JOURNAL DE PHYSIQUE Colloque C6, supplkment au no 12, Tome 37, Dtcembre 1976, page C6-317

D. C . PRICE
Dept. of Solid State Physics, Research School of Physical Sciences
Australian National University, Canberra, Australia

J. D. RUSH, C. E. JOHNSON, M. F. THOMAS

Dept. of Physics, University of Liverpool, Liverpool, U. K.

and

P. J. WEBSTER
Dept. of Pure and Applied Physics, University of Salford, Salford, U. K.

RBsum6. - Ce travail discute les modeles thkoriques qui sont couramrnent utilises pour decrire
le couplage magnktique entre les ions Mn, ainsi que les champs magnktiques hyperfins au siege de
ions non magnktiques S, p prksents dans les alliages XzMnZ de Heusler. I1 montre que les donnks
expkrimentales sur la dkpendance spatiale de l'interaction d'echange entre ions Mn-Mn dans
PdzMnSn, sont consistantes avec le mkcanisme oscillatoire a longue distance d'action, qui rksulte
de la diffusion des moments magnktiques de deux ions Mn par le potentiel d'un Bectron de conduc-
tion de type s (couplage ? double
i resonance). Aucune kvidence prkcise n'a Btk trouvke, qui soit en
faveur d'un autre mkcanisme plausible de couplage.
Le travail passe en revue les modeles actuels, qui impliquent des effets ioniques locaux dans la
production de champs magnktiques hyperfins au siege des ions Z non magnktiques. L'ktude se
concentre tout sp6cialement sur deux modeles qui decrivent les champs en termes de la structure
klectronique locale du ion, et des raisons sont avancks pour justifier cette approche. La nature
approximative de ces modbles est mise en evidence, ainsi que l'inopportunitk conskquente de leur
usage au vue d'ktablir des comparaisons dktaillkes avec l'experience. L'accent est mis sur le besoin
de davantage dY6tudesthkoriques dktaillkes, en vue d'ktablir ces modkles sur une base rigoureuse
pratique et susceptible d'6tre perfectionnke.
Les mesures de l'effet Mossbauer qui sont rapportkes dans ce travail concernent Ies ions "9Sn et
121Sndans les solutions solides des alliages PdzMnIn, Pd zMnSn et PdzMnSb de Heusler. La compa-
raison de ces donnkes Mossbauer avec celles qui sont obtenues par diffraction de neutrons dans
Ies alliagesrenfermant moins d'eiectrons que PdzMnSn, montre que les diffkrentes structures magnk-
tiques qui sont observkes dependent tres sensiblement du nombre moyen d'dectrons au siege de
I'ion Z. Le mkcanisme de couplage magnetique double resonance s'avhe &re, au moins qualitati-
vement, consistant avec ces observations.
Dans les alliages de composition PdzMnSnzSbl-z, les champs hyperfins, au siege des ions "9Sn
et 121Sb, changent rapidement avec le paramMre de composition X, quand X s'approche de zkro.
Ces proprietks sont discutks en termes des modeles mentionnes prkddemment. Quant aux alliages
dont la densitk klectronique approche celle de PdzMnSb, on n'observe ni champ hyperfin, ni varia-
tion de la tempkrature de Curie en fonction de X, qui soit explicable en termes du modble des Blec-
trons libres des bandes de conduction. Des conclusions similaires sont dkduites des measures de
l'effet Mossbauer, concernant les ions 121Sb dans les alliages Pdl+,,MnSb, quand 0 < y < 1.

Abstract. - Theoretical models that are currently used to describe the magnetic coupling
between Mn ions and the magnetic hyperfine interactions at non-magnetic S, p ions in XzMnZ
Heusler alloys are discussed. Experimental data on the spatial dependence of the Mn-Mn exchange
interaction in PdzMnSn is shown to be consistent with the long-range oscillatory mechanism that
arises from the potential scattering of an S-like conduction electron by two Mn moments - the
double resonance coupling. No detinite evidence is found for any other significant coupling mecha-
nism.
Present models that involve local ionic effects in the production of the magnetic hyperfine fields
at non-magnetic Z site ions are reviewed. Attention is concentrated on two models that describe the
fields in terms of the local electronic structure of the ion, and reasons for this are given. Emphasis is
placed on the approximate nature of these models and on the consequent inadvisability of using
them for detailed comparisons with experiment. The need for more detailed theoretical work
aimed at providing a rigorous basis for application and subsequent refining of these models is
stressed.
119Sn and l2lSb Mossbauer measurements have been made on solid solutions of the Heusler
alloys PdzMnIn, PdzMnSn and PdzMnSb. Comparison of Mossbauer data with neutron diffrac-
tion data for alloys with fewer electrons than PdzMnSn shows that the different magnetic structures

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphyscol:1976665

 C6-318 D. C. PRICE, J. D. RUSH, C. E. JOHNSON, M. F. THOMAS AND P. J. WEBSTER

that are observed depend very sensitively on the average number of electrons at a Z site ion. The
double resonance magnetic coupling mechanism is seen to be at least qualitatively consistent with
these observations.
For alloys with electron densities approaching that of PdzMnSb, rapid changes of both the
ll9Sn and 121Sb hyperfine fields with composition are observed. These are discussed in terms of
the hyperfine field models previously described and, along with changes in this composition range in
other quantities such as the Curie temperature, cannot be explained using the normal freeelectron-
like description for the conduction band. Similar conclusions are drawn from 121Sb Mossbauer
measurements on solid solutions of PdMnSb and PdzMnSb.

1. Introduction. - In 1903 F. Heusler reported that and it was shown that the magnetic properties were
it was possible to make ferromagnetic alloys from the both structure and composition-dependent (see [l]).
non-ferromagnetic constituents copper-manganese We shall restrict our attention in this paper to alloys
bronze and group B elements such as aluminium and with Mn on the Y sites as the only magnetic constituent
tin. These alloys, some of whose properties were (i. e. alloys containing no CO)in order to have a specific
reviewed by Webster [l], are of interest from several magnetic system on which to base our discussions.
points of view. They are ternary compounds with a The magnetic properties of these alloys are of interest
simple structure (Fig. ' l) that consists of four inter- because the Mn atoms are well separated, so that direct
exchange between them must be negligible, and appear
to have well-localized magnetic moments [21. Magnetic
order has traditionally been assumed to be due to long-
range interactions via the conduction electrons, but
recent experimental results have been interpreted as
indicating the presence of strong short-range interac-
tions. The ordering temperatures, and in some cases
even the type of order, depend strongly on the par-
ticular atoms occupying both the X and Z sites.
Consequently, this class of compounds provides a
simple and structurally well-defined system of magnetic
Structure O @ 0
moments in a widely variable environment in which to
'-21 X X Y Z
Clb X vacant Y Z study the magnetic interactions.
B2 X X Y,Z Y.2 Perhaps more important in the context of this
FIG. 1. - The Heusler structure, with the four f. c. c. sublattices Conference is the relevance of the large number of
indicated. possible constituent atoms to the study of hyperfine
field systematics and the origins of hyperfine fields in
metals. Of particular interest is the number of S, p
penetrating f. c. c. lattices which may be thought of as a elements that can occupy the Z site and the apparent
b. c. c. CsC1-type structure in which alternate positions ease with which these can be substituted for one
on each sublattice may be occupied by different atoms. another, at least in small concentrations, in some alloys.
Within the general Heusler classification there are three There seems little reason to doubt that such substitu-
common structures corresponding to different arran- tion is restricted to the Z sites except, of course, in
gements of atoms on the four f. c. c. sublattices. These alloys of B2 structure. Campbell [3] has recently
are indicated in figure 1 with their Strukturbericht published a summary of hyperfine field measurements
labels. The stoichiometry of the L2, (which is the ori- in pure and substituted Heusler alloys. We shall not
ginal Heusler structure) and B2 structures is X,YZ attempt to duplicate this here.
while that for the Cl, structure is XYZ. The constituents Our aim in this paper will be to examine current
X, Y and Z may be any of a large number of elements theories of both the magnetic and hyperfine interactions
and consequently there is a wide variety of different in Heusler alloys, and to present some of our recent
Heusler alloys. In general, and there are exceptions, Mossbauer results on a system that has important
X is a noble metal or a transition metal (3d, 4d or 5d) implications in both of these areas. Although much of
atom with an almost full outer d shell, Y is a transition the interest in Heusler alloys stems from their well-
metal atom with fewer outer d electrons than X, and defined structural characteristics we will not comment
Z is a non-transition metal atom. on possible reasons for these metallurgical properties :
Many Heusler alloys, in particular those with COin this appears to be an area that is not yet well
the X site and/or Mn in the Y site, exhibit magnetic understood.
order with transition temperatures ranging up to The following two sections of this paper will deal
several hundred K. Heusler's original ferromagnetic briefly with theories of the magnetic properties of alloys
alloys contained Mn on the Y site of the L2, structure of the type X,MnZ and XMnZ and of hyperfine
interactions in these alloys, particularly at Z site atoms.
renewed interest in the theory of the magnetic coupling
We shall then describe the results of some recent in Heusler alloys. The results show that coupling
Mossbauer and magnetic measurements on the systems between Mn ions at relatively large separations (further
Pd,Mn(In, Sn, Sb) and Pd, ,,MnSb and relate them apart than 3rd nearest neighbours) can be satisfactorily
where possible to the theories described, described by the CB model, but that at smaller separa-
tions it cannot : exchange of the wrong sign is predicted
It is our opinion that there is now a good qualitative
understanding, derived mainly from experiment, of the by the CB model between near neighbour (Inn) and
electronic and magnetic interactions in Heusler alloys. second nearest neighbour (2nn) Mn ions. This prompt-
We feel that further significant advances will depend oned Kasuya 1101 to suggest that a short-range ferro-
purposefully planned experiments and on more detailed magnetic virtual double exchange mechanism might
theoretical work being done in order to obtain a more be present in addition to the long-range double
rigorous basis for comparing current models with resonance coupling.
experiment. Campbell and Stager [l l ] have measured the satura-
tion magnetization and Curie temperatures of alloys
based on Ni,MnSn in which some of the Mn ions are
2. Magnetic coupling in Heusler alloys. - replaced by Ti, V or Cr ions. They interpreted their
2.1 INTRODUCTION. - Prior to 1966 discussions of results to indicate that in these alloys the Ti ions
the magnetic interactions in Heusler alloys were either
empirical or were qualitative and based, in general, on
carried a net moment of - - 1 p,, the V ions were

