformation equivalent ‘‘stored strain’’,
which was used in a 1D small deformation constitutive model (Liu et al., 2006).

3.4. Summary of the model

As demonstrated in the previous section, SMPs demonstrate very complicated thermo-mechanical


behaviors. To capture these behaviors, a comprehensive model is necessary. Table 1 summarizes the model.
One important feature of this model is that it considers separately material behaviors at temperatures above
and below of Tg. This allows the use of other type of models for rubbery behavior and glassy behavior and
capture the shape memory effects. This feature also simplifies the process of material parameter identification.
For example, as shown in Appendix A, using the experimental stress–strain curve at the temperature well
above Tg, the material parameters associated with the RP can be identified. Also, using the experimental
stress–strain curve at the temperature well below Tg, the material parameters associated with the glassy phase
can be identified.

4. Results

The constitutive model was implemented in ABAQUS as a UMAT. In this section, finite element
simulations using this UMAT are compared with the experimental results presented in Section 2. Following
the procedure described in Appendix A, the material parameters in the model were identified and listed in
Table 2.
ARTICLE IN PRESS
H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751 1743

Table 2
Material parameters used in the simulations

Model components Material parameters Values

Thermal expansion coefficients a1(K1) 2.53  104


a2(K1) 1.48  104
Volume fraction evolution A 7.0
Tr (K) 297
Rubbery phase mr (MPa) 0.8
Nr 17
kr (MPa) 1  103
Glassy phase mg (MPa) 0.1
Hyperelastic spring Ng 17
kg (MPa) 1  103

Viscoplastic component Gg (MPa) 0.47  103


lg (MPa) 1.89  103
DG (  1019) 0.92
g_ 0 (s1) 52
s0 (MPa) 56
ss (MPa) 28
h0 (MPa) 400

4.1. Isothermal uniaxial compression simulation

Fig. 8 shows the comparison of the results from numerical simulations and experiments of isothermal
uniaxial compressions at different temperatures and strain rates. In these simulations, four asymmetric four
node elements were used. The uniaxial deformation was imposed by using a rigid surface to compress the top
surface of the SMP. It is noted that since the deformation in this case is homogenous, the choice of number of
elements does not affect the simulation results. It can be seen that the model captures the isothermal
stress–strain behavior of the material at different temperatures and different strain rates.

4.2. DMA simulations to determine T g

The temperature-dependent mechanical properties of the proposed model were evaluated by simulating a
DMA experiment dynamic under uniaxial compression and tension. In the simulations, a cylindrical sample
(modeled by four axisymmetric four node elements) was subjected to an isothermal uniaxial cyclic loading
following a sinusoidal waveform at 1 Hz with a maximum strain of 4%. The imposed strain was achieved by
imposing a uniform vertical displacement on the top surface of the model. Fig. 9 shows tan d vs temperature
behavior from the experiment and the numerical simulations. tan d was obtained from the phase difference
between the imposed strain and the resulting stress. Due to the difference between the model (uniaxial loading)
setup and experimental setup (three point bending), tan d values were normalized by the maximum values in
the experiment and numerical simulation. From Fig. 9, the model shows the general features of a tan d curve
which starts from a small value and increases to a peak then decreases. The numerical simulations predict a Tg
of 53 1C, which is slightly higher than the experimentally obtained Tg of 49 1C. Note that DMA test result
was not used to obtain the material parameters in the model and hence provides a verification of the model
prediction.

4.3. Free recovery

The numerical simulation of a free recovery experiment as described above was conducted to demonstrate
the shape memory effect of the model. Here, the same model for uniaxial compression simulations was used.
Fig. 10 shows the comparison of images from experiment and numerical simulations at different points in the
ARTICLE IN PRESS
1744 H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751

70
Experiments
Simulations
60

50

-True Stress (MPa)


T = 0°C
40
T = 10°C
30
T = 20°C
20
T = 30°C

10
T = 60°C
0
0 0.25 0.5 0.75 1
-True Strain

70
Experiments
Simulations
60
.
T = 0°C  = 0.1s-1
50
-True Stress (MPa)

.
T = 0°C  = 0.01s-1
40

30

20

.
10 .
T = 20°C  = 0.01s-1 T = 20°C  = 0.1s-1

0
0 0.25 0.5 0.75 1
-True Strain

Fig. 8. Numerical simulations of stress–strain behavior of the SMP from isothermal uniaxial compressions: (A) at different temperatures
and (B) at different strain rates. The figures also show the comparisons with experiments.

thermo-mechanical history. Fig. 11 shows the comparison of the deformation recovery measured from the
images taken during the experiment and from numerical simulations. It can be seen that the constitutive model
captures the shape memory effects after cooling and the shape recovery after reheating.

