PH VS ICAL REVIEW VOLUM E 129, NUMBER 5 1 MARCH 1963

Magnetic Structures in the MnSb-CrSb System

W. J. TAI KI, D. E. COX, AND G. SIIIRANE
H estistghouse Research Laboratories, Pittsburgh, Peri&ssylvauia
{Received 16 August 1962)

A magnetic and neutron di8raction investigation has been made on MnSb, CrSb, and their solid solutions,
which crystallize in the hexagonal ¹As-type structure. MnSb is ferromagnetic and CrSb antiferromagnetic,
As CrSb is substituted into MnSb, a canted spin structure is formed, the spins lying within ferromagnetic
{001) sheets with the spins in adjacent sheets alternating in direction. The angle of cant increases with
increasing Cr content in agreement with the model of de Gennes who considered the effects of double ex-
change on an antiferromagnetic system. The magnetic structure of CrSb consists of ferromagnetic (001)
planes which are coupled antiferromagnetically. The spin axis is parallel to c. Between this structure and the
canted spin structure with the spins perpendicular to c is an intermediate region. This antiferromagnetic
region has either a tilted spin axis structure or two coexisting structures.
MnSb was found to have a change in the spin direction from a perpendicular to a parallel orientation with
respect to c as the temperature was increased. The temperature of the spin flip was found to depend on the
preparative conditions as did the Curie temperature and saturation moment.

I. INTRODUCTION I,otgering and Gorter' are essentially in agreement with

I 'HK nickel arsenide structure is often found in com- each other and the transition temperatures found by the
pounds of the transition metals with metalloids. ' former authors are shown in Fig. 2(a). Solutions with
This hexagonal structure, shown in I ig. 1, has the metal large Mn contents show a Curie temperature and their
atoms arranged in a simple hexagonal lattice while the saturation moments decrease rapidly with increasing Cr
metalloids, considered alone, form a hexagonal close- content. Compounds with large Cr contents show an

many of these compounds. "

packed arrangement. Magnetic ordering is found in
Of particular interest for
the study of the magnetic interactions are investigations
anomaly in the susceptibility which was attributed to
an antiferromagnetic transition. Solutions near 75%%u& Cr
seem to show both a Curie and a Neel temperature.
of solid solutions between a ferromagnetic and an anti- Several models were proposed for the magnetic struc-
ferromagnetic NiAs-type compound. One such system tures in this system. I,otgering and Gorter~ suggested
which has been studied previously is the solid solution
between ferromagnetic MnSb and antiferromagnetic I l t

CrSb. A neutron diGraction investigation by Snow'

600(
reported that CrSb had ferromagnetic (001) planes with
alternate planes coupled antiferromagnetically, and a O

spin axis parallel to c. Pickart and Nathans' reported 4P

400—
MnSb to have the same spin direction. The results of h
CL
magnetic measurements by Hirone et cl. s and by ~47

200—
I
I
ransihpn
Metal
0 OZS 0.50 Q75 I.o
X in Mnl &Cr&sb Crsb
Metallpi4
fo) (b)
Fxo. 1. Hexagonal
unit cell of NiAs Fxo. 2. Phase diagrams for Mn~ Cr Sb solid solutions. {a) Hirone
structure. et at. (reference 6); {b}de Gennes {reference 10).

that the Mn and Cr ordered independently, the Mn

forming a ferromagnetic system and the Cr an anti-
ferromagnetic one. Hirone and Adachi' made an analysis
of NiAs compounds using the molecular Geld approxima-
~ Present address: Brookhaven National Laboratory, Upton, tion and including three diferent interactions. They
New York.
' A. Cornish, Acta Met. 6, 371 concluded that, according to the relative values of the
' K. J. (1958).
' T. Hirone, J.
Adachi, Phys. Soc. Japan 16, 2187 (1961). interactions, various magnetic structures would be
L,andolt J3orrIstd rI (Springer-verlag, Berlin, 1962), stable. They proposed that in the MnSb-CrSb solid
Anfage 6, Band II.
4
I. Snow, Phys. Rev. SS, 365 (1952).
A. solutions, a "triangular" spin structure with multispin
' S. J. Pickart and R. Nathans, Suppl. J. Appl. Phys. 30, 280S
(1959) ~

s T. Hirone, S. Maeda, I. Tsubokawa, and N.

F. K. Lotgering and E. W. Gorter, J. Phys. Chem. Solids 3, 238
Tsuya, J. Phys. (1957).
Soc. Japan ll, 1083 (1956). T. Hirone and K. Adachi, J. Phys. Soc. Japan 12, 156 (1957).
 MAGN ETIC S I RUCTU RES I N M nSb —CrSb SYSTEM

