Download as pdf or txt
Download as pdf or txt
You are on page 1of 28

PHILOSOPHICAL MAGAZINE A, 1998, VOL. 77, NO.

5, 1273 1299

M agnetostriction of martensite
By R. D. James
Department of Aerospace Engineering and Mechanics, University of Minnesota,
Minneapolis, Minnesota 55455, USA

and Manfred Wuttig


Department of Materials and Nuclear Engineering, University of Maryland,
College Park, Maryland 20742 2115, USA
[Received 17 March 1997 and accepted in revised form 16 September 1997]

Abstract

A general strategy is described for inducing magnetostriction in ferromagnetic


martensitic materials. An analysis of domain redistribution caused by a magnetic
eld is given, and certain relations between material constants that promote this
e ect are described. These relations suggest a `constrained theory of magnetostriction which is used to predict strain against eld in a tetragonal martensite
subject to an orthogonal biaxial magnetic eld and uniaxial stress. These
predictions are compared with the corresponding experiments in Fe70 Pd30 .
Reversible eld-induced strains of 0.6% are exhibited in this system.
Microstructural observations con rm that these strains are caused by a eldinduced redistribution of martensitic twins.
1. Introduction
Magnetostriction is the spontaneous deformation of a solid in response to its
magnetization. Martensitic transformations also produce a spontaneous deformation of a solid, upon lowering the temperature. If, in addition, the martensitic material is also ferromagnetic, there exists the possibility of inducing the martensitic
transformation or of rearranging the variants of martensite, by applying a magnetic
eld. Since the spontaneous strain in martensitic materials is commonly one order of
magnitude larger than that of (giant ) magnetostrictive materials, the eld-induced
strain available from a martensitic or ferromagnetic material is potentially much
larger than giant magnetostrictive materials. The purpose of this paper is to
investigate this phenomenon: placement of transformation and Curie temperatures, desirable domain structures, micromagnetic theory, fundamental criteria
for the existence of this e ect. We also report the existence of this e ect in the
Fe Pd system.
In ordinary magnetostrictive materials the magnetostrictive strain has a volume
component and a shear component determined by the existence of the spontaneous
magnetization and its anisotropy respectively. In ferrous materials both are normally
5
4
in the 10- range, but they can be as large as 10- in Fe Co alloys (du Tre molet de
Lacheisserie 1993). Much higher values of the shear magnetostrictive strain can be
found in the binary rare earth iron compounds RFe2 . This large shear strain is
accompanied by large magnetocrystalline energy, which limits the technical value
of macroscopic magnetostrictive strain that can be obtained. This limitation can
be overcome by designing a solid solution ( R1 )Fe2 ( R2 )Fe2 of compounds having
0141 8610/98 $12. 00

1998 Taylor & Francis Ltd.

1274

R. D. James and M. Wuttig

magnetocrystalline anisotropy constants of opposite sign such that the ternary alloy
has nearly zero magnetocrystalline anisotropy. This recognition led to the discovery
3
of Terfenol-D (Clark 1980) with a magnetostrictive strain of 10- and zero fourthorder magnetocrystalline anisotropy energy at room temperature.
Another interesting alloy system for the study of the magnetostriction of martensite are Heusler alloys near the composition Ni2 MnGa, studied recently by David
(1991 ), Chernenko and Kokorin (1992 ), K. Ullakko and R. C. O Handley (1996,
private communication) and Ullakko et al. (1996 ). Ullakko et al. demonstrate eldinduced strains of 0.2% in Ni2 MnGa. From the available information, the martensitic crystallography and magnetic structure of Ni2 MnGa are similar to those of Fe3 Pd,
but with apparently di erent easy axes. However, preliminary calculations suggest
that the theory to be presented in 5 is applicable. Because of its large saturation
magnetization and favourable anisotropy (see 4 and James and Wuttig (1996 )), the
Fe Pd alloy system was chosen for the studies reported in this paper.
The evolution of the magnetoelastic strains in martensitic materials depends on
the relative positions of the Curie and martensite start temperatures. In 2 we
describe qualitatively the di erent responses possible with the six di erent relative
positions of transition temperatures. These are eld-induced variant rearrangements
and eld-induced austenite martensite transformation.
In 3 we take a closer look at the possible domain structures, with the goal of
assessing whether there are low-energy transformation paths between the initial and
nal states. A key result states that, even with large deformations (i.e. geometrically
nonlinear strains), the typical strains observed in martensite, together with the typical easy axes observed in ferromagnetic materials, lead to layered domain structures
that are simultaneously mechanically and magnetically compatible. (This is a generalization of a result of James and Kinderlehrer (1993, 6.2). ) These provide lowenergy transformation paths.
In 4 we draw on the results of 2 and 3 to formulate speci c criteria that can
be used to screen promising ferromagnetic shape memory alloys. These criteria are
also useful as a theoretical tool and are satis ed by a `constrained theory of micromagnetics introduced recently by DeSimone and James (1996 ). This theory can be
used to predict the e ect of constant eld and constant stress on behaviour, via
relatively easy calculations. In 5 we use the theory to predict the overall strain
against eld, and associated energy-minimizing microstructures, in several situations
of interest in the Fe Pd system.
In 6 we report the eld-induced rearrangement of martensite variants among
promising alloys in the Fe Pd system, near the composition Fe70 Pd30 . These alloys
have a high value of the saturation magnetization at room temperature and undergo
a cubic-to-tetragonal transformation just below room temperature. We show eldinduced strains of 0.5% at modest elds with a two- eld arrangement. There is
conceptual agreement between the measured and predicted dependences of strain
against eld, but quantitative discrepancies indicate that the precise mechanism in
this system needs further study.
While our results are presented in the context of ferromagnetic martensites, they
apply equally to ferroelectric martensites. In particular, most of the results of 2 5
hold with the magnetization m replaced by the polarization p. Exceptions are, in the
few cases where we invoke the evenness of the free energy in m, our results apply
only to electrostrictive materials, and the use of conductive electrodes to apply the
electric eld to ferroelectric materials leads to di erent boundary conditions from

Magnetostriction of martensite

1275

those assumed in 5. The results can be easily modi ed to account for these di erences.
A preliminary version of several of the results presented here was given by James
and Wuttig (1996 ).
2. Martensite and Magnetism
The material of interest will have two transformations: a ferromagnetic transition and martensitic transformation. Depending on how the Curie temperature and
the martensitic transformation temperature are ordered, the material can be expected
to have di erent behaviours. In this section we consider the e ects of di erent
orderings on qualitative behaviours.
We assume that the ferromagnetic transition is second order, but we allow the
martensitic transformation to be either rst or second order. (The terms `transition
and `transformation are chosen here to re ect this distinction. ) We rst observe that,
in general, martensite and austenite are expected to have di erent Curie temperatures. From a theoretical viewpoint the Curie temperatures are obtained by linearized theory (in the spirit of the Landau (1965 ) theory of second-order phase
transitions), while the martensitic transformation is of rst order, that is associated
with an exchange of stability among remote `energy wells .
There are then three relevant temperatures: the martensitic transformation
mart
temperature aus - mart , the Curie point aus
C for austenite and the Curie point C
for martensite. Excluding equalities, there are six di erent orderings passible;