the RKKY theory of polarization of conduction elec-

trons by the S-d exchange interaction [l]. The first
moment of -
non-magnetic (zero moment) and Cr ions had a
+ 1 p,. As a result of these measure-
ments Campbell [l21 has suggested, following Goode-
model to give reasonable quantitative estimates of
nough 1131, that the dominant exchange process is a
transition temperatures, even though employing the
super-exchange interaction with the Z site S, p ion
molecular field approximation, was that of Caroli
acting as the intermediary.
and Blandin [4] (to be referred to as CB hereafter).
A shortcoming of the CB double resonance coupling
The application of this model to Heusler alloys was
model, the effects of which do not appear to have been
also examined by Geldart and Ganguly [5], Brooks
quantitatively evaluated so far, is that it involves
and Williams [6] and others. The model is based on the
approximations that limit its range of applicability to
work of Caroli [7] and the coupling effectively arises
ions that are separated by large distances i. e. it involves
from the scattering of a conduction electron by diffe-
asymptotic formulae in the evaluation of certain inte-
rent Mn local moments. The scattering occurs as a
grals. We have recently done a calculation in the same
result of the S-d mixing term in the Friedel-Anderson spirit as CB but without the use of these asymptotic
model [8] of the Mn localized moment. Caroli [7] formulae. This calculation and its results will be
estimated that, for the case of Mn in copper, this described briefly below. As a result of it we believe that
double resonance exchange coupling would be approxi- there are no experimental results presently available
mately two orders of magnitude larger than coupling
that cannot be explained (at least in principle) using
due to S-d exchange. double resonance coupling as the dominant exchange
Long-range magnetic coupling of localized moments mechanism.
via the conduction electrons was widely accepted as
being the dominant exchange mechanism in Heusler 2.2 << EXACT>> CALCULATION OF THE DOUBLE RESO-
alloys for some time. Recently, however, Ishikawa NANCE COUPLING. - Our model, which is based on the
and Noda [g] have, using inelastic neutron scattering, Hartree-Fock approximation, is the same as that of
measured the magnon dispersion curves of Pd,MnSn Caroli [7] and very similar to those discussed in some
and Ni,MnSn, and derived the distance dependence detail by Moriya [l41 and Kim and Nagaoka [l 51. We
of the effective exchange integral J(r) in these com- start from the Anderson Hamiltonian [8] generalized
pounds. This excellent series of measurements has to the case of two impurities :

where the notation is standard (see (71, [14]). m, p are localized orbitals on impurity i. V,,, is the S-d mixing
matrix element between the localized state m on impurity i and the state k of the conduction band. A direct d-d
interaction term should also be included in (1) but we have neglected it on the grounds that the magnetic atoms
are well separated. This ignores the possibility that d-d admixture with the neighbouring transition metal ions on
the X sites is important, and this should be borne in mind.
In the limit of isolated magnetic ions (i. e. no interaction between them) this Hamiltonian gives rise to the
familiar virtual bound state picture for the localized orbitals, and a magnetic moment is formed if the effective
+
exchange energy (U 4 J) is sufficiently large compared with the width A , of this state ([8], [l41 and [16]). The
C6-320 D. C. PRICE, J. D. RUSH, C . E. JOHNSON, M. F. THOMAS AND P. J. WEBSTER

interaction between two magnetic ions is manifested firstly as a change in the magnitude of the localized moment
on each and secondly as a coupling between them. The former has been considered qualitatively by Kim and
Nagaoka [l51 and by Moriya [l41 and may be relevant to the interpretation of the results of Campbell and
Stager [ll]. We will concentrate on the latter effect here.
If 6p;(~) is the change in the total density of electron states of spin o caused by the interaction between
magnetic ions, then the interaction will produce a decrease in the total energy of the system given by [7] :

where n"oi is the occupation number of the localized state of spin a in orbital m on ion i in the absence of the
ion-ion interaction and 6nG, is its change due to the interaction. The following assumptions are made in the
evaluation of Eint:
(i) For the case of a pure Mn magnetic lattice we take all of the magnetic ions to be identical.
(ii) Following Caroli [7] we take the 6nG to be small and neglect the second term in (2).
(iii) The different d orbitals on a magnetic ion are taken to be equivalent, apart from their angular momen-
tum properties.
Then, in terms of the Green's functions G::(&, n) for the Mn d states in the absence of the interaction between
ions, the expression (2) reduces to :
EF
Ei(R) - I Tr G!%, no)F.(&, R) G:(&, no) de
spins m -m

where

a"(&) is the d-like phase shift of an electron of spin o t o the integral in (3) come from the region near E,.)
and energy e that is scattered by the Mn ion potential. Caroli [7] evaluated it analytically in the asymptotic
The evaluation of the matrix elements V,, and V, limit, which is equivalent to ignoring the spherical
is important as it is from them that the R-dependence of harmonics in (4) i. e. no account was taken of the d
E,,, is derived. If the ion i is taken to be at the origin symmetry of the scattering of the Mn ion. Alloul [l61
and j is at R, then pointed out that this was also the important difference
between asymptotic and non-asymptotic expressions for
the spin polarization induced by a magnetic moment in
the conduction band. The results obtained are exact
only in the sense that the evaluation of (3) is exact
As qm(r) is an atomic-like function and is written (apart from the k-dependence of V). The model from
as R,(r) Y T ~ )and
, the conduction band states are which the expression (3) is derived contains a number
approximated by plane waves, the expressions of approximations, only some of which have been
for Vnltkand Vmjkreduce to : explicitly mentioned.
The effective exchange integral J is derived from Eint
as

J = - - E1 int Y
4 s2
where S is the effective spin of a Mn ion.
with
2.3 RESULTSOF CALCULATION. - The solid line
in figure 2 was calculated as described above, and the
experimental results of Ishikawa and Noda [g] for
The integral in (3) is then evaluated numerically Pd,MnSn are seen to be satisfactorily described by it.
assuming that V(k) can be approximated by its value at It was assumed in the calculation of this curve that the
k,. (This can be justified if the dominant contributions spin-up 3d states of each Mn ion were fully occupied
 The dashed curve in figure 2 is the exchange para-
meter J(r) calculated from the asymptotic formula of
CB [4] for the same values of the parameters used for
the solid curve. The asymptotic formula neglects terms
of higher powers of llr which is why it predicts a much
larger coupling at small r values than does the non-
asymptotic calculation. The arrows at the bottom of
figure 2 indicate the positions, for the value of k,
used, of the successive Mn coordination shells.
We hope to give further details of this calculation
and its results in a future publication.

