You are on page 1of 35

International Journal of Plasticity

16 (2000) 1309±1343
www.elsevier.com/locate/ijplas

On thermomechanics and transformation


surfaces of polycrystalline NiTi shape
memory alloy material
M.A. Qidwai, D.C. Lagoudas *
Center for Mechanics of Composites, Aerospace Engineering Department, Texas A&M University,
College Station, TX 77843-3141, USA

Received in ®nal revised form 9 December 1999

Abstract
A thermomechanical description of the martenstitic phase transformation and the asso-
ciated shape memory e€ect in polycrystalline shape memory alloys (SMAs) is presented. The
rate-independent constitutive relations are derived in the stress-temperature space using a
Lagrangian formulation. The Kuhn±Tucker optimality conditions, constraints on evolution
equations for transformations strain and shape of transformation function in thermodynamic
force space are obtained naturally through the principle of maximum transformation dis-
sipation. Various transformation functions are investigated and a generalized type transfor-
mation function is proposed. Numerical results of the model based on di€erent
transformation functions are compared with experimental results to determine their accuracy
to predict SMA characteristics like tension±compression asymmetry, negative volumetric
transformation strain and pressure dependence. # 2000 Elsevier Science Ltd. All rights
reserved.
Keywords: A. Phase transformation; A. Thermomechanical processes; B. Constitutive behavior; B. Poly-
crystalline materials; Shape memory alloy (SMA)

1. Introduction

Over the last decade shape memory alloy (SMA) materials have seen widespread
usage in mechanical, automotive, aerospace, nuclear, medical and domestic

* Corresponding author. Tel.: +1-409-845-1604; fax: +1-409-845-6051.


E-mail addresses: dlagoudas@aero.tamu.edu (D.C. Lagoudas), qidwai@anvil.nrl.navy.mil (M.A. Qidwai).

0749-6419/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PII: S0749-6419(00)00012-7
1310 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343

appliance industries (Birman, 1997). Most of the earlier SMA applications were 1-D
in nature where SMA wires, bars, rods and beams had been employed, e.g. as
orthodontic wires (Sachdeva and Miyazaki, 1990) and as actuators in robotic sys-
tems and self-expanding microstructures (Melzer and StoÈkel, 1994). Recently, the
potential of SMAs as 3-D load bearing structures and actuators has also been stu-
died, e.g. in active metal matrix composites to be used in helicopter rotor blades
(Liang et al., 1996) and as torque tube actuators in aircraft wings (Davidson et al.,
1997; Jardine et al., 1997). The earlier reluctance to use SMA materials in 3-D
applications despite their obvious capability to generate large actuating forces was
primarily due to a lack of proper understanding and lack of a 3-D modeling cap-
ability of the inherently nonlinear nature of these materials.
Until 1995, most of the 3-D SMA constitutive modeling was phenomenological
and generalized from 1-D experimental results (Liang and Rogers, 1992; Sun and
Hwang, 1993; Boyd and Lagoudas, 1994). Some of the early and more comprehen-
sive phenomenological models attempted to account for reorientation, kinematic
and isotropic hardening to include the dependence on applied stress state (Tanaka et
al., 1995; Boyd and Lagoudas, 1996). For a model to be successful it is necessary to
take into account the underlying deformation mechanisms responsible for the com-
plex SMA behavior.
To this end micromechanics come into the picture and two di€erent methodolo-
gies have appeared. In one case, the crystallographic modeling of a single crystal or
grain is performed and the results are averaged over a representative volume element
(RVE) to obtain a polycrystalline response (Patoor et al., 1988; Sun and Hwang,
1993; Tokuda et al., 1998; Lim and McDowell, 1999). In the second case, a macro
free energy potential is obtained directly from the micromechanical modeling and
the Coleman±Noll procedure is employed (Raniecki and Lexcellent, 1998; Reisner et
al., 1998; Bo and Lagoudas, 1999a,b,c; Gillet et al., 1999; Lagoudas and Bo, 1999)
to derive constraints on the material constitutive behavior. Most of these macroscale
SMA thermomechanical constitutive models mainly di€er on the form of transfor-
mation strain induced strain hardening term in the free energy. Lagoudas et al.
(1996) presented a uni®ed thermodynamic framework which incorporated di€erent
constitutive models. Another major di€erence arises from the choice of the trans-
formation function, which de®nes the boundaries of the thermoelastic domain (or
the transformation criterion), e.g. J2 type (Boyd and Lagoudas, 1996; Bo and
Lagoudas, 1999a), J2 ÿ I1 type (Auricchio et al., 1997) or J2 ÿ J3 type (Gillet et al.,
1999).
Out of the many intermetallic alloys which possess shape memory e€ect, two of
them Cu±Zn±Al and NiTi have attracted more interest than any other SMA (Gall et
al., 1999). Even between these two alloys, NiTi applications have dominated the
scienti®c scene. The extensive use of NiTi over Cu±Al±Zn is due to its superior
material characteristics such as larger recoverable strains, large recoverable forces,
cyclic path memorization, cyclic stability, fatigue resistance, corrosion resistance
and wear resistance (Gall et al., 1999). The SMA material response is highly depen-
dent on the temperature due to latent heat generation. The faster the loading rate
the lesser the time for the material to dissipate the resulting heat causing higher
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1311

transformations stress levels. Its behavior includes low temperature martensite


reorientation, medium temperature stress induced martensitic transformation and
high temperature austenitic plastic slip.
One of the earlier experimental analyses of the 3-D constitutive behavior of SMAs
was performed by SÏittner et al. (1995). They investigated the stabilization of the
transformation behavior in stress induced martensite in NiTi hollow bar SMA
cylinders. A number of loading cycles were performed in combined tension±torsion
and the specimen was then loaded in either tension or torsion revealed strain aniso-
tropy, i.e. the appearance of other strain components. Jacobus et al. (1996) studied
the e€ect of stress state on the stress induced martensitic phase transformation. NiTi
bar specimens were used to perform tension, compression, hydrostatic compression
and various triaxial compression tests. One of the key results was the evidence that
the SMA phase transformation is pressure dependent, i.e. the e€ective stress for
phase transformation increases with increasing hydrostatic pressure. The small
negative transformation strain volume change (ÿ0.34%) was also corroborated
experimentally. These results were in addition to tension±compression asymmetrical
behavior that had already been observed in 1-D SMA bars and had been attributed
to specimen texture (Gall and Sehitoglu, 1999). Several studies have also concluded
that compressive SMA behavior results in lower recoverable strains, steeper hard-
ening behavior and higher transformation stress levels compared to tensile behavior
(Adler et al., 1990; Melton, 1990). Proportional and non-proportional loading
experiments on SMA torque tubes were conducted by Lim and McDowell (1999).
They investigated the stress-temperature coupling in tension, compression and tor-
sion. They showed that there was more latent heat available in compression as evi-
denced by larger temperature increase in compression than in tension for complete
phase transformation. The thermomechanical behavior in positive and negative
torsion was observed to be symmteric. Non-proportional tension±torsion testing
revealed a high degree of non-associativity of the transformation strain ¯ow rule in
stress space. This has been attributed to the reorientation of the existing martensitic
variants during stress-induced phase transformations (Lim and McDowell, 1999).
The SMA constitutive models based on a J2 invariant type transformation func-
tion are incapable of modeling the asymmetrical behavior (at least initially) and the
dependence of transformation behavior on hydrostatic pressure. Moreover, the
transformation volumetric change also cannot be modeled if associative transfor-
mation strain ¯ow rule is assumed. Constitutive models which use a J2 ÿ I1 invar-
iants type transformation function can model the tension±compression asymmetry
and the pressure dependence. Also, constitutive models with generalized transfor-
mation function based on J2 ÿ J3 ÿ I1 invariants are capable of demonstrating the
volumetric transformation strain and asymmetrical behavior plus the pressure
dependence in a more general sense by distortion of the transformation surface.
The idea of maximum transformation dissipation in inelastic material is not new.
It was proposed independently for plastic materials by Mises (1928) and Hill (1948).
Ziegler (1963) and Ziegler (1983) derived an orthogonality condition which implied
a criterion of maximum dissipation rate. Rajagopal and Srinivasa (1998b) further
derived certain normality and convexity conditions, which are both necessary and
1312 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343

sucient for maximum dissipation rate. In this work, the framework developed by
Simo and Hughes (1998) for plastic materials employing Lagrange multipliers and
optimality conditions is used to derived the conditions for the satisfaction of the
maximum dissipation due to phase transformation in SMAs.
The purpose behind the study of the thermomechanics of SMA in this paper is
two-fold: (1) to extend the thermodynamic framework presented by Lagoudas et al.
(1996) to include the principle of maximum transformation dissipation, (2) to pro-
pose a generalized phase transformation function motivated by the experimental
results discussed earlier. The maximum dissipation principle imposes constraints on
the form of internal state variable ¯ow rules and establishes the shape of the trans-
formation function in generalized thermodynamic force space. Green±Lagrange
strain measure is employed in modeling to account for motions with large rotations
and small strains, which may normally be expected in a SMA based device or
structure, e.g. an active beam in bending.1 Based on the above mentioned thermo-
mechanical framework, a SMA thermomechanical constitutive model is proposed.
The model is developed to capture stable stress-induced phase transformation which
can be recovered by heating the material. That is, it will be assumed that the plastic
strains are not developed during phase transformation and no reorientation of
martensitic variants takes place. Moreover, various transformation functions sug-
gested in the literature are reviewed in addition to the generalized (J2 ÿ J3 ÿ I1 type)
transformation function.
The paper is organized as follows. The general thermomechanical theory of SMAs
is reviewed and constraints on the evolution of internal state variables are obtained
as a result of principle of maximum transformation dissipation in Section 2. In
Section 3, a SMA constitutive model is proposed and various transformation func-
tions are discussed based on the development in Section 2. A generalized transfor-
mation function is also introduced here based on recent experimental results.
Numerical results, discussion and conclusion are given in Section 4.