< aus
< mart
C
C ,
mart
aus
aus- mart < C
< C ,
aus
C
< aus- mart < mart
C ,
mart
C
< aus- mart < aus
C ,
aus
mart
C
< C
< aus- mart ,
mart
aus
C
< C
< aus- mart .
aus- mart

(1)

To illustrate the qualitative response, let us picture the austenite martensite


transformation as the `square-to-rectangle distortion illustrated in gure 1. The six
di erent orderings lead to the four di erent qualitative behaviours shown there. In
the case when the austenite martensite transformation temperature is the lowest
( gure 1 (a)), the non-ferromagnetic austenite rst undergoes a ferromagnetic transiaus
tion upon cooling to C . One would then expect divergence-free magnetization as
pictured qualitatively for the square structure in gure 1 (a). There will also be a very
small conventional magnetostriction associated with this magnetic ordering which is
not illustrated (and is not of main interest here). Upon further cooling, the martensitic transformation will take place at aus - mart . With the ordering of temperatures
shown in gure 1 (a) the martensite will necessarily be ferromagnetic but, owing to its
low symmetry, is expected to have uniaxial magnetic anisotropy, as shown.
For the purpose of this section we ignore the thermal hysterises of the martensitic transformation, commonly de ned by the temperatures Ms , Mf , As and Af . For de niteness,
aus- mart can be considered as the temeprature at which unstressed austenite and martensite
have equal free energies.

1276

R. D. James and M. Wuttig

Figure 1. Qualitative pictures of domain structures with di erent orderings of austenite


martensite transformation temperature aus - mart , Curie temperature aus
C of austenite
and Curie temperature mart
of martensite.
C

Depending on the domain structure, there is the expectation that the easy axes
corresponding to the two variants of martensite point in divergent directions.
Thus by alternately applying a eld in various directions it should be possible to
favour one variant over another, leading to a relative change in shape (of the order of
the eigenstrain of the martensite). Alternatively, one can have a stress favour one
variant and a eld favour the other, as discussed more precisely in 4. An example of
a material corresponding to gure 1 (a) is Ni2 MnGa.

Magnetostriction of martensite

1277
mart

From gure 2 it can be appreciated that, even though a Curie temperature C


for martensite exists at a temperature greater than aus - mart , it may not be observable
under ordinary conditions, because (upon heating from a low temperature) the
crystal transforms to austenite before mart
is reached. In such situations it may be
C
observable under conditions of stress-induced transformation. That is, it is generally
possible to stabilize the martensite at temperatures greater than aus- mart by applying
mart
a suitable stress. Then one can search for C by conventional methods, with the
caveat that the crystal has been slightly biased by the stress, for example its response
is expected to be comparable with the ordinary magnetostrictive material heated
under stress through its Curie point, which has been well studied (du Tre molet de
Lacheisserie 1993).
Note that orderings shown in gure 1 lead to the same multivariant state at low
temperatures. Thus, irrespective of the ordering of temperatures, the potential for
eld-induced variant rearrangement is possible. A schematic arrangement for achieving the full eigenstrain, involving the competition between eld and stress, is shown
in gure 2 (a). However, there is another method suggested by gure 1. Suppose that

Figure 2. Schematic methods of producing a change in shape: (a) eld-induced rearrangement


of martensite variants; (b) eld-induced austenite martensite transformation.

1278

R. D. James and M. Wuttig

there is a jump of the magnetization at the austenite martensite transformation.


From gure 1 this occurs in the rst four orderings of 1. In these cases there is
the potential of causing a eld-induced austenite-martensite transformation (see
Kakeshita, et al. (1995) and references therein). There are several ways to conceptualize such a transformation; one possibility is illustrated in gure 2 (b) in the case
aus
mart
mart
aus
when either aus- mart < C < C or aus- mart < C < C , the austenite martensite transformation is rst order, and the saturation magnetization of martensite
exceeds that of austenite. Suppose that a compressive stress is applied to the austenite as shown on the left of gure 2 (b). The ordinary magnetomechanical coupling of
the austenite is then likely to lead to the uniaxial domain structure shown. Now
suppose that the eld is increased from zero. If the anisotropy constants of austenite
are low, the eld will rst rotate the magnetization into the direction of the eld,
accompanied by ordinary magnetostriction. However, since the saturation magnetization of martensite exceeds that of austenite, it is expected that su ciently high
elds will induce a transformation from austenite to martensite. Alternatively, if the
anisotropy constants of austenite are high, a direct transformation from austenite to
martensite is expected, not preceeded by magnetization rotation. An estimate of the
eld required to transform the sample at a given temperature is given by a Clausius
Clapeyron equation derived from a suitable expression for the free energy ( 4).
3. Doma in structures
The qualitative remarks above suggest that it will be possible to have a eldinduced rearrangement of martensite variants at low temperatures, caused by the
passage from one state to another state of lower free energy. Whether this actually
will happen depends on wall mobility and crucially on the existence of a low energy
transformation path between the two states, that is, a mechanism. In this section we
show that, even with arbitrarily large deformations and the typical crystallographic
changes exhibited by shape memory alloys, together with the easy axes expected
from ferromagnetism, low-energy transformation paths exist between pairs of single
variant states. This is a generalization to general crystallographies of a result of
James and Kinderlehrer (1993, 6.3) in the cubic-to-trigonal case.
We rst describe the strains observed in martensitic materials below the transformation temperature. We use here a nite strain formulation, so as to avoid errors
associated with linearized strains. (As shown by Bhattacharya (1993), the use of
linearized strains can lead to substantial errors in calculations of microstructure,
when dealing with shape memory materials having larger strains, such as NiTi. ) In
this format, the elementary distortions associated with n variants of martensite,
measured relative to undistorted austenite, are given by matrices

, , . . . , U n,

U1 U2

( 2)

that is each U i - I is the ( nite) Bain strain or eigenstrain associated with variant i.
These matrices can always be chosen to be symmetric and positive-de nite. The
number n of such matrices is given by the order of the austenite point group divided
by the order of the martensite point group, and the precise forms of the matrices for
various symmetry changes have been given by Ball and James (1992) (see also Pitteri
and Zanzotto (1996a,b ) for more details about the structure of these distortions ). By
de nition, each matrix U i maps lattice vectors of austenite to lattice vectors for the
ith variant of martensite.

1279

Magnetostriction of martensite
T

Given a pair of variants i and j, it may be true that U i = RU j R , for some 180
rotation matrix R in the point group of the austenite. In fact, in an overwhelming
number of cases, among all possible symmetry changes, this is true. There are rare
cases in low-symmetry systems, for example some cubic-to-monoclinic transformations (but not that represented by NiTi) and some tetragonal-to-monoclinic transformations (represented, for example by LaNbO4 ), which have pairs of variants not
related in his way. Compatible pairs of variants that do not satisfy this condition
form martensitic domains instead of twins (Li and James 1996). We omit such cases
and assume this condition.
If U j = RU i RT , for some 180 rotation matrix R in the point group of the
austenite, it is known that the two variants i and j are compatible and exhibit a
type I twin and a reciprocal type II twin (these can both degenerate to compound
twins in special cases). That is, there is a rotation matrix Q and vectors a and n such
that

QU j

Ui = a

( 3)

n.

