Professional Documents
Culture Documents
Martensitic Transformations in
Shape Memory Alloys
KARL-HEINZ HOFFMANN
Institute of Mathematics
University of Augsburg
Universitätstraβe 2,
D-8900 Augsburg
Germany
MAREK NIEZGÓDKA
Institute of Applied Mathematics and Mechanics
Warsaw University
PKiN, 00-901 Warsaw
Poland
INTRODUCTION
effect
SHAPE MEMORY physical property is a characteristic of numerous
HE including various
solids, metallic alloys and non-metallic solid materials like
This
polymers. of solid,
property consists in an ability a subject to plastic defor-
mation, to recover its original shape after an appropriate thermal treatment (pos-
sibly complemented by a mechanical loading). The effect had already been dis-
covered in the mid-thirties, but an explosive development of interest in it, as well
as an understanding of its enormous applicability range, date from the late sixties
and are related to the discovery of the extraordinarily strong and, equally, pre-
served in time, shape memory property of Ti-Ni alloy (Nitinol), cf. [6,29]. The
same type of effect has also been discovered in many other metallic alloys. There
exists, in-between, an extensive literature devoted to the physics of shape memory
effect (cf. [6,17,28]) and its applications (cf. [3,6,12]). It is not our objective to
give an overview of these aspects. What we are going to expose is related to the
most common characteristic features of the dynamical processes in materials ex-
hibiting shape memory, and then to construct phenomenological models capable
of forecasting the developments in space and time both qualitatively and quantita-
tively. Since the behavior is strongly affected by the choice of a specific class of
materials, we shall focus on metallic alloys, with Nitinol in mind, in particular.
We shall discuss the applicability range of the models proposed and shall show
some typical results of numerical experiments which visualize their forecasting
value.
,1. of INTELL MATER. SYST. AND STRUCT, Yol. I-July 1990 355
1045-389X/90/03 0355-20 $4.50/0
@ 1990 Technomic Publishing Co , Inc
the thermal activation can be applied in either direction, both as a heating and
cooling action (at least in many cases).
In accordance with the traditional metallurgical terminology, the high-
temperature phase is usually referred as austenite and the low-temperature phase
is called martensite. Certainly, there are many specific transformation
mechanisms applicable to different materials. An extensive discussion of the
mechanisms driving the transformations in Nitinol is offered in [6,17-19,28].
There is a strong dependence of the transformation mechanism on chemical com-
ponents and their ratio.
Regardless of its nature, the martensitic transformation not only contributes to
changes in the symmetry of a crystal lattice, but also produces its homogeneous
structure. Another difference, compared to the pure twinning, is concerned with
the role of non-elastic deformations E which can now have non-vanishing trace
(on the contrary to the twins where always trE = 0); the resulting phases differ
from the parent not only in orientation but also in their crystal symmetry.
Such symmetry breaking can be exploited in various systems of energy conver-
sion. This is one of the most significant applications areas for shape memory
materials. Thermally activated (as a rule a small temperature increase suffices
there), the systems are capable of performing mechanical work. In this connec-
tion, let us mention the whole class of thermal engines, so-called thermobiles
[3,6,12], with the energy conversion efficiency up to nearly 10%. In these engines,
cyclic (quasi-periodic) conversion of heat into mechanical work proceeds on ac-
count of the two-way shape memory effect and the martensitic transformation
within the crystal lattice. Such conversion implies, of course, the necessity of a
simultaneous treatment of the coupled energy and momentum balances in con-
structing the corresponding phenomenological models.
Besides, energy dissipation should be taken into account. In order to provide
good performance of such converting systems, the possibility of applying a con-
trolled activating action should be ensured. A typical way of the controlled acti-
vation for the systems under considerations is either via boundary thermal treat-
ment, a mechanical forcing, or via distributed action that uses some external
physical fields. We shall discuss this by constructing the appropriate mathe-
matical models and by performing experiments of numerical simulation in the
following sections.
A number of other application areas of shape memory alloys are discussed in
[3,6,29] .
