Home Search Collections Journals About Contact us My IOPscience

Three-dimensional Landau theory describing the martensitic phase transformation of shape-

memory alloys

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

1990 J. Phys.: Condens. Matter 2 61

(http://iopscience.iop.org/0953-8984/2/1/005)

View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 130.15.241.167
The article was downloaded on 19/04/2013 at 15:13

Please note that terms and conditions apply.

J . Phys.: Condens. Matter 2 (1990) 61-77. Printed in the UK

Three-dimensional Landau theory describing the

martensitic phase transformation of shape-memory
alloys

F Falki and P Konopka

FB Physik, Universitat-GH-Paderborn. D-4790 Paderborn, Federal Republic of Germany

Received 13 December 1988, in final form 3 July 1989

Abstract. In shape-memory alloys a first-order martensitic phase transition is responsible for

pseudo-elastic and for ferro-elastic stress-strain relations. To describe this behaviour a
modified Landau theory is proposed in which the free energy of the crystal depends on the
temperature and on the full strain tensor. The energy is invariant with respect to the cubic
point group Ohof the high-temperature phase. To predict the cubic-to-monoclinic phase
transitionofb-phaseshape-memory alloysanexpansion uptosixthorder instrain isnecessary
for which, for the class of alloys considered, odd terms may be neglected. For a CuAlNi alloy
the expansion coefficients are determined by comparison with experimental results. In
contrast to classical Landau theory of second-order phase transitions, not only a single
second-order but also a fourth-order expansion coefficient depend on temperature.

1. Introduction

The shape-memory effect in certain metallic systems called ‘shape-memory’ alloys has
been the subject of considerable experimental effort. In some temperature ranges these
alloys show pseudo-elastic and ferro-elastic stress-strain curves. It transpires that these
peculiarities are the consequence of a first-order martensitic phase transition which is
connected with a nearly-volume-conserving spontaneous deformation of the crystal
lattice (Delaey et a1 1974). In addition to temperature, external stress may induce the
phase transition from the high-temperature austenitic phase to the low-temperature
martensitic phase. At low temperature stress causes deformation twinning between
crystallographically equivalent martensite variants.
In spite of some interesting applications of these alloys, up to now there has been no
conclusive theoretical description of their mechanical or thermomechanical behaviour.
We propose here to use a modified Landau theory to derive a free-energy function
dependent on strain and on temperature which, as a thermodynamic potential, enables
calculation of the equilibrium properties. For some alloys there have been attempts in
this direction. Nittono and Koyama (1982) and Koyama and Nittono (1982) derived a
free-energy function for indium alloys such as InTl, InCd, InPb and InSn undergoing a
cubic-to-orthorhombic or cubic-to-tetragonal phase transition. Group theory was used
in order to guarantee the correct symmetry of the free energy. Later on Barsch and
t Present address: Battelle-Institut e.V., Am Roemerhof 35, D-6000 Frankfurt 90, Federal Republic of
Germany.

0953-8984/90/010061 + 17 $03.50 @ 1990 I O P Publishing Ltd 61

62 F Falk and P Konopka

Krumhansl (1984) included strain-gradient terms to calculate the structure of moving

twin boundaries between tetragonal martensite variants in InTl. NiTi or copper-based
shape-memory alloys, which are more important for applications, undergo a cubic-to-
monoclinic phase transition. Within a one-dimensional model Falk (1980) proposed a
free-energy expansion of sixth order in a single shear strain component. An expansion
up to fourth order in the full strain tensor invariant with respect to the cubic group has
been given by Liakos and Saunders (1982). As will be seen below, for shape-memory
alloys an expansion up to sixth order is necessary. It is the aim of this paper to construct
such a free-energy function invariant with respect to the cubic symmetry group Oh of
the high-temperature phase describing the first-order phase transition to monoclinic
martensite. The expansion coefficients are determined for a CuAlNi alloy. Furthermore
the stress-strain relations and the heat of transformation are derived and compared with
experiments.