axes was present. A neutron diGraction study of anti- TABLE I. Lattice parameters fox Mnl, Cr, Sb.
ferromagnetic CrSe by Corliss et al,.' revealed a spin
C
arrangement very similar to this type of structure. (i.) c/a
De Gennes" made a theoretical analysis of the eA'ects
MnSb (annealed) 4.130 5.789 1.402
of double exchange on an antiferromagnetic system. MnSb {quenched) 4.15$ 5 771 1.390
Double exchange is a ferromagnetic interaction which Mnp. 76Crp 26Sb 4.106 5.778 1.407
was postulated by Zener" as arising from the presence of Mnp. 4oCl p 61Sb 4.097 5.697 1.391
Mnp. $oCro. voSb 4.104 5.625 1.371
mobile carriers (electrons or holes). As the carriers Mnp. q6Cro. 76Sb 4.107 5.598 1.363
"hop" from one atom to another, the energy is Mno. 2oCro. 8oSb 4.108 5.584 1.359
minimized when the atomic spins are parallel, i.e., there Mno. 16Crp. 86Sb 4.111 5.560 1.352
CrSb 4.121 5.467 1.327
is ferromagnetic ordering. Anderson and Hasegawa"
showed that the energy varied as cos(O/2), where is 0
the angle between the atomic spins. De Gennes can-
cluded that since the antiferromagnetic exchange energy Nathans' investigation, would not explain these results
varied as coso, the ground-state energy is lowered when either. In an attempt to clarify our understanding of this
the two antiferromagnetic sublattices are canted with system, it was decided to make a more extensive
respect to each other. This wouM result in a ferromag- investigation.
netic component. He proposed a generalized magnetic
phase diagram for such a system as shown in Fig. 2(b)
II. EXPERIMENTAL PROCEDURE
and suggested that it would apply to the Mn& „Cr Sb A. Sample Preparation and X-Ray Examination
system if electron transfer is allowed between the Mn
and Cr atoms. For most compositions, two transition The starting materials were 99.9% Cr from Electro-
temperatures were proposed, a high-temperature Curie metallurgical Associates, 99+% Mn from Belmont
or Neel point and, at lower temperatures, a transition Smelting and Refining Company, and 99.999% Sb from
to a canted structure. This theory has been applied to American Smelting and Rehning Company. After
the magnetic properties of the' La&, Ca, MnO3 and" grinding, stoichiometric amounts of the elements were
.
Li,Mn& Se systems. One difI'erence of this canted
model compared to the triangular model of Hirone and
pressed and sealed in evacuated quartz tubes. The
samples were heated slowly to 650'C, held at this
Adachi is in the periodicity of the spin structure. temperature for 1 hr, then annealed at a temperature
De Gennes' model has parallel spins within each (001) between 650 and 900'C for 24 h. The annealing tem-
plane, the cant occurring between planes. Therefore, the peratures were chosen so that the samples did not melt.
magnetic unit cell is the same size as the chemical cell. The samples were reground and the annealing repeated.
Hirone and Adachi s model has three spin directions They were then cooled at the rate of 50'C/h except for
within each (001) plane and, thus, requires a larger unit the annealed MnSb and Mnp, 25Clp. 7gSb which were
cell. cooled at 15'C/h. Another sample of MnSb was pre-
Pickart and Nathans' made a neutron diGrac- pared for us by A. Cornish of these Laboratories. This
sample was removed from the furnace while at the
tion investigation of one composition in the series,
preparative temperature and allowed to air cool. For
Mnp 2Crp, 8Sb, which was reported to have Neel and
convenience, this sample will hereafter be called
Curie temperatures of 600 and 130'K, respectively. At
"quenched" MnSb.
room temperature, they found, as expected, an anti-
ferromagnetic structure like that found in CrSb. How-
X-ray powder patterns were taken with CuE
radiation using a thin Al foil to reduce the fluorescent
ever, at 4.2'K several weak superlattice rejections were
radiation. All the patterns were extremely sharp for the
present which could be indexed by tripling a and b.
These were in addition to the magnetic contributions samples reported in Table I except for the quenched
from a CrSb-type structure which had a moment de- MnSb. %hen MnSb and Mnp, 75Clp, 25Sb were prepared
creased in magnitude compared with that at room tem- by cooling at 50'/h, they were found to give broad lines
for 28 greater than 120'. No trace of free metal or Sb
perature. Their conclusion was that neither the model
was detectable in any of the samples.
of Lotgering and Gorter nor that of Hirone and Adachi
was in agreement with their results. The model of de
The lattice parameters are shown in Fig. 3 and listed
in Table I. For comparison, the results of Lotgering and
Gennes, which was published subsequent to Pickart and
Gorter, 7 who reported a possible error of "several
hundredths of an angstrom, "
are also included. The
o L. M. Corliss, N. Elliot, J. M. Hastings, and R. L. Sass, Phys. present results are in agreement v ith the previous ones.
Rev. 122, 1402 (1961). Since they are determined with higher accuracy„ the
' P.-G. de Gennes, Phys. Rev. 118, 141 (1960).
" C. Zener, Phys. Rev. 82, 403 (1951). trends are nom more clearly established. It should be
"P. W. Anderson and H. Hasegavva, Phys. Rev. 100, 675 noted that the scales in the figure for a and c are diGerent
(1955).
"R. R, Heikes, T. R. McGuire, and R. J. Happel, Jr. , Phys. so that the total variation of c is an order of magnitude
Rev. 121, 703 (1961). larger than that of u. A diGerence of the same order in
2010 TAKE I, COX, AN D SH I RANE

Despite these variations in the dimensions of the unit

580- o o
cells, Fig. 4 shows that the volume is a linear function
C5

of the composition.

B. Magnetic and Neutron Di6'raction

5.60
Techniques
c: (A) The magnetic measurements were made on a balance"
based on the Faraday method. Measurements were
5.50—
made in 6elds up to 10 000 Oe and at temperatures from
Lotger~ng and Gorter 77 to approximately 900'K. The Neel temperatures were
5.40— taken as the inRection points of the susceptibility curves.
As far as possible, the Curie temperatures were deter-
mined by the method of gneiss and Forrer" utilizing
plots of field versus temperature at constant moment.
A Oo
The moments were extrapolated to in6nite 6eld from
4.12 moment versus 1/P curves.
The neutron diffraction experiments were conducted
&. 08— at the Westinghouse Testing Reactor using a beam of
I I I 1.14 A wavelength. The samples, which were all
0 0.25 0.50 Q75 I.O
MnSb
X in Mn, &Cr&sb CrSb powders, were placed in 3/g-in. diameter, thin-walled,
aluminum or vanadium tubes for the low-temperature
F&G. 3. Lattice parameters of Mn1, Cr Sb solid solutions. Solid studies and 0.01-in. wall quartz tubes for some room-
circles are results of the present study and open circles, those of
Lotgering and Gorter (reference 7). temperature runs.