2.4 DISCUSSION. - Because the results reported

here contain a number of potentially serious approxi-
FIG.2. - The effective exchange integral J(r) between two Mn
ions separated by distance r. The experimental points were
mations it is probably not reasonable to base any firm
determined from the magnon dispersion curves of Pd2MnSn conclusions on them. However, we feel that they give
by Ishikawa and Noda [g], and are plotted as a function of an indication that the S-d mixing mechanism may be
k~ r using the value k~ = 1.0 a. U. The solid line is calculated, primarily responsible for the exchange interactions in
as described in the text, from the double resonance model of Heusler alloys such as Pd2MnSn, or at least that more
Caroli [7] but without making any asymptotic approximations.
The parameters used to describe local Mn states are given in detailed calculations should be carried out before other
the text. The dashed curve is calculated, using the same set of exchange mechanisms are assumed to be necessary.
parameters, from the asymptotic formula of Caroli. Note the We feel that, at least in principle, it is possible to
change in the scale of the ordinate that occurs at the vertical
dashed line. qualitatively account for the results of Campbell and
Stager Ill] in terms of this exchange mechanism
although no numerical results have yet been obtained.
An isolated Ti ion, for example, would be assumed to
and that the net moment per ion was 4.2 ,uBas observed
carry no localized moment, as is the case in its dilute
for Pd2MnSn [2]. The width A of the spin down 3d
states was taken rather arbitrarily to be -
EF/10
with E, -- 0.40 a. U. (= 10.9 eV). Using these parame-
alloys with noble and other metals [IS], [19], and the
effect of its interaction with surrounding Mn ions via
the S-d mixing mechanism would be to reduce their
points best for a value of k, of -
ters the calculated curve described the experimental
1.0 a. U. (= 1.9
and V(kF), which was used to scale the curve vertically,
a-') moment. This sort of effect has been discussed by
Moriya [l41 and Kim and Nagaoka [15]. These latter
authors show qualitatively that for a case such as Mn
about 7 % larger than its value calculated from a free
electron approximation [16]. The value of k, is 20 %
larger than is normally assumed from standard pres-
- substituted for Ti in Ni2TiSn, the Mn moment should
increase as the Mn concentration increases, as is
criptions of the number of conduction electrons observed [l l].
contributed by each atom (see, e.g., [4], [5], [6], [10], It will be shown below (Section 5) that the S-d
1171) but we do not regard this as being serious for two mixing exchange as described here can account for
reasons. Firstly, and most importantly, it must be changes in magnetic order with conduction electron
emphasized that the calculation reported here contains density that are qualitatively similar to those observed
a number of potentially important approximations. in the alloy systems Pd,MnIn,-,Sn, (0 X 1)
Apart from those mentioned both explicitly and impli- and Pd2MnIn,-,Shy (0 d y d 0.5).
citly in the last section it will be appreciated that the
coupling considered is that between a single pair of 3. Hyperfine fieIds in Heusler alloys. - 3.1 INTRO-
magnetic ions and as such takes no account of the DUCTION. - At the current stage of development of
effects of other neighbouring and intervening ions. It understanding of the origins of magnetic hyperiine
should be more appropriate to a dilute alloy situation fields in metals it is desirable to establish the systema-
than to the case here of a concentrated magnetic tics of these fields in systems that are as simple as
material. Because of all the approximations involved possible. The S, p ions in the Z sites of X2MnZ (or
we feel that any numerical results obtained should not XMnZ) Heusler alloys constitute such a simple sys-
be taken too seriously. Secondly, in the absence of any tem : the alloys are cubic, their magnetization is
calculation or direct measurement of the structure of localized on the Mn ions and there would appear to be
the conduction band in these alloys it is difficult to very little direct overlap of the Z site ion with the Mn 3d
regard this value of k, as being obviously unreasonable. electrons that carry the magnetic moment. Because we
One interpretation of the results that will be presented feel that this is the main reason for studying hyperfine
below (Section 6) is that the conduction band is much fields in Heusler alloys instead of, say, dilute binary
more complicated, at least in some alloys, than a free alloys or other intermetallic compounds, we shall
electron band. restrict our attention here to this system. There is,
C6-322 D. C. PRICE, J. D. RUSH, C . E. JOHNSON, M. F. THOMAS AND P. J. WEBSTER

however, a growing interest in hyperfine fields at a necessity to account for the local properties of a
transition metal ions in the X sites of Heusler alloys particular ion (which had been recognized for some
and also in the fields at all ions in alloys containing time in the case of dilute impurities in binary
cobalt when it: carries a magnetic moment. Such measu- alloys [23]) although it was not immediately obvious
rements would, we feel, be more properly included which local properties had to be taken into account.
in a more general review of hyperfine fields in Three different models have subsequently been
metals. proposed that incorporate local ionic effects. In one,
A large number of measurements have now been Stearns [24] proposes that it is the ionic volume, or
ma,de of hyperfine fields of non-magnetic ions in more strictly its misfit in the host lattice, that is impor-
Heusler alloys. These have recently been summarized tant. In the others of Blandin and Campbell [25] and
by Campbell [3], and definite trends are becoming Sena and Geldart [l71 it is considered that the effects
evident. Campbell's paper should be consulted for of the local electronic structure dominate. These models
references to specific measurements where they are and their applicability will be discussed below.
not given here.
Historically, measurements were first made 3.2 STEARNS' VOLUME MIS-FIT MODEL. - This model
on pure Heusler alloys such as Cu,MnAl, Cu,MnSn, was originally proposed [24] to account for the hyper-
Ni,MnSn, Pd2MnSn and the first attempt at explaining fine field variations at impurity ions in Fe, CO and Ni,
them was made by Caroli and Blandin [41 in the same although we understand that its applicabilityto Heusler
paper in which they discussed the magnetic coupling. alloys is under consideration [26]. It is an empirical
They assumed that the hyperfine field at a non-magnetic model that makes use of a similarity between the
ion reflected only the polarization of the conduction variations of impurity hyperfine fields and atomic
electrons at its site through the Fermi contact interac- volumes. The hyperfine field at a non-magnetic
tion. However, following the important step of Leiper impurity is assumed to consist of two contributions.
and Campbell [20], 1211 and Swartzendruber and One is from the host spin polarization in the same
Evans [22] of measuring the fields at non-magnetic spirit as the Caroli-Blandin model while the other is
ions substituted into Heusler alloy Z sites, it rapidly proportional to the volume mis-fit of the impurity
became clear to most people that there was a strong in the host.
dependence of the hyperfine field on the particular ion We doubt that this model represents the physical
at which it was measured, and that this could not be situation in either binary alloys or Heusler alloys as it
des~ribed by the Caroli-Blandin model. Figure 3 does not appear capable of explaining in a consistent
way several experimental results. Some inconsistencies
have been mentioned by Campbell and Vincze 1271
although Stearns has replied to these [28]. Other points
of difficulty for this model would appear to be expla-
nations of :
(i) the pressure dependence of the hypefine field at
Sn in iron. Moller 1291 showed that the negative
hyperfine field became much more strongly negative as
the lattice was compressed ;
(ii) the temperature dependence of the hyperfine
fields at non-magnetic impurities in iron, and in parti-
- L0
1 '"
X