2. Thermomechanical description of shape memory alloys using thermodynamic


potentials

Shape memory alloys are materials capable of changing crystallographic structure


due to changes of temperature and/or stress. These microstructural changes corre-
spond to a phase transformation from a high symmetry austenitic phase into a lower
symmetry martensitic phase in the forward transformation and from martensite into
austenite in the reverse transformation. Martensitic volume fraction can be
employed as an indicator of the extent of this phase transformation, collectively
accounting for all di€erent variants of martensite. Phase transformations in SMA
are displacive in nature, i.e. they exhibit themselves in macroscopic strains called the
transformation strains. Due to irreversible changes in microstructure (e.g. dislocation

1
This particular choice of deformation measure for modeling of polycrystalline SMA has also been
adopted by Govindjee and Hall (1999).
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1313

formation and motion), transformation induced internal stresses may also appear
(Bo and Lagoudas, 1999a). Also, the phase transformations induce thermo-
mechanically coupled nonlinear rate-independent material response with loading
history dependence. This loading history dependence requires the introduction of
internal state variables like the martensitic volume fraction, transformation strain,
transformation induced internal stresses along with the loading variables stress and
temperature to the constitutive theory.
For purposes of development here, we prescribe a general Gibbs free energy
potential, G, for a polycrystalline SMA as

G :ˆ G…S; ; †; …2:1†

where  is a set of Lagrangian internal state variables some of which are already
mentioned. For each set of ®xed values of these variables, a thermoelastic SMA
material is de®ned.
Phase transformation in SMAs is a dissipative process or a process which involves
entropy change. Dissipative processes require the application of the laws of ther-
modynamics to obtain constraints on the material response. The most popular
methodology is the one proposed by Coleman and Noll (1963)} for inelastic mate-
rials. In this methodology, the second law of thermodynamics is assumed to be
represented by the strong form of the Clausius±Duhem inequality to derive con-
straints on the material behavior, which is given as
: : :
loc ˆ ÿ0 …G ‡  † ÿ E : S50; …2:2†

where loc is called the local entropy production rate per unit mass, 0 denotes the
mass density in the reference con®guration,  is the entropy, E is the total Green±
Lagrange strain tensor and the dot above the quantity (.) represents the rate of
change of that quantity. The total strain tensor, E, is de®ned in terms of the defor-
mation gradient, F, as follows
1ÿ T 
E :ˆ F Fÿ1 : …2:3†
2
The superscript ``T'' in (2.3) denotes the transpose of a tensor and 1 is the second
order identity tensor.
Taking into account (2.1), the rate of change of Gibbs free energy is given by

: @G : @G : @G :
Gˆ :S‡ ‡ : : …2:4†
@S @ @

Substituting (2.4) into (2.2), we obtain

@G : @G : @G :
loc ˆ ÿ…0 ‡ E† : S ÿ 0 … ‡ † ÿ 0 :  50: …2:5†
@S @ @
1314 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343

The existence of a thermoelastic domain is assumed here. In the: thermoelastic


:
:
region, is zero and since the above relation should be valid for all S;  necessarily,
we obtain the following constitutive relations

@G
E ˆ ÿ0 ; …2:6†
@S

@G
ˆÿ : …2:7†
@

The above constitutive relations are assumed to be valid everywhere at the


boundary of the thermoelastic region also. A detailed discussion on this assumption
is presented by Rajagopal and Srinivasa (1998a), where they employed the notion of
a family of thermoelastic responses associated with the overall inelastic material
behavior.
Now, using (2.5), (2.6) and (2.7), we de®ne transformation dissipation, Dt , to be
the reduced form of loc as
 
t @G : @G :
D 0 ;  :ˆ ÿ0 : 50: …2:8†
@ @

If we de®ne

@G
C:ˆ ÿ0 …2:9†
@
:
as the set of generalized thermodynamic forces conjugate to , then the transforma-
tion dissipation is simply given by
: :
Dt …C; † ˆ C:  50: …2:10†

The above relation is the rate: of work done by phase transformation due to changes
in microstructure induced by  in the generalized thermodynamic force space.

2.1. Transformation function and criterion

The existence of a thermoelastic region or transformation surface, F , bounded by


smooth hypersurface, @F , was assumed earlier. To determine the thermoelastic
region, one assumes the existence of a transformation function, …C†, which is used to
de®ne the hypersurface, @F , by the equation

…C† ˆ 0: …2:11†

The transformation function, , is not assumed to have an explicit dependence on


fS; ; g. The implicit dependence of  on fS; ; g comes through the dependence
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1315

of the generalized thermodynamic forces, C, on fS; ; g. Connectivity with the


thermoelastic domain for at least one loading path requires

…CS;
… ; ††40: …2:12†

The transformation loading and unloading conditions (conditions for the beginning
of phase transformation) are given based on the stress-temperature space loading
criteria as
( : :
@
@S : S ‡ @
@  > 0 forward transformation
…C† ˆ 0; 0441 and @ : @ : …2:13†
@S : S ‡ @  < 0 reverse transformation:

2.2. Principle of maximum transformation dissipation

The principle of maximum transformation dissipation is based on the premise that


if the SMA material at some point during its thermomechanical loading is to trans-
form, then this transformation will be characterized by maximum dissipation. Simo
and Hughes (1998) (Section 2.6) employed the maximum dissipation principle in
plasticity to demonstrate the convexity of the yield criterion in stress-internal state
variable space, the normality (and associativity) of the internal state variable ¯ow
rules and the presence of Kuhn±Tucker loading conditions.2 However, it is noted
that the discussion here is more general in the sense that a broader generalized
thermodynamic force-internal state variable space is considered compared to the
stress and internal state variable space used by the previous authors. It will be shown
that the principle implies associative ¯ow rules and a convex thermoelastic region in
the larger space of thermodynamic force-internal state variable, at the same time
allowing for non-associative ¯ow rules and non-convexity of the thermoelastic
region in the stress-temperature space.
We assume the principle of maximum transformation dissipation as:

For a given set of internal state variables, , among the set of all admissible
generalized thermodynamic forces, Call , satisfying the transformation criterion
(2.12) the transformation dissipation, Dt , de®ned by (2.10) is the global max-
imum for the actual generalized thermodynamic forces, C.

We can write the above principle as


: :
Dt …C; † ˆ maxDt …Ca ;  †; …2:14†
C
a 2C
all

2
Several historical references are cited in Section 1, which contributed to the development of the
maximum dissipation principle in inelastic materials. More recently, Rajagopal and Srinivasa (1998b)
have derived necessary and sucient conditions as criteria of the maximum rate of dissipation in inelastic
materials. Here, the method of Lagrange multipliers is used to maximize the dissipation rate and to derive
the necessary conditions as was employed by Simo and Hughes.
1316 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343

where
 
@G
Ca :ˆ ÿ0 ; …2:15†
@ admissible

is the admissible thermodynamic force and



Call :ˆ fCa …Ca †40g …2:16†

is the set of all admissible generalized thermodynamic forces.


It is important to note that implications of maximum transformation dissipation
de®ned using Lagrangian variables are not generalizable to other formulations, e.g.
Eulerian formulation. The discussion in the next section is valid only in the
Lagrangian generalized thermodynamic force space. The Lagrangian variables are
chosen because ordinarily the recoverable strains associated with shape memory
e€ect are not more than 6±8% and only large rotations are expected in SMA based
devices. The e€ects of the employment of a particular formulation on the con-
sequences of maximum rate of dissipation condition in inelastic materials are
described by Rajagopal and Srinivasa (1998b).

2.2.1. Consequences of the application of principle of maximum transformation


dissipation
To show the impact of the principle of maximum transformation dissipation, the
Lagrange multiplier method is employed. It is assumed here that S;  are not ®xed
because in that case Cwould also be ®xed, which is not permissible. Following the
standard optimization method,3 the maximization principle is transformed into a
minimization principle by changing the sign of the objective function, i.e. ÿ Dt .
Then using (2.10) and (2.14), the Lagrangian associated with the unconstrained
optimization problem is given by
: :
Lt …Ca ; l;  † :ˆ ÿCa :  ‡ l…Ca †; …2:17†

where l is the Lagrange multiplier. The solution to the above optimization problem
is then given by fC; lg, satisfying the following relations
:
@Lt …Ca ; l;  † : @…C†
Ca ˆC
ˆ ÿ ‡ l ˆ 0; …2:18†
@Ca @C

along with Kuhn±Tucker optimality conditions given by

l40; …C†40; l…C† ˆ 0: …2:19†

3
For a pertinent reading on Lagrange multipliers and Kuhn±Tucker theorem, see Beveridge and
Schechter (1970), pages 274±287 and McMillan (1970), pages 104±131.
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1317

The relations (2.18)±(2.19) are necessary but not sucient conditions for the exis-
tence of a global minimum for the dissipation potential, ÿ Dt . Suciency is guar-
anteed for a linear function if the constraint region is convex, its global minimum
being always directly found at the boundary (Beveridge and Schechter, 1970, page
119). Note that the dissipation potential, ÿ Dt , is a linear function and for its global
minimum to exist, the transformation function, …CS; … ; ††, has to be convex in the
generalized thermodynamic force space (C-space). Non-convexity in the stress±tem-
perature fS; g space is still allowed.
The requirement of convexity of  for the stationary point to be a global mini-
mum is now derived. Based on a result found in Simo and Hughes (1998) (Lemma
2.6.1), the function …C† is convex if and only if