the variation of c and u with temperature was found for III. ANALYSIS OF THE DATA
pure MnSb by Willis and Rooksby. The variation of " For convenience, before presentation of the data, the
the c/u ratio for the solid solutions is shown in Fig. 4. pertinent intensity relationships and details of interpre-
There is an initial increase in the ratio as Cr is added tation will be summarized.
followed by a rapid decrease to the value for CrSb. The magnetic structure factor is given by'~:
I.42 F=pg p,q; exp22ri (hx;+hy;+ls;),
I40 where p; is the atomic scattering amplitude, propor-
0
(2 0 tional to the magnetic moment p, q is a function of the
unit vectors e and K, parallel, respectively, to the
c/o scattering vector and magnetic moment.
21= e(e. K) —K.
l34 o The intensity is proportional to the square of F averaged
over the equivalent rejections and for a single spin
l. 32 i- axis structure can be written as

86
F '=(F F)=(q')~; p; xp2 i(hx;+hy;+l;)]' (3).
(q2) is the average, over the equivalent reflections, of
sin'e, where 0. is the angle between the spin axis and the
scattering vector. For convenience, p has now been
assigned a sign determined by q and is positive for
moments parallel to the spin axis and negative for those
80— antiparallel to it.
0
I

0.25 QSO CL?5 IO

In the NiAs structure, the metal atoms are located at
MnSb X in Mn, „Cr&sb CrSb (0,0,0) and (0,0, 1/2). Equation (3) reduces to
Fzo. 4. Volume and c/t'a ratio of Mn1 Cr, Sb sohd solutions. F 2
(~2) (pA~ pB)2 (4)
Solid circles are results of the present study and open circles, those
of Lotgering and Gorter, (reference 7). " D. K. Cox, W. J. Takei, R. C. Miller, and G. Shirane, J. Phys.
Chem. Solids 23, 863 (1962).
'4 B.T. M. Willis and H. P. Rooksby, Proc. Phys. Soc. (London) "P. Weiss and R. Forrer, Compt. Rend. 1?8, 1670 {1924).
Ser 8 67, 290 (1954). » 0. Halpern and M. H. Johnson, Phys. Rev. 55, 898 {1939),
 MAGNETIC STRUCTURES IN MnSb —Cr Sb SYSTEM 20ii

p" and pa refer to the atoms at s equal to 0 and 2,

respectively. The positive sign is used. for reflections
B
with l even and the negative for / odd. Therefore, a ferro- Pg
magnetic structure with the p's equal and positive gives
a magnetic contribution only to the reflections with l ~A ~A 4
0 I &o
even. Conversely, an antiferromagnetic structure has
magnetic intensity only for reflections with l odd. since
the p's differ in sign. The value of (q') is'a:

2 Pd' (c)
(q') = 1 — (h'+h'+hh)d' sin% — cos%, —
(5) FIG. 5. Structures obtained by combining ferromagnetic and
3g2 c antiferromagnetic components. The angles between the spin axes
of the two components are assumed to be: {a}0', {b) 90', {c}be-
where C is the angle between the spin axis and c. tween 0' and 90'.
In some structures, a ferromagnetic or antiferro-
magnetic arrangement was derived which had a spin However, an ambiguity arises in combining these
axis tilted from c. The intensity is given by components to obtain the actual structure. The angles,
F2= (1—A sin% — 8 cos%)(P"a P). (6)
4, between the spin axes and c are known but not their
rotational position about c. Consider the situation when
(q') has been simpli6ed for convenience. For the cases the ferromagnetic and antiferromagnetic components
with 4 between 0' and 90', the possibility of another have their spin axes perpendicular to c. Both the axes lie
type of structure was considered. This was to assume within (001) planes but their relative position within
that the sample consisted of two coexisting structures. the plane is not known. There are three cases to be con-
One would have its spin axis parallel to c ((q') equal to sidered. The angle between the spin axes of the com-
—
1 8) and the other a spin axis perpendicular to c ponents could be: (1), 0'; (2), 90'; and (3) between 0'
—
((q') equal to 1 A). If the mole fraction, X&, of the and 90'. For (1), since the spin axes of the components
perpendicularly oriented portion was sin% and Xi/, are parallel to each other, they combine as illustrated
cos%, the intensity of this mixture would be propor- in Fig. 5(a). The resultant structure has parallel but
tional to unequal moments alternating along c. If the antiferro-
magnetic component had been the larger one, the struc-
F'=X.(q'). (1 "~1")+& (q') (P"~P') ture would have been ferrimagnetic. For case (2), the
= L(1 —A) sin'4+ (1—8) cos'41(P"&Ps) axes are within the (001) plane but perpendicular to
= L1 —2 sin'4 —8 cos'4$(P" &Pa). each other with a resultant structure as shown in
This equation is identical to Eq. (6) for the tilted axis Fig. 5(b). The spins are now equal in magnitude but
alternate by an angle O~/2 with respect to the ferro-
model. Thus, the two models cannot be differentiated
magnetic component axis. This is a canted spin struc-
and the magnitude of the atomic moments derived from
ture. For case (3) with an intermediate angle between
the data would be the same. For convenience, the data
the component axes, the resultant structure, shown in
will be discussed using the tilted axis model but it must
be remembered that there is no a priori reason for Fig. 5(c), has canted unequal moments. These three
cases cannot be differentiated by the present diffraction
eliminating the other model.
Equation (4) was derived for a single spin axis struc-
data. However, there is no evidence for the presence of
unequal moments in this system and the assumption of
ture. However, there is another type of structure to
which it can also be applied directly. Consider a mag-
parallel spin axes results in unreasonable moment mag-
nitudes in some cases. Therefore, it was concluded that
netic structure in which the spins can be resolved into
case (2), the canted spin structure with equal spins, was
ferromagnetic components of magnitude p. o and anti-
the most probable one and the results will be presented
ferromagnetic components p, j along two independent
in terms of this structure.
axes. This is not the same as the mixture case considered
in the preceding paragraph, since there is now only one IV. MAGNETIC AND NEUTRON DIFFRACTION
structure present in the sample. Because of the atomic RESULTS
positions, the structure factors are such that the two
components will diGract independently of each other. A. Cr81
The results of the magnetic measurements of CrSb,
P
Fmsg
2
(q 2)(p
= (qAF )(p
A+p B)2

p ) &= 2N+1
(7) shown in Table II, agree with previous reports. The
structure obtained by room-temperature
'
neutron dif-
By application of these equations to the data, the fraction experiments is in agreement with that of Snow. 4
moment and C of the two components can be derived. Ferromagnetic (001) layers are coupled antiferromag-
netically with the spin axis parallel to c. If, following
' G. Shirane, Acta Cryst. 12, 282 (1959), Snow, the Mn+' form factor ig used, a mo~qnt of 2,6 pp
 TAKE I, COX, AN 0 SH I RANE