+
PdZMnSb
PdMnSb
Nl2MnSb
Host
o NiMnSb
* Ni2Mnln
CuZMnln
cular that of the field at As compared with other
impurities [30] ;
(iii) the calculations .of Sondhi [31] of the overlap
FIG. 3. - Hyperfine fields at 5s, p ions in the Z sites of the between the orbitals of solute S electrons and 3d elec-
Heusler alloy hosts indicated. This shows that, with the possible trons of iron in dilute iron alloys. These indicate that
exception of PdzMnSb, there is only a relatively weak depen- the direct overlap is an order of magnitude too small to
dence of the field on the particular host compared with the explain the volume mis-fit field in binary alloys. In
dependence on the ion at which the field is measured. These
measurements have been taken directly from the literature
Heuslers, where the separation between the Mn ions
(see [3] for references) and no corrections (for example to and Z site ions is even larger, it is difficult to see that
T = 0)have been made. direct overlap could be significant ;
(iv) it will be shown in section 6 below that some of
our measurements on alloys near to the composition
illustrates this dependence : it can be seen that the Pd,MnSb cannot reasonably be explained using the
hyperfine fields vary in a qualitatively similar way volume mis-fit model. Other measurements in
from ion to ion with relatively little change with host Pd,MnSb [42], [49], 1501, [51], [52] also do not vary in
(except in the case of Pd2MnSb). There was, therefore, the same way as the impurity atomic volume.
 3.3 BLANDIN-CAMPBELL MODEL. - This model model applied to a typical X,MnZ Heusler alloy are
represents a modification of the Caroli-Blandin given for impurity phase shifts calculated by
model [4] in order to take account of the effect of the
electronic structure ,in the conduction band in the
immediate vicinity of the impurity. It should be
recognized at once that in principle it cannot be where Z,, is the excess charge to be screened, and for
regarded as being strictly applicable to a concentrated phase shifts calculated if the impurity potential was a
magnetic material since, in common with the Caroli- spherically symmetric square well satisfying the Friedel
Blandin model, it considers isolated pairs of atoms in sum rule. The differences between these curves
an unpolarized electron sea (see Section 2 above). illustrate the inadvisability of making detailed compari-
However, since the effects of coherent scattering and sons of experimental results with qualitative model
other effects of neighbouring ions are not well known, calculations.
the model may be empirically useful. It has the
advantage that the effect of the non-magnetic ion is 3.4 JENA-GELDART MODEL. - In 1963 Daniel
included explicitly as the phase shifts it produces in and Friedel [23] calculated the electron spin polariza-
scattered electrons, so that realistic impurity potentials tion at the nucleus of a non-magnetic ion dissolved in
can be readily employed. an explicitly spin-polarized electron sea, and applied it
As originally formulated by Blandin and Camp- to the case of dilute impurities in ferromagnetic binary
bell [25] the model contained several numerical alloys. The principle on which their calculation was
approximations. These are not essential to the spirit based was that, since the impurity ion potential was
of the model and were presumably made in order to non-magnetic and the density of conduction electrons
give a qualitative indication in simple terms of the in the host was different for the two spin directions,
expectations of the model. In particular, Blandin electrons of opposite spin would be scattered differently
and Campbell only accounted in a very approximate by the non-magnetic ion. They used the simplest model
way for the spin polarization produced by a Mn ion possible to illustrate this principle : the impurity ion
that is a near neighbour to a non-magnetic ion and they was represented by a spherically symmetric square-well
used only a rough estimate of the I = 0 phase shifts potential whose radius was that of the Wigner-Seitz
produced by the non-magnetic ion. Unfortunately the sphere of the ion and whose depth was calculated to
model has sometimes been criticized on the basis of satisfy the Friedel sum rule. The conduction electrons
these numerical approximations (e. g. [3], [51]) and, of the host metal, which were tagen to behave as free
whilst acknowledging that the usefulness of a model electrons, were assumed to have a uniform spin polari-
calculation lies in it being simpler than an ab initio one, zation within this Wigner-Seitz sphere. The conduction
we do not believe this criticism to be justified in this band contained all the effectsof the host. Jena and Gel-
case in the absence of a more detailed numerical dart [32] have given a more fundamental derivation
examination. An example of the effect of a simple of these approximations.
variation in one of these aproximations is shown in Swartzendruber and Evans f33] first suggested that
figure 4. The predictions of the Blandin-Campbell this model might be applicable to Heusler alloys but it
was left to Jena and Geldart [17], [32] to make detailed
comparisons of its predictions with measured hyperfine
fields. The 3ena-Geldart model is identical to that of
Daniel and Friedel in principle and only varies
numerically by the use of an othogonalized plane wave
approximation for calculating the wave functions of the
Bloch electrons in the vicinity of the nucleus. Some
results of the model are illustrated in figure 5 (we note
some minor numerical discrepancies with the results of
Jena and Geldart in the caption). The quantity f that is
plotted in figure 5 against the valence Z , of the ion is a
dimensionless quantity that is defined by Daniel
and Friedel [23]. It is proportional to the spin polariza-
FIG. 4. - The effect of varying one of the numerical approxima- tion at the nucleus of the ion and its value depends
tions used in the Blandin-Campbell model for hyperfine fields only on Z,,, the charge to be screened that depends on
of S, p ions in Heusler alloys. All parameters other than the
l = 0 phase shifts at the Fermi level, 60, are the same as those the ionic potential, and k, a, where k, is the wave
used by Campbell and Blandin [25]. The calculated electronic number for conduction electrons at the Fermi surface
spin polarization at the non-magnetic ion nuclei is shown when and a is the radius of the square potential well used to
their simple approximation 6 0 = a Zsc/4,where Z,, is the impu- approximate the ion's potential. The values of k, a
rity charge to be screened, is used and when the phase shifts
are calculated assuming that the ionic potential is a spherically appropriate to each curve are indicated on the figure.
symmetric square well that satisfies the Friedel sum rule. This model has the conceptual advantage that it is
C6-324 D. C. PRICE, J. D. RUSH, C. E. JOHNSON, M. F. THOMAS AND P. J. WEBSTER

impurity ion is related to its electronic structure, and

therefore to its valence and screening characteristics,
because both depend on the nuclear charge and the
electron-electron interactions within the ion. In her
papers, Stearns [24] has simply drawn attention to the
correlation between hyperfine field and impurity
volume and has only by implication suggested that it
represents directly the physical mechanism responsible
for the observed hyperfine fields. We disagree for the
reasons mentioned in 3.2 above.
These considerations also mean, of course, that
because the models based on local electronic structure
can show the same qualitative trends as the experi-
mental results, they do not necessarily represent the
underlying physical origins of the hyperfine fields
FIG. 5. -Spin polarization calculated from the Daniel-
Friedel model. The quantity f is proportional to the electronic either. However, these models predict, in a qualitatively
spin ~olarizationat the nucleus of an ion with valence ZI, correct way, the form that the hyperfine field vs. ionic
and is defined by Daniel and Friedel [23]. The value of the valence correlation should take, and they also incorpo-
parameter k~ a, defined in the text, that is appropriate to each rate what appears to be a plausible physical process
curve is indicated. In comparing these curves with those in
the paper of Jena and Geldart [l71 it will be seen that there is which, for lack of a reasonable alternative, we shall take
a difference in the points at which the curves cross the ZI axis. to represent the actual mechanism that gives rise to the
This appears to be due to a numerical discrepancy that is still observed hyperfine fields. It should be appreciated that
unresolved, although we do not regard it too seriously in view a realistic calculation of hyperfine fields at ions in
of our comments in the text about the purely qualitative applica-
bility of these models. It may be noted that our calculations, metals based on the principle that local electronic
some of which are shown above, agree with those of Daniel screening and its interaction with the spin-polarized
and Friedel [23] and of Koenig 1351. Another point concerning electrons in the host is the process responsible for them,
the paper of Jena and Geldart 1171is that their published polari- would be extremely complex so, at least in the near
zation curves (their Fig. 1) are calculated using the value
~ / E= F 0.01 for the relative splitting of the conduction band. future, it is unlikely that the supposedly dominant role
This is not mentioned in their paper and account must be taken played by this mechanism will be conclusively confirm-
of it in calculating hyperfine fields from their work. This is ed or denied.
done by normalizing to the phenomenologically determined
value of A/EF. We now turn our attention to the models describing
this process proposed by Blandin and Campbell [25]
and Jena and Geldart [l71 and briefly described above.
formulated specifically for a concentrated magnetic These two models are complementary in that they start
material and includes all the effects of coherent from rather diverse propositions. Both, at their present
scattering, exchange, hybridization, etc. automatically stage of development, appear to describe the observed
through a parameter that is determined phenomeno- trends in hyperfine field systematics for non-magnetic
logically. A disadvantage is that it requires a rather ions in a qualitatively reasonable way. However, we
detailed knowledge of the ion's potential and conse- wish to emphasize that in view of the approximations
quently is difficult to adapt to more realistic ionic employed in their derivation it is difficult to justify any
potentials : the form of the potential is expected to detailed comparison of their predictions with the
become increasingly important for higher values of Z, experimental results. Some of these approximations
when bound states are present. The model may also be have already been mentioned. A central problem is the
expected to have difficulty for very small values of Z,, lack of a sound basis for assigning values for the
when the potential is of the same order as the spin parameters used in the models, such as k, and Z,,,
splitting of the conduction band. Jena and Geldart [32] particularly for materials such as Heusler alloys that
have examined this region using perturbation tech- might be expected to have a fairly complicated band
niques, however, and find qualitative agreement with structure and in which the S, p ions are constituents
the Daniel-Friedel results. of the lattice. This problem has been pointed out by
Campbell and Blandin [25]. Because of the relatively
3.5 COMPARISONS AND APPLICABILITY OF THE MODELS. large body of data that exists (e. g. [3]) and because
- Campbell and Vincze [27] and Watson and Ben- models that can qualitatively account for it exist,
nett [34] have pointed to the ambiguity involved in the we feel that a theoretical effort aimed at characterizing
correlation of impurity hyperfine field values with the electronic energy bands in these alloys can now be
various atomic quantities, since this may simply justified. Experimental work may also be better directed
indicate a relationship between these quantities without towards the determination of properties that would
implying which one is responsible for the basic phy- provide other points of contact for theory e. g. trans-
sical mechanism. For example, the atomic volume of an port properties, specific heats, spectroscopicproperties.
 4. Some recent experimental results : general Examples of P(H) distributions derived from l19Sn
remarks. - The remainder of the paper will be devoted and lZ1Sb Mossbauer spectra are shown in figure 6.
to a rather brief review of some of our recent '19Sn The l19Sn spectra are from the alloy Pd,MnIno~,Sno~,
and lZ1sb Mossbauer measurements on the Heusler at 4.2 K with magnetic fields of 0, 3 T and 6 T applied
systems Pd,Mn(In, Sn, Sb) and Pdl,,MnSb. This
work will soon be submitted for publication in more W,Mnin,, Sn,,

complete form so we will give here only those experi-

mental details that are necessary to this discussion.
There is a certain amount of work on Heusler alloys
in the literature that appears to be in error because the
alloys measured were not properly characterized. This
is most important, as many alloys do not form a
Heusler phase at all, or exist only as a mixture of
phases. Mixed alloy systems do not always form good
solid solutions even when the end members are well-
behaved. Consequently, measurements should only be l
-80
, , ,
-LO
, ,
o ' io 80
I I
20 LO 60 ao
reported when there are sound reasons, backed by Veioc~tylmrn/sl H lTeslo1

X-ray and, where possible, neutron diffraction evidence

to indicate that the alloys concerned are single phase
Pd,Mnlna, Sb,,
Heusler structure and, where appropriate, are good
solid solutions. The alloys that we will discuss below I
l
have been characterized in this manner. Details will
be reported [36].