Here, a n is the 3 3 matrix with components ( a n) km = ak nm. In fact, given U i


I
I
I
II
II
II
and U j , there are precisely two solutions ( Q , a , n ) and ( Q , a , n ) of equation (3)
corresponding to the type I and type II twins respectively. Formulae for the solutions
are given by

n = e,
I

nII = 2 e -

(|
,
|)

a =2
2

Ui e

Ui e

U -i e
1

|2

U -i 1 e

),

Ui e

( 4)

aII = U i e.

Here e is a unit vector on the axis of R: Re = e. For future reference note that, from
equation (4), RnI = nI and RnII = - nII . Using equation (4), the corresponding
I
II
rotation matrices Q and Q can be obtained from equation (3).
The condition (3) is just the Hadamard compatibility condition for a deformation y( x) having alternating gradients y( x) = U i /QU j /U i /QU j /U i /Q U j on layers
with normal n. For a picture of a deformation of this type see gure 3; this gure
was obtained by integrating y( x) = U i /QU j /Ui /QU j /U i /Q U j to get y( x) and
then plotting the deformed positions y( x1 ) , y( x2 ) , y( x3 ) , . . . of an array of dots,
which in the reference con guration occupied positions x1 , x2 , x3 , . . . lying on a
square grid. The particular variants in this case were chosen to be variants 1 and
2 of the cubic-to-tetragonal case, in which U 1 = diag ( 1 + e 2 , 1 + e 1 , 1 + e 1 ) ,
U 2 = diag ( 1 + e 1 , 1 + e 2 , 1 + e 1 ) and U 3 = diag ( 1 + e 1 , 1 + e 1 , 1 + e 2 ) . These two
T
variants satisfy U 2 = RU 1 R for a suitable 180 rotation matrix R and have (compound twinned) solutions given by equation (4).
Now we turn to the magnetic part. For this we must focus on the deformed
con guration (e.g. the pictures of the dots in gures 3 (b) and (c)). The variants
represent low-symmetry phases, and in the majority of cases in ferromagnetism
the directions of easy axes are along low-index directions. The low-index directions
in the martensite variants i and j are the eigenvectors (corresponding to the same
eigenvalues) of the matrices U i and U j . More precisely, it follows from Ball and
James (1992, equation (2.42)) that each element of the point group of variant i
maps the eigenspace of U i corresponding to a xed eigenvalue to itself.

1280

R. D. James and M. Wuttig

Figure 3. Illustration of the result that mechanical compatibility implies magnetic compatibility at a large strain (a) reference con guration; (b) solution I of equations (3)
and (4); (c) solution II of equations (30) and (4). Solutions I and II are for two
tetragonal variants of martensite.

Let us assume that this is true, that is that the preferred magnetizations 6 m1 ,
6 m2 , . . . , 6 mn corresponding to variants U 1 , U 2 , . . . , U n respectively satisfy
U i mi = mi , i = 1, 2, . . . , n. Consistent with this, we assume crystallographic equivaT
lence of mi and mj ; if U j = RU i R , then mj = Rmi . The latter assumption follows
from U i mi = mi if the eigenvalues of U i are distinct; otherwise, it serves harmlessly
to x a degeneracy. According to these assumptions, when we consider a deformation with gradient y( x) = QU i we must simultaneously consider a magnetization
6 Q mi and plot the arrow 6 Q mi on the deformed con guration of variant i. This
was done to make the pictures of magnetic domain structures shown in gure 3. Of
course, the exibility implied by 6 allows one to subdivide a martensitic domain into
180 magnetic domains.
Now we consider energetics. The deformation de ned by y( x) =
U i /QU j /U i /QU j with corresponding magnetization m( y( x)) = 6 mi / 6 Q mj /6
mi / 6 Q mj provides a continuous path between the single variant i and the single
variant j, obtained by continuously changing the volume fraction, as is clear
from gure 3. A reasonable assumption consistent with general principles of micromagnetics (James and Kinderlehrer 1993) is that the magnetoelastic energy density is
minimized exactly at the orbits ( RU i , Rmi ) , i = 1, . . . , n, where R is any rotation
matrix. If we omit 180 walls within a variant, the total exchange energy is proportional to the total area of interfaces between variants. The stray eld energy arises
from two sources: rstly from the internal divergence of magnetization,
6 Q mj - ( 6 ) mi n , where n is the deformed normal of the interfaces, and

secondly the boundary divergence, 6 Q mj n or 6 mi n , where n is the normal to

the deformed boundary of the specimen. The boundary divergence can be reduced by
the choice of appropriate geometries, or by using various yokes of soft magnetic

1281

Magnetostriction of martensite

material. However, the internal divergence is clearly material dependent, that is


dependent on the distortions U i and U j (through the normal and the solution for
Q via equation (4)) and the corresponding easy magnetizations. Thus it is of fundamental interest to know whether, with the assumptions above (which should be
regarded as supremely typical from both the martensitic and magnetic viewpoints),
the internal divergence is zero, that is, whether, with Q and n obtained as solutions
of equations (3) and (4),

[6

Q mj

U -i n
1

mi n = 0,

n =

|U-i 1 n|

( 5)

Here, we have, without loss of generality, omitted one 6 , and we have used the
standard formula from continuum mechanics for the deformed normal n . If one
looks carefully at gure 3, one sees that indeed equation (5) holds for a certain choice
of ( 6 ). We show below that, under our assumptions, this is always true.
Before doing so, we note what would be the consequence of not satisfying
equation (5). There would then be poles on every interface in the laminate, which
could produce substantial stray eld energy, depending on the actual value of the
left-hand side of equation (5). Assuming also typical values of the exchange constant,
it would be unfavourable for the material to have such a large magnetostatic energy,
and therefore each layer would break into the 180 domains whose density would be
determined by a competition between wall and magnetostatic energies. The energy
then would be the sum of the remaining stray eld energy and the total energy of the
180 and the original interfaces. These are not exactly 90 walls, as we are allowing
nite deformations: cf. gure 3 and equation (4). These statements could be made
quantitative by micromagnetic calculations along classical lines.
To show that equation (5) holds under the stated assumptions, we rst use the
integral form of equation (5):

y( X )

m( z) h ( z) dz = 0.

( 6)

Here y( X ) is the deformed con guration. The condition (5) is equivalent to equation
(6) holding for all smooth test functions h that are supported on a subregion of
~ ( x) =
y( X ) . Now change the variables in equation (6), z = y( x) , and de ne m
6 mi / 6 mj / 6 mi / 6 mj on the corresponding layers in X with normal n. We get

y( X )

m( z) h ( z) dz = ( det U i )

= ( det U i )

m( y( x)) h ( y( x)) dx,


X

y( x) - 1m( y( x)) x( h ( y( x))) dx,

= ( det U i ) - 1

z ( x) dx.