2. The martensitic phase transition in shape-memory alloys

Before constructing the free-energy function some remarks on the martensitic phase
transition in shape-memory alloys are appropriate. In this section we deal with the
temperature-induced transition, i.e. both the phases are assumed to be in a stress-free
state. In NiTi or copper-based alloys such as CuZn, CuAlNi, CuAlZn and CuAlGa and
other /?-phase alloys such as NiAl, AuCd and AgCd, the austenite shows, at temperatures
not much above the martensitic transition temperature, an ordered BCC structure of
point group Oh(m3m). The space groups differ in the various alloys but they are not
relevant for the symmetry of the strain tensor.
When austenite transforms to martensite on cooling a spontaneous strain appears.
On the scale of the lattice cell of austenite, that is over a few Angstrom, the Bain strain
deforms the cubic unit cell to an orthorhombic one. As a consequence the following
misfit problem arises. Imagine an inclusion of orthorhombic martensite in a cubic
austenitic matrix. Even if there is no volume change then across the interface between
both the phases a lattice misfit occurs leading to a long-range stress field of high energy.
One way of reducing this stress is to build an incoherent interface, i.e. an amorphous
boundary layer. This is not the way shape-memory alloys solve the problem since it is
observed that the phase transition proceeds by rather rapid interface motion which is
possible only for coherent interfaces. The only possibility of vanishing lattice misfit stress
across a coherent interface occurs if it is invariant with respect to the strain. This means
that any material vector in the interface must not be deformed. In order that an invariant
plane exists at least one eigenvalue of the strain tensor has to vanish. In addition, the
remaining two eigenvalues must be of opposite sign. However, the orthorhombic Bain
strain does not fulfil this condition. As a consequence, in shape-memory alloys mar-
tensite forms as an internally microtwinned or faulted phase, that is, layers of twins of
orthorhombic martensite alternate on a scale of a few atomic distances. In the alloys
mentioned the twinning or basal plane is (110) if referred to cubic axes. The relative
amount of twins is such that the total strain resulting from Bain and microtwinning strain
does not deform the so called habit plane which therefore acts as the contact plane
between the austenite and martensite. The orientation of twinning and habit plane, the
Bain strain, the total strain and the amount of twins are related by WLR theory (Wechsler
eta1 1953). A review on this topic is given in the book by Wayman (1964).
In the shape-memory alloys mentioned the microtwinning occurs in a regular pattern
 Landau theory of shape-memory alloys 63

on a scale of a few cubic unit cells, so that a monoclinic lattice containing many atoms
per unit cell results forming a martensite variant. Typically, those variants show a 2H,
3R, 9R or 18R structure, depending on the alloy under consideration. On a length scale
roughly four orders of magnitude bigger it is observed by light microscopy that shape-
memory alloys form a rnacrotwinned structure in the martensite state. This is because a
plate-like inclusion of martensite variant in an austenite matrix generates, due to a shape
misfit, a stress field at its circumference even if there is no lattice misfit across the habit
plane. This stress field decreases almost to zero if an arrangement of twins of monoclinic
martensite variants forms (self-accommodating groups: see, e.g., Tas et a1 (1973) and
de Vos et a1 (1978)).
The free-energy function to be constructed in the present paper deals with the
deformation of single crystals on a scale between the monoclinic lattice and the self-
accommodating groups. The appropriate strain is therefore defined on a scale of a few
nanometres. As a consequence the Bain strain is below the resolution of the theory and
does not occur here. On the other hand the deformation of a macroscopic sample
consisting of self-accommodating groups of martensite is not dealt with either. Instead
the theory describes the phase-transition phenomena between cubic austenite single
crystals and monoclinic single-crystalline martensite variants on a mesoscale. In order
to describe the macroscopic behaviour of shape-memory alloys one could proceed by
some averaging over the different martensite variants forming the macroscopic sample
in the way sketched by Falk (1989) for a one-dimensional model. However, this is not
the objective of the present paper.
To define mathematically the appropriate mesoscale strain measure consider a triad
of materially fixed vectors u1which in stress-free austenite are chosen to be orthogonal
and of the same mesoscale length a. of a few nm. For simplicity take vectors parallel to
the cubic axes. As a result of deformation which may be due to the martensitic phase
transition or to an applied load the vectors change to U’. The symmetric strain tensor is
defined by
e , = f(U’ * U, . .
- a’ a l ) / a ; = +(cl a / a ; - 811). (1)
The product phase on the mesoscale is called a martensite variant. In the shape-memory
alloys mentioned, a martensite variant shows monoclinic symmetry of point group C2(2)
whereas the space groups are different for the various alloys. Since the order of the cubic
symmetry group 0,of the parent phase is 48 and that of the monoclinic group Czof the
product phase is two, 24 possible orientations of the product occur, i.e. there exist 24
martensite variants. All of them follow from a single one by applying symmetry opera-
tions of the cubic group. Since they are equivalent crystallographically their energies
coincide.

3. Free-energy expansion

The equilibrium properties of a system are determined by the free-energy function. In

Landau’s theory of second-order phase transitions one restricts the variables on which
F depends to the temperature and to the so called order parameter which characterises
the difference between the phases. Furthermore, it is argued that the order parameter
transforms as one of the irreducible representations (IRS) of the more symmetric high-
temperature phase (Landau and Lifshitz 1980). In order to describe not only the first-
order martensitic phase transition but the complete thermomechanical behaviour of the
64 F Falk and P Konopka

system, one has to take into account the full strain tensor. The dependence of F on
strain is governed by the point group Oh of the high-temperature phase. If a symmetry
operation in three-dimensional space is represented by the 3 X 3 matrix D with com-
ponents D ,with respect to the cubic orthogonal base vectors ui, then the strain matrix
or strain tensor transforms according to

