Review Article

phys. stat. sol. (b) 88, 13 (1978)

Subject classification: 18.2 and 18.4; 18.1; 18.3; 21.1

Univereity of C a l q o r i h , Los AwgeZesl)

Exchange Interactions and the Coupled Spin-Lattice

Phase Transitions in Manganese Pnictides
BY
K. B ~ N E R ~ )

Contents
1. Zntroduction

2. The first- and second-order phase tmneitions

2.1 The first-order phase transition in MnAs
2.2 The first-order phase transition in MnBi
2.3 The second-order phase transitions
2.4 Microscopic model interactions for the coupled spin-lattice phase transitions

3. Exchange interaction8
3.1 Exchange interactions in high-spin compounds
3.2 Exchange in low-spin cornpounds

1. Introduction
Any model of the coupling of spin moments in MnV-type Heusler alloys has t o
account for their nearly metallic conductivity (o = 3 x 109 f2-l cm-l) as well. On
that basis, conventional superexchange (SE) couplings are to be ruled out since they
do not provide charge transport. However, a spin-polarized energy band description
cannot be applied either, since strong d-d electron correlations and pd interactions are
obkerved [ 13.
In this work experimental evidence is presented that the magnetic behaviour has
to be interpreted in terms of (triangular) resonant configurations 3dn-l-3dn-3dn+l
constructed from localized d- and extended p- or d-type states. This approach suggests
a hybrid model for manganese pnictides similar to that proposed by Koiller and Fali-
cov [2] for transiti6n metal chalcogenides where 3dn-l-3dn+l resonant states can be
constructed using localized 3d and extended s-type band states. With extended
resonant states or dissociated resonant electron-hole pairs, electronic charge transport,
i.e. a high electrical conductivity can result. The spin moments and magnetic ordering
temperatures Toi should differ from those predicted from superexchange or spin-polariz-
l) Lo8 Angeles, California 90024, USA.
2, Permanent address: IV. Physikalisches Institut der Universitiit Gottingen, BRD.
14 K. BLRNER

ed energy band descriptions. However, empirically they will be deduced from the
temperature dependence of the susceptibility x( T)and the magnetization M ( T)of the
stoichiometric compounds and of the mixed crystals of the quasi-binary systems MnP-
MnAs-MnSb-MnBi using a molecular field (MF) model. As even the applicability of
a M F model is not evident the mutual influence of the lattice and magnetic phase
transitions has t o be known. Therefore, a short discussion of three typical cases is
presented before the magnetic data is analyzed.

2. The First- and Second-Order Phase Transitions

2.1 The first-order phase transition in MnAs
At the first-order ferromagnetic (fm)-paramagnetic (pm) coupled spin-lattice
transition in MnAs (T,= 378 K) the lattice symmetry changes (B8, + B31) and with
it the ionic distances (a, = 5.724 A, b, = 3.676 A, c, = 6.379 A at 328 K) + (a, = 5.72 A,
b, = 3.72 A, c, = 6.44 A at 313 K) [3] and the parameters of the spin system, the
magnetic moment per molecule pa: 3 . 4 + ~ 2~ . 6 ~[4,~51, the magnetic ordering
temperature To: +PO0 K.+ -220 K [6]. The latent heat L = 1.8 cal/g [7], i.e. the
driving force of the transition arises almost completely from the destruction of spin
order connected with the change in To [7,8]; that is why the cooperative transition
has been treated in terms of a volume magnetostrictive effect [8, 91. However, the
highly anisotropic volume change (Aa,/a, 5 + 7 x 10-4, Ab,/b, = -12 x lo+,
Aco/co= -10 x lods) suggests that there exists a quasi-dipolar interaction which
aligns the individual MnX, (X anion) slightly distorted octahedra (point group of
cation D3h + D2h). A significant contribution to the internal energy u, per MnX,
complex should arise from a pne2 hybridization. This is suggested from the changes
of the spin system parameters at T,,:
1) A high-spin (hs)-low-spin (1s) transition without orbital angular momentum
<Lz)= 0 would be 4,uB + 2pB. I n terms of pure crystal field states: (e# (a# (e2)2+
+ (e!33 (a:)l. If the ei orbital is filled a t the expense of el the crystal field splitting
energy ECFor the intra-atomic exchange J has been reduced. A reduction in J(e2e2)
can only be due to a pne2mixing, since cationic e2 orbits mix with pn anionic orbitals
only (ezn)in octahedral symmetry point group 0,.
2) The change in Tois an exchange inversion; a pair of localized low-spin Mn3+
cations would have antiferromagnetic (afm) SE coupling with a large eX-p&g contri-
bution [lo].
>
For T T,(fm state) pa and Toare constant and can be determined simultaneously
from M(T)using a M F model: pa= 3.4,uB, To = 400 K [6,8]. For T, T < < T,=
= 400 K this is not possible because the pte! hybridization varies with the tempera-
ature T [ll]. It is again possible from x-l( T)in the high-temperature B8, phase, i.e.
for T > T,= 400 K : peff = 4.95pB, 0, = 280 K [12] (compare with Section 2.3).