~ ( x)
m
X

( 7)

Here, we have used the speci c form of y and the assumptions on the easy axes, and
~ we see that a
we put z ( x) = h ( y( x)) . From equation (7) and the de nition of m
condition equivalent to equation (5) is the simpler condition:

[6

mj - mi n = 0.

( 8)

1282

R. D. James and M. Wuttig

Now, using that mj = Rmi , we have the equivalent condition 6 Rmi - mi n = 0,


but now recall (cf equation (4)) that there is a relation between R and n implied by
mechanical compatibility: that is, Rn = n (for the type I twin) and Rn = - n (for the
type II twin). Thus, equation (8) is equivalent to

[6

]
[6 ( - m ) - m ]n = 0

mi - mi n = 0 ( f or a type I twin) ,
i

( f or a type II twin) .

( 9)

Hence, by making an appropriate choice of 6 in each case, equation (9) is satis ed


(and therefore also equation (5)). The result can be stated simply (with some loss of
precision ) that, for type I or type II twins with typical easy axes, mechanical compatibility implies magnetic compatibility, even with arbitrarily large deformations.
Note that the, result is true even though type II twins generally have irrational
interfaces ( gure 3, however, shows a case of compound twins).
4. Criteria for screening alloys
From the conceptualization presented above, several features of the desired alloy
become evident. In this section we collect these features and explain how they are
related to measurable material constants. Certain relations among these material
constants promote the behaviour that we seek. For de niteness, we concentrate of
the eld-induced rearrangement of martensite variants.
A crucial feature implied by gure 2 (a) is that values of strain and magnetization
are always associated; when the magnetization changes for some reason, so does the
strain and vice versa. This kind of behaviour can be expected for martensitic and
magnetostrictive materials, because they are governed by magnetoelastic energies
that have energy wells, and the bottoms of these wells are de ned by certain special
strain magnetization pairs, as described in 3. However, to satisfy this condition the
material should have the property that the magnetoelastic energy rises steeply away
from its energy well minima. Below, we translate this statement into conditions on
moduli.
Suppose that this were not true, and reconsider gure 2 (a). Upon application of
the vertical eld, a competing possibility would be that the magnetization simply
rotates into the direction of the eld, while the shape does not change. This kind of
pure rotation of magnetization at xed strain is governed partly by magnetic anisotropy constants of the martensite. Magnetic anisotropy constants have been widely
studied for perpendicular recording media, and especially the conditions that promote very high magnetic anisotropy (Victoria and MacLaren 1993). The principal
qualitative nding of this work is that high magnetic anisotropy is favoured by high
crystalline anisotropy ( R. H. Victoria 1996, private communication), produced by
anisotropic crystal structure, anisotropic ordering or very- ne-scale multilayer geometry. This qualitative feature, at least, is consistent with our proposal, as nearly all
martensites are of low symmetry. In fact, high-symmetry austenite and low-symmetry martensite also gives many variants of martensite, which is crucial for the shape
memory e ect in random polycrystals (Bhattacharya and Kohn 1996), and also
should be bene cial in the present case.
Similar issues arise when considering elastic sti ness at constant magnetization,
which is the consideration of a di erent path away from an energy well minimum.
Referring to the right of gure 2 (a), a crystal that is too elastically soft would simply
compress under the stress, regardless of its magnetization.

1283

Magnetostriction of martensite

High anisotropy, that is steep energy wells, implies large energy barriers between
the wells. At rst, these barriers might be considered to produce such large metastability that it might be regarded as di cult to move twin boundaries. In fact, Clark s
(1980 ) well known strategy that led to the discovery of the giant magnetostrictive
material Terfenol-D involved partly the mixing of appropriate concentrations of
terbium and dysprosium so as to have a nearly ambiguous easy axis. In terms of
the energy landscape, this corresponds to replacing an energy barrier (or at least a
saddle) by an energy well. The e ect of this is to atten the energy surface on
average. On the other hand, there exist large-deformation shape memory alloys,
with rather hard moduli, whose twin boundaries move under small stress. For example, Cu-14.0 wt% Al 3.9 wt% Ni, which has relatively hard moduli, undergoes a 6%
shear strain under a critical resolved shear stress on twin boundaries of just 1.5 MPa,
for a suitably oriented specimen (Chu 1993). In this particular case, the crystal
essentially `jumps between energy wells. Orientation and texture, and metallurgical
factors such as precipitates and dislocations, may also play a crucial role. A predictive understanding on metastability and hysteresis is surely missing for the materials considered here, but particular examples show that it is possible to have the
coexistence of high anisotropy, large strains and mobile twin boundaries.
To formalize these criteria, let us consider micromagnetic theory. To be able to
state criteria in terms of conventional material constants, we choose a conventional
expression for the energy, using geometrically linear strains. Here there are two
caveats: rstly with the larger martensitic strains (e.g. 5 10% range) this can lead
to signi cant errors, and it does not immediately relate to the nite-strain considerations of 3, and secondly, as pointed out by Brown (1966 ), the conventional smallstrain formulation is missing certain terms that should be present according to a
straightforward linearization starting from geometrically nonlinear theory. With
these reservations, we write the standard micromagnetic energy
F=

{u
1
8p

] [

( m) + 12 E - E 0 ( m) C E - E 0 ( m)

] - h m - r E} dx
0

| z m|2 dx.

( 10)

Here, we have omitted exchange energy and strain gradient energies, as we are
dealing with large bodies (DeSimone 1993). u is the magnetic anisotropy energy.
The applied eld is h0 and the applied stress is r 0 . The strain tensor E is given by the
T
usual formula in terms of the displacement gradient: E = 12 ( u + ( u) ) . The potential z m in the magnetostatic energy is the unique solution (among potentials with
nite magnetostatic energy) of the magnetostatic equation
div ( -

+ 4p m) = 0,

( 11)

which itself arises from the conditions curl h = 0( h = - z m ) , div b = 0,


b = h + 4p m. In equation (10) the form of u depends on symmetry, the elastic
modulus tensor C is positive de nite, and E 0 ( m) is the preferred strain tensor
corresponding to the magnetization m; in the cubic case the most general quadratic
form is
E 0 ( m) =

3
2

^
100 ( m

^
m
- 13 I) + ( 111 - 100 )

^ m
^
m
i j ( ei ej )
i =/ j

),

( 12)

1284

R. D. James and M. Wuttig

^ = m m , e e e is an orthonormal basis parallel to the cubic axes, and


where m
/ s { 1, 2, 3}
^
^
mi = m ei are the direction cosines of the magnetization.
We assume that u has the usual cubic form

^ 2m
^2
^2^2
^2^2
^2^2^2
( m) = 1 ( m
1 2 + m 2 m 3 + m 1 m 3 ) + 2 m 1 m 2 m 3 .
u

( 13)