Symmetry requires that

F(e, T ) = F(e', T ) (3)
for each of the 48 matrices D. Necessary group-theoretical tools for exploiting the
symmetry requirements are found in Lyubarsky (1960). For the purpose of generating
symmetry-adapted free-energy functions it is more convenient to look at the symmetric
strain tensor as a vector in a six-dimensional strain space S spanned by the dyads

Q11 = a1 @a1 472 = a2 @a2 Q13 = a3 @a3

2 q 4 = a2 @ a3 + a3 @ a2
(4)
2 9 , = a l @ a 3 + a3 @ a l
2 q 6 = a , @ a 2 + a2@a,.
The components of strain with respect to q K ( K = 1 . . . 6) are denoted by eK:

el = e11 e? = e22 e3 = e33

e4 = 2eZ3 e5 = 2e13 e6 = 2e12
describing the strain in Voigt notation. In strain space S spanned by q K ,symmetry
operations of the group Oh are represented by the 6 X 6 matrices dKLfollowing from the
representation Dmnso that the strain transforms according to

e k = dKLe,. (6)
The representation d,, of group Oh is reducible into the rRs rl,r12
and r25'. Accordingly
the strain space S decomposes into the invariant subspaces SI(one-dimensional), SI2
(two-dimensional) and S 2 5 (three-dimensional):
S = S1 CB Si2 CB S25 (7)
which are spanned by @, (K = 1 . . . 6):
S1 @i = qi + ~ 1 +2 q 3
s12 @5 = 2 q 3 - 91 - v2 @Pj = 91- 92 (8)
s25 CP, = 2 q 4 @ j = 2Q5 0;= 2 9 6 .
If referred' to the symmetry-adapted base vectors @i, the components of strain are
denoted by e i .
ei = ( e , + e2 + e 3 ) / 3 = tr(e,,)/3 (9)
 Landau theory of shape-memory alloys 65

in the linear approximation describes the volume change during deformation.

represent shear deformations on the (110) planes in the (i10) directions. The remaining
components
e: = e& = e23 e ; = e 5 / 2 = e13 e; = e 6 / 2 = e12 (11)
denote shear deformations on the (100) planes in the (010) directions.
As in Landau theory it is assumed that the free energy is an analytic function so that
it can be expanded into a power series with respect to strain:
F ( e , T ) = F o ( T )+ F'(e,T )+ F 2 ( e ,T ) + . . . (12)
with F" an expression of the nth degree:
F n ( e ,T ) = CLl L n ( T ) e L , . . . e L n L,=1 . . .6 , n a l . (13)
Here and in the following the Einstein convention is used, i.e. over indices occurring
twice summation is implied. Each F" has to meet the symmetry requirements (3) and
(6) individually. The expressions (13) can be interpreted as scalar products of two
vectors, namely C and e 8 . . . 8 e, both elements of a symmetrised product space
S!y,,, spanned by the symmetrical n-fold tensorial products
(vLI , v , ~ , .u
8 .. . 8 V L , ) ~ ~ , ,= . u. v L n 1 G L1 C L2 C . . . C L , =Z 6. (14)
The symmetry requirement on F"(e, T ) explicitly reads
CLl L n e L I . , e L , = CL1 LndLIM,. . . d L , M , e M I . . . e M , (15)
and is valid for arbitrary strain e and every symmetry operation d of group Oh.Therefore
it follows that
CLl Ln = d L I M l.. . dLnM,CM1 M n . (16)
The products d L I M l. . . dLnM,
generate a representation d&, of the group Oh acting on
space S&,, which, in general, is reducible even if the representation generated by dLMin
space S is irreducible. Equation (16) says that the coefficients C in order to obey
symmetry have to be built up by invariant directions n, of space Sa,, according to
I"

where the sum goes over all the invariant directions. Fjn are independent material
parameters of nth order. As invariants, the nj transform according to the identical
representation of oh. There are as many invariant directions as the identical rep-
resentation occurs in the decomposition of the representation d& into IRS. This number
is given by
l g
j" = - tr(d&m(i))
g,=1

where d;y,(i) is the matrix of group element number i in the representation d:y,,,. i runs
over all the group elements, that is from 1to group orderg which is 48 for group Oh.The
66 F Falk and P Konopka

Table 1. Strain invariants J:‘ of first and up to fourth order of the cubic group Oh. Strain
components in symmetry-adapted coordinates (equations (9)-(11)).