4.4. Constrained recovery

The numerical simulation of a constrained recovery experiment as described above was also conducted to
characterize the force recovery feature of the constitutive model. Fig. 12 shows the comparison between the
numerical simulation and the experiment. The stress was normalized by the maximum compressive stress
immediately after loading at the high temperature. Numerical simulation predicts the decrease of the force
during cooling and the increase of the force to the level before cooling during heating. It also captures that the
stress recovery after the cooling-heating cycle. However, the simulation predicts the zero reaction force at a
ARTICLE IN PRESS
H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751 1745

1.2 Experiment
Simulation

tan / (tan)max
0.8

0.6

0.4

0.2

0
0 20 40 60 80 100
Temperature (°C)

Fig. 9. tan d vs temperature behavior from the experiment and the numerical simulations.

Rigid Plate

SMP

Fig. 10. Snapshot images of SMP deformations during the free recovery experiment and simulation. A and A0 : before compression,
T ¼ 100 1C; B and B0 : immediately after compression, T ¼ 100 1C; C and C0 : after cooling, T ¼ 0 1C; D and D0 : after the platen was
removed, T ¼ 0 1C; E and E0 : after reheating, T ¼ 100 1C. The prime denote images from the simulation.

higher temperature (33 1C) than the temperature observed in the experiment (7 1C). In addition, during
heating, a large overshoot in the stress was observed in the experiments but is not captured by the model. As
discussed in Section 2.5, the stress decrease during cooling may be delayed by the heat transfer and the stress
overshoot observed in the experiments is conjectured to be caused by structural relaxation during glass
transition, which are not considered in current model. Heat transfer and structural relaxation are currently
under investigations by the authors and will be included into the model in the future.

5. Conclusions

The finite deformation thermo-mechanical behaviors of thermally induced SMPs were investigated in this
paper. It was shown experimentally that SMPs undergo dramatic stress–strain behavior change from a
rubbery hyperelastic behavior at temperatures above Tg to a glassy viscoplastic behavior at temperatures
below Tg. These results were applied to guide the development of a 3D finite deformation constitutive model.
The model employed the concept of the first order phase transition to describe the change in the deformation
mechanism from entropic elasticity of the rubbery state to viscoplasticity of the glassy state as the temperature
ARTICLE IN PRESS
1746 H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751

1 Experiment
Simulation

0.96

Stretch Ratio
0.92 Heating

0.88

0.84

0.8
0 20 40 60 80 100
Temperature (°C)

Fig. 11. Deformation recovered during reheating in the free recovery case.

Experiment
Simulation
1.6
Normalized True Stress

1.2 Heating

Heating
0.8

0.4 Cooling
Cooling

0
0 20 40 60 80 100
Temperature (°°C)

Fig. 12. Stress response in the constrained recovery case. The stress is normalized by the maximum compressive stress immediately after
loading.

traverses Tg. It was assumed further that the newly formed FGP does not inherit the deformation of the
transforming rubber phase; the new deformation incurred by the FGP was modeled as a redistribution of total
deformation. The advantage of this assumption is that it eliminates the requirement of defining a ‘‘stored
strain’’ used in a previous 1D small deformation constitutive model. The model was shown to capture the
shape memory effects and shape recovery of SMPs. A significant feature of this constitutive framework is that
one is not limited to the models for elastomers and glassy polymers used in this paper; in fact, one can use
other models that are suitable for the specific material systems. This feature also allows modeling many other
material systems that demonstrate the shape memory effect.
The model in the paper was implemented into a UMAT in ABAQUS. Numerical simulations of uniaxial
compression experiments under isothermal conditions using this UMAT agreed with the experimentally
obtained stress–strain behavior. For the case with thermal cooling/heating, the model captured the shape
memory effects and shape recovery. However, the model did not fully capture the stress recovery, implying a
complicated transition behavior that was not captured by the simple volume fraction evolution rule proposed
ARTICLE IN PRESS
H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751 1747

in the current model. In addition, the model does not account the temperature rate effects to the thermo-
mechanical behaviors of SMPs. The effects of thermal load are currently under investigation and will be
reported in the future.
This model enables the investigations of more complicated 3D deformation behaviors of SMPs, which are
common in most of material applications. We are currently investigating a few representative 3D deformation
cases, such as indentation of SMPs; the results from these studies will be presented in the future.