The energy wells are associated with the minimizers of the total anisotropy energy.
These minimizers are the pairs ( E 1 , 6 m1 ) , ( E 2 , 6 m2 ) , . . . , ( E n, 6 mn) , where
6 m1 , 6 m2 , . . . , 6 mn are minimizers of u and E i = E 0 ( mi ) . The numbers and
types of variants are determined by the particular values of 1 and 2 (Bozorth
1978, table 4, p. 582). It follows from this form of the micromagnetic energy that
pairs of energy wells, that is pairs of variants of the martensite, are related by
T
E j = QE i Q , mj = Q mi , for some Q belonging to the point group of austenite (of
the Fe Pd alloys studied in this work the point group m3m) but not belonging to the
point group of the ith variant of martensite. These conditions are the translation to
geometrically linear theory of the conditions relating variants given in 3.
As discussed above, a meaningful criterion for screening alloys is that the magnetoelastic energy density grows rapidly as we depart from an energy well in any
direction. It is essential to formulate criteria in terms of dimensionless parameters.
To non-dimensionalize the magnetostatic equation (11), we divide by ms . Thus
z m /ms is dimensionless;2 so we make the energy (10) dimensionless by dividing the
whole expression by ms . The departure from a well (caused by an applied stress or
eld) should be compared with something having the same dimensions. Following
the qualitative argument given in 2 and based on gure 2 (a) the departure from an
energy well should be small compared with a typical distance between the wells.
To calculate this departure we rst rewrite the non-dimensionalized micromagnetic energy (10) in an alternative form by completing the square on the elastic term.
This gives
1
1
2F =
ms
m 2s

(| |
| ) ]-

( m) + 12 E - |E 0 |

- 1r

- |E 0 | |E0| + C |E 0

E 0

1
2r 0

C
r

E0
E0

1
8p m2s

|E 0|

)]

dx +

C - 1r

h0 m

R3

| z m|2 dx.

( 14)

Consider the term containing square brackets in the integrand and suppose that m is
assumed close to one of the values 6 m1 , 6 m2 , . . . , 6 mn. Then the strain E will be
1
close to the corresponding value in the list E 1 , E 2 , . . . , E n if the quantity C - r 0 /|E 0 |
is small compared with the unit tensor E 0 /|E 0 |. This leads to the criterion
max |r 0 |
!
( min C )

where min C

1,

is the minimum elastic modulus and

( 15)

= E 0 ( m) =

()
3
2

1 /2

|11002 |

(
2 100 +

|111 |

1 2
32111 ) /

[ ]
[ ]
[ ]

mi 100 ,
mi 110 ,
mi 111 .

( 16)

1285

Magnetostriction of martensite

Now suppose that (15) holds and consider a departure of m from one of the values
6 m1 , 6 m2 , . . . , 6 mn. This departure is found by calculating the minimizer of the
quantity u ( m) - h0 m - r 0 E 0 ( m) , which gives rise to lengthy expressions that
simplify upon linearization. We nd that (to rst order in h0 and r 0 ) the minimizers
of u - h0 m - r 0 E 0 remain close to the values 6 m1 , 6 m2 , . . . , 6 mn if the following conditions hold:

|msh|
min ( q 1 , q

2)

1,

|100 s ii | !
q

1( no sum) ,

|111 s ij | !
q

1 ( i =/ j ) ,

( 17)

where

(q

( |1 |, |1 |)
( |1 |, |21 + 2 |)
1, q 2) =
( |31 + 2 |, |31 + 2 |)

[100] easy axes,

[110] easy axes,


[111] easy axes.

( 18)

A calculation based on equations (15) and (17) shows that these conditions are
su cient to ensure that the magnetoelastic energy density is negligibly perturbed
by the eld and stress.
There is a more general possibility consistent with the behaviour shown in gure
2 (a). That is, the anisotropy energy could have certain low energy valleys (in ( E , m)
space) leading away from its minima, but with these valleys isolated from each other
by large barriers so as to allow the process shown in gure 2 (a) to occur. As an
example, suppose that the criterion (15) holds, but not criterion (17). Suppose also
that E 0 ( m) exhibits high anisotropy. Then, equation (15) would imply that any state
( E ( x) , m( x)) having low energy would tend to satisfy E ( x) = E 0 ( m( x)) , at least on
most of X . Hence, even though m( x) might be quite free to range over the constraint
set |m( x) | = ms with little energetic penalty, there would be a strong tendency for
strain and magnetization to be coupled, leading to the possibility of a large eldinduced strain. This possibility is not essentially di erent from the one above; it is
just that the energy `wells form an extended set. This idea seems to play a role in the
Fe Pd system (see 6).
Finally, we note that equations (15) and (17) have been derived under the
hypothesis that every stress and eld satisfying the restrictions only slightly perturbs
the energy-well minima. If one goes back and puts in a special eld and stress, these
conditions can be relaxed considerably. Thus, if the magnetoelastic energy density
rises steeply in certain directions but not others, it may be possible to design special
applied elds and stresses that take advantage of the particular feature, again leading
to the behaviour shown in gure 2 (a).
5. Evolution of variants u nder combined stress and field:
micromagnetics of martensite
The criteria (15) and (12) lead naturally to a `constrained theory (DeSimone and
James 1996) which has proved to be extremely useful for predicting the behaviour of
emerging ferromagnetic shape memory alloys. Here, we brie y describe this theory
and then apply it to the behaviour of Fex Pd1- x ( x < 0. 7) .
If equations (15) and (17) hold, it is natural to expect that, except for small
regions of the specimen, the strain magnetization pair ( E ( x) , m( x)) will lie close

1286

R. D. James and M. Wuttig

to the energy wells. This leads to the constrained theory min( E ( x) , m( x))
where
Fconstr =

- h0 m( x) r

wells

E ( x) dx + 81p | z | dx + excess.
R

Fconstr

( 19)

Here, the wells refer to the strain magnetization pairs ( E 1 , 6 m1 ) ,


( E 2 , 6 m2 ) , . . . , ( E n, 6 mn) , and Fconstr is the constrained free energy. The term
`excess is a non-negative and vanishes for a magnetization that is locally divergence
free; excess is the energy of internal poles. In reality, the constraint
`( E ( m) , m( x)) wells is interpreted in a slightly weaker sense that allows for departures from the well on small transition layers. See DeSimone and James (1996 ) for
the precise statements and a rigorous derivation of this theory starting from general
micromagnetics. The energy Fconstr is still somewhat di cult to use for general
shapes, but it can be simpli ed by assuming that X is an ellipsoid. To see this,
note that the rst two terms of equation (20) depend only on the average strain
and the average magnetization, since the applied eld h0 and applied stress are
assumed to be constant. Hence, without loss of generality, one can x the average
magnetization and minimize just the magnetostatic energy. It is well known that,
with the average magnetization xed, and with X an ellipsoid, the magnetostatic
energy alone is minimized by a constant magnetization m( x) = k ml =
( 1 /volume X ) X m( x) dx. The value of the minimum is 12 k ml D k ml where D is
the demagnetization matrix for the ellipsoid X . Finally, it is possible to show by
explicit constructions involving layered (and layers within layers, etc. ) geometries,
together with suitable divergence-free magnetic substructures, that, if the wells are
pairwise compatible, any average magnetization strain in the convex hull of the
wells is achievable. In other words, the average strain magnetization is given by
the expressions

k ml
k El

= 1 m1 + 2 ( - m1 ) + 3 m2 + + 2n ( - mn ) ,
= ( 1 + 2 ) E 1 + ( 3 + 4 ) E 2 + + ( 2n- 1 + 2n ) E n ,