Invariant Strain space

invariant directions q, can be calculated by Wigner projection operators according to

K
q;“~ . M n = z
r=l
dLIMl(4. . . dL,M,(i)(PL,U. . . VcpL, (19)

where d L I M , ( iis
) the representation matrix of group element number i in strain space S.
To get all the invariant directions one has to insert the different base vectors
cp L I , ~. . vqL,(1 s L , s . . . s L, s 6) of space S& in turn into the right-hand side of
equation (19). Some of them coincide, but the number of different ones is given by
equation (18). Performing the calculations one finds 1,3, 6 and 11 invariant directions
of first, second, third and fourth order, respectively. To get the invariants of the nth
order in strain one has to multiply q,”l L n by a general eLI. . . eL,. The result is the
invariant J; ( j = 1 . , .y) of nth order corresponding to material parameter F/”so that
the nth degree in the free-energy expansion reads
in
Fn = 2 Fj’Jj’
j= 1

The invariants up to fourth order are completely listed in table 1.For later convenience,
the fifth- and sixth-order invariants which are from the subspace S1@ S12are listed in
table 2.
 Landau theory of shape-memory alloys 67

Table 2. Fifth- and sixth-order strain invariants of group Ohfrom space SI 6

3SI?,
Invariant Strain space

4. Adaptation of the free-energy function to copper-based alloys

Up to fourth order there are 21 material parameters necessary to determine the free-
energy function. If the mentioned fifth- and sixth-order terms are included this number
increases to 32. In order to handle the free-energy expansion one has to reduce this
number by assumptions motivated by experimental results. The first observation is that
stress-free austenite exists at high temperature. This means that for e = 0 the free energy
must have a minimum at least at those temperatures. This is achieved by vanishing of
the first-order term, i.e. F' = 0. The next observation is that at low-temperature the
free energy must have 24 symmetry-related monoclinic minima. Hence there must exist
a minimum in a monoclinic direction which is not invariant with respect to a subgroup
of Ohricher than the monoclinic group C2.Translated into group-theoretical language
this requirement is called the subduction and chain condition (Birman 1978). Further-
more, experiments suggest an approximation for copper-based and for some other
alloys. One observes that in CuAlNi, CuZnGa, CuZn, CuAlZn, AgCd and NiAl, to
each martensitic minimum of strain eManother one with nearly -eM is associated (Saburi
and Wayman 1979, Okamoto et a1 1986). We assume this relation to hold exactly. Hence
one is led to the restriction
F(e, 7') = F(-e, 7') (21)
which can be obeyed only if invariants of odd degree do not occur in F. Looking at the
values of spontaneous strain of martensite in those alloys, one finds moreover that el,
e2and e3 are in the order of magnitude of 0.1 whereas e4, e5 and e6are roughly one order
of magnitude smaller. The volume change, in the linear approximation given by 3ei =
e, + e2 + e3, is even smaller. Thus the main contribution to the spontaneous strain is
from the irreducible space SI2of orthorhombic symmetry, whereas the contributions
from space S25 are smaller by a factor of 10 and that from space S1 is even smaller.
To obtain a first-order phase transition from an even free-energy function one has
to include sixth-order terms. From the observation just described it is suggested that
only sixth-order contributions from space S I 2are taken, that is the sixth-order invariants
68 F Falk and P Konopka

Jf = (J:)3 and J$ = (1;)’are taken into account. In fourth order we take into account
every invariant of space SI2€3 S25,i.e. only volume changes (space S I )are disregarded.
In second order every invariant is included. This results in the free-energy function
3 5 2

F = C. F:J~+ C. F;J;
1=1 r=l
+ C. F P J ;
I= 1
(22)

with 10 material constants F; which may depend on temperature. The contribution

Fo( T ) is suppressed since it only determines the specific heat and does not influence the
thermomechanical properties of the system. In order to find the minima of F one has to
solve the necessary condition dF/de, = 0. This equation has six types of solutions of
invariant plane strain. The first type is the trivial one e = 0 corresponding to austenite
which is stable only if the matrix J2F/deKdeL is positive definite at e = 0 or F: > 0,
F: > 0 and F: > 0. Of the other five types only one type agrees with the subduction and
chain condition. Only solutions of this type yield a monoclinic spontaneous strain which
has the structure
ei=O e i = -a ej = a eg = -/3
ej = 0 e; =P e, = 2 a e2 = 0 (23)
e3 = - 2 a e4 = - 2 p e5 = 0 e6 = 2p.

a and p are the solutions of

4 8 d e + 8 a 2 F j + 2P2(F; + FZ) + F$ = 0
4 a 2 ( F : + F i ) + 2P2(F; + 2 F i ) + FZ = 0.
Generally there are two solutions for a’ and p2. Only one of them has positive values
and in addition yields a minimum of the free energy. This solution is identified as one of
the martensitic minima. The others follow by applying the symmetry operations of the
cubic group Oh.Irrespective of a and /3, i.e. of the material parameters F:, the strain
given by equation (23) always represents an invariant plane strain. The corresponding
invariant plane is identified with the habit plane. Some algebra yields for their normal n
and for the shear vectors, if referred to cubic axes,
n = ( l / k ) ( aB, , a> (25)
s = ( l / k ) [ 2 k 2+ a(7 - l),p(7 - l), -2k2 + a(? - I)] (26)
k = Vp2+ 2 a 2 y= Vl - 8k2 = 1 - /s12/2.
The structure of the habit-plane normal (equation ( 2 5 ) ) is confirmed by experimental
observations, for example in the alloys CuZnGa, CuAlZn, CuZn, NiAl (Saburi and
Wayman 1979), AgCd (Krishnan and Brown 1973) and CuAlNi (Okamoto et af 1986).
If the values of a and j? given by these authors are inserted into the shear of equation
(26) then again the result describes the experimental values rather well with deviations
in the shear direction of a few degrees and in the amount of 1 to 2% of the measured
values.