PHENOMENOLOGICAL MODELING
The construction of mathematical models of the dynamical processes in shape
memory alloys goes back to I. Muller et al. [1,2,20-22] who have developed a
class of the rate models based on localization (or averaging) in space and on ap-
plication of statistical mechanics arguments for estimating coefficients of the
model. The basic component of that class of models was a crystal lattice cell; all
developments were reduced to its level. With the values of coefficients dependent
only on the space variables, the models described the dynamics of processes in
crystal lattice exclusively in terms of the phase ratios, with temperature playing
the role of an additional internal parameter.
The construction we are going to expose is based on assuming the free energy
as a function of temperature and an order parameter. This is an underlying
hypothesis of Landau’s theory (cf. [6,9]). The order parameter is a variable (of
scalar, vector or tensor type) which uniquely specifies the local phase structure
at a fixed temperature. In the case of processes in shape memory alloys, the strain
tensor can be taken as the order parameter (cf. [6,9]).
The classical Landau theory, making use of the 4th order representation of the
free energy with respect to the order parameter, was applicable to the processes
in two-phase systems, without twinning phenomena in either of phases and only
in the case of 2nd order (continuous) phase transitions, if any.
First, its extension referred to as the Landau-Devonshire theory (cf. [6,9,21])
became applicable to modeling phase transitions in shape memory alloys. Includ-
ing terms up to the 6th order with respect to the order parameter, it equally
reflected the twinning of martensitic M+- and M--phases, and the discontinuous
character of the corresponding phase transition.
In the sequel we confine ourselves to discussing the case of one-dimensional
structures where the Landau-Devonshire approach has proved successful,
delivering physically relevant results. All physical variables will further be ex-
pressed in their specific volumetric values.
The Landau-Devonshire model is based on assuming the thermodynamic
potential in the form of the Helmholtz free energy * defined as a function of
Kelvin temperature 0 and the strain E,
common for all involvedphases. The free energy function * is postulated to sat-
isfy the following hypotheses:
(a 1) * is C &dquo;-function in both variables
(a2) ’Ir is adjusted so that for the autonomous isothermal system it is an even
function of E
(a3) for the autonoumous isothermal system, minima of ~Y(~9,E) (with respect to
E) correspond to the stationary equilibrium states
(a4) there exists a Curie point z9, such that sharply divided thermodynamically
stable phases may exist only below it (above this critical temperature, just
a single phase, high-symmetric austenite, remains thermodynamically
stable)
(a5) in the temperature range below Curie point ~9~ the following behavior is to
be reproducible:
1° the martensitic phases, in particular M~-twins, are the only stable
phases at low temperatures; for z9 < 61 the isothermal free energy.
(~9,E) has exactly two symmetrically located minima as function of E
2° for temperatures between the critical points, ~9 E (~91,~9~), both marten-
sites and austenite are thermodynamically stable; ~Y(~,E) has two lateral
minima and the central one with respect to E
3since above the Curie point ~9~ only the austenite remains thermo-
dynamically stable, the corresponding free energy exhibits just a single
minimum (central) with respect to E
The original autonomous Landau free energy, assumed as a polynomial of the 4th
order in E, was unable to reproduce the twinning effects and successive modifica-
tion of the stable structure at the increase of temperature. All the above
hypotheses are fulfilled by the Landau-Devonshire free energy (cf. [6,9]). In the
case of materials like Nitinol, where exactly two low-temperature phases exist
and for whom one-dimensional geometry can be postulated, the Landau-Devon-
shire energy can be assumed in the form (cf. [4,6,23,24])
where ~Yo(~9) is the term referring to pure heat conduction, defined in the standard
form
e- internal energy
s - entropy
q- heat flux
k- heat conductivity
X- rate of distributed heat sources
f- distributed forces
he Gibbs-Thompson relation then takes the form
With all introduced constitutive hypotheses and relations, the dynamics of the
martensitic phase transitions in shape memory alloys can be described by the fol-
lowing system of balance equations over S2 x (0,T) (we remain at the case of
one-dimensional structures 0 C IR1):
Energy Balance
Mass Balance
The heat flow is assumed to satisfy the Fourier heat conduction law
This enables us to reduce the model to the energy and momentum balance equa-
tions only.
