PH YSICAL REVIEW VOLUM E 83, NUM BER 2 JULY iS, 1951

Neutron Diffraction by Paramagnetic and Antiferromagnetic Substances

C. G. SHULL, W. [Link], AID E. O. WOLLAN
Oak Ridge Notkma/ Laboratory, Oak Ridge, Tennessee
(Received March 2, 1951)
Neutron scattering and diffraction studies on a series of paramagnetic and antiferromagnetic substances
are reported in the present paper. The paramagnetic diffuse scattering predicted by Halpern and Johnson
has been studied, resulting in the determination of the magnetic form factor for Mn++ ions. From the form
factor, the radial distribution of the electrons in the 3d-shell of Mn~ has been determined, and this is com-
pared with a theoretical distribution of Dancoff. Antiferromagnetic substances are shown to produce strong,
coherent scattering effects in the diffraction pattern. The antiferromagnetic reflections have been used to
determine the magnetic structure of the material below the antiferromagnetic Curie temperature. For some
substances the magnetic unit cell is found to be larger than the chemical unit cell. The temperature de-
pendence of the antiferromagnetic intensities has been studied, and the directional effects which characterize
neutron scattering by aligned atomic moments have been used to determine the moment alignment with
respect to crystallographic axes. From studies with magnetic ions possessing both orbital and spin moments,
it is found that the antiferromagnetic intensities contain partial orbital moment components along with the
spin moment component. The degree of orbital moment contribution agrees satisfactorily with that pre-
dicted by models of lattice quenching.

INTRODUCTION The present paper is concerned with measurements on

AGNKTIC scattering effects with neutrons the scattering or diffraction pattern which is obtained
- ~ were 6rst investigated theoretically by Bloch, ' when monochromatic neutrons are incident upon a sub-
Schwinger, ' and Halpern and co-workers. ' The early stance whose atoms possess magnetic moments. Experi-
theoretical developments focused upon the interpreta- mental data will be presented for the scattering by
tion of experiments designed to determine the value of various paramagnetic and antiferromagnetic materials,
the neutron's magnetic moment through its interaction and the interpretation will be given in terms of a mag-
with the experimentally known and theoretically netic lattice with spin and orbital moment alignment.
understood magnetic moments of various atoms and In a second paper in preparation, we shall extend the
ions in paramagnetic and ferromagnetic substances. discussion to include scattering by magnetized and
Since this time, however, more powerful resonance unmagnetized ferromagnetic materials, treating also
techniques have been applied which give high precision some of the neutron polarization phenomena which are
in the determination of the neutron magnetic moment, associated with ferromagnetic scattering.
and, in consequence, present-day interest in magnetic SCATTERING OF NEUTRONS BY PARAMAGNETIC
scattering effects with neutrons has been directed SUBSTANCES
toward a more complete understanding of the basic
The theory of neutron scattering by a true paramag-
phenomena which characterize magnetic media.
netic substance has been given in detail by Halpern and
Early experimentation on neutron scattering by
Johnson. In such a substance, there exist permanent
paramagnetic materials was performed by Whitaker
and co-workers4 in a series of transmission and scat- magnetic moments at individual atomic sites caused by
unbalanced electronic moments within the atoms, and
tering type experiments. Because of the relative weak-
these atomic moments are completely uncoupled to
ness and heterogeneous nature of the neutron beams
each other and directed in random orientation. Because
from their Ra™Besource, coupled with the complexity
of this randomness the substance will display no per-
of crystal scattering effects, only general, qualitative
manent macroscopic magnetic moment. Application
conclusions couM be drawn from the data. The more
recent work at this laboratory' and at Columbia Uni- of an external magnetic 6eld will, however, disturb the
randomness and cause partial alignment of the atomic
versity by Ruderman' has shown unambiguously the
moments, so that an induced macroscopic magnetic
general features of paramagnetic scattering as predicted
moment is evidenced. Thermal disordering effects tend
theoretically by Halpern and Johnson' and has yielded
information on the nature of the magnetic form factor.
to oppose the alignment caused by the magnetic GeM,
and Langevin showed that the magnetic susceptibility
' F. Bloch, Phys. Rev. 50, 259 (1936). (induced moment per unit Geld) should vary inversely
' J. S. Schwinger,Phys. Rev. 51, 544 (1937) with absolute temperature, as erst determined experi-
'O. Halpern and M. H. Johnson, Phys. Rev. 55, 898 (1939);
O. Halpern and T. Holstein, Phys. Rev. 59, 960 (1941); Halpern, mentally by Curie. This can be described by the well-
Hamermesh, and Johnson, Phys. Rev. 59, 981 (1941). known Curie law,
4 M. D. Khitaker,
Phys. Rev. 52, 384 (1937); Whitalmr, Beyer,
and Dunning, Phys. Rev. 54, 771 (1938); M. D. Whitaker and
W. C. Bright, Phys. Rev. 57„1076 (1940); 60, 280 (1941).
where x is the magnetic susceptibility, T the absolute
' I. W. Ruderman, J. Phys.
S. Smart, Phys. Rev. 76, 1256 (1949).
~ C. G. Shull and