5 . CuAlNi alloys

In this section we determine the 10 material parameters F; of the free-energy function

 Landau theory of shape-memory alloys 69

(equation (22)) for the alloy CuAlNi by comparison with experimental results. Because
of a lack of appropriate data this is not so easily done. One would like to know the elastic
moduli of austenite single crystals and of martensite single variant crystals together with
their temperature dependence. Unfortunately there has been, as far as we know, only
a single experiment on martensitic elastic moduli of shape-memory alloys, namely on
the y ; martensite of Cu-14 wt%Al-3 wt%Ni by Yasunaga et a1 (1983). We thus decided
to adapt the free energy to this material. However, the experiment was not done on a
single variant crystal but on untwinned martensite. To recalculate the data we have to
use some other theoretical idea described later. Furthermore no temperature depen-
dence was reported. Instead we use data on the heat of transformation.
The first step is to determine F ! , F i and F: from the elastic moduli C t l , Cfi and
C$ of cubic austenite which are the only independent ones and are defined by
C i L = d 2 F / d e K a e L1 e = O . (27)
From equation (22) together with table 1 one finds
- 3Z(C;: + 2 C 5 )
F21 -
F22 - C A - CA12
- I1 (28)
F: = 2C$.
The elastic moduli of a Cu-14.1 wt %A1-3 wt %Ni alloy were measured by Yasunaga et a1
(1982) and by Suezawa and Sumino (1976) for an alloy of a slightly different composition.
Yasunaga et a1 report the temperature dependence of Ctl - C$ and of C$ whereas
from the data of Suezawa and Sumino the bulk modulus may be calculated: this turns
out to be nearly independent of temperature. As a result we have
F: = 592 GN m-'
F; = (14.1 + ( T - 300 K) X 4.6 X K-')GN m-* (29)
F: = (148 - ( T - 300 K) X 9.4 X K-1)GNm-2.
One observes that austenite is highly anisotropic with a very soft shear modulus
Cfi - C$. The seven parameters FP and at 300 K follow from five of the monoclinic
elastic moduli of martensite together with values of the spontaneous strain measured by
Okamoto et a1 (1986). From their data on habit-plane orientation n and on spontaneous
shear 1 s 1 the values
CY = 0.023 /3 = 0.0068 (30)
follow (equations (25) and (26)). There is some scatter in the reported value of y1 which
propagates to 8. It has been mentioned that no data on the elastic moduli of internally
microtwinned martensite exist. Yasunaga et a1 (1983) detwinned y ; martensite prior to
their experiments by applying an appropriate external stress. Their elastic moduli data
referring to orthorhombic single crystals are listed in table 3. We deal here with an
internally microtwinned monoclinic martensite variant which consists of layers of ortho-
rhombic twins. Their amount and their orientation with respect to the austenitic parent
crystal has been reported by Okamoto et a1 (1986). We assume that the elastic moduli
of the microtwinned variant follow as an average over those of the single crystals. To
this end the moduli of the orthorhombic twins are referred to a common system of
coordinates, e.g. to the cubic axes of the parent phase. Then they are averaged according
to their relative amount. The averaging has been done on the elastic moduli themselves
70 F Falk and P K o n o p k a

Table 3. Elastic moduli (in GPa) of orthorhombic y j, martensite single crystals (untwinned)
due to Yasunaga et a1 (1983) in Voigt notation.

189 141 205 54.9 19.7 62.6 124 45.5 115

(as proposed by Voigt (1910)) and on their inverses which are the elastic compliances
(Reuss 1929). For further calculations we use the mean values of both the results listed
in table 4 if referred to base @,or to the symmetry-adapted base @i.These are compared
with
C i , = a’F/ae,aeL 1 ,M (31)
M
takenat the spontaneousstrain ( e x ) = (0, -a, a , -/3,0,/3) of martensite. Fromequation
(22) together with tables 1 and 2 one finds
CE = 2FT CiR = O K>1
C: = 6F: + 120a’F: + 12/3*F; + 10p’F: + 1152a4Ff + 8a4@
C: = 2F: + 24n2F f + 48’ F i + 2/3’ F i + 1 9 2 d Ft + 8a4F9
c-.- C# = 2F: + 12p2Fi + 16/3’F; + 8a’F; + 8 a 2 F :
M
44