2.2 The first-order phase transition in MnBi

MnBi also exhibits a first-order pm + fm phase transition. At T, = 633 K 10 to
15% of the cations change from their normal to a neighbouring trigonal-bipyramidal
(TB) interstitial site (Mn*) [ 13, 141 (Fig. 1). This cation displacement does not involve
a change in symmetry (B8, + B8,, point group of cations D3h + D3h) [13] but dis-
continuous changes in the ionic distances (a= 4.305 A, c = 6.118 A) + ( a = 4.34 A,
c = 5.96 A) [13, 141 and in pa: 3 . 9 5 + ~ ~ to 171, To:730 K + 460 K.
~ ~2 . 5 , [15
I n this case the latent heat arises from the change in configuration entropy, spin
entropy, and the entropy change due to the modified phonon spectrum whereby the
 Exchange Interactions and the Coupled Spin-Lattice Phase Transitions 16
Fig. 1. Hexagonal unit cell (B8,);1,2,3 exchange paths between
Mn pairs on normal sites; Cmajor exchange path when interstitial
cations (Mn*) are present. Hatched circles: normal Mn sites,
empty circles: anion sites, double circle: TB interstice
t
i
relative proportions have not been estimated as yet. A cooperative transition is
possible for the fm + pm transition only: the cations may simultaneously occupy
positions in one of the TB interstitial rows. The pm + fm transition probably occurs
by diffusion.
The internal energy U, has to be compared between a normal-site Mn cation
neighbouring and an empty TB site and vice versa, i.e. MnX,-OX, + OX,-Mn*X,.
U,should contain a significant part due to p-d hybridization, too, because
+
1) A hs-1s transition is observed; for MnX,-Mn*X, pairs 4pB 2pB if 6(e1)’ (a$ X
x (ez)2+ 5(e,)a (a#
on TB sites;
+ +
l(a.,)l (el)a, i.e. if of the cations are optimally distributed

2) a drop of AT, = 270 K is observed between the fm and conical spin arrange-
ment [16, 171; it could arise from an afm coupling &ntribution between low-spin
pairs. For T < T, (fm state)pa = 3 . 9 5 , ~and~ ~To = 730 K [13, 141 can be determined
from N(T )using a M F model; this is true also for x-l( T)of the high-temperature
pm phase: 0, = 440 K. However, pa = 1 . 9 [13] ~ ~ as obtained from M ( T ) of the
quenched high-temperature phase measures only the fm component 1 . 7 3 ~ [ 171
~ of
the conical array. Also, the one ordering Bemperature To = 460 K does not reveal
the two competing exchange couplings.
In order to extract S from pathe gyromagnetic ratio 9-1 has to be known; for MnBi
-
g(293 K) = 2 4 [18] which could indicate rather large <Lz)contributions or an-
isotropic exchange [ 191.