@…C†
…Ca † ÿ …C†5…Ca ÿ C† : 8Ca 6ˆ C …2:20†
@C

The satisfaction of this inequality is shown as follows. Based on de®nition (2.14),


we have
: :
Dt …C; †5Dt …Ca ; † …2:21†

: :
C: 5Ca :  …2:21a†

Substituting the relation from (2.18) and bringing all the terms on the right hand
side, we obtain

@…C†
l…Ca ÿ C† : 40 …2:22†
@C

Then, using the Kuhn±Tucker condition, i.e. l…Ca † ˆ 0, we write

@…C†
l…Ca ÿ C† : 4…Ca †: …2:23†
@C

For the non-trivial case, when l > 0 and noting that …C† ˆ 0, we obtain

@…C†
…Ca ÿ C† : 4…Ca † ÿ …C†; …2:24†
@C

which coincides with (2.20), and hence convexity in the generalized thermodynamic
force space follows.
At the end, the signi®cance of the results obtained due to the application of the
maximum dissipation principle is explained. The relations (2.19) describe the
(transformation) loading and (elastic) unloading conditions. If l ˆ 0, the solution C
may exist either inside the thermoelastic region, , or at its boundary. The material
1318 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343
:
behaves thermoelastically because  ˆ 0 in the light of (2.18). If l > 0, the solution
Cexists at the boundary of the thermoelastic region, . The material transform
because the set of internal state variables, , evolves. The requirement of associa-
tivity and normality of internal state variable evolution equations in the generalized
thermodynamic force space is implied by (2.18). An internal state variable evolution
equation is called associative when the scalar potential used to describe its evolution
is the transformation function itself. It is said to follow normality when its direction
is described normal to the transformation function in the thermodynamic force space.
This result does not preclude non-associative ¯ow rules in the stress±temperature
space. The explicit dependence of Con fS; ; g still allows for the possibility of non-
associative ¯ow rules in stress-temperature space.

3. Constitutive modeling of 3-D stress induced phase transformation in polycrystalline


NiTi SMA

We now present a constitutive model based on the thermomechanical framework


developed in Section 2 and the earlier work of Boyd and Lagoudas (1996) and
Lagoudas et al. (1996). The model is capable of modeling the pseudoelastic and the
shape memory e€ects, while reorientation is ignored. The Gibbs free energy poten-
tial, G, for such a material is proposed as follows.4 The Gibbs free energy potential
is now proposed as follows

1 1  
G…S; ; Et ; † ˆ ÿ S : S : S ÿ S : … ÿ 0 † ‡ Et …3:1†
20 0
  

‡C … ÿ 0 † ÿ  ln ÿ 0  ‡ U0 ‡ F…†;
0

where S; ; C; 0 and U0 are the e€ective compliance tensor, e€ective thermal


expansion coecient tensor, e€ective speci®c heat, e€ective speci®c entropy at the
reference state, and e€ective speci®c internal energy at the reference state, respec-
tively. Also, the changes in the microstructure of the material are represented by the
set of internal state variables,  ˆ fEt ; g, where Et is the Lagrangian transforma-
tion strain tensor and  is the martensitic volume fraction. The function, F…†, in Eq.
(3.1) accounts for transformation hardening due to interaction between martensite
and austenite and the martensitic variants themselves. The e€ective material prop-
erties are assumed to vary with the martensitic volume fraction, , as follows5

4
The choice of G is made assuming that the material retains its symmetry under loading cycles and
undergoes large displacements, large rotations and small strains. Also, since the second Piola±Kirchho€
stress and Green±Lagrange strain components are unaltered when the material is subjected to rigid body
motions, any material description developed for in®nitesimal deformation analysis using engineering
measures (Lagoudas et al., 1996) can be directly employed for large displacement, large rotation and small
strain analysis.
5
See Boyd and Lagoudas (1994) for Mori±Tanaka estimates of these e€ective material properties.
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1319

S :ˆ S A ‡ …S M ÿ S A †; :ˆ A ‡ … M ÿ A †;
C :ˆ CA ‡ …CM ÿ CA †; 0 :ˆ A M A
0 ‡ …0 ÿ 0 †; …3:2†
U0 :ˆ UA M A
0 ‡ …U0 ÿ U0 †:

In (3.2), the superscripts M and A stand for martensite and austenite, respectively.
The constitutive relations are obtained using (2.6) and (2.7) with the speci®c form
of G given by (3.1), i.e.

E ˆ S : S ‡ … ÿ 0 † ‡ Et ; …3:3†
 
1 
 ˆ S : ‡ C ln ‡ 0 :
0 0

Additive strain decomposition appears in (3.3) and the di€erence Ee :ˆ


E ÿ … ÿ 0 † ÿ Et can be identi®ed as the elastic strain.6 Note that the elastic strain
depends on the degree of phase transformation through the dependence of the elas-
tic compliance on . The nonlinearity in the constitutive behavior is due to the use of
nonlinear kinematics, the evolution equations for Et and the transformation func-
tion (to be discussed later).
Using (2.10), the local dissipation rate or the transformation dissipation, Dt , gen-
eralized: thermodynamic forces, C, and their generalized conjugate internal variable
rates, , respectively, are given by
 : : : :
Dt S; p; Et ;  :ˆ S : Et ‡ p 50: …3:5†

C:ˆ fS; pg; …3:6†

: : :
 :ˆ fEt ;  g: …3:7†

In the above equations, the generalized thermodynamic force S is given by the


identity

@G
S ˆ ÿ0 ; …3:8†
@Et

and the generalized thermodynamic force p is derived from the relation7

6
The additive decomposition is not infrequently used, e.g. Simo and Ortiz (1985), Casey and Naghdi
(1992) and Govindjee and Hall (1999). It has been justi®ed by Casey (1985) for the following cases: (1)
small inelastic deformations and moderate elastic strain, (2) small elastic strain and moderate inelastic
deformation, (3) small strain and moderate rotations. Case number 3 is what we ordinarily expect in a
SMA application, for example during bending of an active beam.
7
It is obvious from the development that the generalized thermodynamic forces can be dependent on
each other. For example, note that p is actually a function of S;  and .
1320 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343

@G 1
p ˆ ÿ0 ˆ S : S : S ‡  : S… ÿ 0 †
@ 2
  
 @F…†
‡ 0 C … ÿ 0 † ÿ  ln ‡ 0 0  ÿ ÿ 0 U0 : …3:9†
0 @

The terms S;  ; C; 0 ; U0 in (3.9) indicate the di€erence of compliance
tensor, thermal expansion coecient tensor, speci®c heat, reference entropy and
reference internal energy between the martensitic and austenitic phases, respectively.
At this point, we propose the existence of a thermoelastic region or a transfor-
mation surface, F , bounded by a smooth hypersurface, @F , for a given set of
 :ˆ fEt ; g. This region is de®ned by a transformation function …S; p† and its
boundary is determined by the relation

…S; p† ˆ …C† ˆ 0: …3:10†

It is further assumed that the construction of  will be such that all admissible
material states will satisfy the constraint

…S; p†40: …3:11†

Then, in the light of (2.18) and (2.19), the application of the principle of maximum
transformation dissipation on this material results in the following ¯ow rules and
Kuhn±Tucker conditions
: @…S; p†
Et ˆ l ; …3:12†
@S

: @…S; p†
ˆl ; …3:13†
@p

l50; …S; p†40; l…S; p† ˆ 0: …3:14†

Since (3.13) is a scalar equation, it is obvious that the Lagrange


: parameter, l, is
proportional to the rate of martensitic volume fraction,  . An important result is
obtained here by substituting the expression for l from (3.13) into (3.12), that is, the
evolution of transformation strain is dependent on the evolution of martensitic
volume fraction as given by
: : 1 @…S; p†
Et ˆ  @…S;p† : …3:15†
@p
@S

Conclusively, the application of the principle of maximum transformation dis-


sipation implies that for a SMA constitutive model which admits the martensitic
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1321

volume fraction as an internal state variable, the evolution of all the internal state
variables depend on the evolution of the martensitic volume fraction. Now, recall
the expression for transformation dissipation given by (3.5) and substitute the
transformation strain ¯ow rule given by (3.15) in it to obtain
!
t 1 @…S; p† : :
D ˆ @…S;p† S : ‡ p  :ˆ  50: …3:16†
@p
@S

The above expression states that all the dissipation is caused by a change in the
martensitic volume fraction, a result already mentioned earlier. From this develop-
ment, an e€ective thermodynamic force, , is de®ned, which is conjugate to  and a
combination of the generalized thermodynamic forces fS; pg. From (3.16) and the
de®nition of p in (3.9), it is further seen that if zero stress state is assumed at auste-
nitic state, then a change in the martensitic volume fraction can still be brought
about by decreasing the temperature resulting in self-accommodated martensite.
At this stage of the constitutive modeling, we have two choices. Either based on
experimental results we deduce the ¯ow rules for both transformation strain and
martensitic volume fraction, or the transformation function is suggested. In practice,
it is easier to approximate the shape of the transformation surface from the lesser
number of experiments usually available than to approximate the evolution of
transformation strain and martensitic volume fraction.8
Recalling the implications of (3.15), we propose the following general form of the
transformation function.9

h i2
…S; p† :ˆ ^…S† ‡ p ÿY2 …3:17†

h ih i
ˆ ^…S† ‡ p ÿ Y ^…S† ‡ p ‡ Y :

where Y is the measure of internal dissipation due to microstructural changes during


phase transformation and describes the onset of the phase transformation. Note in
(3.9) that the thermodynamic force, p, contains a combination of hardening terms
due to change in material properties with phase transformation, and transformation
strain hardening due to interaction between martensite variants and austenite as
represented by F…†. If the de®nition of p is substituted into (3.17), the transforma-
tion function, …S; p†, will be of the form, …S; ; †.