Rev. 76, 1572 (1949). temperature, and C~ the Curie constant which is
333
 SHULL, STRAUSER, AND WOLLAN
related to the strength of the individual atomic mag- ion should possess a diGerential magnetic scattering
netic moments. Many substances are known to obey this cross section of 1.69 barns/steradian in the forward
relationship, particularly at high temperatures, but direction, and this exceeds the coherent nuclear diGer-
there are frequent departures from the idealized state at ential scattering cross section for Mn of 0.109 barns/
low temperatures. In such cases the agreement can be ste radian.
extended in temperature range by the introduction of %ith regard to paramagnetic scattering eGects, we
the Curie-Weiss constant 8 into Eq. (1) so that there have investigated the angular dependence of neutron
results the Curie-gneiss law, scattering by several Mn++ paramagnetic salts, with the
idea of establishing quantitative agreement with the
(2)
theoretical predictions and also of determining experi-
This modification is considered to be an indication that mentally the magnetic form factor. A beam of mono-
the magnetic fields of surrounding atoms are interacting chromatic neutrons of wavelength 1.057A was scattered
to the extent that the magnetic moments are no longer by the sample under study in one of the Oak Ridge
uncoupled, as premised in the development of the neutron diGraction spectrometers. Since the paramag-
simple Curie law. This interaction may become so netic eGects appear in the diGuse scattering, it is
strong in certain cases that even Eq. (2) fails, and the imperative that the scattering sample be very free of
material may show ferromagnetic or antiferromagnetic any combined or occluded water, because the very large
behavior. spin diGuse scattering of hydrogen can easily mask the
Since the neutron possesses a magnetic moment, it desired scattering. The hydrogen diGuse scattering,
would be expected to experience a simple dipole-dipole furthermore, possesses an angular variation not unlike
interaction in addition to the usual short-range nuclear the magnetic form factor scattering, thus adding to the
interaction upon being scattered by an atom possessing difhculties. Very pure and anhydrous samples of MnO,
a magnetic moment. Halpern and Johnson have given MnSO4, and MnF2 were prepared by Mr. D. Lavalle of
a convenient formulation of this interaction in terms of this laboratory for this study, and these were sufficiently
diGerential scattering cross section and scattering am- anhydrous that the contaminant hydrogen scattering
plitude. For the case of complete randomness of atomic was negligibly small.
dipole orientation, they show that the magnetic scat- The neutron diGraction patterns for all these sub-
tering is completely zricoherent and that the scattering stances contain various components including (1)
by an assemblage of such atoms in a crystal lattice is nuclear coherent scattering, (2) nuclear spin incoherent
given by the sum of the scattered intensities from each scattering, (3) thermal diffuse scattering, (4) multiple
of the atoms. The diGerential cross section for magnetic scattering in the specimen block, and (5) the para-
scattering per atom is given by [their Eq. (5.41)] magnetic diGuse scattering. The first of these is repre-
do„=gS(S+1)(e'y/mc')f'd0, sented by the Debye-Scherrer diGraction peaks, and,
(3)
since these are localized in angular position in the
where S is the spin quantum number of the scattering pattern, there is no difBculty in resolving them from
f
atom, is the amplitude form factor, y is the neutron the diGuse scattering. The nuclear spin incoherent scat-
magnetic moment expressed in nuclear Bohr magnetons, tering arises principally from the Mn scattering, since
and the other terms have their conventional significance.
none of the other nuclei in these compounds exhibits
nuclear incoherent scattering, ' and this should be com-
In contrast to nuclear scattering cross sections which are
without a form factor angular dependence (the nuclear
pletely isotropic. By calculation, the spin incoherent
size being so much smaller than the neutron wave-
scattering is small, and one can reliably make correc-
length), the magnetic scattering effects are always
characterized by their angular dependence, since the tions for it. Thermal diGuse scattering is also rather
electrons which are responsible for the atomic magnetic small, vanishingly so for small angles, and can be
moment are distributed in a volume whose linear extent calculated with the Debye formula making use of
is quite comparable to the neutron wavelength ( 1A). experimental data on the elastic constants of the speci-
f
The magnetic form factor will depend upon the radial men lattice. Among the diffuse scattering contaminants,
the largest and most dificult of correction is the multiple
distribution within the atom of only those electrons
which are magnetic donors and in this respect will diGer scattering in the specimen block. This has been deter-
materially from the usual x-ray scattering form factor mined in an empirical fashion by studying the diGuse
which is representative of all of the atomic electrons. scattering with samples having equivalent scattering
The magnetic cross section discussed becomes quite and absorption properties but having no magnetic
sizeable and even exceeds nuclear cross sections in scattering. In general, the total contribution of spin
many cases. An example of this is given in paramag- incoherent, thermal diffuse, and multiple scattering for
netic salts containing Mn++ ions. This ion is in a spec- the specimen amounted to about 20 or 30 percent of the
troscopic S-state, with 5 electrons in the 3d-shell whose total diGuse scattering in the forward direction, so that
spins are all aligned parallel, so that the spin quantum
number of the ion is 5/2. According to Eq. (3), such an ' C. G. Shull and E. O. %'ollan, Phys. Rev. 81, 527 (1951).
 NEUTRON DIFF RACTION 335

the resultant paramagnetic scattering could be evalu-

ated fairly well.
After correcting the disuse scattering in the diffrac-
tion patterns in the aforementioned fashion, the resul-
tant paramagnetic scattering for the three Mn++ salts
was obtained as shown in Fig. 1. The data are on an
absolute scale and are shown as diGerential scattering I I I
&Oe Roe «Oo
cross sections. Because of the relatively large magnetic IXWNTER AHOLE