( 20)

where 1 , 2 , . . . , 2n take values in the interval (0, 1) and i i = 1. The fact that
the magnetic substructure can always be constructed using a divergence-free magnetization implies that excess equals zero. Hence, the minimization problem (20)
becomes a quadratic programming problem
min

( E ( x) ,m( x) ) wells

( Fconstr ) =

k E l ,k ml

min

satisfy ( 20 )

- r k E l

( - h0 k ml

+ 12 k ml D k ml ) ,

( 21)

solvable by standard methods. The solution is a set of minimizing weights 1 ,


2 , . . . , 2n; from these the theory guarantees at least one compatible magnetoelastic
domain structure, which we have drawn in various cases below. Sometimes there is
non-uniqueness of the domain structure; in such cases we have determined a unique
structure by forcing compatibility with a given initial state (see below).
There is a super cial resemblance between the nal form (21) and the domain
rotation model of Jiles and Thoelke (1992, 1994). The starting point of the domain
rotation model is quite di erent, namely a collection of non-interacting particles, and
the resulting energy and predictions are also di erent. A discussion of the relation
between the two models has been given by DeSimone and James (1996 ).
Now we specialize the theory to Fe70 Pd30 and predict its response under various
combinations of applied eld and stress. By comparing the predictions of the theory

Magnetostriction of martensite

1287

based on several choices of easy axes with experiments described in 6, it was


inferred that the easy axes of the martensite are the pair of k 100l directions perpendicular to the c axis. This results in energy wells de ned by

( E 1 , 6 ma1 ) ,

( E 1 , 6 mb1 ) ,

( E 2 , 6 ma2 ) ,

( E 2 , 6 mb2 ) ,

( E 3 , 6 ma3 ) ,

( E 3 , 6 mb1 ) ,

( 22)

where
E1

E2

ma1 = ms ( 010) ,

mb1 = ms ( 001) ,

ma2 = ms ( 100) ,

mb2 = ms ( 001) ,

ma3 = ms ( 100) ,

mb3 = ms ( 010) .

E3 =

1
2

( 23)

Note that there are two easy axes corresponding to each easy strain. (This does
not cause problems for the theory above, although equation (12) for E 0 ( m) is too
simple to give the wells (23). ) We consider three kinds of experiment on a [100] rod.
(1) Fixed axial eld; variable transverse eld. Here we x the axial along [100]
and gradually increase from zero the transverse eld along [010]. For each
pair of elds we calculate the solution of the quadratic programming problem. The input to the calculations is the wells (22) evaluated with the data
given in table 1. All particular graphs below use the data at 40 C below Ms .
The demagnetization matrix is taken to be that appropriate to an in nitely
long rod with axis [100], for simplicity. In solving this quadratic programming problem, it is found that the solution may lie in the interior of
the constraint set (all i between 0 and 1), or on a `face (one i = 0 or 1),
or `edge (two of the i = 0 or 1), etc. Thus, at special values of the two elds
there is a change in the type of the solution. These transitions correspond to
either a jump or a change in slope of the graph of strain against eld. For the
present case the calculated transitions are potted in gure 4. The region
between the two Vs is the region where demagnetization energy competes,
neither entirely winning nor losing, with the energy of the applied eld. It is
found from the solution that, as expected, the volume fractions corresponding to 6 mb1, 6 mb2,- ma1 and - mb3 are always zero for any non-zero values of
the elds. We also nd that, even though the values of the transition elds
are uniquely determined, the remaining individual volume fractions are not.
(The essential reason for this is the plethora of easy axes available in equation (22)) and the rather symmetric problem chosen. ) In particular, there is
freedom to assign arbitrarily the ratio of the volume fractions associated

1288

R. D. James and M. Wuttig

Table 1. Input data for the calculation of strain against field modelled after Fe70 Pd30 (magnetic data from Kussmann and Jessen (1963 ) and Matsui and Adachi (1989), and
strains from the lattice parameter measurements of Oshima and Sugiyama (1982)).
ms is taken to be independent of temperature.
Temperature interval
( C) below Ms

Strain 2
on c axis

Strain 2
on a axis

Saturation magnetization mS
(emu cm- 1 )

10
25
40
100

- 0.010
- 0.015
- 0.021
- 0.029

0.0044
0.0055
0.0090
0.011

1390
1390
1390
1390

with ma1 and mb3. By checking interface normals during the corresponding
tests ( 6), we found that, upon cooling through the austenite martensite
transformation in the absence of a eld, the specimen exhibited a typical
habit plane transformation, leading to a nely twinned specimen with strains
E 1 and E 3 . It is well known from the `crystallographic theory of martensite
that the expected ratio of volume fractions of these two variants should be
either 13 or 23. We observed that the subsequent domain structure evolution
seen during the present experiments took place within the already established twinned laminate of martensite. Thus it seems reasonable to put the
a
b
ratio of volume fractions corresponding to m1 and m3 equal to 13 or 23 and it
was found that, by comparison with the measured curves, 13 provided a better
match. Thus, the curves of strain against eld shown in gure 5 were plotted
under this restriction. We emphasize that since the transitions of eld ( gure
4) are uniquely determined, the only e ect of di erent choices of this volume
fraction is to scale the strain axis shown in gure 5.

Figure 4. Transition elds corresponding to cases (1) and (2). The axial eld is h1 and the
transverse eld is h2 .

Magnetostriction of martensite

1289

Figure 5. Predicted axial strain against axial eld at a xed transverse eld of 2000 Oe and no
stress.

An evolution of domain structure corresponding to various points on the


graph of strain against eld is shown in gure 6. All strains in all directions have been multiplied by a factor of ten so that the macroscopic and
microscopic strains are clearly visible. In gure 6 (as well as in gures 8
and 11 later) the scale of the microstructure should be much smaller than
the dimensions of the specimen, in keeping with the large-body limit
assumed in the theory. Note that the considerations 3 are applicable
here.
(2) Fixed transverse eld; variable axial eld. Here, we x the transverse eld
along [010] and increase the axial eld along [100] from zero. The discussion
of case (1) applies also to this case. Strain against eld is shown in gure
7, while a corresponding evolution of domains is shown in gure 8. The
same ratio for the volume fraction of variant 1 to that of 3 as above was
used here.
(3) Axial compressive stress; axial eld. Here we x an axial compressive
stress along [100] and increase the eld along [100] from zero.
Intuitively, this should be the most e ective way to produce a large strain
at a small eld because a compressive stress prefers strain E 1 , an axial
eld prefers magnetizations m3 or m5 (which are associated with E 2 and

1290

R. D. James and M. Wuttig

Figure 6. Predicted microstructures and domain structures corresponding to the black full
circles in gure 5, with a xed transverse eld of 2000 Oe, no stress, and axial elds of
(a) 0 Oe, (b) 1000 Oe, (c) 2000 Oe and (d) 4000 Oe: small black full circles, variant 3;
large grey full circles, variant 2; large black full circles, variant 1. All strains in all
directions multiplied by a factor of ten.