8
In literature, we ®nd both examples. For example, Boyd and Lagoudas (1996) proposed the trans-
formation strain ¯ow rule before proposing a J2 type transformation function in stress-temperature space.
On the other hand, Gillet et al. (1999) proposed a J2 ÿ J3 type transformation function and derived the
transformation strain ¯ow rule assuming associativity in stress space.
9
A scalar coecient of p di€erent than 1 can be used here but it will only serve as scaling all the
equations simultaneously without a€ecting their functional form.
1322 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343

Substituting (3.17) into (3.12)±(3.14), we obtain the following set of equations

: h i @^…S†
Et ˆ 2l ^…S† ‡ p ; …3:18†
@S

: h i
 ˆ 2l ^ …S† ‡ p …3:19†

Inverting (3.19) for l and substituting it in (3.18), the transformation strain evo-
lution is shown to be dependent on the evolution of the martensitic volume fraction
as

: : @^…S†
Et ˆ  : …3:20†
@S

Note that the transformation strain ¯ow rule is not associative in stress space.
Also, using the Kuhn±Tucker optimality conditions, (3.14), and (3.17), the following
conditions are obtained for the onset of phase transformation
:
ÿY < ^…S† ‡ p < Y ) l ˆ  ˆ 0; …3:21†

 :
l 6ˆ 0 ) ^ …S† ‡ p ˆ
 Y ) : > 0 : …3:22†
ÿY ) <0

Finally, recalling (2.13), the above conditions for the onset of phase transforma-
tion can be given in the form of loading/unloading conditions in the stress-tem-
perature space as follows
8
< @……S†‡p† : @…^ …S†‡p† :
Y and :S‡ >0
^…S† ‡ p ˆ @S @
: @…^ …S†‡p† : : …3:23†
: ÿY and @…^ …S†‡p†
@S :S‡ @  < 0

:
The evolution of martensitic volume fraction,  , can be :calculated either from
(3.19) where l is obtained using the consistency: condition … ˆ 0† or directly from
the consistency condition itself. That is why,  may be considered the consistency
parameter for this particular model. Finally, this model is fully represented by (3.3),
(3.20) and (3.23).
Now before deciding on the form of ^…S†, it is important to mention the experi-
mental results available for consideration:10

10
See Jacobus et al. (1996).
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1323

1. Polycrystalline NiTi exhibit negative volume change during phase transforma-


tr
tion, i.e. tr…Et † ˆ V
V ˆ ÿ0:0034.
2. Polycrystalline NiTi loaded under compression demonstrates
. smaller recoverable transformation strain levels,
. higher critical transformation stress levels,
. steeper transformation stress±strain slopes,
in comparison with polycrystalline NiTi loaded under tension.
3. With increasing hydrostatic pressure, P :ˆ 13 tr…S†, the e€ective stress required
to transform increase.

The abovementioned results are for the proportional loading cases. Lim and
McDowell (1999) recently cited results from non-proportional loading cases and
reported reorientation of martensitic variants and high non-associativity of trans-
formation strain ¯ow rule in the stress space. Presently, we intend to model pro-
portional loading cases only. However, the non-proportional case can be
accommodated in the present thermomechanical framework by introducing addi-
tional internal state variables related to reorientation along the line of work done by
Boyd and Lagoudas (1996) for in®nitesimal deformation. Here the emphasis will be
given to the selection of the function, ^…S†, and its e€ectiveness in describing the
experimental results mentioned in the last paragraph.

3.1. J2 Based transformation function

Let ^…S† be de®ned by11


p
^…S† :ˆ 3J2 ; …3:24†

where the material parameter corresponds to the maximum transformation strain


obtained during forward transformation, J2 ˆ 12 S0 : S0 ˆ: 12 kS0 k2 is the second
deviatoric stress invariant and S0 ˆ S ÿ 13 tr…S† is the deviatoric stress. The transfor-
mation strain ¯ow rule is then obtained using (3.20) as follows
r
:t 3 S0 : :
E ˆ 0  ˆ: K ; …3:25†
2 kS k
q S0
where Kˆ 32 0 . Using the transformation direction tensor, K;  ^ …S† can also
kS k
be described as12
@^…S†
^…S† ˆ S : Kˆ S : : …3:26†
@S

11
Used by Boyd and Lagoudas (1996), Bo and Lagoudas (1999a), etc.
12
All the choices of ^…S† in this work can be written in this form. This result can be e€ectively used
during the numerical implementation of the constitutive model.
1324 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343

The complete transformation surface, , is plotted in S11±S22 space in Fig. 1 for


forward and reverse transformations at  ˆ 0 and  ˆ 1, respectively. The transfor-
mation surface does not contain the asymmetry of uniaxial tensile and compressive
behavior. It does not exhibit pressure e€ects due to sole dependence on deviatoric
stress. Transformation induced volume change cannot be captured as veri®ed by
taking the trace of the transformation strain ¯ow rule. Furthermore, it can be seen
from Fig. 1 that there is an upper and lower bound of transformation surface at  ˆ
0 and  ˆ 1 for forward and reverse transformations, respectively. Any stress state
not on or between each pair of transformation surfaces is thermoelastic. For exam-
ple, if SMA is loaded isothermally in the austenitic state ( ˆ 0), it behaves ther-
moelastically until it hits the lower surface of martensite ( ˆ 0), then transforms
fully into martensite at ( ˆ 1), and then again exhibits thermoelastic behavior.

3.2. J2ÿI1 Based transformation function

For this type of function, we de®ne ^…S† to be13


p
^…S† :ˆ 3J2 ‡ I1 ; …3:27†

where I1 ˆ tr…S† is the ®rst stress invariant. In (3.27), and are material para-
meters, which are used to demonstrate the tension-compression asymmetry. The
transformation strain ¯ow rule is then given by
r !
:t 3 S0 : :
E ˆ ‡ 1  ˆ: K ; …3:28†
2 kS0 k

Fig. 1. Plot of J2 and J2 ÿ I1 transformation functions, , in S11±S22 space for both forward and reverse
phase transformations above Aof .

13
Used by Auricchio et al. (1997) to model the tension±compression asymmetry.
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1325
q
0
where Kˆ 32 kSS0 k ‡ 1. If is used to model the volumetric transformation strain
then the accompanying tension-compression asymmetry is negligible. On the other
hand, if it is used to model the asymmetry then the predicted volumetric transfor-
mation strain may be erroneous as shown in Section 4.2. For the later choice of
simulation, the transformation surface, , is plotted in S11±S22 space in Fig. 1.

3.3. J2ÿJ3ÿI1 based (generalized) transformation function

Another choice of ^…S† which captures the tension±compression asymmetry in


addition to the volumetric transformation strain is
" #
p J 3
^…S† :ˆ  3J2 1 ‡  3 ‡!I1 ; …3:29†
…3J2 †2

where J3 ˆ Det…S0 † is the third deviatoric stress invariant,  and  are related to
asymmetry, ! is introduced to capture the volumetric e€ect and  is a constant. For
 ˆ 0 or  ˆ 0 and ! ˆ 0, the form of J2 function is recovered. For  ˆ 0 or  ˆ 0,
the form of J2 ÿI1 transformation function is recovered. However, they are still dif-
ferent functions because the material parameters are calibrated for di€erent material
behaviors depending upon the function. The J2 and J2 ÿ I1 transformation functions
are recoverable from the generalized transformation functions only when pertinent
constants are calibrated for the same material behavior
 in each function. In inelastic
literature, di€erent values of  has been used, e.g. 13 ; 12 ; 1 . For  ˆ 12 and ! ˆ 0, the
transformation function proposed by Gillet et al. (1999) based on J2 ÿ J3 is
obtained. The transformation surface, , for this choice of constants is plotted in
S11±S22 space in Fig. 2. The tension-compression asymmetry is observed but note
that for certain choices of  the surface may be non-convex.

Fig. 2. Plot of J2 ÿ3 and J2 ÿ J3 ÿ I1 transformation functions, , S11±S22 space for both forward and
reverse phase transformations above Aof .
1326 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343

In order to account for volumetric change due to transformation, ! should be non


zero. In addition, the value of  is assumed to be 1 in the present work for simplicity.
This modi®cation retains the original shape of the function, as seen in Fig. 2, and
simpli®es the derivation of the transformation strain ¯ow rule and the calibration of
the parameters  and  from the uniaxial case. The transformation strain ¯ow rule is
then obtained for  ˆ 1 as

  pÿ 0 0 2   
:t 3S0 3J2 S S ÿ 3 J2 1 ÿ 3J3 S0 : :
E ˆ  p ‡  2
‡ !1  ˆ: K ; …3:30†
2 3J2 …3J2 †

 pÿ 0 0 2  
0
3S 3J2 S S ÿ 3 J2 1 ÿ3J3 S0
where Kˆ  p  ‡ ‡ !1. For comparison purposes, all the
2 3J2 …3J2 †2
transformation functions are plotted together for forward transformation at  ˆ 0 in
Fig. 3.
To capture the di€erent hardening behaviors during tension and compression, the
following form is proposed for the transformation strain hardening function, F…†,
( : :
1 M 3
3 0 b3  ‡ 12 0 bM 2
2  ‡ …1 ‡ 2 †
@
@S : S ‡ @
@ : > 0
F…† ˆ 1 A 3
: …3:31†
3 0 b3  ‡ 12 0 bA 2
2  ‡ …1 ÿ 2 †
@ @
@S : S ‡ @S  < 0;

Fig. 3. Plot of various transformation functions, , in S11±S22 space for both forward phase transfor-
mation at  ˆ 1 above Aof .
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1327

where 0 bM A M A
3 ; 0 b3 ; 0 b2 ; 0 b2 ; 1 and 2 are transformation strain hardening material
constants. Any of these constants can be taken to be zero to obtain the desired
transformation strain hardening. An example is shown in Fig. 4, where uniaxial
pseudoelastic tensile and compressive stress±strain responses are plotted for the J2 ÿ
J3 ÿ I1 transformation function. Three di€erent combination of the hardening con-
stants lead to three di€erent hardening e€ects during compression. In Table 1, the
list of material constants that appear in the constitutive model is given.
It is important to note that the branching of the hardening function has not vio-
lated any assumptions here because the direction of loading is independent of the
value of . If the transformation function, , contained any term containing a
combination of ffS; g; g then the above arrangement would be invalid. The above
choice of F…† leads to cubic dependence of the Gibbs free energy, G, on the mar-
tensitic volume fraction, . Note that this can be replaced by trigonometric (Liang
and Rogers, 1992; Brinson and Lammering, 1993), logarithmic (Tanaka, 1986),
polynomial (Boyd and Lagoudas, 1996), or any other function to accommodate
other constitutive models available in the literature.