dilution in MnSO4 as compared to MnF~ and MnO and Fro. 1. Paramagnetic diffuse scattering for neutrons by several
the consequent low scattering intensity, the MnSO4 paramagnetic salts containing Mn++ ions.
curve is not considered to be as reliable as the other
two. The MnF2 magnetic scattering shows the expected within the atom of only those electrons which con-
form factor distribution with angle, whereas the MnO tribute to the atomic magnetic moment in the case —
curve shows a deanite peaking or residual coherence. under discussion, the 5 electrons in the 3d-shell. The
This suggests that in 3EnO ai room temPeraiure, the experimentally determined form factor, which is repro-
magnetic moments are not comPleiefy randomly arranged, duced in Fig. 2 as the amplitude form factor (along with
buf that there exists a short-range riMgneHc order with an estimate of the experimental uncertainty), can be
neighboring moments Iemhng to be ahgeed so that there used to obtain the radial distribution curve of the
is partial coherence in the scattering. This is somewhat 3d-shell within the atom. For this purpose the usual
analogous to the scattering of x-rays by liquids where, fourier inversion integral is employed as discussed, for
because of positional short-range order, residual co- instance, in Compton and Allison. ' Here
herence is obtained in the scattering pattern. In the
magnetic short-range order state, the degree of mag- f
netic alignment order is the determining factor rather u(r) = (2r/rr)~ kf(k) sinkrdk,
than the geometrical positional order. Presumably, it
will be possible to interpret the magnetic short-range where u(r) is the number of electrons within the shell
order patterns in terms of short-range order coefEcients
or parameters just as has been done in the case of
f
of thickness dr at radius r, and is the amplitude form
factor in terms of k=4+(sin8/X), with 28 the scattering
chemical short-range order in superlattices. '
This has angle. This inversion of the experimental form factor
not been attempted as yet. has been performed and the resultant radial distribution
In contrast to the MnO scattering data, that of function u(r) determined as shown in Fig. 3. In order to
MnF2 does not indicate the presence of residual co- determine the quantitative significance of u(r), several
herence. As a test of this, the pattern was studied when diferent experimental form factors were inverted, all
the sample temperature was raised to 400'C, and no
signi6cant change in the magnetic pattern was found.
Raising the temperature should certainly reduce any
short-range magnetic order in the lattice, and the fact
that no change was observed indicates that the MnF2
magnetic pattern corresponds to the true magnetic 0.8-
form factor. According to Eq. (3), the paramagnetic LIJ