Acknowledgments

The authors gratefully acknowledge the support from a NIH Grant (EB 004481), a NSF career award to
HJQ (CMMI-0645219), NSF-Sandia initiative (Sandia National Laboratories, 618780), Laboratory Directed
Research and Development program at Sandia National Laboratories (105951), and US Army Research
Office (W31PQ-06-C-0406). Discussions with Professors Martin Dunn, Kenneth Gall, and Drs. Richard Vaia,
Jeffery Baur, Richard Hall, and Mr. Jason Hermiller are gratefully appreciated.

Appendix A. Guideline for material parameter identification

In the model developed in this paper, there are 17 material parameters, among which 15 are fitting
parameters. Here, we provide a procedure to estimate these parameters, which serve as initial values. The final
values of these parameters should be subject to fitting experimental curves, but should be close to the initial
estimated values.
In order to identify the material parameters, three types of experiments should be conducted. The first
experiment is a DMA test to determine the glass transition temperature Tg of the material. Although
determining Tg does not directly identify the material parameters in the model, it provides a guide to decide
the temperature ranges used in the following two types of experiments. The second experiment is to determine
the thermal expansion coefficient. The third type of experiment is isothermal uniaxial compression experi-
ments at different strain rates and at different temperatures ranging from well below Tg to well above Tg.

A.1. Coefficients of thermal expansion (CTE)

As discussed above, CTE experiences a change as the temperature traverses through the Tg. The CTE at
temperatures above Tg gives a1, and the CTE at temperatures below Tg gives a2.

A.2. mr and N r

mr And Nr are the parameters that characterize the hyperelastic behavior of the material at high
temperatures, therefore can be determined by fitting the isothermal uniaxial compression stress–strain curve at
temperatures above Tg. For example, in this paper, we used the stress–strain curve at T ¼ 80 1C to obtain
mr ¼ 0.8 MPa and Nr ¼ 17.

A.3. G g , lg , and mg

Lamé constants Gg and lg are related to Young’s modulus Eg and Poisson’s ratio vg by Gg ¼ Eg/(2+2vg)
and lg ¼ Egvg/b(1+vg)(12vg)c. For polymers, Poisson’s ratio is typically 0.4, we therefore take vg ¼ 0.4 so
that Gg ¼ 0.36Eg and lg ¼ 1.43Eg. Since Eg and mg characterize the initial modulus of glassy behavior, the
slope of the isothermal uniaxial compression curve at the temperature well below Tg can be used. In this paper,
the curve at T ¼ 0 1C was used to obtain Eg+3mgE1300 Mpa.

A.4. A and T r

The parameters A and Tr in the volume fraction evolution rule Eq. (4) can be determined from the initial
modulus measurements. From Eqs. (3), (9), and (14a), the initial Young’s modulus under the isothermal
ARTICLE IN PRESS
1748 H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751

103 Experiment
Simulation

Initial Modulus (MPa)


102

101

100
0 25 50 75 100
Temperature (°C)

Fig. A1. Dependence of initial modulus on temperatures.

condition is
E ¼ 3f r mr þ f g ðE g þ 3mg Þ. (A.1)
Combining with Eq. (4), we have
 
1 1
E¼ 3m þ 1  ðE g þ 3mg Þ. (A.2)
1 þ exp½ðT  T r Þ=A r 1 þ exp½ðT  T r Þ=A
Fig. A1 shows a plot of the initial Young modulus (small-strain) as a function of the temperature for the
material studied in this paper. Tr and A can then be identified by fitting the curve using Eq. (A.2), which yield
Tr ¼ 24 1C, and A ¼ 7 (see Fig. A1). It is also noticed that Eq. (A.2) cannot provide a best fit for the initial
modulus measured from experiments, implying Eq. (4) might not provide a satisfactory approximation of the
transition. Other models may be used to replace Eq. (4) to further improve the predictability of the model.