E 3 respectively, both of which have the same strain in the [100] direction),

no demagnetization energy is associated with mixtures of m3 or m5 with


the present long-rod idealization, and the largest possible uniaxial strain
e ( E i - E j ) e is obtained by choosing e = 100 , i = 1 and j = 2 or 3.
Analogous to gure 4, there is a uniquely determined graph of transition
stress against applied eld ( gure 9), where a jump of strain occurs on the
graphs of strain against eld. Unlike the case above, here the graph of
strain against eld is uniquely determined ( gure 10). There is still nonuniqueness of some of the volume fractions (i.e. those associated with m3
and m5 at large elds), but this does not a ect strain against eld. An
evolution of domain structure corresponding to gure 10 is shown in
gure 11.

[ ]

Magnetostriction of martensite

1291

Figure 7. Predicted axial strain against transverse eld at a xed axial eld of 2000 Oe and no
stress.
6. Experiments on Fe Pd alloys
In this section we collect our preliminary experimental ndings on Fe70 Pd30 . A
complete magnetomechanical characterization of this alloy, together with a detailed
description of the experimental set-up, will be given in the forthcoming paper by
James et al. (1997), The purpose of this section is to con rm the expected magnetomechanical response in a promising martensite.
Fe Pd alloys near the composition Fe70 Pd30 undergo an almost second-order
f cc f ct transition with c /a < 1 (Sugiyama et al. 1984) and display transitional
states within about 5 of the transformation temperature (Seto et al. 1990a, b,c),
which in the present alloy was measured to be about - 10 C. The lattice parameters
exhibit a rapid variation with temperature in the fct phase (Oshima and Sugiyama
1982), leading to relatively large eigenstrains several tens of degrees Celsius below
transformation. The ordering of transition temperatures is as in gure 1 (a), as both
austenite and martensite are ferromagnetic. In the temperature range of interest the
saturation magnetization is only weakly dependent on temperature but has large
values near 1400 emu cm- 3 (Matsui and Adachi 1989). Rather large values of magnetocrystalline anisotropy constant 1 of the order of 107 - 108 erg cm- 3 have been
reported (Klemmer et al. 1995). Hence, from what is known, this system satis es the
conditions that favour the e ect studied in this paper.

1292

R. D. James and M. Wuttig

Figure 8. Predicted microstructures and domain structures corresponding to the black full
circles in gure 7, with a xed axial eld of 2000 Oe, no stress and transverse elds of
(a) 2000 Oe, (b) 5000 Oe, (c) 8000 Oe and (d) 10 000 Oe: small black full circles, variant
3; large grey full circles, variant 2; large black full circles, variant 1. All strains in all
directions multiplied by a factor of ten.

The specimen was a single crystal of approximate dimensions 1. 5


1. 5 mm 6. 6 mm oriented with [100] pointing parallel to the long axial direction.
The sample was cut from an available crystal (Li et al. 1993, Saxena and Barsch
1993, Chopra and Wuttig 1995). It was mounted so that three di erent elds could
be applied to it at various temperatures: an axial eld parallel to [100], a transverse
eld parallel to [0, 0.95, 0.31] and a compressive stress parallel to [100]. Except for
the transverse magnetic eld being tilted away from [010] and the specimen not being
in nitely long (recall the choice of the demagnetization matrix in 5), the experimental conditions correspond to the situations analysed in 5. The strains were

Magnetostriction of martensite

1293

Figure 9. Transition stresses corresponding to case (3) with no transverse eld.

Figure 10. Predicted strain against axial eld with a stress of

- 5 MPa and no transverse

eld.

Figure 11. Predicted microstructures and domain structures corresponding to the black full
circles in gure 10 with no transverse eld: small black full circles, variant 3; large grey
full circles, variant 2; large black full circles, variant 1. All strains in all directions
multiplied by a factor of ten.

1294

R. D. James and M. Wuttig

measured with an axial strain gauge. The xed elds were applied by a set of permanent magnets. The microstructural evolution of the martensite variants, upon
application of a magnetic eld or stress, was observed from surface relief on the
specimen by an optical microscope with di erential interference contrast. Three
experiments are reported here which correspond, respectively, to the predictions in 5.
(1) Fixed transverse magnetic eld; variable axial magnetic eld; no stress
( gure 12). Here, a transverse eld of 2300 Oe was applied during cooling
and during the experiment performed at - 36 C and the axial eld was cycled
about zero, starting from zero. The dependence of the strain on the axial

Figure 12. Strain against axial eld in Fe70 Pd30 at - 36 C. The transverse eld was xed at
2300 Oe and the axial eld was cycled as shown. No stress was applied.

Magnetostriction of martensite

1295

eld is shown in gure 12. Upon cooling through the austenite martensite
transformation, axial bands of martensite appeared (cf. 5, case (1)). As the
axial eld was increased, there was large-scale movement of these bands and
some new bands appeared, all having interface normals consistent with those
shown in gure 6. There was clear evidence of large-scale eld-induced
movement of martensitic twins. The left-hand side of the curve in gure
12 conceptually agrees with the corresponding predicted dependence ( gure
5). Starting from the saturated state created during cooling in the transverse
eld of 2300 Oe, the strain decreases approximately linearly as the axial eld
drops below this value. However, the strain does not increase in a symmetrical fashion beyond zero as expected for a perfectly oriented [100][010]
sample. Instead, it increases asymmetrically towards a di erent asymptotic
strain possibly re ecting magnetization rotation.
(2) Fixed axial magnetic eld, variable transverse magnetic eld, no stress
( gure 13). Here, an axial eld of 2300 Oe was applied during cooling to

Figure 13. Strain against transverse eld in Fe70 Pd30 at - 17 C. The axial eld was xed at
2300 Oe and the transverse eld was cycled as shown. No stress was applied.

