VOLUME 79, NUMBER 20 PHYSICAL REVIEW LETTERS 17 NOVEMBER 1997

Premartensitic Transition Driven by Magnetoelastic Interaction

in bcc Ferromagnetic Ni2 MnGa
Antoni Planes, Eduard Obradó, Alfons Gonzàlez-Comas, and Lluı́s Mañosa
Departament d’Estructura i Constituents de la Matèria, Facultat de Fı́sica, Universitat de Barcelona,
Diagonal 647, 08028 Barcelona, Catalonia, Spain
(Received 12 June 1997)
We show that the magnetoelastic coupling between the magnetization and the amplitude of a short
wavelength phonon enables the existence of a first order premartensitic transition from a bcc to a
micromodulated phase in Ni2 MnGa. Such a magnetoelastic coupling has been experimentally evidenced
by ac susceptibility and ultrasonic measurements under an applied magnetic field. A latent heat around
9 Jymol has been measured using a highly sensitive calorimeter. This value is in very good agreement
with the value predicted by a proposed model. [S0031-9007(97)04508-0]

PACS numbers: 64.70.Kb, 81.30.Kf

Martensitic transitions (MT) are first order displacive samples subjected to external stresses (which can result in
structural phase transitions accompanied by a significant a modification of the characteristics of the transition).
strain of the unit cell. In general, homogeneous and in- In this Letter we present a phenomenological model
tracell (short-wavelength phonons) strains are necessary in for the intermediate transition based on a Landau ex-
order to describe the path followed by the atoms at the pansion, which includes a magnetoelastic coupling.
transformation. An interesting feature displayed by the The primary order parameter is the amplitude h of a
systems undergoing this kind of first order transitions is TA2 f110g phonon, and secondary order parameters are
the existence of precursor effects [1,2]. They reflect that, ´, a s110d f110g homogeneous shear suitable to describe
in a sense, the system prepares for the phase transition be- a cubic to tetragonal change of symmetry, and M, the
fore it actually takes place. For instance, in bcc materials magnetization (considered here to be a scalar). In terms
the TA2 f110g phonon branch has low energy, the corre- of these three order parameters we assume the free energy
sponding elastic constant (C 0 ) is very low and softening of function to have the following general form:
all these phonons and C 0 occurs on cooling [2]. F sh, ´, Md ­ Fstr sh, ´d 1 Fmag sMd 1 Fme sh, ´, Md ,
The Ni2 MnGa Heusler alloy is investigated in the (1)
present Letter. At high temperature it is ferromagnetic
(the Curie temperature is Tc ­ 381 K), displays a bcc which includes a purely structural term Fstr , a magnetic
structure with an L21 atomic order (space group Fm3m), term Fmag , and a mixed term Fme accounting for the
and transforms martensitically at TM ­ 175 K. This al- magnetoelastic interplay. Considering the symmetries of
loy is unique in the sense that (i) it is the only known the system, the following expansions are proposed for the
bcc ferromagnetic material undergoing a MT and (ii) the three contributions:
MT is preceded by a structural phase transition (inter- 1 p 2 2 1
Fstr sh, ´d ­ m v h 1 bh 4
mediate transition) to a micromodulated phase (the cu- 2 4
bic symmetry is preserved) resulting from the freezing 1 1
of a q ­ 0.33TA2 phonon which becomes the intracell 1 gh 6 1 c´2 ,
6 2
strain characterizing the new phase. Such a phase tran-
sition has been evidenced by neutron scattering [3], x-ray 1 1
Fmag sMd ­ 2 AM 2 1 BM 4
[4], electron microscopy [5], and ultrasonic measurements 2 4
[6,7]. We have recently suggested that this transition to 1 A2
the premonitory (intermediate) phase is a consequence of . AsM 2 M0 d2 2 ,
4 B
the magnetoelastic interplay between the phonon and the
magnetization [7]. At the MT the system transforms to 1 1
Fme sh, ´, Md ­ k1 M 2 h 2 1 k2 M 2 ´2 .
a modulated structure with tetragonal symmetry (homo- 2 2
geneous strain) [8]. The modulation of the martensitic All the coefficients in the above expansions are positive
structure is different from that of the premartensitic phase. and only v 2 is supposed to be temperature dependent
It has been argued [7] that the intermediate transition has [11]. Actually v is identified as the frequency of the
to be first order because there is no complete softening of anomalous phonon which condensates at the intermediate
the frequency of the soft phonon; nevertheless, attempts in transition. Experimentally, the square of this frequency
measuring a latent heat have not been successful [9] and has been shown to exhibit a marked linear decrease on
a small thermal hysteresis has been detected [10] only in approaching the intermediate transition [3]; hence we