4.1 FITTING THE MOSSBAUER SPECTRA : P(H) DISTRI-

BrmoNs. - The spectra we will discuss are assumed,
in general, to result from nuclei (either l19Sn or lZ1Sb)
that experience a distribution P ( H ) of magnetic hyper-
fine fields H owing to their different local environments I I . , , , , . r
-L0 -20 0 20 L0 10 20 30
in a particular alloy. The forms of these distributions Veloofy Irnrn/sl H ITesla I
were determined from the Mossbauer spectra in the
manner described by Window [37] i. e. the distribution
was approximated by a sum of harmonic components FIG. 6. - Examples of 119Sn and 121Sb Mossbauer spectra
and of the P(H) distributions derived from them as described
with fixed boundary conditions. The number of har- in the text. The solid lines drawn through the spectra are pro-
monics that are ,required to satisfactorily approximate duced by the P(H) distributions shown : a) IlgSn measurements
the distribution depends on its shape, while the number on the alloy PdnMnIn0.9no.6 at 4.2 K. The magnitudes of
that can be meaningfully determined depends on the magnetic fields Ha applied perpendicular to the y-ray direction
are indicated. The two main peaks in the P(H) curve are taken
statistical quality of.the measured spectra. These two to indicate two regions of different magnetic order in the alloy
competing considerations often mean that what should and the type of order in each is deduced from the way in which
be sharp P(H) distributions will contain oscillations the P ( H ) peaks depend on the applied field ;b) 121Sb measure-
whose frequency will be that of the highest harmonic ments on the alloy Pd~MnIno.7Sbo.3at 4.2 K. The magnitudes
of the magnetic fields Ha applied perpendicular to the y-ray
used : care must be taken that such oscillations are not direction are indicated. As for the ll9Sn results in (a) the two
interpreted as genuine features of the P ( H ) distri- peaks in P(H) are associated with regions of different magnetic
bution. order in the sample. The relatively poor statistics of the spec-
trum run in a magnetic field are due to the large source-detector
In determining these P(H) distributions it was distance in this configuration and to the relatively low l2lSb
assumed, in general agreement with measurements content of this sample. While care must be taken in the inter-
made at room temperature when the alloys are para- pretation of P(H) curves derived from spectra of this quality
magnetic, that all similar nuclei in a sample have the we believe that the two P(H) peaks shown are clearly recogni-
zable in the Mossbauer spectrum : this is our criterion in discuss-
same isomer shift and experience zero electric field ing features of P ( H ) curves.
gradient. This latter condition was probably not
rigorously satisfied in some alloys, but the effects of
quadrupole interactions were generally small enough perpendicular to the y-ray direction. The source was
that the P(H) distribution provided a meaningful Ca llgmSnO3 at room temperature. The lZ1Sb
description of the major features of the Mossbauer spectra (Fig. 6b) are from the isoelectronic alloy
spectrum from which it was derived. It should be PdzMnIno.,Sbo~, also at 4.2 K and with the same
emphasized that our interest here will be only in the
broad qualitative features of the spectra and, therefore,
of the P(H) distributions obtained from them.
-
experimental geometry. The source was Ca 121mSn~,
at 80 K. This source had a satellite line of approxi-
mately 9 % of the intensity of the main line and
C6-326 D. C . PRICE, J. D. RUSH, C . E. JOHNSON, M. F. THOMAS AND P. J. WEBSTER

-
shifted by 18.7 mm/s with respect to it. This satellite
line was included in the fitting of the spectra.
Pd,MnSn or Pd2MnSb were added, the structure
changed to antiferromagnetic type 3 A (A 3 A) and
The solid lines drawn through the experimental finally to ferromagnetic (F). As mentioned in section 2,
spectra in figure 6 are those calculated from the the magnetic moments were localized at the Mn sites
corresponding P ( H ) distributions. (see also [45]). The spin directions of the Mn ions for
the A 2 and A 3 A structures are shown in figure 7a
4.2 MAGNETIC HYPERFINE INTERACTION IN 12'sb and b. The magnetic phase, or the proportions of two
MOSSBAUER SPECTRA. - While the magnetic moment
of the ground state of the 12'Sb nucleus is known
(pg = 3.359 0 + 0.000 2 p, [38]), that of the 37.15 keV
excited state is a little uncertain. Ruby [39] showed
that the ratio R of the nuclear g factors of the excited
(I = 712) and ground (i= 512) states is approximately
0.5. Subsequent determinations of R have produced
values of 0.533 [40], 0.520 [41] and recently 0.535 [42].
Because of this apparent uncertainty in the exact value
of R (we are not entirely convinced of the accuracy
of the previous determinations), and because our
interest here will be restricted to composition-depen-
dent changes in the gross features of the spectra, we will
be satisfied to fit our spectra using the value R = 0.50,
bearing in mind that the correct value is almost
certainly higher than this. The dominant effect of a
+
small change in R, to 0.50 AR, is to increase the
value of the magnetic hyperfine field by a factor of
approximately (1 - 3 AR)-l. This must be taken into
account in comparing hyperfine field values reported
in this paper with those already existing in the literature.

5. Mifssbaner and neutron scattering experiments on

Pd,MnIn,-,Sn, and Pd2MnInl-$by. - The experi-
ments described in this section provide an example of
the complementarity of Mossbauer and neutron elastic
scattering measurements in the determination of the
magnetic structures of a series of Heusler alloys. The
alloys studied were the two isoelectronic series
Pd2MnInl-,Shy (0 4 y g 0.5) and Pd,MnInl-,Sn,
(0 X 1) whose end members are antiferromagnetic
Pd,MnIn and the ferromagnet Pd,MnSn (see 1231).
Particularly at the Pd2MnIn end of the series the
structures are known to depend on heat treatment [2] :
we will restrict our attention to alloys that have been
slow-cooled from high temperatures as these show the
largest degree of order in the L2, structure.
The object of this study was to determine the nature
of the magnetic phase change and its dependence on
local stoichiometry in the alloy. The neutron results,
of course, provide the primary magnetic structure
determination. They have been reported [43] and will
also be the subject of a forthcoming publication [36].
We summarize the major conclusions below.
FIG.7. - Schematic illustration of the antiferromagnetic
5.1 NEUTRON DIFFRACTION RESULTS. - The neutron
structures derived from neutron diffraction measurements on
L21 structure alloys in the series P d ~ M n I n l - ~ S nand ~
measurements indicate that two diEerent magnetic P d ~ M n I n l - ~ S bOnly
~ . the Mn ions are shown for clarity :
phase changes occur in each of the two alloy systems a) The f. c. c. antiferromagnetic type 2 structure (A 2) [44] ;
studied. The structures of the phases were identified as b) The f. c. c. antiferromagnetic type 3 A structure (A 3 A) [44] ;
c) The magnetic phase diagram derived for these two alloy
being, in the notation of ter Haar and Lines 1441, anti-
series. The proportion of each phase existing in an alloy is
ferromagnetic type 2 (which will be referred to as A 2) plotted as a function of Nts,sp, the average number of 5s,5p
at the Pd2MnIn end of both series [2], and, as either electrons on a Z site ion.
 MOSSBAUER AND MAGNETIC PROPERTIES O F HEUSLER ALLOYS C6-327

magnetic phases, existing in an alloy were shown to results for the ferromagnetic alloys whilst figure 6
depend on the electron concentration in the alloy, i. e. shows the spectra of alloys of mixed F and A 3 A
isoelectronic alloys from the two series studied had the phases. The P(H) curves in figures 8 and 9 are arranged
same magnetic structure. It is therefore possible to so that results for isoelectronic alloys are on the same
draw an approximate magnetic phase diagram show- horizontal level.
ing the structure of an alloy as a function of the average The main features of the results for the antiferro-
number, NZ, of 5 S, 5 p electrons on a Z site ion, magnetic alloys in zero applied field (Fig. 8) are a broad
and this is done in figure 7c. peak near H = 1 T and a smaller peak near 3 T in the
l19Sn spectra, and a broad peak near H = 0 in the
5.2 MOSSBAUER RESULTS. - The 19Sn Mossbauer lZ1Sbspectra. One observation that can be made is that
results for the series Pd,MnIn,-,Sn, and the Iz1Sb there appears to be no significant difference in the
results for Pd,MnIn, -,Shy are represented by the P(H) results for alloys of A 2 and A 3 A structure, and this is
curves shown in figures 6,8 and 9. hose in figure 8 are consistent with the structures. In both structures
for the antiferromagnetic alloys, figure 9 contains the (see Fig. 7a and b) every co-ordination shell of Mn ions
around a Z site contains equal numbers of moments in
each direction. However, this would lead us to expect
zero hyperfine field at the Z site nuclei in both struc-