3.4. Material constants

Isotropy for thermoelastic response can be reasonably assumed for polycrystalline


SMA. In that case, the e€ective compliance tensor, S, can be fully represented by the
Young's moduli, EA and EM , and the Poisson ratios, A and M , for austenite and
martensite, respectively. Similarly, isotropy reduces the determination of the e€ective
thermal expansion coecient tensor, , to the two thermal expansion coecients, A

Fig. 4. Pseudoelastic uniaxial stress±strain response in tension and compression for J2 ÿ J3 ÿ I1 (gen-
eralized) transformation function exhibiting di€erent hardening responses during compression.
1328 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343

Table 1
Shape memory alloy material parameters

Material constantsa Valuesb Model variables

EA 72.0109 Pa Calculate isotropic compliance tensors, SA and SM


EM 30.0109 Pa
A ˆ  M 0.42

A Calculate isotropic thermal expansion coecient tensors


M A and M

0 C ˆ CM ÿ CA 0.0 J/(m3K)
 A
0 0 dS
Ht 0.05 ˆ ÿ
Ht d t
 A
0 0 dS
Hc ÿ0.035 ˆÿ
 A  M jHc j d c
dS dS Ht ‡ jHc j Ht ÿ jHc j
ˆ 8.4106 J/(m3 K) t
J2 : ˆ H ; J2 ÿ I1: ˆ ; ˆ
d t d t 2 2
Vtr Ht ‡ jHc j 27 Ht ÿ jHc j ÿ 2 Vtr
ÿ0.0034 J2 ÿ J3 ÿ I1 :  ˆ ; ˆ ;! ˆ
V 2 2 H ‡ jH j
t c 3V
ÿ 
Aof 281.6K  ˆ 0 U0 ‡ 1 ˆ 12 0 0 Mos ‡ Aof
ÿ  ÿ of 
Aos 272.7K 0 bA A
3 ‡ b2 ˆ ÿ0 0 A ÿ A
os

ÿ  ÿ 
Mos 254.9K 0 bM 3 ‡ b2
M
ˆ ÿ0 0 Mos ÿ Mof
of 0 1 ÿ A  1ÿ A 
M 238.8K 2 ˆ b ÿ bM3 ‡ 2 b2 ÿ b2
M
2 3 3
ÿ 
Y ˆ ÿ 12 0 0 Aof ÿ Mos ÿ 2
a
A=austenite, M=martensite, t=tension, c=compression.
b
Taken from the work of Jacobus et al. (1996).

and M . In the current model, only the di€erence in speci®c heat of both phases,
C, appears. Many calorimetry experiments show that the speci®c heat of both
phases is almost equal, hence it is assumed that c ˆ 0 (Lagoudas and Bo, 1999).
The maximum uniaxial transformation strains, Ht and Hc , for tension and com-
pression, respectively, are the recoverable strains after the material is fully loaded to
the martensitic state at a temperature below Aos and then unloaded completely.
One way of calibrating the material parameters ; ; ; ;  and ! is to use the
strains associated with transformation. For the J2 transformation function can be
identi®ed with either the maximum tensile transformation strain, Ht , or maximum
compressive transformation strain Hc ; Ht is chosen here. Recall, that the J2 ÿ I1
model can either be used to model the volume change during phase transformation
or the tension-compression asymmetry. In the former case, is chosen as mentioned
for the J2 model and is simply one-third of the percentage volume change, i.e.
tr
ˆ V3V . In the later case, and are obtained using the maximum transformation
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1329

strains, Ht and Hc . For the J2 ÿ J3 ÿ I1 transformation function, ! is equal to one-


third of the volumetric transformation strain, and  and  are calibrated using the
maximum transformation strains, Ht and Hc .
The speci®c entropy di€erence per unit volume, 0 0 , can be calculated from the
area under the curve of a Di€erential Scanning Calorimetry (DSC) test.14 Another
method is to use the slope of the transformation curves in stress-temperature space. For
the uniaxial case, the initiation condition for phase transformation (3.22) can be plotted
to obtain forward and reverse phase transformation curves in stress-temperature space.
The slope of these curves at zero stress determine the speci®c entropy di€erence per
unit volume between the phases. The slope of these curves is approximated to be
ÿ0 0
Ht and ÿjH0 
cj
0
for uniaxial tension and compression, respectively.
The transformation temperatures Mos ; Mof ; Aos and Aof are calculated from the
strain vs. temperature graph at zero stress, or can be obtained from a DSC test. The
dissipation constant, Y, can be obtained from an isothermal uniaxial pseudo-elasti-
city test. It is directly correlated to the total hysteresis area enclosed by a stress-
strain curve obtained from such an experiment. Y can also be obtained in terms of
speci®c entropy di€erence and transformation temperatures by calibrating the
model as shown in Table 1.
The speci®c internal energy di€erence per unit volume, 0 U0 , and the six trans-
formation hardening parameters, 0 bA M A M
3 ; 0 b3 ; 0 b2 ; 0 b2 and 2 can be directly
calibrated from the model in terms of speci®c entropy di€erence and transformation
temperatures. It is not necessary to evaluate 0 U0 and 1 separately because they
both contribute to the linear part of the Gibbs free energy with respect to . There-
fore, another constant,  :ˆ 0 U0 ‡ 1 , is de®ned. Note in Table 1 that only the
summation of hardening constants, 0 b3 and 0 b2 is determined. Therefore, the

Fig. 5. (a) Uniaxial tensile and compressive tests to obtain the maximum transformation strains and the
volumetric change due to phase transformation. (b) Di€erential scanning calorimetry (DSC) test to mea-
sure the speci®c entropy di€erence and the transformation temperatures.

14
Bo et al. (1999), pages 79, 80, carry explanation of this method in considerable detail.
1330 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343

choice of one of the parameters has to be made based on the shape of hardening in
the uniaxial stress-strain curve. Depending on that, a linear, quadratic or a mixture
of hardening e€ects are possible.
In summary, the least number of tests necessary to obtain the material constants
and parameters beside the thermoelastic constants, A and M , is 3. Two uniaxial
tensile and compressive tests in the temperature range Mos 44Aos are required to
obtain Ht and Hc as shown in Fig. 5a. The Young's moduli, EA and EM can be
obtained from the tensile or compressive test. The diametral and axial strain mea-
tr
surements during the uniaxial test can be used to obtain the volumetric change, V3V ,
and Poisson's ratios, A ; M . A DSC measurement can be used to obtain 0 0 , C,
and Mos ; Mof ; Aos and Aof (see Fig. 5b). The remainder of the material parameters in
Table 1 are obtained in terms of these constants.

4. Implementation and results

The experimental results in the paper by Jacobus et al. (1996) are employed for
comparison purposes in this section. They conducted various uniaxial and triaxial
loading experiments on SMA NiTi bars. The stress state was homogeneous in all of
the tests. Five types of experiments were performed: (1) uniaxial tension (UT), (2)
uniaxial compression (UC), (3) hydrostatic compression (HC), (4) overall zero
hydrostatic pressure (ZH) and (5) triaxial compression (TC). The loading matrices
of these tests are given in Fig. 6. Three di€erent specimen were used in the tests with
the following geometries: (1) L ˆ 25:4 mm, D ˆ 7:37 mm for UT, (2) L ˆ 19:05 mm,
D ˆ 11:13 mm for UC, and (3) L ˆ 12:7 mm, D ˆ 5:59 mm for triaxial cases, where
L is the length and D is the diameter of the SMA bar. The tests were carried out at
very slow strain rate of 10ÿ4/s to ensure that there was no signi®cant temperature
change due to latent heat.