3926 0031-9007y97y79(20)y3926(4)$10.00 © 1997 The American Physical Society

VOLUME 79, NUMBER 20 PHYSICAL REVIEW LETTERS 17 NOVEMBER 1997

assume that mp v 2 ­ asT 2 Tu d. We have not included rendering a transition latent heat: L ­ TI DS. It has
any direct coupling between h and ´, which is supposed been reported that no latent heat has been detected
to be negligible in comparison with the magnetoelastic using differential scanning calorimetric techniques [9].
coupling. Concerning the purely magnetic contribution Nevertheless, a heat capacity anomaly (jump) has recently
Fmag , considering that the intermediate phase appears well been observed at the premartensitic transition [7]. Such a
below the Curie point, the changes in the magnetization jump can be calculated from our Landau model as
are expected to be small; therefore, it is reasonable to ∑ ∏ µ 2∂
DC ≠DS 1 ≠h a2
linearize Fmag around the value M0 (the equilibrium ­ ­2 a ­2 .
TI ≠T T ­TI 2 ≠T T­TI e
b
magnetization close to the intermediate transition).
Minimization of F sh, ´, Md with respect to ´ and M (4)
leads to an effective free energy function Feff along From (3) and (4) and using the expression of TI , a
a given transformation path (´ ­ 0 and M ­ M0 yf1 1 latent heat L ­ 2sT0 2 TI dDC is obtained. T0 , TI , and
k1 2
2A h g) in termsk
of the phonon amplitude h. Expanding DC have been measured experimentally, thus enabling
the term f1 1 2A1 h 2 g21 in power series and keeping only an evaluation of the latent heat. Taking T0 ­ 225 K,
the terms up to sixth order, we obtain TI ­ 230 K, and DC . 0.7 Jymol K, the values L .
1 p 2 2 1 e 4 1 27 Jymol and DS . 20.03 Jymol K are obtained. These
Feff ­ e h 1
m v bh 1 e h6,
g (2) values are quite small (for comparison note that at the
2 4 6
MT the latent heat L is around 2100 Jymol and DS .
k M2
where mp v e 2 ­ mp v 2 1 k1 M02 ­ asT 2 fTu 2 1a 0 gd ­ 20.5 Jymol K).
2 2
e ­ b 2 k1 M0 , and g 3 k3 M 2 In order to corroborate the validity of the values pre-
asT 2 T0 d, b A e ­ g 1 4 1A2 0 .
k2 M 2 dicted by the model, calorimetric measurements have
The interesting point to be stressed is, provided that 1A 0 been carried out using a highly sensitive scanning mi-
is large enough, b e can become negative, and in this case a crocalorimeter [12]. This special calorimeter enabled the
first order transition can take place before the system be- use of large samples (this is not usual in standard differ-
k M2
comes linearly unstable at T0 s­ Tu 2 1a 0 d. It is also ential scanning calorimeters which are designed to operate
worth noting that, in case there was not magnetoelastic with very small samples of a few milligrams). A single
coupling, only a continuous transition at T ­ Tu would crystal of Ni2 MnGa of 2.9 g of mass and a Cu reference
be possible. In general, a first order transition is predicted with the same mass have been used. An example of the
by the Landau theory when the system symmetries enable measured thermogram is shown in Fig. 1. The large ther-
the existence of a cubic invariant in the free energy expan- mal effect corresponds to the MT. This transition has
sion. However, in our case the possibility of the first order
character for the premartensitic transition is a consequence
of the nonlinear coupling between the symmetry-breaking
order parameter and the magnetization. There is some ex-
perimental evidence that this transition has a first order
character. The main argument supporting this point is the
fact that the frequency of the anomalous phonon does not
reach zero value at any temperature [3]. Actually, v 2 de-
creases linearly with temperature and reaches a minimum
(finite) value at the transition temperature. Extrapolation
down to v ­ 0 predicts that a complete phonon softening
would occur 5 K below the actual first order transition.
Based on this experimental evidence, in the following we
will assume that in Ni2 MnGa the premartensitic transition
e , 0.
is first order, that is, b
The equilibrium first order transition temperature TI
is obtained from the two conditions: ≠Feff y≠h ­ 0 and
1
Feff shI d ­ Feff sh ­ 0d, where hI ­ 6s23by4e e g d 2 is
the value of the order parameter of the distorted phase
at the transition temperature. These conditions lead to a
e2
3 b
transition temperature TI ­ 16a 1 T0 , higher than T0 .
e
g FIG. 1. Thermograms recorded during cooling and heating
The corresponding entropy change at TI is obtained from calorimetric runs in a Ni2 MnGa sample. The small peak
µ ∂ µ ∂ e
≠Feff ≠Feff 1 3ab corresponds to the premartensitic transition and the large one to
DS ­ 2 ­ 2 ahI2 ­ , the martensitic transition. The inset shows the premartensitic
≠T 0 ≠T hI 2 g
8e peak with the measured base line for evaluation of the
(3) latent heat.
3927
VOLUME 79, NUMBER 20 PHYSICAL REVIEW LETTERS 17 NOVEMBER 1997