TABLE II. Magnetic properties of Mnl, Cr, Sb. Tz and Tz are axes perpendicular and parallel to c. In addition to the
the Curie and Noel temperatures; 8„, the asymptotic Curie tem-
perature from the susceptibility measurements; 2S, twice the expected, ferromagnetic reflections, there are very small
number of unpaired spins in the paramagnetic region as derived peaks which cannot be indexed on the basis of the
from the Cur)e constant; p o, the measured atomic saturation chemical unit cell or small multiples of it. Accurate
moment extrapolated to infinite fieM and O'K.
positional and intensity measurements could not be
made but they seem to occur in pairs associated with the
TQ T+ Hy p, o
(001) and (101) positions as would be expected for a
('K} ('K) ('K) 2S (p~)
sinusoidally modulated antiferromagnetic structure.
0 (annealed) 600 ~ ~
600 3.43 3.53
0 (quenched) 565 ~ ~
525 3-9o 3.3z The magnetic component involved is extremely small,
0, 25 505 ~ ' ~ ~
505 3.23 2.8z representing only a small perturbation of the basic
0.51 390 ~ ~ ~
410 2.93 1.8&
0.7 255 500 260 3.1g 0.9 ferromagnetic structure and does not materially afI'ect
0.75 175 550 185 3.30 0.7 the main conclusions to be drawn.
0.8 130 590 175 3.20 0.4 During the initial intensity calculations, the Mn+'
0.85 (&77) 620 45 3.2g
1.0 ~ ~ ~
705 —625 3.8o form factor was used. It was noted that this resulted in
an atomic ferromagnetic moment larger than that
obtained from the magnetic saturation measurements.
is calculated compared to his 2.7 p, g. However, the use Therefore, new form-factor values were computed using
of the Cr+' form factor as obtained from Cr203" made
it evident that the proper moment was 2.84 p, ~ at room
temperature which extrapolates to 3.0 p& at O'K by the
Brillouin function. Figure 6 shows the form factors
cakulated using this Inoment as well as those from
Cr203. The Mn+' form factor curve is also shown for
comparison. A composite form factor curve was drawn
from the two sets of points and used in the subsequent
calculations for the solid solutions. It may be noted that
3 p, ~ is smaller than the 3.89 pg expected on the basis of
the susceptibility measurements. A possible reason for
this discrepancy may be that the susceptibility measure-
ments could not be extended suKciently into the para-
0.4
magnetic region to permit determination of the proper
slope. It is also possible that there is a component of the
Cr moment which does not order at 300'K.
0.2
B. MnSb
The magnetic structure derived from the neutron
diBraction data for both the annealed and quenched I

MnSb at 77 K was basically that of a ferromagnetic G2

structure with the spin axis perpendicular to c. The Sing /1
data for the quenched MnSb are given in Table III and Fro. 6. Observed form factors for Cr. Moments are assumed to
compared with the calculated intensities assuming spin be 2.99 p, g for Cr203 (reference 15) and 2.84 pg for CrSb.

Tax III. Neutron diffraction intensities for quenched MnSb. the intensity data from the ferromagnetic reQections
Calculated intensities include a temperature factor J3 of 0.4X10 '6
'
A. at 77'K and 0.8X 10 ~ X' at 300'K. Magnetic contributions and the atomic moments obtained from the magnetiza-
calculated utilizing the measured saturation moments of 3.35 p~ tion measurements. They are shown in Fig. 7. It can be
at 77'K and 2.90 p, g at 300'K. C is the angle between the spin
axis and c. seen that the points are above the Mn+' curve. Also
shown is the form factor for Mn obtained by Roberts"
77'K 300'K from MnBi. A smooth curve was drawn through the
Inuo
talc
Itot
{4 ~0) {4 90)
Itoc
Iobo
Ioalo
(4 0) (4
Ioalo Ioale
90) (4 =60) Iobo points and used for the subsequent calculations. It
{100) 188 436 3)i 329 368 276 299 306
should be noted that this is a small change and does not
{101) 152 150 150 146 146 146 146 149 aHect the conclusions as to the spin directions.
(OD2I 83 82 126 1)4 79 1ii 103 87
4 46 100 102 66 58 70 At room temperature, the results from the two MnSb
(110) 5 48 27 22 36 20 24 21
(200) 50 73 58 57 samples difI'ered from each other. They also di8ered
(20)) 49
{103) 49 283 283 262 from those reported by Pickart and Nathans, 5 who
(112) 180 reported a spin axis parallel to c. The annealed MnSb,
(202) 2 21 )8 15
whose diGraction pattern is shown in Fig. 8, was found

"D. E. Cox, %'. J. Takei, and G. Shirane (to be published). ~ B. %. Roberts, Phys. Rev. 104, 607 (1956}.
 MAGNETIC STRUCTURES IN MnSb-CrSb SYSTEM
4700'
3000-

2000-
z {ioo)
tooo

0.6 ~~ 3000- 1 1

46M MngmC~o. aSb 298'K Oo»

~8 2000-

—
-NnSI
t l t t
20 25 R) 35
P, e

FIG. 8. Neutron diGraction pattern for annealed MnSb and

Mno. 49Cro. &&Sb at 300'K. Thin-walled, 3/8-in. -diam, cylindrical
ILl CR quartz sample holder. Wavelength, 1.14 L.
Stn@L