A.5. DG, g_ 0 , s0 , ss , h0

The material parameters DG, g_ 0 , s0, ss, h0 in the viscoplastic component for the glassy polymer behavior can
be estimated from the isothermal uniaxial compression curves at different strain rates. Here, we use the curves
at T ¼ 0 1C to illustrate this process. In a uniaxial compression test, the equivalent shear stress t and shear
strain g are related to the uniaxial stress and strain by
pffiffiffi pffiffiffi
t ¼ s= 3 and g ¼ 3. (A.3)
From Fig. 4b, the yielding stresses are s1 ¼ 62:4 MPa for _1 ¼ 0:1=s, and s2 ¼ 57:6 MPa for _ 2 ¼ 0:01=s.
Therefore, the corresponding equivalent shear yield stresses and equivalent shear strain rates are: t1 ¼
36:0 MPa for g_ 1 ¼ 0:173=s, and t2 ¼ 33:3 MPa for g_ 2 ¼ 0:0173=s. Since the yielding point in the stress–strain
curve corresponds to the onset of significant viscoplastic flow, and thus the onset for the evolution of athermal
shear strength s, it is reasonable that at the yielding point, s ¼ s0. From Eq. (19), we have
   
DG t1
g_ 1 ¼ g_ 0 exp  1 , (A.4a)
kT s0
   
DG t2
g_ 2 ¼ g_ 0 exp  1 . (A.4b)
kT s0
ARTICLE IN PRESS
H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751 1749

From Eq. (A.4), we obtain


ln g_ 1  ln g_ 2
DG ¼ kT s0 ¼ 310  1029 s0 . (A.5)
t̄1  t̄2
From Fig. 4b, at e ¼ 0.5, the stress–strain curve reached a plateau, implying the athermal shear strength s
approached the saturation value ss, or sss. At e ¼ 0.5, s3 ¼ 35:7 MPa for _1 ¼ 0:1=s, and s̄4 ¼ 33:3 MPa for
_2 ¼ 0:01=s. The corresponding equivalent shear yield stresses and equivalent shear strain rates are: t̄3 ¼
20:6 MPa for g_ 1 ¼ 0:173=s, and t4 ¼ 19:2 MPa for g_ 2 ¼ 0:0173=s. Following the procedure similar to the one
to obtain Eq. (A.5), we have
ln g_ 1  ln g_ 2
DG ¼ kT ss ¼ 621  1029 ss . (A.6)
t̄4  t̄3
From Eqs. (A.5) and (A.6), we have
s0  2ss . (A.7)
In order to estimate h0, we obtain from Eq. (20a)
ds
¼ h0 dg. (A.8)
s=ss  1
Integrating Eq. (A.8),
    
s s0
ss ln  1  ln 1 ¼ h0 ðg  g1 Þ. (A.9)
ss ss
We assume that s decreases to (s0+ss)/2 as the stress in the stress–strain curve decreases to ðs1 þ s3 Þ=2.
At s ¼ ðs1 þ s3 Þ=2 ¼ 49:1 MPa, e ¼ 0.13, or g ¼ 0.23, and at s1 ¼ 62:4 MPa, e1 ¼ 0.07 or g1 ¼ 0.12. From
Eqs. (A.9) and (A.7), we have

ln sss  1  ln ss0s  1
h0 ¼ ss  6:3ss . (A.10)
g  g1
Eqs. (A.7), (A.6), and (A.10) provide the guideline to estimate s0, DG, and h0, once ss is known. To start
with, we can assume ss is slightly higher than t3 , for example, ss25 MPa. From Eqs. (A.7), (A.6), (A.10), we
have s050 MPa DG1.5  1019 J, h0150 MPa.