a jerky character which is especially remarkable in the tic constants under application of a magnetic field at room
exothermic forward transition. Also an intense acoustic temperature (ferromagnetic state) by the use of ultrasonic
emission has been detected during this MT [13]. A very techniques. These experiments will be reported in full
small peak is clearly observed above the MT. It corre- elsewhere [13]. In Fig. 3 we show an example of the be-
sponds to the premartensitic transition. The dots shown havior exhibited by the three independent elastic constants
in the inset below the peak correspond to a measurement for a cubic symmetry. In all cases, prior to each mea-
of the difference in the specific heats of a smaller sample surement, the sample was heated up to a temperature well
of the same crystal [14] and a Cu reference, carried above the Curie point and cooled down to room tempera-
out using a modulated differential scanning calorimeter. ture so that each ultrasonic measurement corresponds to
This measurement determines the base line that enables the first magnetization process. The ultrasonic waves as-
a correct integration of the calorimetric peak leading to sociated with C 0 were affected by a strong scattering by the
the transition latent heat. We have obtained L ­ 29 6 magnetic domains. This resulted in a larger error in the de-
3 Jymol. This value is in very good agreement with our termination of the field dependence of C 0 . For this reason,
numerical prediction based on the Landau model. we also measured C11 , which enabled us to compute C 0
For a first order transition, the Clausius-Clapeyron law as C11 1 C44 2 CL ; the result is shown as a continuous
must hold. In the present investigation, the magnetoelas- line in Fig. 3. We have found that all elastic constants
tic interaction must lead to a change in the transition tem- increase with the magnetic field up to a saturation value
perature (TI ) under application of a magnetic field. We [15]. For the same range of magnetic fields, an increase
have checked the field dependence of TI by measuring the of the magnetization from zero to a saturation value has
magnetic susceptibility across the intermediate transition been reported in a similar sample [16]; this indicates that
with different dc fields applied along the f100g direction, the change in the elastic constants presented here is asso-
using an ac susceptometer. Results are shown in Fig. 2. ciated to a change in the value of the magnetization.
An unambiguous decrease in TI with increasing magnetic Across the intermediate phase transition, M undergoes
field has been found. The change in TI becomes more a change (we recall that M depends on h), which results
pronounced at larger values of the field. Such a nonlinear in a jump of C 0 at the transition, DC 0 ­ k2 sMI2 2 M02 d,
behavior seems to be due to a decrease in DS as the field where MI is the magnetization in the h fi 0 modulated
increases; this decrease would be consistent with the fact phase. Since MI , M0 , it is expected that DC 0 ,
that with the application of magnetic field, the transition 0. It is worth noting that this is in agreement with
temperature approaches the unstability value. Moreover, measurements of the temperature dependence of C 0 which
the observed decrease is in concordance with the behavior show a marked decrease of this elastic constant at the
obtained by computing dT ydH from the model. premartensitic transition [6,7].
The elastic constant C 0 corresponding to a s110d f110g The model presented in this Letter accounts for the
shear can also be obtained from the model as C 0 ­ c 1 magnetoelastic interplay between the structural order pa-
k2 M 2 . With the aim of providing experimental evidence rameters and the magnetization. Since in the tempera-
for this magnetoelastic effect we have measured the elas- ture range of interest Ni2 MnGa shows soft magnetic
properties, this interplay is mostly related to the re-
orientation of the magnetic moments. The interaction