C z = 2F: + 8P2F! + 8a’Fi

Table 4. Elastic moduli (in GPa) of monoclinic internally twinned y ; martensite (referred
to symmetry-adapted base vectors). Mean value of Voigt and of R e u s averaging procedure
applied to the elastic constants of the orthorhombic martensite of table 3 referred to base
OKand to symmetry-adapted base O K .

c,, c
2 2 C33 c
44 cis C66 ClZ c,? c23

172 162 172 44 31 44 103 92 103

-2.6 5 1.5 3.7 -20 -2 -0.3 5 0.8 1 -15 -3

1102 447 127 177 123 177 -21 -2 -33

1.5 -21 0.3 -3.6 51 4.8 -12 50 7.3 3.9 -61 -12
 Landau theory of shape-memory alloys 71

Table 5. Elastic moduli (in GPa) of monoclinic internally twinned y ; martensite (referred to
base Q K ) as calculated from the free-energy function.

181 170 181 44 31 44 112 101 112

~~

ci4 CIT cl6 c

24 c?S c26 c34 c3S c36 c4T c46 c56

35 0 -26 -9 0 -9 -26 0 35 0 -15 0

There follows the linear dependence 3Cjj - Cii - 2C5j = 0, which is obeyed by the
values of table 4. Together with equations (29) and (30) the five values Cji, Cjj, C,,,
C;, and C,, are used to calculate the seven parameters FP and FP at 300 K:
F l = -1.182 x lo4 G N m-* F! = 3.13 x lo5 GN m-2 F i = 1.64 x l o 5 GN m-2
F: = -5.53 x lo4 GN m-* F i = -4.27 X lo4 GN m-2 (33)
Ff = 3.35 x lo6 GN m-* 6 = 3.71 X lo7 GN m-2.
In the last step their temperature dependence is investigated. The only hint is the
equilibrium phase transition temperature and the heat of transformation. Todepends
strongly on composition and, consequently, there is a large scatter in the values reported
by different authors. We decided to use the heat of transformation reported by Otsuka
et a1 (1976) to be -48.3 MJ m-3 (or -86.2 cal mol-') at 300 K. Irrespective of the value
chosen here the temperature dependence of F: and F: alone never sufficiently influences
the value of the free energy at the martensitic minimum to produce a reasonable heat of
transformation. At least one of the parameters F: and FF also has to depend on
temperature. Since the major part of spontaneous strain belongs to the space SI, we
chose the only fourth-order parameter of this space, F';, to depend linearly on the
temperature:
F l ( T ) = Flo +(T- 300K)FjT. (34)
The heat of transformation Q follows from the entropy S ( e , T ) defined by
S(e, T ) = -8F/dT
according to
Q ( T ) = TAS
where Tis the temperature of transformation and AS represents the difference in entropy
of both the phases. For a temperature-induced transformation of a stress-free crystal
one finds

Q ( T ) = T ( S ( e A ( T ) ,T ) - s ( e ' ( T ) , T ) ) = -T(4a2F$, + 2P2F& 16Lu4Fj~).

(35)
Together with equation (29) this yields
FtT =35.5GNm-2. (36)
Equations (29), (33), (34) and (36) give the full set of parameters F:' for the free-energy
function F(e, T ) of equation (22).
72 F Falk and P Konopka

6. Consequences of the free-energy function

The free-energy function with the parameters just determined completely characterises
the equilibrium properties of the system. In the following some of these are discussed.
At any temperature the austenitic minimum is at e = 0. The martensitic minima, one
of them determined by equations (23) and (24), the other 23 by symmetry operations,
exist below T, = 303 K. Their positions in strain space, i.e. the spontaneous strain of
martensite with respect to austenite, depend on temperature as shown in figure 1.

e1
..................................................................
0.04 "

-0.04
........ Figure 1. Temperature dependence of the
...... ............... spontaneous strain of martensite. In the
-0.06 dotted range martensite is metastable.

Between T, and To = 271 K the martensitic minima have a higher free energy than the
austenitic minimum. Hence in this temperature range martensite is metastable. Below
Tomartensite is stable, having a lower free energy than austenite. However, since the
austenitic minimum does not vanish the austenite remains metastable. In figure 2 the
free energy is plotted for different temperatures along a straight line in strain space
joining the austenitic minimum with one of the martensitic ones. Note that the magnitude

-20 1
Figure 2. Free-energy curves for different temperatures along a straight line in strain space
joining the austenitic and martensitic minima (see text for details). To avoid intersections,
the different free-energy curves are shifted along they axis.
 Landau theory of shape-memory alloys 73

of the strain at the austenitic minima as well as the direction of the section plotted slightly
depends o n temperature. T h e lowest energy barrier does not exactly lie o n the joining
straight line. Instead, it is given by the energy of a saddle point which, however, is a little
bit lower than the maximum energy along the straight line.
The stress-strain law derives from the free energy according t o
OK = dF/deK (37)
where oKand eK refer to Voigt notation. In figure 3 the derivative of F i n the direction
of the line joining the austenitic and martensitic minima is plotted. It describes a stress-
strain curve during a shear deformation leading to one of the martensite variants. The
part showing a negative slope is unstable since here an elastic modulus is negative.