1296

R. D. James and M. Wuttig

- 17 C and the transverse

eld, applied parallel to [0, 0.95, 0.31] was cycled


about zero, starting from zero. The observed dependence of the strain
against transverse eld is shown in gure 13. Again, there was clear evidence
of magnetic eld-induced motion of martensitic twins as shown in gure 14.
Note particularly the change in the volume fraction. The observed curve
shows an approximate linear dependence of the strain on the magnitude
of the axial magnetic eld in conceptual agreement with the corresponding
predicted curve ( gure 7). However, the observed curve is not symmetrical
with respect to the axial magnetic eld; it is shifted to the right. There are
two reasons for this shift. They are, rst, the lack of symmetry of the investigated sample and, second, the asymmetry of the eld-cooled initial state.
Both asymmetries apparently lead to magnetization rotation as indicated by
the nonlinearity at large positive axial elds. Figure 14 shows the observed
microstructure at two elds corresponding to the graph in gure 13. The
temperature value is nominal. Note the clear evidence of the movement of
martensitic domains, especially the change of volume fractions. If we
account for the relation of the normal (0, - 0. 31, 0.95) of the observed
face relative to (001) , the orientations of interfaces shown in gure 14
agree well with those in gure 8. In particular, a comparison of the two
microstructures reveals a decrease in the population of variants whose
traces are approximately perpendicular to the [100] sample axis as the
magnetic eld is increased from a large negative value, as expected
from gure 8.
(3) Fixed compressive stress; variable axial magnetic eld; no transverse magnetic
eld ( gure 15). Here, a constant compressive stress of - 5 MPa parallel to
[100] was applied with a soft loading device and held xed during cooling to
- 16 C. Subsequently, an axial magnetic eld was applied, also parallel to
[100], and was cycled about zero, starting from zero. As predicted by theory
(see gure 10), the evolving magnetic eld induced strain is symmetrical in
the axial magnetic eld and saturates. However, the predicted sudden variation in the strain at a critical eld is smoothed out and the values of the jump
in strain are signi cantly smaller than those predicted. The former can
be attributed to the variations in the internal magnetic eld due to
the inhomogeneous demagnetization eld. The latter may re ect the
same inhomogeneity, giving rise to local rotation propagating through
the sample.
In summary, there is conceptual agreement between theory and experiment in
all three cases. The eld-generated strain is symmetrical with the respect to the
applied varying eld when the sample eld con guration is symmetrical and it is
asymmetrical otherwise. In addition, the eld-generated strain is linear over signi cant ranges of the generating eld as predicted. The major limitation of the
present experiments has to do with the lack of symmetry of the sample chosen
because of its ready availability. This lack of symmetry will lead to a signi cant
torque on the spontaneous magnetization which has not been accounted for in
the theory. The torque will lead to magnetization rotation which will compete
with the elastic twinning and lead to non-linearities in the eld-generated strain.
Future studies of approximately cut single-crystal ellipsoids will eliminate rotation.

Magnetostriction of martensite

1297

(a)

(b)

Figure 14. Twin structure of the Fe70 Pd30 sample at two transverse elds as indicated by the
arrows in gure 13: (a) - 6500 Oe; (b) 4800 Oe. The temperature is nominal; the eld of
view is 1.9 mm 1. 7 mm. The upper and lower borders of the sample are parallel to its
[100] axis.

1298

R. D. James and M. Wuttig

Figure 15. Strain against axial eld in Fe Pd at - 16 C. A stress of 5 MPa was applied and the
axial eld was cycled as shown. No transverse eld was applied.

ACKNOWLEDGEMENTS
The authors gratefully acknowledge the experimental assistance of T. Shield, P.
Schumacher, R. Tickle and Y. Zhang. This study was supported by O ce Naval
Research Advanced Research Projects Agency (for M.W. grants N00014-95-11071
and N00014-93-10506; for R.D.J., grants N00014-95-1-1145 and N00014-91-J-4034).
It also bene ted from the support of the Air Force O ce Scienti c Research (for
R.D.J., grant 49620-96-1-0057), the Army Research O ce (for M.W., grant
DAAL03-92-G-0121) and the Natural Science Foundation (for M.W., grant
DMR-93-21185; for R.D.J. grant DMS-9505077).

Magnetostriction of martensite
References

1299

Ball , J. M., and James, R. D., 1992, Phil. Trans. R. Soc. A, 338 , 389.
Bhattacharya , K., 1993, Contrib. Mech. Thermodyn., 5 , 205.
Bhattacharya , K., and Kohn, R. V., 1996, Acta mater., 44 , 529.
Boz orth, R. M., 1978, Ferromagnetism (Piscataway, New Jersey: IEEE).
Brown, W. F., 1966, Magnetoelastic Interactions, Springer Tracts in Natural Philosophy, Vol.
9, edited by C. Truesdell (New York: Springer ).
Chernenko, V. A., and V. V. Kokorin, 1992, Proceedings of the International Conference on
Martensitic Transformations, edited by C. M. Wayman and J. Perkins (Monterey, CA:
Institute for Advanced Studies), p, 1205.
Chopra, H. D., and Wuttig, M., 1995, J. Phys., Paris, C8 157.
Chu , C., 1993. Thesis, University of Minnesota.
Clark , A. E., 1980, Ferromagnetic Materials, Vol. 1, edited by E. P. Wohlfarth (Amsterdam:
North-Holland), p. 531.
David, S., 1991, Thesis, Laboratoire de Magnetisme Louis Ne el, Grenoble.
DeSimone, A., 1993, Arch. Ration. mech. Anal., 125, 99.
DeSimone, A., and James, R. D., 1996, preprint.
Du Tremolet De Lacheisserie, E., 1993, Magnetostriction Theory and Applications of
Magnetoelasticity (Boca Raton, Florida: CRC Press).
James, R. D., and Kinderlehrer , D., 1993, Phil. Mag. B, 68 , 237.
James, R. D., Schumacher, P., Shield, T. W., Tickle, R., Wuttig , M., and Zheng , Y.,
1997 (to be published ).
James, R. D., and Wuttig, M., 1996, Mathematics and Control in Smart Structures, Proceedings of SPIE, Vol. 2715, edited by V. V. Varadan and J. Chandra (Bellingham,
Washington: SPIE), p. 420.
Jiles, D. C., and Thoelke, J. B., 1992, J. Magn. magn. Mater., 104 107 , 1453; 1994, ibid.,
134 , 143.
Kakeshita , T., Saburi, T., and Shimizu , K., 1995, J. Phys., Paris, 5 , C8 367.
Klemmerr , T., Hoydick, D., Okumura , H., Zhang , B., and Soffa , W. A., 1995, Scripta
metall. Mater., 33, 1793.
Kussman, A., and Jessen, K., 1963, Z. Metallk., 54, 504.
Landau , L. D., 1965, Collected Papers of L . D. L andau edited by D. TerHaar (New York,
Gordon and Breach (Oxford: Pergamon).
Li, J., Chopra, H. D., and Wuttig, M., 1993, Proceedings of the Conference on Shape
Memory Alloys, Trans. Mater. Res. Soc. Japan, 18B , 821.
Li, J., and James, R. D., 1997, Acta mater (submitted).
Matsui, M., and A dachi, K., 1989, Physica B, 161 , 53.
Oshima, R., and Sugiyama , M., 1982. J. Phys., Paris, 42 , C4 383.
Pitteri, M., and Zanzotto, G., 1996a, preprint; 1996b, preprint.
Saxena, A., and Barsch, G. R., 1993, Physica D, 66 , 195.
Seto, H., Noda , Y., and Y amada , Y., 1990a, Mater. Sci. Forum, 56 , 77; 1990b, J. phys. Soc.
Japan, 57 , 3668; 1990c, ibid., 59 .
Sugiyama, M., Oshima, O., and Fujuta , E., 1984, Trans. Japan Inst. Metals, 25 , 585.
Victoria , R. H., and Mac Laren, J. M., 1993, Phys. Rev. B, 47 , 11 583.
Ullakko, K., Huang, J. K., Jantner, C., O Handley, R. C., and Kokorin, v . V., 1996,
Appl. Phys. L ett., 69 , 1967.

You might also like