diffuse scattering should amount to 1.69 barns/steradian I-

in the forward direction for Mn++, where S=5/2. The 0 0.6-
0
extrapolated MnF~ curve falls perhaps j.0 percent lower
CD
than the calculated value, and this may be the result of 0.4-
uncertainties in the standardization procedure or in the $5
determination of the angular distribution in the small X
angle region. The magnetic susceptibility data for 0.2-
Mn++ salts are in excellent agreement with calculated
values assuming 5 electrons to contribute to the atomic I I
I.0 2.0 3.0 4.0 5.0
moment, and the above cross section agreement within
10 percent indicates agreement in the number of con- 4s. (sin S)
tributory electron spins to within 5 percent.
Fxo. 2. Magnetic amplitude form factor for Mn++ ions. The
RADIAL DISTRIBUTION OF THE ELECTRONS curve is that obtained from paramagnetic disuse scattering with
IN THE 3d-SHELL OF Mn++ estimated error as shown. The points represent values of the form
factor obtained from the low temperature antiferromagnetic re-
As mentioned in an earlier section, the magnetic jections of MnO.
form factor is representative of the spatial distribution
~A. H. Compton and S. K. Allison, X-rays aed
s
J. M. Cowley, J. Appl. Phys. 21, 24 (1950). ExPeriraent (D. Van Nostrand Company, Inc. . ¹wYork, 1946).
iw Theory
 SHULL, STRAUSER, AND WOLLAN
transition temperature. '~ This anomalous behavior was
5d ELECTRON DSTRISUTIOM N Mn
interpreted as being caused by the development of
magnetic order at low temperatures, with an antiferro-
magnetic lattice existing below the transition tempera-
EXPERIMENTAL ture or antiferromagnetic Curie point. In such a mag-
netic lattice some of the neighboring atomic moments
RETICAL are coupled in antiparallel orientation by the quantum
AN COFF)
mechanical exchange forces. Macroscopically there are
as many moments directed in one orientation as in the
opposite, so that the susceptibility does not experience
large values as in the case of a ferromagnetic lattice
below the ferromagnetic Curie point in which the
moments are locked in parallel orientation. Theoretical
RAOIUS (A)
treatment of such antiferromagnetic lattices has been
0.4 QS I.e " "
given by Noel, Bitter, '4 Van Vleck, and recently by
Flo. 3. Comparison between the theoretical distribution of
electron radial density in the 3d-shell of Mn++ and that experi-
Anderson" and Li. "
mentally determined from the magnetic scattering of neutrons. In this picture it would not be unexpected that a
The error representation on the experiment. tal curve is that sug- short-range magnetic order could exist in the lattice at
gested from inversions performed on several possible experimental
form factor curves. The ordinate values are not on an absolute temperatures not very elevated with respect to the
scale. Curie temperature. The fact that MnO exhibits this
property at room temperature, which is some 180'
of which were contained within the suspected uncer- above the Curie temperature of 120'K, shows that the
tainty in the data. The resultant errors in Fig. 3 are a short-range order characteristics can persist over quite
measure of this signi6cance. For comparison purposes a sizeable temperature range. Below the Curie point an
a theoretical curve has been calculated for the 3d-shell entirely different type of magnetic pattern would be
in Mn++ by Dancoff, ' using a self-consistent 6eld expected in the neutron scattering, since the magnetic
analysis with exchange effects included. Both the ex- scattering amplitudes would now show coherent scat-
perirnental and theoretical curves show maxima in the tering effects. A preliminary report on some of these
vicinity of radius 0.6A, but a rather pronounced dif- observations has been given in an earlier letter, ~ and
ference shows up at larger radii values, One possible we now discuss in detail these studies.
explanation for this discrepancy might be that the The theory of neutron scattering by ferromagnetic
computation was based on an isolated ion, whereas the lattices as developed by Halpern and Johnson' can be
experimental data were obtained for an ion in a crystal used in the interpretation of the antiferromagnetic
lattice where the distribution might be compressed scattering effects. Halpern and Johnson (H-J) show that
under the influence of surrounding neighbor ions. the differential scattering cross section P' per magnetic
atom for unpolarized neutron scattering by an oriented
SCATTEMNG OF NEUTRONS BY ANTI- magnetic lattice is given as
PERROMAGNETIC SUBSTANCES
(5)
Many paramagnetic substances such as MnO and
MnF2 exhibit anomalous magnetic susceptibility and where C is the coherent nuclear scattering amplitude,
heat capacity behavior at low temperatures. In the and D is the magnetic scattering amplitude. The
former case, for instance, the magnetic susceptibility nuclear amplitude cannot be calculated theoretically
shows a reversal in its temperature dependence in the because of our ignorance of nuclear force 6elds, but
vicinity of 120'K. As the temperature is lowered values of C for many nuclei have been determined
towards 120'K the susceptibility increases according to experimentally. ~ On the other hand, the magnetic
Eq. (2) with a Curie-Weiss 8 of —510'K, but in going interaction can be treated theoretically, and H-J give
through 120'K the susceptibility trend reverses itself for D the formulation,
and decreases as the temperature is further reduced. " D= (e'/~e') ySf,
A marked change ia. speci6c heat is also found" at this
temperature. Early x-ray diffraction studies" had where the terms have the same signi6cance as for
shown no change in lattice structure, with the MnO Eq. (5). The absolute sign of the magnetic scattering
structure being isomorphous with the NaC1 face- amplitude D can be either positive or negative, de-
centered-cubic structure both above and below the ~' See later discussion for more recent x-ray diGraction 6ndings.
» L. Noel, Ann. phys. 17, 64 (1932); 5, 256 (1936).
""Bizette,
Unpublished calculations by S. M. Banco'.
Squire, and Tsai, Compt. rend. 207, 449 (1938); C. F.
'4 F. Bitter, Phys. Rev. 54, 79 (1938).
"J.H. Van Vleck, J. Chem. Phys. 9, 85 (1941).
Squire, Phys. Rev. 56, 922 (1939). ' P. W. Anderson, Phys. Rev. 79, 350); Phys. Rev, 79, 705
~ B. S. Ellefson and N. %. Taylor, J. Chem. Phys. 2, 58 (1934). (1950).
(with references to earlier work by Millar and Anderson). '~ Yin-Yuan Li, Phys. Rey, SO, 457 (1950).
 NEUTRON D I F F RACT I ON 337

pending upon the relative orientations of the atomic and dered sample was contained in a thin walled cylindrical
neutron magnetic moments. It is to be emphasized that capsule held within a low temperature cryostat. Both
the square of D in Eq. (5) is a classical or numerical patterns were taken of the same sample before and
square, in contrast to the quantum mechanical square after introduction of liquid nitrogen coolant. The room
which appeared in Eq. (3) describing paramagnetic temperature pattern shows both magnetic diffuse scat-
scattering. In oriented, magnetic lattice scattering, only tering and the Debye-Scherrer diffraction peaks at
a single-spin state is existent, and, hence, the square positions indicated for nuclear scattering. There should
of the amplitude involves S rather than S(S+1). be coherent nuclear scattering at both all-odd and
The term q' in Eq. (5) depends upon the relative all-even reQection positions from this NaCl-type lattice,
orientation of the two unit vectors e and x, where e is and since the signs of the nuclear scattering amplitudes
the scattering vector given by are opposite for Mn and for 0, the odd reflections, (111)
and (311), are strong whereas the even reflections, (200)
and (220), are very weak. When the material is cooled
where h and k' are the incident and scattered wave to a low temperature, there is no change in the nuclear
vectors, and x is a unit vector along the direction of scattering pattern, '"
but the magnetic scattering has
alignment of the atomic magnetic moments. H-J show now become concentrated in Debye-Scherrer peaks at
that new positions. As can be seen from the 6gure, these
q= eX (eXx), extra magnetic reQections cannot be indexed on the
so that basis of the conventional chemical unit cell of edge
q'=1 —(e x)'. length 4.426A. The innermost reQection for this cell is
the (100), falling at about 132" in angle, and there exists
It is seen that q' can attain values between 0 and 1 and, a strong magnetic reQection inside of this angle at about
for the particular case where x is randomly directed, 11~". It is possible to index the magnetic reQections,
q' (random) = .
-', however, on the basis of a cubic unit cell whose axial
length is just twice the above, or 8.85A. For this cell
This dependence of q' upon the relative directions of the magnetic reQections are all-odd, intensity being
scattering and magnetization has been given a direct observed at the (111), (311), (331), and (511) positions.
experimental test in the scattering from magnetized,
ferromagnetic substances, " and these data show the
The (311) ~ is on the shoulder of the (111) „,~, as can
be seen from the asymmetry of this reQection.
correctness of the above formulation. This twice-enlarged magnetic unit cell indicates that
The differential scattering cross section F' determines successive manganese ions along the cube axis directions
what is available for coherent neutron scattering but
are oriented differently, so that the repetition distance
tells nothing about the angular distribution of scattered
(for identical scattering power) along the axis is 8.85A
intensity from a magnetic lattice. Details of the scat-
tered intensity in the diGraction pattern will be deter-
BSI) (58)
mined (as in x-ray or electron diffraction) by the crystal 100 f os~8. 85K