'19Sn spectra occurs at -

tures. We are not certain why the main peak of the
1 T rather than at H = 0
but a similar result was obtained for FeSn, [46]. The
results of figure 8 clearly do not preclude the possibility
of a simiIar field being present at the lZ1Sbnuclei in the
antiferromagnetic Pd2MnInl -$by alloys because the
peaks in the P(H) curves are much broader than 1 T.
This peak width may be ascribed to a number of causes
including the use of a relatively small number of
harmonics to describe the P(H) function, the presence
of small unresolved quadrupole interactions, and metal-
lurgical disorder in the alloys. In particular, Mn-In
disorder is known to exist, even in slow-cooled alloys,
at the Pd,MnIn end of both alloy series [2], [36], [43].
FIG. 8. - P ( H ) distributions derived from the 119Sn and l2lSb
Mossbauer spectra of PdzMnInl-zSnz and PdzMnInl-l/Sbv We attribute the smaller peaks near 3 T in the P(H)
alloy at 4.2 K. Results for isoelectronic alloys are on the same at l19Sn nuclei in Pd,MnIn,-,Sn, to the presence of
horizontal level. Curves for various values of magnetic field Ha the disordered B 2 .phase.
applied perpendicular to the y-ray direction are shown : all When an external magnetic field H, is applied, the
solid curves are from spectra measured in the absence of an
applied field. The changes in the peak positions of the 119Sn peak at HO T in the '''Sn P(H) curves
P ( H ) distributions with applied field indicate that these alloys as JHi+ H: 8) as would be expected
are antiferromagnetic, as deduced from neutron diffraction
measuren~ents.
for a cubic antiferrornagnet with the internal field
parallel to the direction of the Mn magnetic moments.
T h e results for the isoelectronic alloys
Pd2MnIno.,Sno., and Pd2MnIno.,Sb, ., given in
figure 6 are of particular interest. The two large
peaks in both the '19sn and lZ1SbP(H) curves indi-
cate the presence of two separate well-defined envi-
ronments for the Z site ions. The magnetic phase
diagram for the systems (Fig. 7c) shows that the F
and A 3 A phases are present in both of these alloys,
and we assign the two P(H) peaks in each case to Z site
ions in F and A 3 A magnetic regions of the crystals.
Comparison with the results of figures 8 and 9 leads us
FIG. 9. - P ( H ) distributions derived from the l 9 S n and to associate the higher field P(H) peak in each case to
121Sb Mossbauer spectra of Pd~MnTnl-*Sns and
PdzMnInl-uSby alloys a t 4.2 K. Results for isoelectronic
nuclei in ferromagnetic regions. The applied field
alloys are on the same horizontal level. Curves for various dependences are consistent with this assignment. The
values of magnetic field Ha applied perpendicular to the y-ray area ratios of the two P(H) peaks in each case appear
direction are shown : all solid curves are from spectra measured to be qualitatively similar to the proportions of the two
in zero applied field. The changes in the peak positions of the
P(H) distributions with applied field indicate that these alloys
phases present in the respective alloys (Fig. 7c).
are ferromagnetic as deduced from neutron diffraction mea- Therefore, when considered together in this way, the
surements. Mossbauer and neutron results appear to indicate that
 C6-328 D. C. PRICE, J. D. RUSH, C. E. JOHNSON, M. F. THOMAS AND P. J. WEBSTER

the A 3 A-F magnetic phase boundary in these alloys the magnetic transition temperature is given in the
depends very sensitively on the average density of molecular field approximation as :
S, p electrons in a region of crystal, and that two phases
can occur in a single sample because of rather small
(probably statistical) fluctuations in this electronic
density within the crystal. The coexistence of two where k, is the Boltzmann constant, y i is the net
magnetic phases cannot be associated with any metal- number of Mn moments in the ithco-ordination shell at
lurgical segregation in the alloy. a radial distance ri from the Mn moment at the origin.
The results for the ferromagnetic alloys in the range The summation in (5) was extended over N = 12 shells,
under consideration (i. e. X ,< 1, y 0.5) are illustrated to include a total of 248 Mn neighbours. The results of
by the P ( H ) curves in figure 9. The curves for both this calculation for four different magnetic structures
l19Sn and 121sbnuclei show one main peak, broadened (F, A 2, A 3 A, A 3 B - see [44]) are shown in
perhaps by quadrupole interactions and disorder figure 10. They indicate that a small decrease in k,
effects. The linear dependence of the peak positions on
applied field indicates that the alloys are fairly isotropic
ferromagnets, and shows that the '19Sn hyperfine field
in the (In, Sn) alloy series is negative (antiparallel to the
magnetization direction) while that at lZ1Sb nuclei
in the (In, Sb) series is positive. The former result agrees
with previously reported results for Pd,MnSn [47]
and both agree qualitatively with the hyperfine field
systematics discussed above (Section 3).

5.3 Drscuss~o~. - The Mossbauer results for both

alloy systems discussed above are qualitatively consis-
tent with the magnetic structures deduced from neutron
diffraction measurements and summarized in figure 7,
although we could not distinguish between the two
FIG. 10. - The magnetic transition temperature Teas a function
antiferromagnetic phases. of the effective value of k~ calculated for the various magnetic
The fact that the magnetic structures of the two structures indicated (see [44]) using the molecular field model
alloy systems can be represented on a phase diagram described in the text and the exchange interaction shown by
the solid line in figure 2. From this it would be predicted that,
such as figure 7c as a unique function of the average S,p with increasing k ~the , magnetic structure would change from
electron density shows that this density plays a crucial A2+A3B +F.
role in determining the nature of the magnetic order
in these systems. The Mossbauer results presented here
indicate further a rather strong sensitivity of the from its value in Pd,MnSn can explain quite complex
magnetic phase to the local conduction electron changes in the magnetic structure. Using these results
density : two well-defined and quite distinct magnetic we would predict a change from A 2 -, A 3 B -, F
structures were observed in a single sample with no order as k, is increased towards its value in Pd,MnSn,
evidence for anything but small statistical fluctuations whereas the observed sequence is A 2 -,A 3 A -, F.
in the electron density throughout the crystal. In view of the approximations involved in the molecular
It is tempting to interpret this as evidence that the field model, the calculated radial dependence of the
dominant exchange mechanism is the long-range effective exchange integral (see Section 2) and trunca-
double resonance process discussed in section 2. tion of the summation in eq. (5) at N = 12 (see [5])
Magnetic phase changes would then occur as a result of we do not regard the difference between the experimen-
expansion or contraction of the J(r) curve (Fig. 2) due tal results and those shown in figure 10 as necessarily
to changes in the value of k,. Darby and Webster [48] indicating the existence of other large exchange pro-
have calculated the most stable magnetic phase for cesses. Neither of the other mechanisms proposed ([10],
various values of k, and of the Mn moment using the [12]) could correct the predicted sequence of structures.
asymptotic Caroli-Blandin [4] expressions for the
double resonance interaction. In a similar way we have 6. Miissbauer measurements on alloys neat the
calculated the magnetic transition temperature for Pd2MnSb composition. - Recent measurements of
various magnetic phases as a function of k, using a the hyperfine fields at the nuclei of 5s, p elements
molecular field approximation and taking the exchange substituted into the Z site in Pd2MnSb give results
integral to be as calculated in section 2 and shown as that are, in most cases, significantly more positive than
the solid line in figure 2, since this particular curve expected on the basis of the models described in
represents the experimental data for Pd,MnSn quite section 3 ([49], [42], [50] [51], 1521, 1531). In order to
well. In terms of this effective exchange integral J(r), examine the applicability of these models in this case we
have made both "'Sn and I2'Sb Mossbauer measure- conduction electron density. In terms of the hyperfine
ments on alloys in the series Pd2MnSbl-,Sn, field models discussed in section 3 this would imply
(0 G X G l), Pd2MnSbl-,In, (0 G y G 0.5) and a strong dependence either of the effective value of k,
Pd, +.MnSb (0 z 1). These measurements indicate or of the conduction band polarization (or both) on the
that the application to Pd2MnSb of the hyperfine field local environment of a Z site ion. Such an interpretation
models discussed in section 3 is not straight forward, is not consistent with simple free electron-like beha-
possibly because of strong deviations from free viour of the conduction electrons.
electron-like behaviour of the conduction electrons.
6.2 ll'Sn MOSSBAUER RESULTS FOR Pd,MnSb, -,Sn,.
6.1 I2'Sb M~~SSBAUER RESULTS FOR Pd2MnSbl -,Sn, -Figure 12 shows the '19Sn Mossbauer results for the
AND Pd2MnSbl -,In,. - P(H) distributions derived alloys in this series whose 12'Sb spectra were represent-
from 121Sb Mossbauer spectra of the L2, structure
alloys Pd,MnSb, -,Sn, and Pd,MnSb, -,In, are shown
in figure 11. Results for isoelectronic alloys from each