2.3 The second-order phase transitions

The second-order B31 + BS,phase transition a t Tt= 126 K in MnAs presumably
is a soft-mode-type transition [ 111. Since an orthorhombic lattice (B31) introduces
two Mn sublattices in respect to a hexagonal one (B81), the soft phonon mode is of
the “antiferroelectric” type, i.e. it condenses a t the edge of the Brillouin zone (BZ).
Application of the Landau theory of structural phase transitions suggests that between
space groups D&,and Dit a condensation can occur only at MT [20]. If it occurs a t
<
T = Tt for T X t the components of the distortion tensor q, i.e. the deviations of
the ion positions from their value in the B8, phase, exhibit a typical continuous order
parameter variation [21] (Fig. 2,3); the small deviation below the magnetic ordering
temperature T, presumably is of magnetoelastic origin.
At a lower temperature T h l < Ttthere is a gradual ha-1s transition [22] indicating
that a pxe, mixing is involved and that the major reduction of J(e2ea)occurs at another
highly symmetric point of the BZ, say a t I?, where the gradual hybridization changes
<
due to q( T)are less drastic. Aside from pa(T)for T T,,To should change, too; this
is why a Curie-Weiss law is not observed for T, T < < T,;if, however, p,(T) is
known [22], To(T) can be obtainedfrom x-l( T) assuming that a M F model is applicable
in principle; this evaluation is shown in Fig. 4 for the compound MnAso.ozPo.os.
16 K. B ~ N E R

:4

Fig. 2 Fig. 3

Fig. 2. Orthorhombic unit cell (B31; space group Dlt); ei and S, deviations from B8,

Fig, 3. The deviations ei, of the ion positions from their positions in the B8, structure of
MnAs0.92Po.o~below Tt yield the B31 structure [21]. TkB31+B8, transition temperature; TNN b l
temperature; T, ferromagnetic ordering temperature

<
From this tentative analysis it follows that with low spin To 0, i.e. there is a
gradual exchange inversion with inversion temperature T idefined by To = 0. For
MnAs below T,the first-order transition occurs: T, Ti; for MnAso.g~P0.0~ there is
no such transition; however, before the NBel point To = T Nis reached, x-l( T) drops
steeply. As there is no sudden change in S or lattice parameter near this drop (Fig. 3),
it cannot be understood in terms of a direct variation of a coupling strength with
lattice dimensions; it has to be assigned to an electronic effect connected with band
overlap. As will be shown in detail later (Section 3.1) as soon as the low-spin state
(e33 (a:)' is introduced T 5 Tisome mobile e: electrons reappear because of an
increasing band overlap connected with q(T). These electrons establish a competing
double exchange mechanism whose ferromagnetic coupling strength, measured by el,
is proportional to the number n(e$ of e: electrons. Tentatively using a M F model,

t 2uu g
8
I00

-200
a b
C L 300 6 WO 6 500
TIXI -
as a function of temperature (B = 0.8 T).
Fig. 4.a)Reciprocal molar susceptibility of MnAs~.~~Po.os
Thl high-spin-low-spin transition temperature ; TI ferromagnetic ordering temperature. - - - . -
extrapolated %-I( T) using pa(T); - - - extrapolated x-l(T) using p a ( T ) and &(T) from
Fig. 4b. b) Magnetic ordering temperatures Ooi for MnAso.QzPo.08 as functions of temperature re-
constructed from Fig. 4a and pa(T). Of paramagnetic (pm) Curie temperature; ON pm Nbel
temperature; Ti inversion temperature
 Exchange Interactions and the Coupled Spin-Lattice Phase Transitions 17

e,(T) can be calculated from x-l(T) once e,(T) is known; e N ( T )can be obtained
independently by extrapolation and the cusp of x-l(T) a t 8, = T N .Both 8,(T) and
ON( T ) tend to become constant for T ,S T,/2 when q( T)sz const. Since T,/2 lies
near T , the magnetic moment M ( T )for T <
TN should be described in terms of
a constant spin moment S and two constant competing exchange interactions (com-
pare with Section 3.1).