structure factors, and from the experimental deter-

mination of these factors, one can hope to establish the
60
magnetic lattice. It is interesting to note that according
to Eq. (5) there is no coherent interference between the
magnetic and nuclear portions of the scattering, and jK
20
that in essence the two intensities of scattering are
merely additive. This is a consequence of the treatment (I00) (IIO) (III) (200) (sii)
for unpolarized incident neutron radiation and would IOO
' ac*4 43 )L

not be the situation if the neutron magnetic moments MnO

80.
were all aligned in the incident beam. For the latter p
I
Te ~ I 20'K
293 K

case, the differential scattering cross section contains 60

cross terms between the nuclear and magnetic ampli- 40.
tudes in addition to the above square terms.

MaO 20' Rl'

10 50'
SCATTERING ANGLE
As already mentioned, MnO is thought to be anti-
ferromagnetic below its Curie temperature of j.20'K; Fzo. 4. Neutron diGraction patterns for MnO taken at liquid
nitrogen and room temperatures. The patterns have been cor-
and Fig. 4 shows neutron powder diffraction patterns rected for the various forms of extraneous, di6'use scattering
taken for this material at 300'K and at 80'K. The pow- mentioned in the text. Four extra antiferromagnetic rejections
are to be noticed in the low temperature pattern.
"Shull, %'ollan, and Strauser, Phys. Rev. 81, 483 (1951}.See
also discussion by D. J. Hughes and M. T. Surgy, Phys. Rev. 81, " The nuclear intensities will increase by a few percent due
498 (1951}. to a slight increase in the Debye-%aller temperature factor.
 SH ULL, STRAUSE R, AN D WOLLAN

ment with the data. '9 Table I shows the comparison

between the experimental intensities and those cal-
culated on the basis of this model. The intensities shown
in this table are on an absolute scale and should be
compared in absolute as well as in relative values. In the
calculated intensities, the magnetic form factor for the
3d-shell in Mn++ as determined from the above para-
magnetic scattering experiments (see Fig. 2) has been
used. Conversely, on the basis of the suggested MnO
lattice, the antiferromagnetic rejections can be inter-
preted in terms of the magnetic form factor, and when
this is done, the form factor values shown as discrete
points in Fig. 2 are obtained. The intensity agreement
AL
shown in Table I appears quite satisfactory, with the
CELL exception of the (311) intensity for which the observed
Mn ATOMS IN MnO
value is somewhat lower than the calculated value. As
seen in Fig. 4, the (311) magnetic reflection is unre-
Fn. 5. Antiferromagnetic structure existing in MnO below its solved from the (111) nuclear reflection, so that its
Curie temperature of 120'K. The magnetic unit ceQ has twice the
linear dimensions of the chemical unit cell. Only Mn ions are intensity assignment is subject to considerable error,
shown in the diagram. and, thus, the intensity discrepancy may not be sig-
nificant.
rather than 4.426A as seen in x-ray scattering or in The magnetic lattice is seen to consist of parallel
neutron scattering by the nuclei. The intensities in the sheets L(111) planes] within which all the Mn ions
magnetic reflections fall off regularly with increasing are coupled ferromagnetically but with antiferromag-
angle, which is just what the magnetic form factor netic coupling between neighboring sheets. Of the 12
should require. Because of this weakening of magnetic nearest neighbors surrounding any Mn ion, 6 are
intensity with increasing angle, only a limited number oriented parallel to the central ion and 6 are oriented
of magnetic re6ections are measurable; in this case the antiparallel. Among the 8 second-nearest neighbors,
four mentioned above. If nothing were known about however, there is complete antiparallel coupling with
the MnO structure, it would be dificult and uncon- the central ions. In other words, along the cube axes there
vincing to attempt a magnetic structure determination is always antiferromagnetic coupling from one 3& ion to
on the basis of just four intensities. However, x-ray dif- the next, and this coupling persists through, or is caused by,
fraction studies have already shown the positional rela- an intermediate oxygen ion between adjacent Mn ions.
tionship of the various atoms and the general symmetry This phenomenon of superexchange, in which the elec-
of the lattice. Furthermore, the magnetic data and the tronic wave functions of magnetic ions (Mn) overlap
nature of the quantum mechanical exchange forces those of intermediate ions (0) and thereby result in an
responsible for the magnetic alignment limit drastically exchange coupling of the magnetic ions, was first ad-
the number of possible structures which could exist. vanced by Kramers20 in 1934 and later discussed further
Neel" has suggested several possibilities, and the by Anderson. "
results of these, along with some others, have been Studies have also been made on MnS and MnSe,
compared with the experimental neutron intensities. which are isomorphous with MnO, above and below
their antiferromagnetic Curie temperatures. The pat-
MAGNETIC STRUCTURE OF MnO terns obtained were completely similar to those for
Of the various magnetic structure models which were MnO and hence the magnetic structures are considered
tested, only that shown in Fig. 5 gave satisfactory agree- to be the same as for the oxide.
In the lattice of Fig. 5 it is to be noticed that not all
TAmz I. Comparison between the observed antiferromagnetic '" Reasonable
reflection intensities for MnO and those calculated for the mag- neutron intensity agreement was also found with
netic structure model of Fig. 5. a random sublattice model, but this type of structure does not
correlate with the precision x-ray splitting effects to be discussed
later. On this model, the four simple cube sublattices of the face-
Magnetic Observed integrated Calculated intensity centered-cubic Mn array are considered to have nearest neighbor
reflection intensity (neutrons/min) antiferromagnetic coupling within a sublattice but with no
coupling from one sublattice to another. This model predicts
(111) 1072 1038 neutron scattering intensities in satisfactory agreement with ob-
(311) 308 460
132 servation, but then no distortion of the cubic symmetry would
(331) 129 be expected below the Curie temperature, whereas this is always
(511) 70 54 found. The model shown in Fig. 5 is identical to the random
(333) sublattice model with respect to antiferromagnetic coupling and
divers from the latter only in. having the various sublattices
coupled together in a definite fashion.
'9 L. Noel Ann. phys. 3 137 (1948). ~ H. A. Kramers, Physica 1, 182 (1934).
 NEUTRON DIFFRACTION 339