Fig. 6. Loading matrix for the uniaxial tension (UT), uniaxial compression (UC), hydrostatic compres-
sion (HC), zero hydrostatic pressure (ZH) and triaxial compression (TC) cases.
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1331

The material constants, EA ; EM ; A ; M ; Aos ; Aof ; Mos and Mof are obtained from
the above experimental work and given in Table 1. No DSC measurement is pro-
vided for the SMA material used in the experiments. The slope of the e€ective stress
…J2 † versus temperature is given approximately as 8.0 MPa K among many possible
choices. The speci®c entropy di€erence, 0 0 , is then calculated based on a slope
of 8.4 MPa K . It is further assumed that 0 0 is the same for both tensile and com-
pressive cases. From the experimental observation, it is also concluded that trans-
formation hardening is linear for both tension and compression in this material,
therefore, 0 bA M
3 and 0 b3 are assumed to be zero.
The tests were carried out at a temperature above Aof and the plastic deformation
occurred before ®nal fracture and before full martensitic transformation took place.
Therefore, no experimental data is available on the maximum uniaxial transformation
strains. To make a suitable choice of these constants the following procedure is adop-
ted. We note that for the uniaxial case, the transformation direction tensor, K, which
appears in (3.25), (3.28) and (3.30), is equal to Ht …Hc †, for the tensile (compressive) case.
Also, at the given critical transformation stress for both tension and compression,
 ˆ 0, and the transformation function, , also has a value of zero. Furthermore, the
experimental temperature is taken to be 308 K>Aof . We now substitute the values of
all the known quantities into the transformation function in (3.17) and solve for
Ht …Hc †. Values of Ht ' 0:5 > jHc j ' 0:035 are obtained under this procedure.
The boundary value problems are numerically simulated using the commercially
available nonlinear ®nite element code ABAQUS. The incremental (discretized)
SMA constitutive model based on an explicit return mapping algorithm is imple-
mented in the user supplied subroutine UMAT.15 The bars are assumed to be ®xed
at one end in the axial direction. Eight node quadratic axisymmetric elements are
used to geometrically discretize the problem domain. All ®ve experiments are mod-
eled for the J2 ; J2 ÿ I1 and J2 ÿ J3 ÿ I1 transformation functions. The J2 ÿ I1 case is
used to model the tension-compression asymmetry only. The load is applied such
that all austenite transforms into martensite followed by full unloading to the aus-
tenitic state (pseudoelastic e€ect).
The second deviatoric stress and strain invariants were used as e€ective measures
in the post-processing of the experimental data. The two e€ective measures and,
furthermore, an e€ective transformation strain measure are used in the numerical
results also for consistency. These measures can be written in the following form for
the loading matrices given in Fig. 6
p
2 1
eff
S ˆ …S11 ÿ S22 †2 ‡…S11 ÿ S33 †2 ‡…S22 ÿ S33 †2 2 ; …4:1†
2

15
The numerical implementation of the model will not be discussed in this section. The interested
reader is referred to Qidwai and Lagoudas (1999) for full numerical development. That development has
been extended to take into account the multiple transformation functions presented in this paper and the
nonlinear deformation measures. A side note: ABAQUS uses Eulerian deformation measures as input/
output to UMAT, therefore, transformation mappings had to be performed for the correct input to
Lagrangian SMA constitutive model and the output back to Eulerian form for ABAQUS usage. The
details of these mappings are not given here for the sake of conciseness
1332 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343

p
2 1
Eeff ˆ …E11 ÿ E22 †2 ‡…E11 ÿ E33 †2 ‡…E22 ÿ E33 †2 2 ; …4:2†
3
p h
2 ÿ t 2 ÿ 2 ÿ 2 i12
Et eff
ˆ E11 ÿ Et22 ‡ Et11 ÿ Et33 ‡ Et22 ÿ Et33 : …4:3†
3
where the subscripts {11,22,33} denote frr; zz; g, respectively. Let us also de®ne the
critical e€ective transformation stress, Seff
c to be the e€ective stress as de®ned in (4.1)
at the beginning of forward phase transformation. Similarly, the maximum e€ective
transformation strain, Etm eff , is the e€ective transformation strain as de®ned in (4.3)
at the end of the forward phase transformation. And, critical hydrostatic pressure,
Pc , is the hydrostatic pressure when Seff ˆ Seff c

4.1. Results

In this section, the model results for the uniaxial tension (UT) and uniaxial com-
pression (UC) experiments are simulations, since these experiments have been uti-
lized to calibrate the model as discussed in the beginning of this section. The
remaining experimental data for hydrostatic compression (HC), zero hydrostatic
pressure (ZH) and total compression (TC) are model predictions.
In Fig. 7a and b, axial stress is plotted versus axial strain in tension and com-
pression for the three transformation functions. In tension, the material response is
the same for all choices. However, for the compressive case, the simulation based on
the J2 function is di€erent from the other two, because the J2 function is symmetric
about the origin in uniaxial stress (see Fig. 1). Note that the J2 ÿ I1 and J2 ÿ J3 ÿ I1
transformation functions do simulate the increased magnitudes of critical stress and
lower maximum transformation strain for compression, whereas the J2 function
predict the same magnitude of critical transformation stress and maximum trans-
formation strain for tension and compression.
In the hydrostatic compression test, all the normal strains are equal, i.e.
E1 ˆ E2 ˆ E3 ˆ E. The prediction of the strain, E, is plotted against the hydrostatic
pressure for the three transformation functions in Fig. 8. The maximum pressure
applied is ÿ800 MPa. The resulting curves fall on top of each other in a straight line
indicating that SMA does not undergo a phase transformation under pure hydro-
static pressure. The experimental curve (Jacobus et al., 1996) shows a slightly dif-
ferent elastic response in Fig. 8. Simple calculations based on the experimental
Young's modulus of austenite (72 GPa) reveal that for such an elastic response, a
Poisson's ratio of 0.41 is required instead of 0.42 as suggested in the experimental
work. Since the e€ect of a small di€erence in Poisson's ratio is not expected to be
large on the phase transformation behavior, the value of 0.42 is used in the later
results also.
The elastic response predicted by all three models can also be graphically
explained in the 3-D realization of the transformation function, , at  ˆ 0 at the
beginning of the forward transformation, for all choices of  ^ in Fig. 9. We note that
the J2 based transformation function is a cylinder with the hydrostatic pressure as its
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1333

axis. The J2 ÿ I1 based transformation function is a cone with its apex on the posi-
tive side of the hydrostatic pressure axis (not shown here), indicating a pure elastic
response for negative hydrostatic pressure.16 The J2 ÿ J3 ÿ I1 transformation func-
tion is a surface with its major axis along the hydrostatic pressure axis and the sur-
face tapers to an end on both side of the axis (not shown here). For the material
constants used in this analysis, the end on the negative side is at least twice the
applied load in this test, i.e. 1.6 GPa.

Fig. 7. Axial stress versus axial strain for di€erent transformation functions under (a) uniaxial tension
(UT), (b) uniaxial compression (UC).

Fig. 8. Strain, E1 ˆ E2 ˆ E3 ˆ E, versus pressure exhibiting elastic SMA deformation for di€erent
transformation functions.
16
Similar to the case of porous and geological materials, a restrictive surface can be introduced as a
cap on the open-ended side of the cone towards the negative side of the hyrdrostatic pressure axis. How-
ever, this modi®cation to the J2 ÿ I1 function requires more experiments to be performed at very high
negative hydrostatic pressures to observe phase transformation before it can be justi®ed.
1334 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343

Fig. 9. Transformation functions, , in 3-D stress space for: (a) J2 , (b)J2 ÿ I1 , (c) J2 ÿ J3 ÿ I1 (general-
ized).
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1335

The e€ective stress versus e€ective strain response of the material is plotted in Fig.
10a for J2 transformation function. Four loading cases, i.e. UT, UC, ZH and TC are
shown in the ®gure. The e€ective material response for each case is the same due to
the symmetry of the J2 function. That is, the e€ective critical transformation stress
required in each loading case is equal, contrary to the experimental measurements
(Jacobus et al., 1996) as shown in Fig. 10b. Only the forward phase transformation
part of the numerical results is plotted to be consistent with the experimental results.
Also, in Fig. 10a, we note that the maximum e€ective transformation strain is the
same as the maximum transformation strain for the uniaxial case, i.e. Etm eff ˆ Ht .
This means that the amount of recoverable strain is una€ected by the amount of
hydrostatic pressure.
The J2 ÿ I1 transformation function, however, is asymmetric in stress space as it is
observed in Fig. 11a. where the e€ective stress is plotted against the e€ective strain
for the four loading cases. Note that the maximum e€ective transformation strain

Fig. 10. (a) E€ective stress versus e€ective strain under uniaxial tension (UT), uniaxial compression (UC),
zero hydrostatic pressure (ZH) and triaxial compression (TC) for J2 transformation function; (b) com-
parison of experimental and numerical e€ective stress versus e€ective strain material response under var-
ious loading cases for J2 transformation function.

Fig. 11. (a) E€ective stress versus e€ective strain under uniaxial tension (UT), uniaxial compression (UC),
zero hydrostatic pressure (ZH) and triaxial compression (TC) for J2 ÿ I1 transformation function; (b)
comparison of experimental and numerical e€ective stress versus e€ective strain material response under
various loading cases for J2 ÿ I1 transformation function.
1336 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343

for all loading cases in Fig. 11a is within the interval of the maximum transforma-
tion strains for tensile and compressive cases. This result will be explained later in
Section 4.2. The forward phase transformation part of the numerical result is plotted
against the experimental ®ndings (Jacobus et al., 1996) in Fig. 11b and they agree
very well.
In Fig. 12a, the e€ective stress versus e€ective strain response is shown for the
J2 ÿ J3 ÿ I1 transformation function. We observe that in cases for which hydrostatic
pressure is greater than zero, the critical maximum transformation strain is equal to
Ht , whereas it is equal to the magnitude of Hc for cases in which hydrostatic pres-
sure is less than zero. The material response for this transformation function does
not closely follow the experimental results for the ZH and TC as shown in Fig. 12b.
For example, the critical e€ective transformation stress required in the TC case is
lower than the UC case. Similarly, the critical e€ective transformation stress
required in the ZH case is lower than the UT case.
The model prediction of the critical e€ective transformation stress for each load-
ing case versus the critical hydrostatic pressure is plotted in Fig. 13 for all transfor-
mation functions, together with the experimental data points (Jacobus et al., 1996).
For purpose of clarity, the model data points are joined by a solid line for each
transformation function. The best match between the numerical and experimental
results is obtained for the J2 ÿ I1 transformation function. The J2 transformation
function based results show constant e€ective stress versus hydrostatic pressure as
mentioned earlier. The curve due to the J2 ÿ J3 ÿ I1 transformation function fall
between the curves due to J2 ÿ I1 and J2 transformation functions.