2.4 MicroscOpic model interactions for the coupled spin-lattice phase transitions

It has been shown empirically that the magnetic phase transitions are intimately
connected to those of the lattice, which indicates a strong coupling between lattice
and spin system. Certainly one cannot explain these results in terms of the relatively
weak spin-orbit coupling. When in rare-earth compounds similar phase transitions
have been observed [23] they have been attributed to an Anderson mixing inter-
action [23,24] between the quasi-free s electrons and the localized Pf electrons. In
general, these sf interactions involve both the spin and orbital part of the 4f state
and thus potentially allow for coupled spin-lattice phase transitions, provided the
4f state in turn is coupled to the motion of the nuclei. The latter coupling is normally
defined as an ion-lattice interaction without any hybridization (Jahn-Teller effect)
[23]. It is always present to some extent as long as the atoms are partly ionic.
For manganese pnictides we suggest that p d mixing interactions are the cause of
the coupled spin-lattice phase transitions. However, in contrast to s-f interactions
there are no empty extended p states available. Thus, in addition, one has to assume
that during the time where the valence band p electron occupies a state of the open
3d shell, the corresponding hole is bound to the cation site by Coulomb forces. E r s t
[25] has extended the Anderson interaction such that it includes second-order terms
and can be written as a sum of multipolar interactions in the LSJ coupling limit
(spin-orbit coupling is not included). When pure crystal field states can be assumed
and when the total spin momentum S is conserved, for each base function g a classi-
fication is possible using the irreducible representations A( of the point group of the
cation site and the rank (0 tor 1)of the spin tensor. For a single 3d4 ion in an hexagonal
environment the largest coupling constant I involving spin is

where V d k is an interaction potential, k the wave vector of the extended state, and E*
the energies of intermediate configurations of the ion, corresponding to one electron
added or removed. This notation should be equivalent to that for the bonding d
orbitals of a Mn8+cation a t the normal site of the B8, lattice (point I? in the BZ [26]).
However, the pxe2mixing interaction involved in the first-order phase transition in
MnAs cannot be completely described in these terms because of the hs-1s change
(S is not conserved). For this type of transition the formalism has to be extended
...
such as to include discrete changes A S = 1, ,S. Aside from that, a sublattice of
3d4 cations should allow for a larger number of interaction types. However, since
strong mixing does not occur for all k,one might restrict this number again by con-
sidering interactions at points of high symmetry in the BZ only. For example, the
interaction involved a t the displacive phase transition at T , should be of the type
(9,9,A, 1)where A is an irreducible representation and g a base function a t M as the
displacement occurs at M and S does not change [22].
2 physica (b) 88/1
18 K. B ~ N E R

3. Exchange Interactions

3.1 Exchange interactions in high-spin compounds

Empirically a M F model can be used to determine pa, Toin certain temperature
intervals which are defined by phase transition temperatures. However, the magnetic
coupling is not necessarily via superexchange (SE). Following the Goodenough-
Kanamori (GK) rules, the magnitude of conventional SE for a localized high-spin
3d4-X-3d4 octahedral configuration (Jab 2 0, weak) does not account for the strong
fm coupling observed (for example, To(MnBi)= 730 K). This is also true for the
distorted octahedral configuration of the B8, lattice (135" e,-pa-el, a,-%, or 90"
al-pn-a, interactions between half-filled orbitals are all negative).
In an attempt to explain strong fm coupling in insulating 3d4 compounds, Good-
enough proposed to include the coupling of the cation state to the lattice vibrations [ 101.
If X vibrates between the two cations, because of the virtual electron transfer the
average configuration resembles 3d3-X-3d6 rather than 3d4-X-3d4 (quasi-static limit).
Strong fm coupling follows as a 180" el-pa-el interaction between full and empty el
orbitals is strong and positive. Because of the participation of phonons this coupling
should change sign or become weaker for T + 0. Anomalies in the magnetic behaviour
of both MnSb and MnBi are indeed observed for T <
50 K [13,27] but not an ex-
change inversion or a reduced fm coupling. Instead weak afm neutron reflections
appear in addition to the set from simple fm order, indicating that there exist a t
least two qualitatively different exchange mechanisms although there is only one
magnetic sublattice.
The dominating fm coupling presumably is a double exchange (DE) mechanism.
It is well known that if the transfer integral b for one orbital overlap, say el-p,-el
between mixed valence Mn cations is large enough, the cation pair is orbitally degen-
erate in respect to this transfer and a resonant state is introduced with an energy
splitting of E = 2b1, cos (8/2) where 8 Q (Sa,S,) and cos (812) appears because of the
use of spinor states [28 to 301.
Since 3d4 compounds have no mixed valence pairs, a resonant state involving a t
e,
least three cations 3d6-3d4-3ds with S = f, f has to be assumed whereby electron