of the space-diagonal planes ((111) and permutations) TABLE II. Comparison between observed MnO antiferromag-
netic intensities and those calculated for various models of mag-
are equivalent. This results in a reduced multiplicity, so netic orientation with respect to crystallographic axes.
that instead of eight equivalent (111)planes there result
only two which contribute magnetic intensity at the Calculated for various
oriented models
(111) re6ection position. The same situation holds for (a) (b) (c)
Observed
(neutrons/min)
the other magnetic reflections, there being a fourfold
reduction in multiplicity or scattered intensity for each. (111) 1038 0 1072
(311) 460 675 308
(331) 129 109 132
DETERMINATION OF MAGNETIC MOMENT (5»)
DIRECTION IN LATTICE 54 24 70
(333)
It has been mentioned that the magnetic scatter-
ing cross section, and hence the scattered intensity, Scherrer ring under study. The orientation of the mag-
is dependent upon the relative directions of the ionic netic moments relative to this normal will affect the
magnetic moment and the scattering vector. This observed magnetic intensity. All of the other crystallites
dependence has been shown in Eqs. (5) and (9). Since which are not correctly aligned in the above sense con-
the neutron diffraction data are obtained by scanning tribute no intensity and, hence, are not being viewed
across a limited region of the Debye-Scherrer ring, for in the observation.
which the scattering vector is parallel to the normal to
the diffracting planes of the particular crystallites in DEPARTURE FROM CUBIC SYMMETRY IN MIO
the powder sample which happen to be correctly
At room temperature, precison x-ray di6'raction
aligned, one can hope to get information about the studies on MnO have shown it to be strictly cubic in
orientation of the magnetic moments with respect to
symmetry. Recently, however, Tombs and Rooksby"
the diffracting planes.
have shown that, at low temperatures in the vicinity
For the magnetic lattice of MnO shown in Fig. 5 of liquid nitrogen temperature, the x-ray lines are split
there appear three cases of interest concerning the
slightly so that the true symmetry becomes slightly
relative orientation of the moments and the lattice
rhombohedral. This implies that the various (111)
axes, and the observed scattering intensities shouId distances within the unit cell become slightly different
serve to distinguish between these.
at low temperatures. The magnetic structure model of
(a) The magnetic moments are aligned along arbitrary t 100j Fig. 5 is very suggestive of this type of splitting, since,
directions, in other words, along the cube axes, as illustrated in indeed, the lattice forces in the various (111) directions
Fig. 5. For this case it can be sho~n that qs (average) = f for each
would be expected to differ somewhat.
of the four observed magnetic reQections in MnO.
(b} The magnetic moments are aligned perpendicular to the Other oxides such as NiO, FeO, and CoO have been
ferromagnetic (111) sheets. For this case, qs vanishes for the (111) studied, also by Tombs and Rooksby, at various tem-
reQection and becomes 32/33 for the (311) and 32/57 for the peratures, and here again lattice splitting effects are
(331) reQections. found. These oxides all become antiferromagnetic at
(c) The magnetic moments are aligned arbitrarily in the ferro-
magnetic (111) sheets. Here q' should be 1 for the (111) reQection
and various odd values for the other reQections, depending upon
the assignment of moment orientation within the sheet. (III) Mno