To complete the model, we need to introduce appropriate initial and boundary
conditions. In particular, we can assume them in the following form:
where a-positive heat exchange coefficient through the boundary r, ~9e, Qe-ex-
temal temperature and stress, respectively (given as functions of time, hence ap-
plicable as the activating variables).
Let us note that the additional activation of the process is possible via the dis-
tributed source terms in balance equations. Altogether this provides multiple pos-
sibilities of forcing controlled developments in the system.
When the thermal inertion of the system is negligibly small (i.e., the coefficient
a in (14) is sufficiently large, then instead of (14) one can prescribe the boundary
The Landau-Devonshire model in the generalized form, with the free energy
and represents the surface energy at the interphase boundaries, now forming
boundary layers rather than walls that would sharply separate single phases.
The assumed form of the free energy is reflected by an appropriate modifica-
tion of the balance equations. In the one-dimensional case, they differ from the
Landau-Devonshire equations in as much as the momentum equation includes
now the additional term xu.,_,.
The third model (cf. [5,10,13]), also making use of the Landau-Ginzburg free
energy, differs in the way of constructing the total free energy. It is defined here
as a combination of the energies of single phases. As an additional variable, the
vector of phase ratios is introduced and the model is an extension of a variational
inequality that describes the time-evolution of the latter vector by rate laws.
MATHEMATICAL AND NUMERICAL ASPECTS OF THE MODELING
We shall confine here ourselves to discussing the Landau-Devonshire model in
thecase of one-dimensional setting.
This model admits the creation of sharp walls separating different phases. In
this connection, jump discontinuities of the order parameter E are admissible.
Consequently, it does not make sense to interpret the model pointwise in its clas-
sical formulation where smoothness of all components is necessary. Instead, the
model is to be understood in a weak variational sense. Let us be more specific
and consider the one-dimensional situation, with a homogeneous boundary con-
dition for displacement u. In this case, the form of the model globally integrated
over il x (O,T) is the following.
Problem (VM)
Determine a pair I t9,usuch that on a given time interval [0,T] the constitutive
relations hold and the following variational equations are satisfied for any
For this model, global-in-time existence and uniqueness of the weak solution
I t9,u), i.e., a pair of functions that satisfy the constitutive relations and the system
(19), (20), has been proved in [15,24,26].
Results on the Lipschitz continuous dependence of the solution on the data of
the problem have been established in [14]. These results contribute to the possi-
bility of controlling the process development. Hence, a combined activation via
the imposed boundary conditions and distributed source terms may contribute to
the creation of desired phase structures, to a localized precipitation and provide
regular in time propagation of the free boundaries separating different phases.
In the case of the Landau-Ginzburg model, asymptotic results on stable steady
states of the system (their number and localization) have also been obtained (cf.
[11]).
In order to show the character of the predictable system developments, we will
describe some typical results of the performed numerical simulation experi-
ments. Both mechanically and thermally activated processes were simulated. The
activating action was performed at the domain boundary. We shall restrict our-
selves to the one-dimensional case. We will complete the discussion with a pres-
entation of a series of graphically represented results of the performed numerical
experiments.
The first stage of the experiment consisted of observing the developments that
accompany a mechanical loading via boundary conditions imposed on the stress
(with the activation symmetric at both end-points and lasting for 400 time steps).
Figures 3-5 visualize the most characteristic effects obtained from the numerical
simulation. In Figures 3 and 4, the initial patial structure of the system was
assumed homogeneous, with pure M’-phase throughout. In turn, the initial struc-
ture in Figure 5 comprised both martensitic twins forming a domain pattern with
sharp walls in-between. Let us notice the following qualitative features:
~ As long as the temperature remained low (below a critical point), only marten-
sitic twins could be considered thermodynamically stable, hence macroscop-
ically observable (see all Figures 3 through 5).