0 0 \

a w f tMn
\ 1M"

C
Fig. 5. a) (p,,eJs and b)( p+,)J resonant states
and c) a,-pra, indirect exchange
0 /
Fig. 6. Energetically degenerate spin arrange-
ments when antiferromagnetic superexchange
is superimposed to double exchange in anti-
ferromagnetic layer structures
 Exchange Interactions and the Coupled Spin-Lattice Phase Transitions 19

and hole orbit simultaneously. In particular, we propose a (p0e1)3resonant orbital, i.e.

fcf. Fie. 51
V I

(ell2 (a1
(ez)"(edl (ad1(e2)2-(al)1 A*
If the resonance were restricted to a triangular array one would have a bound state;
however, the resonances touch in the hexagonal lattice and one will have stationary
solutions with the electron-hole pair travelling like an exciton. Also, dissociation
might occur leaving both particles in extended Bloch states without the loss of
polarization or local spin moment alignment. Given a dissociation at the two electrodes
or a dissociation energy ED == k T an array of linked resonances is always conducting.
This is different from the classical (La3+Mna+O:-) - (Ca2+Mn4+O!-) system where
there are percolation thresholds (for example, x x 0.15 Mn4+from p (a-phase) [ 10, 311).
DE alone mostly yields fm coupling (0 = 0) whose strength is proportional to b
and the number of transfer electrons. I n competition with SE, fm or complex spin
orders may occur (Fig. 6) depending on the sign and magnitude of the SE integral.
The simultaneous presence of a positive SE coupling turning negative a t low temper-
atures (exchange inversion) might explain the canted spin orders in MnSb, MnBi and
the conical spin order in the quenched high-temperature MnBi phase. This SE coupling
would involve p, and the comparatively localized e2,a1 states only.
Whereas pa (MnSb) and pa(MnBi) are close to 4pB, pa (MnAs) = 3.4pB is too low
to be explained by this particular resonant state. Besides, the variation of pa, To with
composition x of the mixed crystal series MnAsl-,Sb, indicates that for x 2 0.5
another resonant state is introduced. Fig. 7 shows aTo/aa and aT0/ac as obtained
from measurements of To(x),a(x),c(x) [6, 14, 32 to 361 assuming
dT, - aTo aa
dx aa ax
+-aTo -.ac
ax ax
The cusps near &As and MnSb are due to the deformation of the p valence band
connected with the anion (X) disorder [37]. Agreement is obtained when aT,,/aa =
= const but aTo/ac changes sign rather abruptly, i.e. another exchange type prevails
below x x 0.5; also, an unusual strong pressure dependence of Toand strong magneto-
elastic effects are now observed [33 to 35, 38 to 401. Since T,(MnAs) = +400 K, this
<
cannot be ane,-p,-e,SE,coupling with Jab Oas suggested by the pxe2mixing. Instead
we propose a mixed triangular (pnel)8-(p&2)aresonant state :
(el)' (+d1(e2)3-(e1)1(~l)1(e2)2-(~)1(e,)eL\ -
A resonant exchange of one e2(f)electron together with an
+ +
el(?) hole yields a - - a s in configuration, i.e. Fa =
a p.
= 10/3PB = 3 . 3 and ~ ~ a low-spin state is preformed. 4 8 0 ,
since onp, orbital points towards the centre of a triangle
containing three e, orbitals (Fig. 5) this resonant.state should
depend very sensitively on c/4, as is observed. However,

Fig. 7. The derivatives aBf/aa (1) and BB& (2) (in 1V KIA) as ob-
tained from i30f/az (6), aa/ax (4), ac/ax (3) for MnAI-,B, mixed
crystals (A, B = P, As, Sb, Bi); -experiment, - - - re-
constliuction MnAs 05 MnSb
P- xibt%)-- -Bi
2.
20 K. B ~ N E R

this does not explain why the change happens at all; the cause might simply be the
variation of the overlap integral Al, for the el-pa-el path with the anion-cation
distance. Bethe [41]has suggested very early that Aij goes to zero for both very small
and very large distances. This is true also for the transfer integral bcj since bij = EAij
[42] where E is the one-electron energy, i.e. the energy splitting e = 2bIa is severely
reduced. If bza >
blo, the transfer integral of the mixed resonance and with it E is
enhanced. The transition also shows that the (~,,e,)~ resonance can be emptied without
losing the property of spin coupling. This should hold also when only a fraction of
el electrons is available per molecule.