Table II summarizes the calculated magnetic inten- loo-

sities for each of these three cases, for comparison with
the experimentally observed intensities. The (111) in- X gyoK
& so-
tensity is quite sensitive to this selection, and it is seen
that case (a) is to be preferred. The data thus suggest z
Ch

Cl
K
that the moments are ali'gned along the cube axes either g eo- -N o [Link]
perpendicular to or parallel to the direction of antiferro IIOoK

magnetic coupting Since the. theory of exchange coupling

'ill l4o
does not correlate the alignment direction with respect a'
lal IR4o K
to the relative positions of the magnetic ions, the above I
K
directional indication must be representative of some RO-
second-order lattice symmetry effect.
It is to be emphasized that the aforementioned con- I I I I
IO IRo IOo IRo IR4
clusions are derivable from random powder samples IO

without the necessity of single-crystal or single-domain SCATTERING ANGLE

study. Only those crystallites in the powder which Fzo. 6. The (111) antiferromagnetic reQection of MnO as ob-
happen. to be oriented with the normal to the Bragg tained at various temperatures. A noticeable broadening occurs
in the vicinity of the Curie temperature.
reflecting planes parallel to the scattering vector will
contribute intensity to that portion of the Debye- » N. C. Tombs and H. P. Rooksby, Nature 165, 442 (1950).
340 SH ULL, STRAUSE R, AN D %0LLAN

bility data taken for a variety of Mn++ salts support

l6. the contention that the ionic moment is equal to that
of the 5 electronic spin moments in the 3d-shell with no
l2- TL (SPEGFIG HEAT) orbital moment contribution.
In addition to MnO, scattering data on other iso-
l- TL (MAOMETIG) morphous oxides of the transition elements, FeO, CoO
lA
8- ( and NiO, have been obtained above and below their
LL)
l- antiferromagnetic Curie temperatures. All of these
oxides show susceptibility and speci6c heat anomalies
4- similar to those of MnO but at diferent temperatures.
Data for these oxides are of interest because the various
magnetic ions possess orbital momentum components as
40' 804 l204 ISO 4
well as spin components in contrast to the above con-
TEMPERATURE (oK)
sidered Mn++ compounds. Since there is a magnetic
Pro. 'tI'. Temperature dependence of magnetic intensity for moment associated with the orbital momentum values,
MnO. The Curie temperatures suggested by speci6c heat and it might be expected that this, as well as the usual
magnetic susceptibility data are shown.
electron spin moment, would contribute to the scattered
neutron intensity.
suitably Iow temperatures, and undoubtedly the lattice
Figures 8, 9, and 10 show the neutron diftraction
splitting e6'ects are to be associated with the appearance
of an antiferromagnetic lattice. This has been discussed patterns for FeO and CoO at liquid nitrogen tempera-
ture and at room temperature and for NiO at room
recently by Greenwald and Smart. ~
temperature. In the latter case, ~' the antiferromagnetic
TEMPERATURE DEPENDENCE OF THE MAGNETIC Curie temperature of 250'C is well above room tem-
INTENSITIES perature, so that it is unnecessary to cool the substance
to invoke the rigid magnetic lattice. The arttiferrorrtag
A study has been made of the intensity in the anti-
metic patterrts for all of these cubic oxides show the char
ferromagnetic pattern at and below the antiferromag-
ac(eristic dogble-size magnetic cell with reQections being
netic Curie temperature. The (111) magnetic reQection
observable at only all-odd indexed positions. All of the
in MnO was scanned, with the specimen held at various
temperatures. Typical di6'raction peaks are shown in magnetic reBections show the characteristic form factor
Fig. 6, and the data are summarized in Fig. 7. A sharp decline of intensity with increasing angle of scattering.
increase is seen to occur as the temperature is lowered There are, of course, pronounced differences in both
through 120'K, followed by a more gradual approach nuclear and magnetic intensity for the diferent sub-
to a saturation intensity. The Curie temperatures sug-
gested. by magnetic susceptibility and specinc heat data (35I)
qp 8.58 a
are also shown in the 6gure. The temperature depend-
80-
ence of scattered intensity suggests that the magnetic
lattice perfection improves as the temperature is
reduced, with fewer and fewer local mistakes or faults
in alignment appearing. In the peak for 124'K in Fig. 6 40-
there is some evidence of line broadening, and a simple
calculation of magnetic crystallite size which would a
account for this broadening yields the value 50A. I 0
(~ 0) [Link]
In the comparison of the various structural models O
f
q,

with experiment it is necessary to use experimental ILI

& 80-
intensities for the mell-ordered lattice. From the inten-
a)
sity dependence shown in Fig. 7, the saturation inten- ILI

sities at 0'K were estimated, and these are the inten-

I

sities shown in Tables I and II.