FIG. 2. Transition temperature as a function of a dc magnetic

field applied along the f100g direction. The inset shows
an example of the ac magnetic susceptibility x across the
transition, for two different applied fields [0 Oe (dots) and FIG. 3. Relative change of the elastic constants (solid sym-
20 Oe (stars)]. Data are normalized by the measured value bols) as a function of an applied magnetic field H along the
at 260 K (x0 ). Measurements have been conducted with an ac f110g direcion. The solid line corresponds to C 0 computed from
field of 2 Oe and a frequency of 666 Hz. measured values of C11 , C44 , and CL .
3928
VOLUME 79, NUMBER 20 PHYSICAL REVIEW LETTERS 17 NOVEMBER 1997

between the magnetic and structural degrees of freedom MT by the modification of the dynamical response of the
has been demonstrated by the measured field dependence bcc parent lattice. Within a general framework, the re-
of C 0 and of the intermediate transition temperature. The sults presented here evidence that coupling of a secondary
coupling between the amplitude of the anomalous short- field (magnetic in this case) to incipient unstable excita-
wavelength phonon (primary order parameter) and the tions, associated to structural degrees of freedom, can fun-
magnetization gives the possibility for the occurrence of damentally affect the characteristics of a phase transition.
a first order transition before a linear phonon unstability This work has received financial support from the CI-
is reached. The thermodynamic quantities of this tran- CyT (Spain), Project No. MAT95-504 and CIRIT (Cat-
sition have been obtained from the model. They are in alonia), Project No. SGR00119. The single crystal was
good agreement with experimental results. Actually, a kindly provided by V. A. Chernenko. E. O. and A. G. ac-
predicted latent heat of around 27 Jymol is in excellent knowledge financial support from DGICyT (Spain).
agreement with the value obtained from calorimetric mea-
surements. The intermediate transition takes place at a
temperature slightly above the one where the anomalous [1] J. A. Krumhansl and Y. Yamada, Mater. Sci. Eng. A 127,
transverse phonon would become unstable. This indicates 167 (1990).
that the first order character of this transition is weak, in [2] W. Petry, J. Phys. (France) IV, Colloq. 5, C2-15 (1995).
[3] A. Zheludev, S. M. Shapiro, P. Wochner, A. Schwarz,
concordance with the small value found for the latent heat.
M. Wall, and L. E. Tanner, Phys. Rev. B 51, 11 319
The microscopic origin of the phonon anomaly and (1995).
magnetoelastic coupling related to this phase transition are [4] G. Fritsch, V. V. Kokorin, and A. Kempf, J. Phys.
not explained by the proposed phenomenological model. Condens. Matter 6, L107 (1994).
It has been suggested [17] that their origin lies in the [5] E. Cesari, V. A. Chernenko, V. V. Kokorin, J. Pons, and
electron-phonon coupling and specific nesting properties C. Seguı́, Acta Mater. 45, 999 (1997).
of the multiply connected Fermi surfaces. While this [6] J. Worgull, E. Petti, and J. Trivisonno, Phys. Rev. B 54,
explains similar short-wavelength phonon anomalies ob- 15 695 (1996).
served in Ni-Al alloys [18], no specific calculations in [7] Ll. Mañosa, A. Gonzàlez-Comas, E. Obradó, A. Planes,