3.2 Exchange In low-spin compounds

3.2.1 Orthorhombic crystals

The spin moment pa of orthorhombic MnAsl-,P, compounds with 0.03 x 0.15 < <
is near the low-spin value of 2pBs With point group Dzh the ideal low-spin state would
be (e",)z(egz)l(ay)l[43]. Since in that case the e; orbitals are empty and the e; electrons
localized at the cation sites there should be no resonant states involving (pae33 or
(pxeg)3. Negative SE couplings from localized d orbitals is left (eg-p,,-e; and a-:. or
ay-p,,-ay) (Fig. 5 ) ) which should give afm order.
Experimentally it is found that the low-spin magnetic sublattice exhibits canted
spin orders (Fig. 2, 6) [21, 441. Whereas small cant angles could be due to anisotropic
superexchange, the large cant angles observed (for example, 76" in MnAso.gzPo.08)
indicate at least one fm and one afm exchange mechanism. Thus, we conclude that
some ey states are always occupied in MnAsl-,P, compounds and one can only ap-
proach the ideal low-spin state (this is occupation by band overlap in terms of an
uncorrelated energy band description). I n particular, an emptied (pae!)3-(p,,eX2)3
resonant state is proposed :
R, m integer
n(e~)3(a~)'(e~)1-m(eg)sO'-lz(eg)z(a~)1; ,
with bola or b02, >
bl, whereas the electron concentration n(e'$ is less. For n = m = 1
this yields a+-+-+ spin configuration, i.e. pa = = 2.@B and n(e:) = 0.3 per
molecule which well corresponds to pa = 2 . 6 for ~ ~x = 0.08 [21].
The competing afm exchange has been termed SE and has been estimated correctly
so far using the GK rules. However, as the valence band states are extended, this
coupling should be assigned to resonant states such as (p$ay)B or (p:egl)3 involving
valence band holes.
I n a first attempt to describe the observed properties one would decompose the
system into quasi-free ey electrons and quasi-localized e;, a! electrons. This would yield
n S = local spins in between (n +
m) S = 1 spins and a fm DE due to n mobile ey
electrons competing with an afm SE due to the localized electrons.

Fig. 8. Magnetic moment and reciprocal susceptibilityas func-

tions of temperature for MnAs0.92Po.o~( B = 0.36 T),
0 50 4 100
TK-- -
750 i; 100 6250 - _ _ according to theory
 Exchange Interactions and the Coupled Spin-Lattice Phase Transitions 21

-X
X-
Fig. 9. a) Magnetic phase diagram of MnAsl-,P, mixed crystals with z > 0.03 at%P for zero
magnetic field (- ) and for B = 0.8 T (- - -), Tc,T,, TI,TK are magnetic ordering temper-
atures; b) Calculated variation of the magnetic ordering temperatures Tc, TN,TI with concen-
tration of transfer electrons [46]. Tc Curie temperature, TN N6el temperature; T, transition
to a canted spin order