FIG. 7. Observed form factors for Mn. Moments used: 3.53 pg direction for this reQection and thus the magnetic con-
for MnSb at 77'K, 3.26 pg for MnSb at 300'K, and 2.51 pg for
Mno. 7~Cr0. 2~Sb. MnSi values from Roberts (reference 20). tribution would not be completely eliminated.
The room temperature neutron diffraction data for
to have a ferromagnetic structure with a spin axis quenched MnSb are given in Table III. The observed
perpendicular to c. One crucial point leading to this intensities are not in agreement with either a parallel
conclusion was that there was a signiGcant magnetic or perpendicular spin orientation. The best Gt seems to
contribution to the (002) reflection. This would not be be obtained with an intermediate orientation, C equal
allowed by a spin axis parallel to c since (q ) would then to 60'.
be zero. Also, the magnitude of (102), which has a very The neutron diffraction data for the two MnSb
small nuclear component, was much too large. This samples and the results reported by Pickart andNathans
structure was tested by making measurements of the could be explained by a spin Hip occurring at a variable
ratio between the nuclear and total intensity of various temperature. Further evidence for this "spin flip" was
reQections. The nuclear intensities were measured by given by a study of the temperature variation of the
eliminating the magnetic contributions using two differ- magnetization. The results are shown in Fig. 9. The
ent methods. One was by scanning a sample heated to a high-Geld curves for the annealed sample follow a
temperature (625'I) above the Curie point. The second normal Brillouin curve with an extrapolated moment
method was to scan a sample in a magnetic Geld applied
parallel to the scattering vector. The results obtained l20— CO
are shown in Table IV with the calculated. values for
100—
spins parallel and perpendicular to c. The best agree-
ment is obtained by assuming a spin axis perpendicular SO
l.5
to c although application of a magnetic Geld did not 60
reduce the intensity of (102) suKciently. The high- /Pl

temperature experiment did give the proper ratio for

~
F 40—
4%

this peak. This could be explained if the perpendicular b 20—

to the (102) plane was along or close to a "hard. direc- " . m
0
tion of magnetization. If the anisotropy was large, the c sm~
spins would not be oriented by the Geld in the desired gl00

TmLE IV. Ratio, E, between nuclear and total intensities.

Magnetic contribution eliminated for E~by application of a
magnetic ield and for Eh t, by heating above the Curie tempera-
ture, C is the angle between the spin axis and c.

Wic Wlc
ReQ. (C =90 ) (@=0') 400 600
Tenper04re ('K)
(100} 0.63 0.59 0.61 0.44
(201) 0.98 0.99 1.00 1.00 Fro. 9. Magnetization and inverse susceptibility as a function of
(002) 0.72 0.56 0.66 1.00 temperature for quenched and annealed MnSb. Values for
(102) 0.22 0.03 0.03 0.07 quenched MnSb are those obtained on heating. On cooling, the
curves approach those of the annealed sample.
 TA KE I, COX, AN D SH I RANE

of 3.5 p, g at O'K and in6nite 6eld. This is in agreement

"
with the results of Hirone et ul. ' and of Guillaud, who Ql
reported 3.49 and 3.53 p&, respectively. However, the E
low-6eld curves show a pronounced peak at approxi- OP
20—
b
mately 520'K. This behavior is similar to that reported 5
al

by Guillaud" for Mn2Sb which shows a "spin Qip" at O

approximately 240'K. The properties of quenched 2P Q

MnSb are signi6cantly different. The standard magnetic
parameters are given in Table II. The extrapolated
moment, 3.3~ p, g, is smaller than that of the annealed T T T

200 400 600 800 l000

sample, as is the Curie temperature. Of particular Temperoture |'K)
interest is the fact that the peaks in the low-6eld
saturation curves are now broad ones centered at Fto. 10. Magnetization and inverse susceptibility as a function of
temperature for Mnp. 2~Clp, 7QSb.
approximately 320'K. This is in agreement with the
neutron diGraction results which found an intermediate
structure at 300'K. These magnetic properties are not
"
MnBi reported by Roberts. At room temperature,
the same after heating to 850'K, however, as then the
MnBi has a ferromagnetic structure with the spin axis
parallel to c. At 77'K, a structure intermediate between
susceptibility and magnetization curves on cooling
follow those of the annealed sample. The extrapolated
a parallel and perpendicular spin orientation was found.
Differences were noted in the range of this transition for
moment and Curie temperature are higher and the spin
two samples prepared under diferent conditions.
Qip is at approximately 510'K.
This dependence of the magnetic properties could be
explained as a consequence of a composition dependence C. Mn& .Cr,Sb
of the Qipping temperature. The equilibrium composi- The shape of the magnetization and susceptibility
tion of MnSb i.s de6cient in Sb above 8/O'K according
to Hanson. " Thus, rapid cooling would result in a
curves for the solid solutions are the same as those
previously reported. '~ For this reason, the curves will
MnSb phase de6cient in Sb plus free Sb (or Sb-rich not be presented except, as an example, those for
component). Guillaud" has reported the magnetic Mno, 7$CI0.2$Sb in Fig. 10. All samples with x equal to
properties of MnSb de6cient in Sb. The Curie tempera- or greater than 0.7 showed the peak in the inverse
ture and magnetic moment are lower than in the susceptibility curve and the Neel temperature was taken
stoichiometric compound. On the basis of his values, it at the minimum as indicated. The measured magnetic
is estimated that the quenched sample would have a 1% parameters are given in Table II. They are in agreement
de6ciency. The corresponding amount of free Sb would with the previous values taking into consideration the
notbe detected by the x-ray study. Heating the sample to difhculty of making the extrapolations in some cases.
850'K followed by slow cooling would allow the Sb to The compounds and the temperatures at which
disuse into the compound and approach the stoichio- neutron diGraction studies of the solid solutions were
metric composition. This would account for the different made can be determined by the location of the entries
magnetic properties on heating and cooling. in Table V. Most scans were conducted to a 28 value
The sample studied by Pickart and Nathans' was of 70', corresponding to a value of 0.5 for sin8/X, so that
reported to have an extrapolated atomic moment of many reQections were available which did not have a
3.3 p, ~. This low moment could be correlated with a signi6cant magnetic contribution. No indication was
larger Sb de6ciency and, thus, a lower Qipping tempera- obtained of any chemical ordering of the metals in the
ture. Their data would then be explained as resulting solid solutions. This lack of ordering would seem to
from their sample at room temperature being above make the presence of independent Cr and Mn spin
this transition point and, therefore, having a spin axis systems as proposed by I.otgering and Gorter' unlikely.
parallel to c.
The "best" sample to be studied. had a "spin Qip" TAsLE V. Moment in p~ of antiferromagnetic component perpen-
temperature, 520'K, close to the Curie temperature of dicular to c in Mn~, Cr, Sb. S refers to presence of satellites.
600'K. If it assumed that this sample was also slightly
oB-stoichiometry, then the true transition temperature 77'K 4.2'K
might be higher. It may be possible that the spin struc- 0 (annealed) 0
ture with the axis parallel to c is not a stable phase in 0 (quenched) 0
stoichiornetric MnSb. 0.25 0.5
0.51 1.8
It may be noted that this spin Qip for MnSb is of the 0.70 2. 16 2.4p
same type as that found for the isostructural compound 0.75 1.28 2.8g
0.80 0.3
"C. Guillaud,Ann. Phys. (Paris) 4, 671 (1949). 0.85 ~0.2
0.8
~M. Hanson, CoestANfion of Binary Alloys (McGravr-Hill 1.0
Book Company, Inc. , Neer York, 1958), 2nd ed.
 MAGNETIC STRUCTURES IN MnSb —CrSb SYSTEM 2015