/
fa1 fbl

n
IS1 0.05
o, ,- Is1
, /

0.15 0.1 e 0 0.05 0.1 0.15 0.15

':
/ o.8-1-/0.1 0.15

/ -1: / -1 I
i -21 .;-
I
/ -21
I C
'E
icl *I : / id)
z

I -
ll D I
/ \ 0.05
.
- .- . ... .
IS1
0 0.05 0.1 0.15

I
i -I!
I I
-1

I -2'
1

Figure 3. Stress-strain curve in shear along the deformation mode from austenite to mar-
tensite for different temperatures: ( a ) T = 340 K. ( b ) T = 300 K. (c) T = 271 K and ( d ) T =
230 K.

Therefore, with increasing stress the austenite becomes unstable when the maximum of
the stress-strain curve is reached where the stress-induced transformation to martensite
along the upper arrow sets in. O n reducing the stress at the minimum, martensite
becomes unstable and the retransformation to austenite starts. In this way the free-
energy curves give rise to a stress-strain hysteresis which, however, is reduced by
nucleation phenomena not included in the Landau approach. A t temperatures T > T,
the stress-strain curves are of pseudo-elastic type whereas at lower temperature ferro-
elastic curves follow.
The elastic moduli in Voigt notation are defined by
CKL, = d'F/deKde[- (38)
where on the right-hand side the appropriate values of strain have to be inserted. In
figure 4 the moduli of monoclinic martensite are plotted as functions of temperature.
The values at 300 K are given in table 5 . T h e moduli C , , , C l 2 ,C 1 3 C
, z z ,CZ3,Cii. Cj4,
Cj5and Chhplotted in figure 4 ( a ) characterise an orthorhombic crystal whereas the rest
74 F Falk and P Konopka

(figure 4(b)) are due to the monoclinic distortion of martensite. We would like to
emphasise that the orthorhombic moduli calculated from the free energy agree rather
well with experiments (compare tables 4 and 5). The monoclinic moduli are much
smaller, differing from the values of table 4. This may be a consequence either of the
averaging procedure necessary to get the moduli of microtwinned martensite or of
neglecting third- and fifth-order terms in the free-energy expansion. This can be resolved
only when experimental data on the moduli of microtwinned martensite are available.
The stability or metastability range of each phase is the region in strain space in which
every eigenvalue of the matrix C K Lis positive. If one approaches the boundary of stability
then the eigenvector corresponding to the vanishing eigenvalue gives the direction in
strain space into which the system will evolve to a stable configuration.

N
-
E
z
Gl,c33 250 -
%
2
... ,
' .. ......

, .. . ....:

. . ... \

(b)

65 c25 c35 [45 [56

c2L c26

c4 6
__cc Figure 4. ( a ) ( b ) Elastic moduli of n i x -
44 Cl, tensite in Voigt notation agdinst tem-
perdture In the dotted range martensite is
metdstahle 1he curies end where mar-
tensite becomes unstdhle

In figure 5 the heat of transformation is plotted against temperature: it decreases

continuously with increasing temperature. The equilibrium phase transition tem-
perature where the free energy of both the phases coincides can be calculated numerically
to yield the result
To = 271 K. (39)
Salzbrenner and Cohen (1979) suggest that T,,is identified with (M, + Af)/2. Calculated
values over a wide range are reported by different authors (see e.g., Otsuka etal 1976,
Okamoto er a1 1986, Yasunaga et a1 1982).
 Landau theory of shape-memory alloys 75

7. Discussion

As it stands the free energy proposed in the present paper describes homogeneous
shape-memory alloys, i.e. one-phase single crystals in which the strain does not vary. In
order to deal with more involved cases, for example with strain depending on position
or even with multiphase crystals, one has to embed the free energy as a constitutive
equation into the basic laws of continuum mechanics, namely into the balance of mass,
of momentum and of energy. Restrict oneself to isothermal equilibrium, then only the
balance of momentum is relevant which then reads
V*a+f=O (40)
within the body and
n.a=a (41)
at the surface, where f and a denote external volume and surface force densities,
respectively, and n is the outer normal of the surface of the body. Typically, f is due to
gravity and may be neglected. Since the free energy contains terms higher than second
order the stress is a non-linear function of strain, which results in the non-linear balance
equations (40) and (41). It is thus rather cumbersome to solve these equations for
arbitrary surface forces or in the case of defects such as dislocations. In some special
cases, however, the problem reduces to one dimension. Assume for example a par-
allelepiped with edges parallel to the habit plane, the orientation of which follows from
the free energy according to equations (25) and (24). Furthermore, consider constant
external forces acting on the surface parallel to the habit plane only. Then the solution
of the balance equations is a shear deformation parallel to the habit plane varying only
in a direction normal to it. The shear is proportional to that given by equation (23). If
one denotes its amount by s the free energy for this special deformation mode is given
by
F(s, T ) = UhSh - u4s4 + a$ (42)
with temperature-dependent positive coefficients a, following from the parameters Fi' ,
Together with the stress-strain curves in shear the result is plotted in figures 2 and 3.
This free-energy function is exactly of the type proposed by Falk (1980) in a purely
one-dimensional approach. A two-phase system consisting of layers of austenite and
76 F Falk and P Konopka