Appendix B. Contributions to stress–strain behaviors from the hyperelastic models

In the proposed model for SMPs, the hyperelastic branch in the glassy polymer model contributes mainly to
the post-yielding behavior at low temperatures; the hyperelastic model for RP affects the high temperature
behavior. Fig. B1 shows the comparison between the model with one hyperelastic model (with mg ¼ mr
0.8 MPa, and Ng ¼ Nr ¼ 17) for all phases and the model with hyperelastic models with different material
parameters for glassy phase and RP (with mg ¼ 0.1 MPa, mr ¼ 0.8 MPa, and Ng ¼ Nr ¼ 17). From Fig. B1, it
can be seen that when the same material parameters are used for the hyperelastic models, the model predicts a
slightly higher stress at post-yielding at T ¼ 0 1C. At T ¼ 100 1C, there is no difference between the two
because the hyperelastic model in the glassy polymer model does not function at high temperatures (due to
fgE0). Fig. B2 shows the effects of Ng in the glassy polymer model to the stress–strain behaviors at T ¼ 0 1C
(with mg ¼ mr 0.8 MPa, and Nr ¼ 17). As Ng becomes smaller, post-yield hardening can be observed. For the
material studied in this paper, the material does not demonstrate the hardening behavior, which results in the
similar material parameters for the two hyperelastic models. However, it is noted that some other SMPs may
demonstrate the post-yield hardening behavior at low temperatures but do not show dramatic stress increase
at large stretch at high temperatures, the two hyperelastic models offer the capability to capture more general
behaviors of SMPs.
ARTICLE IN PRESS
1750 H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751

70

60

50

-True Stress (MPa)


g = 0.8MPa, T = 0°C
40

30
g = 0.1MPa, T = 0°C
20

g = 0.8MPa, T = 100°C
10
g = 0.1MPa, T = 100°C
0
0 0.25 0.5 0.75 1
-True Strain

Fig. B1. Comparison between the model with one hyperelastic model (with mg ¼ mr 0.8 MPa, and Ng ¼ Nr ¼ 17) for all phases and the
model with hyperelastic models with different material parameters for the glassy phase and the rubbery phase.

70
T = 0°C
60

Ng = 1.6
50
-True Stress (MPa)

40 Ng = 2.0

30
Ng = 17.0

20

10

0
0 0.25 0.5 0.75 1
-True Strain

Fig. B2. Effects of Ng in the glassy polymer model to the stress–strain behaviors at T ¼ 0 1C (with mg ¼ mr 0.8 MPa, and Nr ¼ 17).

References

Bergstrom, J.S., Boyce, M.C., 1998. Constitutive modeling of the large strain time-dependent behavior of elastomers. J. Mech. Phys. Solids
46 (5), 931–954.
Boyce, M.C., Arruda, E.M., 1993. Three-dimensional constitutive model for the large stretch behavior of rubber elastic materials. J. Mech.
Phys. Solids 41 (2), 389–412.
Boyce, M.C., Arruda, E.M., 2000. Constitutive models of rubber elasticity: a review. Rubber Chem. Technol. 73 (3), 504–523.
Boyce, M.C., Parks, D.M., Argon, A.S., 1988a. Computational modeling of large strain plastic-deformation in glassy-polymers. Abstr.
Pap. Am. Chem. Soc. 196 156-POLY.
Boyce, M.C., Parks, D.M., Argon, A.S., 1988b. Large inelastic deformation of glassy-polymers. 1: Rate dependent constitutive model.
Mech. Mater. 7 (1), 15–33.
ARTICLE IN PRESS
H.J. Qi et al. / J. Mech. Phys. Solids 56 (2008) 1730–1751 1751