The additional peaks requiring an increase of the unit

l.Oi
cell size, as predicted by the model of Hirone and (eo)f
Adachi, ' are not present. The canted spin model pre- ~&Norrnoliged.

dicts, in the ferromagnetic region, the occurrence of

antiferromagnetic reQections. This pred. iction is in
agreement with the experimental data as is shown by
the development of (001) reflections. Figure 11 shows
the room temperature intensity of this reflection which
must be entirely magnetic except for an extremely small
Ig~g (000
half-wavelength component which has been subtracted.
A magnetic (001) reflection can arise only from an
antiferromagnetic component perpendicular to c. A
CrSb-type structure with a spin axis along c would have
no contribution to (001) since the "q" value would be
zero. The magnitude of this perpendicular component
is given in Table V.
In addition to the contribution from the perpendicular X~~(el) - Correct
antiferromagnetic component, the diGraction data show
the magnetic intensity which would be expected on the
basis of the magnetic measurements. Thus, in the
ferromagnetic region, each peak with / even had a 1
0 OM 0.50 0.7S l.O
magnetic component corresponding to a ferromagnetic Mnsg X tn Mn1 gcr~Sb Crsh
spin arrangement with an axis perpendicular to c. The FIG. 11.Magnetic intensity of Mn~ Cr Sb solid solutions. Inten-
magnetic intensity of two such peaks at room tempera- sities of (100) and (102) are normalized to equal one for MnSb. Inten-
ture is shomn in Fig. 11. As would be expected from the sity of (101) has been corrected for the contribution due to anti-
ferromagnetic component perpendicular to c as discussed in text.
variation of the saturation moment, the intensity
decreases with increasing Cr content. In the anti-
parallel spin axes, the resultant would have been
ferromagnetic region, the reQections with l odd had a
ferrimagnetic, composed of an unreasonably large
magnetic contribution correspond. ing to a CrSb-type moment of 4.3 p, ~ alternating with one of —0.6pg
structure with the spin axis parallel to c. The variation
at O'K.
of magnetic intensity for (101) is shown in Fig. 11. It
has been corrected for the simultaneous contribution TABLE VI. Neutron diffraction intensities of Mnp. 4pCrp. pySb at
due to the antiferromagnetic component perpendicular room temperature. Temperature factor of 0.6)& 10 'p A~ included.
Ferromagnetic moment of 1.36 pg and antiferromagnetic moment
to c. of 1.80 pg used.
The resultant magnetic structure is obtained by
combining the various components as indicated in Sec. Calculated
III. In the ferromagnetic region, the two components
Ref.
I~
ferro
I~
antiferro It
Observed
Ip
combine to form a canted spin structure. As an example, Inuc t I
the room-temperature data for Mnp. 49Crp. ~~Sb are (001) 68 68 73
shown in Fig. 8 and Table VI. At this composition, only {100) 34 52 70
(101) 159 57 216 209
Sb contributes to the nuclear intensity since the Mn and
Cr scattering lengths of — 0.36 and +0.35X10 cm, " (002)
(102)
29
26
6
12
35
38
34
45
respectively, average to zero. The magnetic intensity {110) 44 2 46 60
(113) 11 13
was calculated using 1.36@,~ for the ferromagnetic (003) 3
component as determined by the magnetization meas- (200) 8 10 13
urements and 1.80 p, ~ for the antiferromagnetic com- {201) 43
(103) 43 16/
ponent. Both components lie in the basal plane. Com- (112) 58
bining these two components on the assumption that (202) 10 12
(004) 7 6
their spin axes are perpendicular gives a canted spin {210) 9
structure composed of moments of 2.26 p, ~ with a cant (104) 9
102
angle of 103'. Application of the observed 0'3oo K/0'o'K {211) 52
(203) 26
ratio of 0.73 gives an extrapolated moment of 3.1 pg (212) 17
at O'K. This extrapolation assumes that both com- {300) 17 67 67
ponents follow the same BriHouin curve in this case. (114) 33
(204) 6
This would seem to be a reasonable approximation on (213) 35,
(302) 24 83
the basis of the study of Mnp, 2QClp, 7$ which will be
{105) 18
presented later. If the ferromagnetic and antiferro- (320) 10 10
magnetic components had been assumed to have
 TAKE I, COX, A N D SH I RAN E
I I
spin axis of approximately 30' avray from c to give the
fgflgg5 CFg75sb ~8 K calculated intensities in Table VII.
Another interesting phenomenon which was studied