4.2. Discussion

The results in the previous section show that the J2 transformation function do
not model the experimental results within a reasonable accuracy. For all the loading
cases, same critical e€ective transformation stress and maximum e€ective transfor-
mation strains are found. That is, the modeled SMA material is not pressure

Fig. 12. (a) E€ective stress versus e€ective strain under uniaxial tension (UT), uniaxial compression (UC),
zero hydrostatic pressure (ZH) and triaxial compression (TC) for J2 ÿ J3 ÿ I1 (generalized) transforma-
tion function; (b) comparison of experimental and numerical e€ective stress versus e€ective strain material
response under various loading cases for J2 ÿ J3 ÿ I1 transformation function.
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1337

Fig. 13. Comparison of experimental and numerical e€ective stress versus pressure under uniaxial tension
(UT), uniaxial compression (UC), zero hydrostatic pressure (ZH) and triaxial compression (TC) for dif-
ferent transformation functions.

dependent, which is against the experimental ®ndings. The model also cannot take
into account the volumetric transformation strain. The maximum e€ective trans-
formation strain is equal to Ht for all the loading cases. The components of trans-
formation strain, the e€ective transformation strain and the transformation
volumetric strain results are summarized in Table 2 for the four loading cases.
The results for the J2 ÿ I1 transformation function agree quite well with the
experimental results, i.e. the e€ective critical transformation stress increases with
increasingly negative hydrostatic pressure and the uniaxial maximum transforma-
tion strain decreases in compression relative to in tension. However, the prediction
of the model of a volumetric transformation strain of 2.25% during phase transfor-
mation is erroneous (see Table 3). Recall that asymmetry was introduced in the J2
transformation function through the ®rst stress invariant, I1 [see (3.27)]. Thus, a

Table 2
Variables related to transformation strain for the J2 transformation function

Loading case Trans. strain comp. E€ective trans. strain Etm eff Trans. vol. strain E vol

Et1 Et2 Et3

UT ÿ0.025 0.050 ÿ0.025 0.05 0.0


UC 0.025 ÿ0.050 0.025 0.05 0.0
ZH ÿ0.025 0.050 ÿ0.025 0.05 0.0
TC 0.025 ÿ0.050 0.025 0.05 0.0
1338 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343

Table 3
Variables related to transformation strain for the J2 ÿ I1 transformation function

Loading case Trans. strain comp. E€ective trans. strain Etm eff Trans. vol. strain E vol

Et1 Et2 Et3

UT ÿ0.0138 0.0500 ÿ0.0138 0.0425 0.0224


UC 0.0288 ÿ0.0350 0.0288 0.0425 0.0226
ZH ÿ0.0138 0.0500 ÿ0.0138 0.0425 0.0224
TC 0.0288 ÿ0.0350 0.0288 0.0425 0.0226

large volumetric transformation strain results due to the dependence of the trans-
formation strain rate on the derivative of the function,  ^ , as shown in (3.20). Per-
haps, this indicates the requirement of an enhanced non-associativity of the
transformation strain ¯ow rule in the stress-space than it already is. A remedy could
be the introduction of physically appropriate terms in the Gibbs free energy, G, such
that the thermodynamic force p…S; ; †, which appears in , carries extra terms in
stress. This would have the e€ect of increasing the non-associativity of the trans-
formation strain ¯ow rule.
Another solution has been given for other pressure dilatant materials, such as,
rate-independent elastic-plastic materials, soils, rocks, etc. When the yield function
includes a dependence on ®rst stress invariant and a normality rule is employed to
de®ne the inelastic strain evolution in these materials, similar incongruity in the
inelastic volumetric strain is obtained. To correct this error, Casey and Jahedmo-
tlagh (1984), for the case of purely mechanical rate-independent elastic±plastic
material, divided the plastic strain into deviatoric and volumetric parts. They pro-
ceeded to propose an associative ¯ow rule for the deviatoric plastic strain satisfying
normality in the deviatoric stress space. However, the volumetric plastic strain ¯ow
rule was assumed to be non-associative. This set-up captured the experimental
volumetric plastic strain both qualitatively and quantitatively. However, it will not
satisfy maximum dissipation principle as it is posed here in the paper.
The e€ective transformation strain at the end of the phase transformation is
4.25% for all the loading cases for J2 ÿ I1 transformation function. Since, the
experimental data is not available on maximum e€ective transformation strain, the
numerical results cannot be veri®ed. We, however, note that the maximum e€ective
transformation strain is the average of the maximum transformation strains in
t c
tension and compression, i.e. Etm eff ˆ H ‡2jH j. This e€ect is due to the presence of
volumetric transformation strain in the components of transformation strain. Table
3 summarizes the various component values pertaining to transformation strain for
the J2 ÿ I1 transformation function. As a result, the axial stress-axial strain response
does not match with the e€ective stress-e€ective strain response for uniaxial tension
for the J2 ÿ I1 function. Fig. 14a and b show this comparison for the J2 and J2 ÿ I1
transformation functions, respectively.
The SMA model based on the J2 ÿ J3 ÿ I1 transformation function is able to
account for the negative transformation volumetric strain of  ÿ0.34%. It also
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1339

correctly simulates the uniaxial results and predicts higher critical e€ective trans-
formation stress and lower maximum transformation strain during compression.
However, the e€ective response does not follow the experiments in all the cases.
That is, the critical e€ective transformation stress does not increase with negative
hydrostatic pressure necessarily. The e€ective transformation strain at the end of
phase transformation is 5.11% for the UT and ZH cases, that is, Etm eff ˆ Ht ‡ . On
the other hand, for the UC and TC cases it is equal to 3.39%, that is, Etm eff ˆ jHc j ÿ .
For comparison, the axial stress-axial strain and e€ective stress-e€ective strain
responses are plotted for the uniaxial tension case in Fig. 14c. The components of
transformation strain, the e€ective transformation strain and the transformation
volumetric strains for each loading case is also summarized in Table 4.
Before we conclude, we remark for completeness on the ability of the present
model to capture the torsion induced and combined tension-torsion induced phase
transformation experiments reported by Lim and McDowell (1999). They carried
out non-isothermal torsional, and proportional and non-proportional combined
tensile-torsional loading tests on untrained polycrystalline NiTi torque tubes. Even
though we have only modeled isothermal pseudoelastic cases for trained NiTi SMA

Fig. 14. Axial/e€ective stress versus axial/e€ective strain under uniaxial tension (UT) for (a) J2 transforma-
tion function, (b) J2 ÿ I1 transformation function, (c) J2 ÿ J3 ÿ I1 (generalized) transformation function.
1340 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343

Table 4
Variables related to transformation strain for the J2 ÿ J3 ÿ I1 (generalized) transformation function

Loading case Trans. strain comp. E€ective trans. strain Etm eff Trans. vol. strain E vol

Et1 Et2 Et3

UT ÿ0.0267 0.0500 ÿ0.0267 0.0511 0.0034


UC 0.0158 ÿ0.0350 0.0158 0.0339 0.0032
ZH ÿ0.0267 0.0500 ÿ0.0267 0.0511 0.0034
TC 0.0158 ÿ0.0350 0.0158 0.0339 0.0032

in this paper, the trends are the same for e€ective critical stress and e€ective max-
imum transformation strain with respect to applied loading under isothermal and
non-isothermal cases. For the torsion induced phase transformation, the experi-
mental stress-strain behavior exhibits symmteric response for positive and negative
shear strains. We note here that the J2 ÿ I1 and J2 ÿ J3 ÿ I1 transformation functions
reduce to J2 transformation function for torsion since J3 ˆ I1 ˆ 0. In that case, sym-
metric torsional stress±strain response will be predicted by all the transformation
functions.
The experimental combined tension-torsion non-isothermal proportional loading
case can also be qualitatively described here. The experiments exhibited an asym-
metric stress±strain behavior similar to the tension±compression asymmetry referred
to earlier. The J2 based transformation function will not predict this asymmetry due
to reason mentioned earlier in the discussion, however, the J2 ÿ I1 and J2 ÿ J3 ÿ I1
transformation functions can be calibrated to account for it in the manner presented
in Section 3.4. The non-proportional loading cases cannot be modeled with the
present model due to reorientation of the martensitic variants in the actual experi-
ments. Additional internal state variables related to reorientation must be added to
the model to account for reorientation due to non-proportional loading.

4.3. Conclusion

In conclusion, the model based on the J2 transformation function provides the least
accurate comparison with the experimental results for polycrystalline NiTi SMA. The
SMA constitutive model based on the J2 ÿ I1 and J2 ÿ J3 ÿ I1 transformation func-
tions capture the uniaxial results correctly. Furthermore, the J2 ÿ J3 ÿ I1 transfor-
mation function based SMA model does not capture the correct critical e€ective
transformation stress for zero hydrostatic stress (ZH) and triaxial compression (TC)
tests; however, it does take into account the volumetric transformation strain accu-
rately. On the other hand, the J2 ÿ I1 transformation function based model captures
the increasing critical e€ective transformation stress with increasingly negative
hydrostatic pressure, but at the cost of predicting a large positive volumetric trans-
formation strain which is physically unrealizable for polycrystalline NiTi SMAs.
This may be remedied by forming a non-associative transformation strain ¯ow rule
in stress space.
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1341

Acknowledgements

The authors acknowledge the ®nancial support of the Air Force Oce of Scienti®c
Research under Grant No. F49620-98-1-0041, monitored by Major Brian Sanders.