FIG. 11. - P(H) distributions derived from the l2lSb Moss-

bauer spectra of Pd~MnSbl-,Iny and PdzMnSbl-zSnz alloys
at 4.2 K. Results for isoelectronic alloys are on the same hori-
zontal level. Curves for various values of magnetic field H a FIG. 12. - P(H) distributions derived from the ll9Sn Moss-
applied perpendicular to the y-ray direction are shown : all bauer spectra of Pd~MnSbl-&% alloys a t 4.2 K. Curves for
solid curves are from spectra measured in zero applied field. various values of magnetic field Ha applied perpendicular to
The changes in'the peak positions of the P(H) distributions the y-ray direction are shown : all solid curves are from spectra
with applied field confirm that these alloys are ferromagnetic measured in zero applied field. The changes in the peak positions
and that the lzlSb hyperfine fields are in the same direction as the of the P(H) distributions with applied field c o n k m that these
alloy magnetization (i. e. are positive). alloys are ferromagnetic and that the ll9Sn hyperfine fields
are negative for large values of x and become positive as x is
decreased.
series are on the same horizontal level. Firstly, it can be
seen that the distributions in isoelectronic alloys are
very similar, which indicates that electronic density is ed in figure 11, and qualitatively similar behaviour is
an important parameter in determining the hyperfine observed. The ll'Sn P(H) distribution changes from a
field, as is expected from the discussion of section 3. The single peak at H -- - 3.7 T in Pd2MnSn through a
second feature of significance is the way in which the series of strongly broadened peaks to a single peak at
probability distribution changes from a single peak H -- + 23 T for Pd,MnSbo~99Sno.,1. The alloys
near H = 20 T for low electron densities ( X --+ 1, whose spectra are shown here (and in Fig. 13) were
y + 0.5) to a single peak near 60 T for pure Pd2MnSb. quenched rapidly from -- 800 OC, but slow-cooling
It is an important point that this change is not achieved gave qualitatively similar results. It may be noted
by a simple shift of the single peak but mainly by the that the P(H) curves derived from 119Sn Mossbauer
probability being redistributed through a series of spectra appear to be better resolved in some cases than
broad, but apparently well-defined, peaks. We interpret those from 12'Sb spectra. This is partly because of a
this as indicating a strong sensitivity of the electronic difference in nuclear properties (excited state lifetime,
spin polarization at a Z site to local fluctuations in the ground and excited state magnetic moments) and
C6-330 D. C. PRICE, 3. D. RUSH, C. E. JOHNSON, M. F. THOMAS AND P. J. WEBSTER

There does not appear to be any evidence for non-

random substitution in more concentrated alloys. A
previous example of the detection of non-random
substitution in solid solutions by hyperfine interaction
measurements has been reported [54].
Figure 14 shows '19Sn spectra f o r t h e
Pd2MnSb,~g,Sno~02 alloy at temperatures of 226 K,
235 K and 240 K. It is apparent from them that the
'19Sn ions are in an environment of reasonable ma-
gnetic homogeneity with a mean Curie temperature
somewhere between 235 and 240 K. We will see
below (Section 6.4) that this is consistent with the
Heusler phase Pd2MnSb,-,Sn, and appears to pre-
clude the possibility that a significant fraction of the Sn
atoms are in an impurity phase.

- 20 0 20 0 10 20 30
Veloc~tylmm/s 1 H l Teslo l

FIG. 13. - 119Sn Mossbauer spectra and P ( H ) distributions

derived from them for PdzMnSbl-&nz alloys in the range
0.01 x 0.10 at 4.2 K. The spectra shown were measured
in zero applied field and the P ( H ) distributions derived from
them are the solid curves. The P ( H ) curves derived from spectra
(not shown) that were measured with fields of 6 T applied per-
pendicular to the y-ray direction indicate that the ll9Sn hyper-
fine fields are mostly positive.

partly because the hyperfine fields at '19Sn nuclei are

normally smaller in magnitude than those observed at
lZ1Sbnuclei.
The composition region near Pd,MnSb is examined
in more detail in figure 13. On the basis of the compo-
sition dependence observed we would associate the
peak in the P(H) distribution at -- + 23 T with 'I9Sn
nuclei that are isolated from the effects of other Sn ions FIG. 14. - 119Sn Mossbauer spectra for the PdzMnSbo.98Sn0.0~
alloy at 226 K, 235 K and 240 K showing the collapse of the
in the Pd2MnSb host. This is in general agreement with spectra from well split, although poorly resolved, magnetic
the results of Boolchand et al. 1521, but is in marked hypefine structure to a narrow single line within a small tem-
disagreement with a previously reported measurement perature range.
of Campbell and Leiper 1211.
If the interpretation given above for the lZ1Sbresults
is applied to the results of figure 13 it would seem 6 . 3 l 2 Sb MOSSBAUER MEASUREMENTS ON
necessary to further conclude that, for concentrations Pd,+,MnSb. - Hyperfine field probability distribu-
of the order of 2 % or less, tin ions do not substitute tions derived from the lZ1Sb Mossbauer spectra of
for Sb ions in a random manner, although it is not Pd, +, MnSb alloys (for 0 < z < 1) are shown in fi-
clear whether this is due to an effective attraction or gure 15 along with those for the isoelectronic series
repulsion between Sn ions. If the interaction was Nil +,MnSb. Similar results have been reported earlier
attractive it would need to be of sufficiently short range by Swartzendruber and Evans [22], [33]. Our intention
to be virtually ineffective in Pd2MnSbo.99Sno.ol, is to point out the differences in the results for the two
whereas if it was repulsive then the broad multi- systems and the qualitative similarity of the
peaked nature of the P(H) distribution for lI9Sn Pd, +,MnSb results with those from the Pd,Mn(Sb, Sn)
in Pd2MnSbo.g,Sno.02would imply that the hyperfine and Pd,Mn(Sb, In) series shown in figure 11. In making
field at a 'I9Sn nucleus can be quite strongly affected this latter comparison we do not wish to make a
by a fairly distant Sn neighbour. However, these results detailed examination of the apparently equivalent
could also be interpreted as implying a concentration- effects of removing Pd atoms from Pd,MnSb and
dependent non-randomness in the distribution of tin substituting Sn or In atoms for Sb, but simply to draw
atoms in the Z sites that may originate in band- attention to the qualitatively similar way in which the
structure effects which we believe may also be respon- 12'Sb hyperfine field appears to depend quite strongly
sible for the large hyperfine field variations observed. on local atomic environment. This can again be inter-
preted to indicate that the spin polarization at a I2'Sb
site in alloys near to the composition Pd2MnSb is very
sensitive to local electron density fluctuations. It is not
clear how such a sensitivity could be explained in terms
of the usual assumption (e. g. [17], [10]) that Pd
(and Ni) atoms contribute only about 0.1 electrons to
a free electron-like conduction band in the alloy. The
results for lZ1Sb in Ni,+,MnSb (Fig. 15) show that

FIG. 16. - The Curie temperatures of ferromagnetic alloys

in the two series Pd2MnIn1-~Sn~ and P d ~ M n S b l - ~ Splotted
n~
as a function of NFs,Sp,the average number of Ss, 5p electrons
on a Z site ion.

this quantity for Pd2MnSn and Pd,MnSb and attribut-

ed the striking variation )) between the two to conduc-
tion band differences.

6.5 DISCUSSION. - It is clear from the broad,

multiple-peaked nature of the P(H) curves in figures 11,
12, 13 and 15 that in alloys near the Pd2MnSb compo-
FIG. 15. - P(H) distributions derived from the 121Sb MOSS-
sition the hyperfine field at a Z site nucleus depends
bauer spectra of Pdl+sMnSb and Nil+=MnSballoys at 4.2 K. strongly on its local environment. We have interpreted
The solid curves are from spectra measured in zero applied this to indicate a strong dependence of the local
field while the dotted curves are from spectra measured when electronic structure on environment. This is in marked
a field of 6 T was applied perpendicular to the y-ray direction. contrast to the results presented in section 5 which
These confirm that the alloys are ferromagnetic and indicate
that the 121Sb hyperfine fields are positive. could be described in terms of the average electronic
density. We suggest that the difference here is either
that there are important structural features in the
conduction band near to the Fermi surface, or that the
most 121Sb nuclei experience a hyperfine field that is
nature of the electronic screening of the Pd ion is
essentially independent of Ni concentration, which is in
strongly dependent on local environment. In this
better agreement with this assumption. The appearance
context it may be noted that the NMR results of Endo
of a subsidiary peak in the l2'Sb P(H) curve for
et al. [56] for the system Nil -,Cu,MnSb indicate 12'Sb
z >, 0.5 has been attributed to Mn-Ni disorder [49].
P(H) distributions that are remarkably similar to those
6.4 CURIE TEMPERATURES. - Figure 16 shows reported here for x 5 0.5 (see also [57]).
the Curie temperatures [36], [43] for the ferroma- Finally, it may be noted that the results presented in
gnetic alloys in the series Pd,MnInl-,Sn, and this section cannot be reconciled in any reasonable way
Pd2MnSbl-,Sn, plotted as a function of N5Zs,sp, with a model for the origin of the hyperfine field that
the average number of 5s, 5p electrons on a Z site depends largely on atomic volume effects, such as that
ion. For smaller values of Ns",,5,the Curie temperature proposed by Stearns [24]. Consequently we assume that
increases gradually and then flattens out with incre- electronic effects are important in determining the
asing number of electrons in a way that is qualita- hypedne fields at non-magnetic ions in Heusler alloys.
tively similar to that predicted by the molecular field
model (Fig. 10). However, as Pd2MnSb is appro- 7. Conclusions. - We have tried to summarize the
ached the Curie temperature starts to increase quite present level of understanding of the magnetic proper-
rapidly. This behaviour cannot be explained in terms ties of the X,MnZ Heusler alloys and of the origins
of the simple molecular field model unless the occu- of the hyperfine interactions at Z site nuclei. It is mainly
pied part of the conduction band is changing signi- based on experimental measurements and on pheno-
ficantly near Pd,MnSb, as has been suggested above. menological models derived from them. It is our feeling
Another point of interest concerns the pressure that there is now sufficient information available to
dependence of T,. Austin and Mishra 1551 measured justify more detailed theoretical work being undertaken
C6-332 D. C. PRICE, J. D. RUSH, C. E. JOHNSON, M. F. THOMAS AND P. J. WEBSTER