was the temperature variation of the (001) peak be-
l000-
tween 300 and 77'K for the 75%%uz Cr compound. It was
found to show a considerable thermal hysteresis. On
77'K cooling, the intensity of the (001) re6ection increased
ItNI
rapidly below 220'K so that at the Curie temperature
of 175' it vras close to its 77'K value. However, on
heating, the (001) intensity decreased only gradually
l000 and did not approach its original value until room tem-
I
perature was reached. The magnetization measurements
20 25 30 on this sample showed no hysteresis of the ferromagnetic
28
ordering. These results can be interpreted as a conse-
FIG. 12. Neutron diffraction patterns of Mno. »Cr0. »Sb at 77'I quence of the following structural changes. As the
and 300 K. Thin-walled, 3/8-in. diameter, cylindrical vanadium
sample holder used with wavelength of 1.14 g.
sample is cooled, the angle betvreen the antiferromag-
netic axis and c starts to increase rapidly at approxi-
mately 50' above the Curie temperature. At the Curie
Figure 11 shows that the (001) reflection also occurs
in the antiferromagnetic region. However, the signi6-
cance of this peak is fundamentally di6erent from that
previously. In the ferromagnetic region, it showed that
an entirely di6erent type of structure was present„
namely, a canted spin arrangement. In this case, how-
ever, it simply condrms that the structure is antiferro-
magnetic but vrith a tilted spin axis. It is a transition
structure between the CrSb type, with the spin axis
parallel to c, and the canted spin structure with the
spins perpendicular to c.
An illustration of a compound which can show
ferromagnetic and antiferromagnetic transitions is
Mno. 2gCr0. 7gSb. This compound has a Curie temperature I I
of 175'K and a Neel temperature of 550'K. The data og5 o.no
X in Mnl &Cr&sb
0.75
taken at 77 and 300'K are shown in Fig. 12 and Table
VII. At 77'K, the magnetic structure is made up of a Frc. 13. Moments of MnI, Cr, Sb solid solutions at O'K. The
small ferromagnetic component of 0.65 p, ~ and a large average atomic moment is obtained by combining the ferromag-
netic component as determined by the magnetization measure-
antiferromagnetic one of 2.81y~. The resultant is a ments with the antiferromagnetic component from the neutron
moment of 2.88 p~ vrith a cant angle of 154'. In the di8raction results.
antiferromagnetic region at 300'K, a small (001) reQec-
tion still remains. This is accounted for by a tilt in the temperature the spin axis is perpendicular to c and at this
point, a ferromagnetic component starts to order result-
TAsLE VII. Neutron di8raction intensities of Mn0. QQCr0. 7f;Sb. ing in a canted structure. On heating, af ter theferromag-
For 300'K, a temperature factor 8 of 0.5X10 ~63.', is included. netic component disappears, the antiferromagnetic axis
approaches the c axis only gradually. A similar explan-
77'K 300'K
IAF ation could be given in terms of the model using the
Inue IF IAF Itot In 2.65 p, s
30' (calc)
Itot In
mixture of tvro structures.
(calc) 0.65 ps 2.81 p, s (calc)
(001) 158 158 157 31 30 28
It vras remarked previously that, in the ferromagnetic
(ioo) 4 4 8 14 4 4 region vrith increasing Cr content, there is a rapid
(101) 149 110 259 261 94 238 245
(002) 13 1 14 14 13 13 decrease in the observed atomic moment as determined
(102} 73 75 72 73 65
(110) 82 2 84 94 7S 84 by magnetization measurements. It can now be seen
(111) 18 18 15 22 20 12
(003) 4 6 1 1 &3 that this decrease is due to the canting of the spins.
(201) 49 9 58 58 11 54 58
(112} 29 Figure 13 shows the total moment obtained for each
(103) 47 100 88 4 71 71
(202) 35 1 36 43 32 31
composition. The values quoted are those appropriate
(QQ4)
(211)
14
61
1 15 15 12 12 to O'K, extrapolated by means of a Brillouin function.
61
(203) 2 94 90 84 The dashed line indicates the moments calculated for a
(104) 31
(122) 48$ 78 ?8 42 43 composition average assuming 3.53 and 3.0 p~ for Mn
(3OO) 3Oj 26 27
(114) 58 58 49 49 40 and Cr, respectively. As can be seen, the observed
values are close to this line, lying slightly below it. This
 MAGNETIC STRUCTURES IN MnSb —CrSb SYSTEM
small "loss of moment" is similar to that found in othe-
solid solution systems such as Fe203-FeTi03, Fe20~r "
"
V203 and MnFe204-MnCr~04 ~'
C'A