For an asr_--bly of local ion spins S which are coupled by an afrn SE J a b and by
fm DE via n transfer electrons near k = 0, de Gennes [45] has calculated x-l(T),
M ( T ) ,and the dependence of To on the carrier concentraton n. The calculation has
to be done for each crystal structure and spin order separately.
The spin order of orthorhombic MnAso.92Po.os is a layer structure with layers
perpendicular to the a,-axis and the spins in one layer parallel but canted in respect
to the next (Fig. 2). If we apply the model to that spin and lattice structure, several
predictions follow:
1) Two different magnetic ordering temperatures are observed for one magnetic
sublattice Tl(fm), T,(afm) because the thermal averages (cos 6) and (cos (6/2)) are
different [46, 301, i.e. a MF model might be used for T <T, and T >
T N but not for
Tl< T < T,.
2) The cant angle 6 is given by cos (6/2) = bn/(4 lJabl 8 8 ) .
3) There is no saturation in high magnetic fields because of a field-induced variation
of the cant angle 6 [32].
4) Neutron diffraction yields fm and afm spin components simultaneously [21, 221
all of which have been verified experimentally. M ( T ) of &Aso.92Po.os is shown in
Fig. 8. Besides T,, T , there is an additional ordering temperature T g , characterized
by a drop in H(T ) which does not follow from the model; this drop has been connected
with an e! electron localization a t low temperatures involving J(eie!) [l].For T >T,
there is a discrepancy of slope. This has been ascribed to the continuous increase of
eO, electrons for T, < < T T i (compare Section 2.3). From conductivity data there
is evidence that n(eb increases with x for x > 0.03. Thus, the variation of T,(n),
T,(n) as calculated can be compared qualitatively with the “magnetic phase diagram”
of the MnAsl-,P, mixed crystals. The two diagrams (Fig. 9) show similar features
except that there is a triple instead of a quadrupole point and an afm region below
TK(4 -
3.2.2 Cation disordered alloy8
Cation disordered MnBil-,Sb, compounds (disordered B8, structure) with 0 x < <
< 0.5 have magnetic properties which are very similar to those of orthorhombic
< <
MnAsl-,P, crystals with 0.03 x 0.15 [14,16]. The cations which occupy the
TB interstitial sites (Mn*) have low spins and cause low spins a t the neighbouring
normal sites (Mn), i.e. (rtl)l(e2)3-X-(rtl)1(el)s.If all pairs are low-spin, an afm coupling
22 K. BLRNER

Fig. 10. Magnetic moment and reciprocal susceptibility of

cation disordered MnBio.sSb0.z as functions of temperature
(B = 0.23 T)
1
a
%a2
9
r n L
aoi

0
L7 100 7; 200
T(KI --
would result (el-p,-e, interaction between half-filled orbitals via the common trian-
gular face). However, canted and conical spin orders have been found [16]. This
suggests that one has non-ideal low-spin states here, too, i.e. some mobile e-type
electrons exist and allow the construction of an emptied mixed (poe)3-(pxe)3resonance;
for example,
7&)3 (e1)l-m
(s)1 (e1)s(a1)1-n(e2)2(~1)1 -
Again n = m = 1 yields p , = 2-6,uLg corresponding to quenched high-temperature
MnBi (pa= 2.5pB), whereas for MnBio.&bo.l pa = 2 . 0 indicating
~ ~ that n(eJ ;r:0.
I f a canted, spiral, or conical spin order is obtained depends on anisotropy, next
neighbour interactions, etc. [45]. Another supporting argument is provided by M ( T).
For MnBio,sSbo.z T,, T,, and the nearly correct slope of x-l(T) for T T, are >
obtained (Fig. 10).

Acknowledgements
The author would like to thank Prof. Dr. H. U. Harten for his kind support and
Prof. Dr. E. Vogt for many valuable discussions.