Boyce, M.C., Parks, D.M., Argon, A.S., 1988c. Large inelastic deformation of glassy-polymers. 2: Numerical-simulation of hydrostatic
extrusion. Mech. Mater. 7 (1), 35–47.
Boyce, M.C., Weber, G.G., Parks, D.M., 1989. On the kinematics of finite strain plasticity. J. Mech. Phys. Solids 37 (5), 647–665.
Boyce, M.C., Kear, K., Socrate, S., Shaw, K., 2001. Deformation of thermoplastic vulcanizates. J. Mech. Phys. Solids 49 (5), 1073–1098.
Fulcher, G.S., 1925. Analysis of recent measurements of the viscosity of glasses. J. Am. Ceram. Soc. 8, 789–794.
Govindjee, S., Reese, S., 1997. A presentation and comparison of two large deformation viscoelasticity models. J. Eng. Mater. Technol.
Trans. ASME 119 (3), 251–255.
Govindjee, S., Simo, J., 1991. A micro-mechanically based continuum damage model for carbon black-filled rubbers incorporating
Mullins effect. J. Mech. Phys. Solids 39 (1), 87–112.
Lagoudas, D.C., Entchev, P.B., Popov, P., Patoor, E., Brinson, L.C., Gao, X.J., 2006. Shape memory alloys, part II: modeling of
polycrystals. Mech. Mater. 38 (5–6), 430–462.
Lendlein, A., Jiang, H.Y., Junger, O., Langer, R., 2005a. Light-induced shape-memory polymers. Nature 434 (7035), 879–882.
Lendlein, A., Kelch, S., Kratz, K., Schulte, J., 2005b. Shape memory polymers. In: Encyclopedia of Materials: Science and Technology.
Lion, A., 2000. Constitutive modelling in finite thermoviscoplasticity: a physical approach based on nonlinear rheological models. Int. J.
Plasticity 16 (5), 469–494.
Liu, Y.P., Gall, K., Dunn, M.L., McCluskey, P., 2004. Thermomechanics of shape memory polymer nanocomposites. Mech. Mater. 36
(10), 929–940.
Liu, Y.P., Gall, K., Dunn, M.L., Greenberg, A.R., Diani, J., 2006. Thermomechanics of shape memory polymers: uniaxial experiments
and constitutive modeling. Int. J. Plasticity 22 (2), 279–313.
McKenna, G.B., 1989. Glass formation and glassy behavior. In: Booth, C.P.C. (Ed.), Comprehensive Polymer Science. Pergamon,
Oxford, pp. 311–362.
Miehe, C., Keck, J., 2000. Superimposed finite elastic–viscoelastic–plastoelastic stress response with damage in filled rubbery polymers.
Experiments, modelling and algorithmic implementation. J. Mech. Phys. Solids 48 (2), 323–365.
Monkman, G.J., 2000. Advances in shape memory polymer actuation. Mechatronics 10 (4–5), 489–498.
Otsuka, K., Wayman, C.M., 1998. Shape Memory Materials. Cambridge University Press, Cambridge, New York.
Qi, H.J., Boyce, M.C., 2005. Stress–strain behavior of thermoplastic polyurethanes. Mech. Mater. 37 (8), 817–839.
Reese, S., Govindjee, S., 1998. A theory of finite viscoelasticity and numerical aspects. Int. J. Solids Struct. 35 (26–27), 3455–3482.
Scott, T.F., Schneider, A.D., Cook, W.D., Bowman, C.N., 2005. Photo-induced plasticity in cross-linked polymers. Science 308 (5728),
1615–1617.
Simo, J.C., 1987. On a fully 3-dimensional finite-strain viscoelastic damage model—formulation and computational aspects. Comput.
Methods Appl. Mech. Eng. 60 (2), 153–173.
Tammann, G., Hesse, W., 1926. Dependence of viscosity on temperature in supercooled liquids. Z. Anorg. Allgem. Chem. 156, 245–257.
Thamburaja, P., Anand, L., 2002. Superelastic behavior in tension–torsion of an initially textured Ti–Ni shape-memory alloy. Int. J.
Plasticity 18 (11), 1607–1617.
Tobushi, H., Hara, H., Yamada, E., Hayashi, S., 1996. Thermomechanical properties in a thin film of shape memory polymer of
polyurethane series. Smart Mater. Struct. 5 (4), 483–491.
Treloar, L.R.G., 1958. The Physics of Rubber Elasticity. Clarendon Press, Oxford.
Vogel, H., 1921. The law of viscosity change with temperature. Physik Z 22, 645–646.
Warner, M., Terentjev, E.M., 2003. Liquid Crystal Elastomers. Oxford University Press, Oxford, New York.
Weber, G., Anand, L., 1990. Finite deformation constitutive-equations and a time integration procedure for isotropic, hyperelastic
viscoplastic solids. Comput. Methods Appl. Mech. Eng. 79 (2), 173–202.
Yakacki, C.M., Shandas, R., Lanning, C., Rech, B., Eckstein, A., Gall, K., 2007. Unconstrained recovery characterization of shape-
memory polymer networks for cardiovascular applications. Biomaterials 28 (14), 2255–2263.

S-ar putea să vă placă și