References
Adler, P.H., Yu, W., Pelton, A.R., Zadno, R., Duerig, R., Barresi, T.W., 1990. On the tensile and tor-
sional properties of pseudoleastic ni-ti. Scripta Metallurgica et Materiaia 24, 943±947.
Auricchio, F., Taylor, R.L., Lubliner, J., 1997. Shape-memory alloys: macromodelling and numerical
simulations of the superelastic behavior. Computer Methods in Applied Mechanics and Engineering
146, 281±312.
Beveridge, G.S., Schechter, R.S., 1970. Optimization: Theory and Practice, McGraw-Hill, New York.
Birman, V., 1997. Review of mechanics of shape memory alloy structures. Applied Mechanics Reviews 50
(11), 629±645.
Bo, Z., Lagoudas, D.C., 1999. Thermomechanical modeling of polycrystalline SMAs under cyclic loading,
part i: theoretical derivations. International Journal of Engineering Science 37, 1089±1140.
Bo, Z., Lagoudas, D.C., 1999. Thermomechanical modeling of polycrystalline SMAs under cyclic loading,
part iii: evolution of plastic strains and two-way memory e€ect. International Journal of Engineering
Sciences 37, 1141±1173.
Bo, Z., Lagoudas, D.C., 1999. Thermomechanical modeling of polycrystalline SMAs under cyclic loading,
part iv: modeling of minor hysteresis loops. International Journal of Engineering Science 37, 1174±
1204.
Bo, Z., Lagoudas, D.C., Miller, D., 1999. Material characterization of SMA actuators under non-
proportional thermomechanical loading. Journal of Engineering Materials and Technology 121, 75±85.
Boyd, J.G., Lagoudas, D.C., 1994. Thermomechanical response of shape memory composities. Journal of
Intelligent Materials and Structures 5, 333±346.
Boyd, J.G., Lagoudas, D.C., 1996. A thermodynamic constitutive model for the shape memory alloy
materials. Part i. the monolithic shape memory alloy. International Journal of Plasticity 12, 805±842.
Brinson, L.C., Lammering, R., 1993. Finite-element analysis of the behavior of shape memory alloys and
their applications. International Journal of Solids and Structures 30, 3261±3280.
Casey, J., 1985. Approximate kinematical relations in plasticity. International Journal of Solids and
Structures 21, 671±682.
Casey, J., Jahedmotlagh, H., 1984. The strength-di€erential e€ect in plasticity. International Journal of
Solids and Structures 20 (4), 377±393.
Casey, J., Naghdi, P.M., 1992. A prescription for the identi®cation of ®nite plastic strains. International
Journal of Engineering Science 30 (10), 1257±1278.
Coleman, B.D., Noll, W., 1963. The thermodynamics of elastic materials with heat conduction. Archives
of Rational Mechanical Analysis 13.
Davidson, F.M., Liang, C., Lobitz, D., 1996 Investigation of torsional shape memory alloy actuators. In:
Mathematics and Controls in Smart Structures, Vol. 2717, pp. 672±682. SPIE.
Gall, K., Sehitoglu, H., 1999. The role of texture in tension-compression asymmetry in polycrystalline
NiTi. International Journal of Plasticity 15, 69±92.
Gall, K., Shehitoglu, H., Chumlyakov, Y.I., Kireeva, I.V., Maier, H.J., 1999. The in¯uence of aging on
critical transformation stress levels and martensite start temperatures in NiTi: part i Ð aged micro-
structure and micro-mechanical modeling. ASME Journal of Engineering Materials and Technology
121, 19±27.
Gillet, Y., Patoor, E., Berveiller, M., 1999. Calculation of pseudoelastic elements using a non symmetrical
thermomechanical criterion and associated rule. Journal of Intelligent Materials and Technology 9,
366±378.
Govindjee, S., Hall, G.J., 1999. Computational aspects of solid-solid phase transformation modeling with
a gibbs function. In: Mathematics and Controls in Smart Structures, pp. 302±313. SPIE.
1342 M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343

Hill, R., 1948. A variational principle of maximum plastic work in classical plasticity. Quaterly Journal of
Mechanics and Applied Mathematics 1, 18±28.
Jacobus, K., Sehitoglu, H., Balzer, M., 1996. E€ect of stress state on the stress-induced martensitic
transformation in polycrystalline Ni±Ti alloy. Metallurgical and Materials Transactions 27A.
Jardine, A.P., Kudva, J.N., Martin, C., Appa, K., 1997. Shape memory alloy TiNi actuators for twist
control of smart wing designs. In Mathematics and Controls in Smart Structures, Vol. 2717, pp. 160±
165. SPIE.
Lagoudas, D.C., Bo, Z., 1999. Thermomechanical modeling of polycrystalline SMAs under cyclic loading,
part ii: material characterization and experimental results for a stable transformation cycle. Interna-
tional Journal of Engineering Sciences 37, 1249±1305.
Lagoudas, D.C., Bo, Z., Qidwai, M.A., 1996. A uni®ed thermodynamic constitutive model for sma and
®nite element analysis of active metal matrix composite. Mechanics of Composite Materials and
Structurs 3, 153±179.
Liang, C., Davidson, F., Schetky, L.M. Straub, F.K., 1996. Applications of torsional shape memory alloy
actuators for active rotor blade control±opportunities and limitations. In: Mathematics and Controls in
Smart Structures, pp. 91±100. SPIE.
Liang, C., Rogers, C.A., 1992. The multi-dimensional constitutive relations of shape memory alloys.
Journal of Engineering Mathematics 26, 429±443.
Lim, J.T., McDowell, D.L., 1999. Mechanical behavior of a Ni-Ti shape memory alloy under axial-
torsional proportional and nonproportional loading. Journal of Engineering Materials and Technology
121, 9±18.
McMillan, C., 1970. Mathematical Programming: An Introduction to the Design and Application of
Optimal Decision Machines, John Wiley & Sons, New York.
Melton, K.N., 1990. Ni-Ti based shape memory alloys. In: Duerig, T.W., Melton, K.N. StoÈkel, D.,
Wayman, C.M. (Eds.), Engineering Aspects of Shape Memory Alloys. pp. 21±35.
Melzer, A., StoÈkel, D., 1994. Performance improvement of surgical instrumentation through the use of
Ni-Ti materials. In: Pelton, A.R., Hodgson, D., Duerig, T. (Eds.), Proceedings of the First Interna-
tional Conference on Shape Memory and Superelastic Technologies. pp. 401±409.
Mises, V., 1928. Mechnik der plastischen formanderung von kristallen. Z. Angen. Math. Mech. 8.
Patoor, E., Barbe, P., Eberhardt, A., Berveiller, M., 1988. Thermomechanical behavior of shape memory
alloys. Arch. Mech. 40, 775±794.
Qidwai, M.A., Lagoudas, D.C., 1999. Numerical implementation of a shape memory alloy thermo-
mechanical constitutive model using return mapping algorithms. Journal of Numerical Methods in
Engineering 47, 1123±1168.
Rajagopal, K.R., Srinivasa, A.R., 1998a. Mechanics of the inelastic behavior of materials. part i: theore-
tical underpinnings. International Journal of Plasticity 14 (10-11), 945±967.
Rajagopal, K.R., Srinivasa, A.R., 1998b. Mechanics of the inelastic behavior of materials. part ii: inelastic
response. International Journal of Plasticity 14 (10-11), 969±995.
Raniecki, B., Lexcellent, C., 1998. Thermodynamics of isotropic pseudoelasticity in shape memory alloys.
European Journal of Mechanics and Solids 17 (2), 185±205.
Reisner, G., Werner, E.A., Fischer, F.D., 1998. Micromechanical modeing of martensitic transformation
in random microstructures. International Journal of Solids and Structures 19, 2457±2473.
Sachdeva, R.C. & Miyazaki, S., 1990. Superelastic Ni±Ti alloys in orthodontics. In: Duerig, T.W. Melton,
K.N., StoÈkel, D., Wayman, C.M. (Eds.), Engineering Aspects of Shape Memory Alloys. pp. 452±469.
Simo, J.C., Hughes, T.J.R., 1998. Computational Inelasticity, Springer-Verlag, New York.
Simo, J.C., Ortiz, M., 1985. A uni®ed approach to ®nite deformation elastoplastic analysis based on the
use of hyperelastic constitutive equations. Computer Methods in Applied Mechanics and Engineering
49, 221±245.
SÆittne, P., Takakura, M., Tokuda, M., 1995. The stabilization of transormation pathway in stress induced
martensite. Scripta Metallurgica et Materilia 32 (12), 2073±2079.
Sun, Q.P., Hwang, K.C., 1993. Micromechanics modeling for the constitutive behavior of polycrystalline
shape memory alloys-i: Derivation of general relations. Journal of Mechanics and Physics of Solids 41,
1±17.
M.A. Qidwai, D.C. Lagoudas / International Journal of Plasticity 16 (2000) 1309±1343 1343

Tanaka, K., 1986. A thermomechanical sketch of shape memory e€ect: one dimensional tensile behavior.
Res Mechanica 18, 251±263.
Tanaka, K., Nishimura, F., Hayashi, T., Tobushi, H., Lexcellent, C., 1995. Phenomenological analysis on
subloops and cyclic behavior in shape memory alloys under mechanical and/or thermal loads.
Mechanics of Materials 19, 281±292.
Tokuda, M., Ye, M., Takakura, M., SÏittner, P., 1998. Calculation of mechanical behaviors of shape
memory alloy under multiaxial loading conditions. International Journal of Mechanical Sciences 40(2±3),
227±235.
Ziegler, H., 1963. Some extremum principles in irreversible thermodynamics. In: Progress in Solid
Mechanics, North Holland, Amsterdam, New York.
Ziegler, H., 1983. An Introduction to Thermodynamics, North Holland, New York.

You might also like