martensite is represented by a one-dimensional deformation of the type just discussed

in which the strain jumps from the value of austenite to that of martensite across the
habit plane. The dynamics of this problem was treated by Falk and Seibel (1987).
Since the free energy proposed in the present paper yields 24 martensitic variants
one can deal with self-accommodating groups of four variants in an austenitic matrix.
For example, four variants with
(eK)I= (0,-2a, 2cu, 0,-28, -28)

(eK)lll= (0,2a,- 2a, -2a, 0 , -28, -28)

(eKIIV = (0,-2a, 2a, 0,28, 28)
yield a vanishing mean value of strain. They have habit planes with normals
nl = (8, a,- 4 all = (8,a,-a)

a,,, = (-8, a,- 4 n I v = (-8, a,-4.

Moreover, the relative strain from one variant to another is an invariant plane strain for
the pairs 1-111, I-IV, 11-111 and 11-IV with normals (0,1,-1), (l,O,O), (1,0,0) and
( O , l , - l), respectively. Therefore in this group every interface is stress-free.
In further investigations heterogeneous nucleation can be treated. Up to now there
have been attempts in this direction by Clapp (1973), Guenin and Clapp (1986) and
Olson and Cohen (1982). In these papers, however, an inappropriate one-dimensional
free energy is used. Taking the three-dimensional free energy proposed for the first time
in this paper, one has to solve the balance equation (40) for defects such as dislocations
of different types and different orientations. Because of the highly non-linear stress-
strain relation this can be done only numerically. It is expected that at least for a
metastable phase around a dislocation core an embryo of the new phase exists in
equilibrium.

References

Barsch G R a n d Krumhansl J A 1984 Phys. Rev. Lett. 53 1069

Birman J L 1978 Group TheoreticalMethodsin Physics, 6th In/. Colloq. ed. P Kramer and A Rieckers (Lecture
Notes in Physics vol79) (Berlin: Springer) p 203
Clapp P C 1973 Phys. Status Solidi b 57 561
Delaey L. Krishnan R V. Tas H and Warlimont H 1974 J . Muter. Sci. 9 1521
Falk F 1980 Acta Metall. 28 1773
- 1989 Int. J . Engng Sci. 27 277
Falk F and Seibel R 1987 Int. J . Eizgng Sci. 25 785
Guinin G and Clapp P C 1986 Proc. Int. Conf. on Martensitic Transformation (Tokyo: Japan Institute of
Metals) p 171
Koyama Y and Nittono 0 1982 J . Physique Coll. 43 C4 145
Krishnan R V and Brown L C 1973 Metall. Trans. 4 423
Landau L D and Lifshitz E M 1980 Statistical Physics (Oxford: Pergamon) Ch. XIV
Liakos J K and Saunders G A 1982 Phil. Mag. A 46 217
Lyubarsky G Y 1960 The Application of Group Theory in Physics (Oxford: Pergamon)
Nittono 0 and Koyama Y 1982 Japan J . Appl. Phys. 21 680
Okamoto K. Ichinose S. Morii K. Otsuka K and Shimizu K 1986Actu Meuzll. 342065
Olson G B and Cohen M 1982 J . Physique Coll. 43 C4 75
Otsuka K , Wayman C M. Nakai K , Sakamoto H and Shimizu K 1976 Acta Metall. 24 207
 Landau theory of shape-memory alloys 77

Rems A 1929 Z. Angew. Math. Mech. 9 49

Saburi T and Wayman C M 1979 Acta Metall. 27 979
Salzbrenner R J and Cohen M 1979 Acta Metall. 27 739
Suezawa M and Sumino K 1976 Scr. Metall. 10 789
Tas H, Delaey L and Deruyttere A 1973 Metall. Trans. 4 2833
Voigt W 1910 Lehrbuch der Kristallphysik (Leipzig: Teubner) p 962 (Reprinted 1960 New York: Johnson
Reprint Corp.)
de Vos J, Delaey L and Aernoudt E 1978 2. Metallk. 69 511
Wayman C M 1964 Introduction to the Crystallography of Marrensitic Transformations (New York: McMillan)
Wechsler M S, Liebermann D S and Read T A 1953 Trans. Am. Inst. Mining Engng 197 1503
Yasunaga M , Funatsu Y. Kojima S , Otsuka K and Suzuki T 1982). Physique Coli. 43 C4 603
- 1983 Scr. Metall. 17 1091