References
[l] K. B ~ N E Rphys.
, stat. sol. (b) 84, 385 (1977).
[2] B. KOIILEE and L. M. FALICOV, J. Phys. C 7, 299 (1974).
[3] R. H. WILSONand J. S. KASPER, Acta cryst. 17, 95 (1964).
[a] J. B. GOODENOUGH and J. A. KAFALAS, Phys. Rev. 167, 389 (1967).
M. MENYUK, J. A. KAFALAS, K. DWIQHT, and J. B. WDENOUQH, Phys. Rev. 177,942 (1969).
K. BLRNEB,phys. stat. sol. (a) 5, 405 (1971).
A. J. P. M @
. YER and P. TAGLANG, J. Phys. Radium 1 4 , 8 2 (1953).
C. P. BEANand D. S. RODBELL, Phys. Rev. 128, 104 (1962).
H. WAGNER and J. SWIW,z. Phys. 239, 182 (1970).
J. B. GOODENOUOH, Magnetism and the Chemical Bond, John Wiley & Sons, New York/
London 1963 (pp. 165, 226).
K. B ~ N E RHabilitation,
, Gottingen 1976.
A. SERRES,J. Phys. Radium 8, 146 (1947).
B. W. ROBERTS, Phys. Rev. 104, 607 (1956).
K. AHLBORN, K. B ~ R N E R , W. SCHROTER,
and phys. stat. sol. (a) 30, 251 (1975).
R. R. HEIKES,Phys. Rev. 99, 446 (1955).
A. F.ANDRESEN and J. E.ENOEBRETSEN, Acta cryst. 20,196 (1972).
TU CHEN and W. E. STUTIUS,IEEE Trans. Magnetics 10, 581 (1974).
G. D. ADAMand K. J. STANDLEY, Proc. Phys. SOC.66A, 823 (1953).
NAILI HUANO LIU,K. J. LING, and R. ORBACH,Phys. Rev. B 14,4087 (1976).
H. F. FRANZEN, C. HAAS,and F. JELLINEK, Phys. Rev. B 10, 1248 (1974).
E. L. HALL,L. H. SCHWARTZ, G. P. FELCHER, and D. H. RIDQELEY, J. appl. Phys. 41, 939
(1970).
 Exchange Interactions and the Coupled Spin-Lattice Phase Transitions 23
[22] L.H.SCHWABTZ, E.L. HALL, and G. P. FELCHER, J. appl. Phys. 42, 1621 (1971).
[23] P. FULDE, Z.Phys. B 20, 89 (1975).
[24] P. W.ANDERSON, Phys. Rev. 124,41 (1961).
[25] L. L. HIRST,Z. Phys. 244,230 (1971).
1261 J. M.TYLERand J. L. FRY,Phys. Rev. B 112,4604 (1970).
[27] W. J. TAKEI,D. E. Cox, and G. SHIBANE, Phys. Rev. lZ9, 2008 (1963).
[28] C. ZENER,Phys. Rev. 82,403 (1951).
[29] P.W.ANDERSON and H. HASEGIAWA, Phys. Rev. 100, 675 (1955).
[30] P. W.ANDERSON, Magnetism I, Ed. G. T. RADOand H. S m , Academic Press, New York/
London 1963.
[31] J. H.VANSANTEN and G. H. JONKER, Physica (Utrecht) 16, 599 (1950).
[32] A. WER and R. FRUCHARDT, Mater. Res. Bull. 3, 253 (1968).
[33] W.KOSTERand E. BRAUN, Ann. Phys. (Leipzig) 7, 66 (1959).
[34] J. B. GOODENOUGIH, D. H. RIDGIELEY, and W. A. NEWMILN, Proc. Internat. Cod. Magnetism,
Nottingham 1964, Institute of Physics and the Physical Society, London 1964 (p. 542).
[35] L. R. E D W A ~ and
D S L. C. BARTEL,Phys. Rev. B 3, 1064 (1972).
[36] K.B ~ N E R phys.
, stat. sol. (a) 8, 119 (1971).
[37] B. VELICKY,S. KIELKPATFZCK, and E. H. EHBENREICH, Phys. Rev. 176,747 (1968).
[38] M.DORFLER and K. B~RNER, phys. stat. sol. (a) 17, 141 (1973).
[39] J. E.SCEIRBER and D. H. ANDERSON, J. appl. Phys. 39,1010 (1968).
[40] N. P. GRAZZHDANKINA and Yrr. S. BERSENEV, Soviet Phys. - J. exper. theor. Phys. Z4, 702
(1967).
[41] H.BETHE,Hdb. Phys., Vol. XXIV, Springer-Verlag, 1933 (p. 595).
[42] J. B. GOODENOUGIH, J. appl. Phys. 39,403 (1908).
[43] J. B. GOODENOUGIH, J. appl. Phys. 36, 1083 (1964).
[44] H. BERGI, phys. stat. sol. (a) 40,4559 (1977).
[45] P. G. DE GENNES, Phys. Rev. 118, 141 (1960).

(Received November 7, 1977)