such a direction have, to our knowledge, been performed V. A. Chernenko, V. V. Kokorin, and E. Cesari, Phys.
in the case of the system studied here. Nevertheless, no Rev. B 55, 11 068 (1997).
[8] V. V. Martynov, J. Phys. (France) IV, Colloq. 5, C8-91
premartensitic transition has been reported to occur in Ni-
(1995).
Al. Actually, the ferromagnetic character of the Ni2 MnGa [9] V. V. Kokorin, V. A. Chernenko, E. Cesari, J. Pons, and
is the main difference between these two alloys. C. Seguı́, J. Phys. Condens. Matter 8, 6457 (1996).
The MT in nonferromagnetic alloys have been de- [10] A. Zheludev and S. M. Shapiro, Solid State Commun. 98,
scribed using Landau-type models. In these models, the 35 (1996).
transition takes place as a consequence of the anharmonic [11] With the exception of v, all coefficients are not expected
coupling between an anomalous transverse phonon and to vary rapidly with temperature near the transition.
a homogeneous strain [19]. Such a coupling (h 2 ´) is [12] G. Guénin et al., in Proceedings of ICOMAT’86 (The
not included in our model. However, we argue that the Japan Institute of Metals, Nara, 1986), p. 794.
magnetoelastic interaction indirectly couples h and ´. At [13] A. Gonzàlez-Comas, E. Obradó, Ll. Mañosa, A. Planes,
the MT, ´ (tetragonal distortion) becomes different from V. A. Chernenko, V. V. Kokorin, and A. Labarta (to be
published).
zero and the periodicity of the transverse modulation with
[14] The small sample has been cut from the single crystal.
wave vector q ­ 0.33 is modified (a five layer modula- It has been verified that the composition of the crystal is
tion is obtained). To account for such a modification in homogeneous. Moreover, when possible, measurements
the periodicity, an explicit wave vector dependence of the have been conducted on the two samples, and the
free energy should be included in the Landau model. It is obtained data have been found to be coincident within
also worth noticing that reorientation of martensitic twin the experimental errors.
variants has been achieved recently by the application of [15] The same behavior is obtained with the magnetic field
magnetic field [16]. applied along the [100] axis. In this case, for each elastic
In conclusion we have presented a model that accounts constant, the same saturation value (within experimental
for the first order phase transition between the bcc and the error) is achieved at a lower value of the magnetic field.
intermediate phases. The first order character has been [16] K. Ullakko, J. K. Huang, C. Kantner, R. C. O’Handley,
and V. V. Kokorin, Appl. Phys. Lett. 69, 1966 (1996).
experimentally demonstrated by the measurement of a la-
[17] A. Zheludev, S. M. Shapiro, and P. Wochner, Phys.
tent heat and by the magnetic field dependence of the tran- Rev. B 54, 15 045 (1996).
sition temperature. In the model, the first order transition [18] G. L. Zhao and B. N. Harmon, Phys. Rev. B 45, 2818
occurs as a consequence of a magnetoelastic coupling. (1992).
Such a coupling has also been experimentally evidenced. [19] P. A. Lindgård and O. G. Mouritsen, Phys. Rev. Lett. 57,
Hence, the premartensitic transition must be considered 2458 (1986); J. A. Krumhansl, Solid State Commun. 84,
as a magnetically driving precursor effect announcing the 251 (1992).

3929