~
Up to and exceeding the yield limit by the activating boundary stress, the pro-
cess evolved smoothly (without any creation of singularities); the spatial struc-
ture of the system remained unchanged-no phase transition occurred [see
Figures 3(a) and 4(a)].
~ After the yield limit had been achieved, a discontinuity interface arose and
began propagating; this interface corresponded to the initiation of the
martensite-austenite phase transition and caused the movement of the free
boundary separating different phases [see Figures 3(a) and 4(a)].
~ An additional nucleation followed in the interior of the domain due to super-
position of the boundary actions applied at both end-points [see Figure 3(b)].
~ Since the boundary activation was extended only over a finite time interval, the
developments could be observed to stabilize, with a sharply defined spatial
phase structure [see Figures 3(c,d) and 4(b)].
~
Regardless of the initial situation (provided thermodynamically stable), the
spatial phase structure remained non-affected by mechanical activation below
the yield limit [see Figures 5(a,b), which demonstrates a continuation of the
experiment presented in Figure 3].
Converting the direction of the mechanical activation at low temperatures
resulted in a reverse development, with alternation between the martensitic twins
as the only phase transition, up to a certain hysteretic nonlinearity (due to the
slipping within the central loop of the material equilibrium curve, cf. Figure 2).
This phenomenon is shown in Figure 6 where an appropriate trajectory for a
fixed point of the domain is drawn in the E (7-plane. -
the Landau-Devonshire models (cf. [4]) and for the Landau-Ginzburg models
(cf. [30]) were discussed.
FINAL REMARKS
The models so far developed are mainly one-dimensional. In the three-
dimensional case, only static situations [24] or rather specific (diagonalized)
dynamic structures were considered.
A complete treatment of spatial transformation groups in the dynamical frame-
work is still to be done.
Another aspect, so far neglected, concerns variations of the mass density and
the corresponding dynamical mass transfer (diffusive effects). A closely related
aspect of this mass density is the accounting for thermal expansions of the
dynamical systems involved.
And after all, a very important factor in the dynamics of shape memory alloys
is concerned with the rheological character of the constitutive laws, of special
significance in cyclically activated processes. This kind of effect has not yet been
studied at all. Except for Nitinol, where rheological properties are of less signifi-
cance, in other shape memory alloys subject to cyclic forcing they represent a
crucial factor manifesting in dynamic developments.
All of the directions of research we have just mentioned still require basic
study, with special accent on the numerical simulation of the processes as an
underlying step at constructing phenomenologically correct models.
ACKNOWLEDGEMENTS
The work partially supported by the DFG, SPP &dquo;Anwendungsbezogene
was
REFERENCES
1 Achenbach, M , I. Muller. 1983 "Creep and Yield in Martensitic Transformation," Ingenieur-
Archiv, 53.73-83
2 Achenbach, M and I Muller 1984 Shape Memory as a Thermally Activated Process,
Hermann-Fottinger Institut, TU Berlin, Preprint
3 Aifantis, E. C and J Gittus 1986 Phase Transformations Amsterdam Elsevier
4 Alt, H W , K -H Hoffmann, M Niezgodka and J Sprekels 1985 "A Numerical Study of Struc-
tural Phase Transitions in Shape Memory Alloys," Inst. of Mathematics, University of Augsburg.
5. Colli, P., M. Frémond and A. Visintin 1988 "Thermo-MechanicalEvolution of Shape Memory
Alloys," Pubbl. No. 607, IAN CNR, Pavia
6. L and M. Chandrasekaran 1987. Martensitic Transformations Les Editions Physique,
Delaey,
Les Ulis.
7 Ericksen, J 1987 "Twinning of Crystals I," in: S S Antmann, et al , Metastability and In-
completely Posed Problems, New York Springer-Verglag, pp. 77-93
8 Falk, F "Martensitic Domain Boundanes in Shape-Memory Alloys as Solitary Waves," in
Martensitic Transformations.
9. Falk, F. 1990 "Elastic Phase Transitions and Nonconvex Energy Functions," K. H. Hoffmann,
J. Sprekels, eds., Free Boundary Problems-Theory and Applications V-VI, London: Longman