and we believe that further significant advances in while than further testing of the fine numerical detaiIs
understanding will require a more rigorous theoretical of these models.
basis with which to compare experimental results.
For reasons given in sections 2 and 3, we believe that Acknowledgments. - We wish to thank P. Jena
the continuation of carefully planned experimental for interesting discussions.
work designed to test the fundamental principles of JDR wishes to acknowledge the receipt of a Research
phenomenological models is likely to be more worth Studentship from the Science Research Council.

References

[l] WEBSTER, P. J., Contemp. Phys. 10 (1969) 559. DO] LEDANGKHOI,VEILLET, P. and CAMPBELL, I. A., J. Phys. F.
121 WEBSTER, P. J. and TEBBLE, R. S., Phil. Mug. 16 (1967) 347. 5 (1975) 2184.
[3] CAMPBELL, C. C. M., J. Phys. F. 5 (1975) 1931. [31] SONDHI,I., J. Chem. Phys. 62 (1975) 1385.
[4] CAROLI,B. and BLANDIN, A., J. Phys. & Chem. Solids 27 [32] JENA,P. and GELDART, D. J. W., private communication,
(1966)
. . 503. to be published.
[5] GELDART, D. J. W. and GANGULY, P., Phys. Rev. B 1 (1970) [33] SWARTZENDRUBER, L. J. and EVANS,B. J., A. I. P. Conf.
3101. Proc. no 5 (1971) 539.
[6] BROOKS, J. S. and WILLIAMS, J. M., J. Phys. F. 4 (1974) 2033. [34] WATSON, R. E. and BENNETT, L. H., preprint, Brookhaven
[7] CAROLI, B., J. Phys. & Chem. Solids28 (1967) 1427. National Laboratory no BNL-21690 (1976).
[S] ANDERSON, P. W., Phys. Rev. 124 (1961) 41. [35] KOENIG,C., J. Phys. F. 3 (1973) 1497.
[g] ISHIKAWA, Y. and NODA,Y., Solid State Commun. 15 (1974) [36] WEBSTER, P. J. and RAMADAN, M. R. I., to be published.
833. [37] WINDOW,B., J. Phys. E. 4 (1971) 401.
ISHIKAWA, Y. and NODA,Y., A. I. P. Conf. Proc. no 24 (1975) [381 See Mossbauer Effect Data Index, eds. J. G. Stevens and
145. V. E. Stevens (Plenum, New York) 1973.
NODA,Y. and ISHIKAWA, Y., J. Phys. SOC.Japan 40 (1976) [39] RUBY,S. L., Miissbauer Effect Methodology 3 (1967) 203.
690. [40] RUBY,S. L. and KALVIUS, G. M., Phys. Rev. 155 (1967) 353.
[l01 KASUYA, T., Solid State Commun. 15 (1974) 1119. RUBY,S. L. and JOHNSON, C. E., Phys. Lett. 26A (1967) 60.
[l11 CAMPBELL, C. C. M. and STAGER, C. V., Can. J. Phys. [41] IKRAEV, S. M., VALIEV,K. K., GOLOVNIN, V. A. and
(to be published). KUZ'MM,R. N., Zh. Eksp. Teor. Fiz. 59 (1970) 419
STAGER, C. V. and CAMPBELL, C. C. M., Can. J. Phys. (Sov. Phys.-J. E. T.P. 32 (1971) 229).
(to be published). [42] LANGOUCHE, G., DIXON,N. S., MAHMUD, Y., TRIPLETT, B.
[l21 CAMPBELL, C. C. M., private communication, to be publish- B., HANNA, S. S. and BOOLCHAND, P., Phys. Rev. C 13
ed. (1976) 2589.
[l31 GOODENOUGH, J. B., Magnetism and the Chemical Bond 1431 RAMADAN, M. R. I., Ph. D. Thesis, University of Salford,
(Wiley, New York) 1963. U. K. (1975).
[l41 MORIYA,T., in Theory of Magnetism in Transition Metals, 1441 TERHAAR,D. and LINES,M. E., Philos. Trans. R. Soc. 254A
ed. W. Marshall (Academic, New York) 1967. (1962) 521.
[l51 KIM,D. J. and NAGAOKA, Y., Prog. Theor. Phys. 30 (1963) 1451 NODA,Y. and IsHIKAWA,Y., J. P h p . Soc. Japan 40 (1976)
743. 699.
[l61 ALLOUL,H., J. Phys. F. 4 (1974) 1501. [46] TRUMPY,G., BOTH, E., DJEGA-MARIADASSOU, C. and
1171 JENA,P. and GELDART, D. J. W., Solid State Commun. 15 LECOCQ, P,, Phys. Rev. B 2 (1970) 3477.
(1974) 139. [47] GELDART, D. J. W., CAMPBELL, C. C. M., POTHIER, P. J.
[l81 HEEGER, A. J., in Solid State Physics eds. H . Ehrenreich, and LEIPER,W., Can. J. Phys. 50 (1972) 206.
F. Seitz and D. Turnbull (Academic, New York) 1969, [48] DARBY,M. I. and WEBSTER, P. J., in 21st Annual Confe-
Vol. 23. rence on Magnetism and Magnetic Materials, Phila-
[l91 GRUNER, G., Adv. Phys. 23 (1974) 941. delphia, Dec. 1975 (to be published in A. I. P. Conf.
[20] LEIPER,W. and CAMPBELL, C. C. M., reported in Perspec- Proc.).
tives in Mossba~ler Spectroscopy, eds. S . G. Cohen [49] SWARTZENDRUBER, L. J. and EVANS,B. J., Phys. Lett. 38A
and M. Pasternak (Plenum, New York) 1973. (1972) 511.
[21] CAMPBELL, C. C. M. and LEIPER,W., A. I. P. Conf. Proc. [50] BOOLCHAND, P., TENHOVER, M., JHA,S., LANGOUCHE, G.,
no 18 (1973) 319. TRIPLETT, B. B., HANNA, S. S. and JENA,P., Phys. Lett.
[22] SWARTZENDRUBER, L. J. and EVANS,B. J., A. I. P. Conf. 54A (1975) 293.
Proc. no 10 (1972) 1369. [51] MACKAY, G. R., BLAAUW, C. and LEIPER,W., J. Phys. F. 5
[23] DANIEL,E. and FRIEDEL,J., J. Phys. & Chem. Solids 24 (1975) L 166.
(1963) 1601. [52] BOOLCHAND, P., TENHOVER, M., JHA, S., LANGOUCHE, G.,
(241 STEARNS, M. B., Phys. Lett. 34A (1971) 146. TRIPLEIT, B. B. and HANNA,S. S., Solid State Commun.
STEARNS, M. B., Phys. Rev. B 4 (1971) 4081. (to be published).
STEARNS, M. B., Phys. Rev. B 8 (1973) 4383. [53] CAMPBELL, C. C. M. and CAMERON, J. A., J. Phys. F. 6
[25] BLANDIN, A. and CAMPBELL, I. A., Phys. Rev. Lett. 31 (1973) (1976) L 221.
51. [54] LE DANGKHOI, VEILLET,P. and CAMPBELL, I. A., J. Phys.
CAMPBELL, I. A. and BLANDIN,A., J. Mag. Mug. Mat. 1 F.4 (1974) 2310.
(1975) 1. [55] AUSTIN,I. G. and MISHRA,P.K., Phil. Mug. 15 (1967) 529.
[26] STEARNS, M. B., private communication to W. Leiper. [56] ENDO,K., SHINOGI, A. and -A, R., J. Phys. Soc. Japan
[27] CAMPBELL, I. A. and VINCZE, I., Phys. Rev. B 13 (1976) 4178. 34 (1973) 268.
[28] STEARNS, M. B., Phys. Rev. B 13 (1976) 4180. [57] SWARTZENDRUBER, L. J. and EVANS, B. J., J. Phys. Colloq. 35
[29] MOLLER, H. S., Solid State Commun. 8 (1970) 527. (1974) C 6-265.