V. DISCUSSION CW CN
The results for MnSb-CrSb solid solutions are sum-
marized by the magnetic phase diagram shown in Fig. 14
and the structures illustrated in Fig. 15. This diagram
has been simpli6ed and is qualitatively similar to that
of de Gennes, Fig. 2(b). On the left, the stable structure
at high temperature is of the MnSb type shown in Fig. (o) (b)
15(a) with a ferromagnetic arrangement of spins per-
pendicular to c. A phase line is indicated by line (A), FIG. 15. Magnetic structures present in
below which the structure is a canted one as in Fig. MnI, Cr~Sb solid solutions.
15(b). The exact location of this phase line has not been
determined. by this study. The corresponding phase line found. in MnSb with the spin parallel to c. If this struc-
between an antiferromagnetic and. a canted spin struc- ture is truly characteristic of stoichiometric MnSb, then
ture is that determined by the magnetization measure- it should extend into the solid solution region. However,
ments. The region between this line and the dash-dot since the Gipping temperature is so close to the Curie
temperature in the present sample of MnSb, it might
be expected to have only a small region of stability.
The third possible structure in the system is that
giving rise to the "satellites" found in MnSb at 77'K.
60$ Since they seem to be associated with the reflections
with / odd, they may correspond to a sinusoidal modula-
tion of an antiferromagnetic component. Clari6cation
p 400 of this structure, however, must await the study of
K
single crystals in order to increase the accuracy of the
data.
Several mechanisms have been proposed to account
for the origin of canted spin structures. Dzialoshinski25
0 gag a50
t
and Moriya" have derived these structures as a conse-
NlSb X tn Mnl. ~+Sb quence of an anisotropic superexchange interaction due
FIG. 14. Proposed magnetic phase diagram for MnI, Cr, Sb solid
to spin-orbit coupling. However, it is a small eGect,
solutions. F and AF refer to ferromagnetic and antiferromagnetic being of the order of ~g/g where g is the gyromagnetic
structures, respectively. Lines (A) and (B) explained in text. ratio and Ag its deviation from the free electron value.
Its application, in general, has been limited to the
line (8) indicates the region in which the transition is explanation of the occurrence of parasitic ferromag-
occurring from the CrSb-type structure, with the spin netism in basically antiferromagnetic compounds such
axis along c as in Fig. 15(d), to the canted structure, as nFe203. Structures similar to the canted spin arrange-
with the spin axis lying in the (001) plane. In this ments are the triangular spin configurations as derived
antiferromagnetic region, the spins are either tilted by Yafet and Kittel, "
Kaplan, "
and Hirone and
away from c as in Fig. 15(c) or there are two coexisting Adachi. ' These authors used the molecular 6eld ap-
structures as discussed previously. proximation and assumed a negative interaction be-
There are several other structures which, although not tween atoms in the feromagnetic sublattice of a ferri-
indicated in Fig. 14, may also be present in this system. magnetic or antiferromagnetic structure.
One is the structure found for Mnp. 2Cip. sSb at 4.2 K The theoretical analysis which has the best agreement
by Pickart and Nathans. ' Attempts to locate this phase with the present results is that of de Gennes. As dis-
in the 80 and 85% Cr samples were not successful. It is cussed. in the introduction, canted spin structures were
possible that preparative variations could a6ect the derived by considering the eGects of double exchange on
stability range of this phase and that if the 75%%u~ Cr an antiferromagnetic structure. Predictions were made
sample or an intermediate composition had been which are in agreement with the experimental results.
examined at 4. 2'I,it might have been found. These are:
Another possible phase is the ferromagnetic structure
"I. Dzialoshinski, J. Phys. Chem. Solids 4, 241 (1958).
~ G. Shirane, D. E. Cox, %'. J. Takei, and S. L. Ruby, J. "T.
'7
Moriya, Phys. Rev. 120, 91 (1960).
Phys. Soc. Japan 17, 1598 {1962). Y. Yafet and C. Kittel, Phys. Rev. 87, 290 (1952).
~ S. J. Pickart and R. Nathans, Phys. Rev. 116, 317 (1959}. '8 T. A. Kaplan, Phys. Rev. 116, 888
(1959).
2018 TAX E I, COX, AN D SH I RANE

It is perhaps somewhat surprising that a linear

variation of cos(8/2) is obtained over such a wide com-
i8 position range in view of the fact that double exchange
theories have generally been con6ned to a rather narrow
range. Moreover, owing to the lack of conductivity data
direct evidence concerning the mobility of the carriers in
this system is not available. The possibility that a
suitable molecular 6eld approach, for example, could
give equally good agreement with the experimental
6ndings should therefore not be overlooked. However,
if double exchange is assumed to occur throughout the
0 0.25 OAO Q.7S LO
hhSb X in Mn& &Cr~ Ccsb system, it may be concluded that the experimental
results provide strong support for the theory of
FIG. 16. Cosine of half of the cant angle, 8, as a function of
composition for Mn1, Cr Sb solid solutions at O'K. de Gennes.
There are several points of interest which could be
studied further by neutron di8raction. These include:
(1) Canted spin structures are found. (a) direct observation of the spin flip in MnSb; (b) the
(2) The generalized magnetic phase diagram,Fig. sinusoidally modulated structure in MnSb at 77'K;
2 (b), is of the same form as that found, Fig. 14. (c) the exact location of the phase line (A) in the mag-
(3) The inverse susceptibility curves have a peak as netic phase diagram, Fig. 14; (d) the possibility of
indicated by the typical example of Mn0. ~5Cro. y~Sb in structures such as in Figs. 5(a) and 5(c); these struc-
Fig. 10. tures, though improbable, could be tested by neutron
(4) A linear variation of cos(0/2) as a function of di8raction data taken with an applied magnetic field;
carrier concentration and therefore of composition was
(e) the phase found by Pickart and Nathans for
predicted. The experimental results are shown in Fig. 16. MQO. SClo. pSb. Unfortunately, before these points could
The values plotted are those appropriate to O'K as the be investigated, the reactor operations were terminated.
study of the 75% Cr sample showed that, after an However, the main features of the MnSb-CrSb systems
initial change, there was only a small variation of angle are now well established and clarilcation of these other
with temperature. features will be undertaken at a later date.
(5) ln the magnetic phase diagram of de Gennes,
Fig. 2(b), there is one composition which passes directly
ACKNOWLEDGMENTS
from a canted spin structure to paramagnetism on heat-
ing. For increasing carrier concentration, this is the The authors wish to thank Dr. R. Nathans of the
composition at which the ordered phase at high tem- Brookhaven National Laboratory, Dr. S. J. Pickart of
perature changes from an antiferromagnetic to a ferro- the Naval Ordnance Laboratory, and Dr. R. R. Heikes
magnetic structure. It was predicted that this composi- and Dr. A. J. Cornish of these Laboratories for helpful
tion would have a cant angle of 103'. From Fig. 14, it is discussions. The preparation of one of the MnSb samples
estimated that this composition is approximately by Dr. Cornish has been mentioned previously. The
Mno. 4Cro. ~Sb. From Fig. 16, this would have a cant assistance of M. Janocko with some of the preparative
angle of about 113', in agreement with the prediction. and x-ray experiments is also gratefully acknowledged.