40-

OTHER CUBIC OXIDES

I
All the data so far presented have been for ma- IO 4)0 50
SCATTERING aNG~
terials containing Mn++ ions. This ion is in a normal
'S&f2 state according to spectroscopic data, and, there- Fzo. 8. Neutron diffraction patterns for FeO taken at liquid
nitrogen and room temperatures.
fore, its magnetic moment should have only electron
spin contributions with no angular momentum com- ~~ Various samples of NiO enriched in the nickel isotopes Ni~,
ponent. As mentioned earlier, the magnetic suscepti- Ni~, and Ni~ have also been examined, and, although the nuclear
reQections show pronounced differences from sample to sample,
~ S. Greenwald and J. S. Smart, Nature 166, 523 (1950). the antiferromagnetic reflections are similar in intensity.
 NEUTRON DI F F RACTION 34i

stances, since the nuclear scattering amplitudes and the

ionic magnetic moments di8er widely.
The intensities and reaction positions in all of the
patterns are consistent with a common structural C
X
model of the type shown in Fig. 5 for MnO, but with Z
O
IBO.
(L
some differences in detail, such as in absolute orientation I
(20.
of the magnetic moments in the lattice and in the
strength of the magnetic moments. These differences Bp.
can be seen most easily when the powder intensities are X
converted to differential scattering cross sections per 4p ~

magnetic ion for all of the magnetic reflections. This

conversion puts the data on an absolute scale and IP' 20' M' 40'
eliminates all of the instrumental and lattice charac- SCATTERIMG AHG&

teristics such as multiplicity, angle of scattering, sample FIG. 10. Neutron diffraction pattern for NiO taken at room
density, etc. A direct comparison is then possible with temperature. The Curie temperature is well above room tem-
the various theoretical predictions of the magnetic cross perature, so the antiferromagnetic re6ections appear here.
section, depending upon the type of spin and angular
momentum coupling in the ion or the degree of orbital axis, i.e. , case (a) was considered to be correct. How-
momentum quenching in the lattice. The di6erential ever, if the magnetic moments in Feg are aligned
scattering cross sections obtained from the experimental perpendicular to the ferromagnetic sheets of the latter
data are summarized in the points of Fig. ii for each (case (b)], then no intensity would be expected in the
of the four simple oxides under discussion. The curves (111) magnetic reflection, whereas the other (311) and
shown in the figure are various theoretical curves; dis- (331) reflections would possess normal intensity. This
cussion of them will be given later. situation could, then, account for the observations on
All of the experimental points in this figure show the FeO. Similar arguments for the CoO and NiO data
expected form factor variation with angle, with the suggest that these are similar to MnO with respect to
single exception of the value characteristic of the inner- moment alignment.
most (111) magnetic reflection of Feo. This particular In order to compare the observed magnetic scattering
reflection is very weak or even absent for Feo (see data with the theory, we note that Halpern and Johnson
diffraction pattern), whereas the outer FeO reflections considered, for convenience, only the case where the
appear normal in intensity and form factor behavior. ionic moment was due to spin only. The generalization
A possible explanation for this follows from the direc- to cases where the magnetic moment is due to combined
spin and orbital angular momentum is formally straight-
tional eGect considerations given in the MnO section.
There it was shown that the magnetic moment align- forward; however, we shall discuss it only brieQy. "
ment in MnO was indicated as being along the cube For the general case we note two important di6'erences
from the spin-only case:
1. The contribution of the orbital moment to the ionic moment
200- will depend among other things upon the type of coupling in the
(5II)
atom, the symmetry of the crystalline field ("quenching"), and
o ~8. 48A the exchange coupling in the lattice (e.g. , ferromagnetism or anti-
ferromagnetism). This will be discussed only for the particular
cases at hand.
120- 2. The form factor to be expected in a case where A&0 will,
in general, be a function not only of (sin8/) ) but also of the orien-
tation of the asymmetric ion with respect to the particular scat-
80- tering planes. For small (sin8/)), this deviation from the form
factor characteristic of spherical symmetry is small, but is not
necessarily small for larger (sin8/X). In the following we shall
I 40-
(III) (200) (RRO) assume the Mn~ form factor (already discussed) for all cases,
~4.R4A
( O
neglecting the deviations from spherical symmetry as well as the
4l
I- I20-
changes expected from diferent values of Z. This procedure,
Z
though justifiable only as a first orientation, is in reasonable
agreement with the experimental data.

In the following we shall, furthermore, assume that

(a) Russell-Saunders coupling holds in all cases, (b) the
50'
I
crystalline fields are stronger than the L S coupling
SCATTERING ANOLE but weaker than the interaction leading to Russell-
Fro. 9. Neutron diffraction patterns for CoO taken at liquid Saunders coupling so that the orbital angular mo-
nitrogen and room temperatures. There is still remnant antifer-
romagnetic intensity at room temperature, which disappears as We are much indebted to our colleague Dr. L. C. Biedenharn
the temperature is increased. for many instructive and illuminating discussions on these points.
 SHULL, STRAUSE R, AND KOLLAN

Fe0 coupling aligned the moments parallel for the free ion,
2.0 the spin and orbital moments are taken to add together
in the crystalline lattice case. For a cubic field the orien-
tation of the spin moment is unimportant to the degree
of quenching, and the eGective magnetic moment for
both Fe++ and Co++ turns out to be just (25+~L).
The magnetic moments predicted from the level
assignments cited [25 for Mn++ and Ni++, (25+s'L)
0 004 CLOS OO4 CLCI 004 OOS OIR 004 for Fe++ and Co++] have been used along with the
(Sle 8) Mn++ form factor to give the theoretical curves for
the magnetic scattering cross section shown in Fig.
Fro. 11. Representation of the magnetic di8erential scattering 11. For comparison, there is also shown in the figure
cross sections corresponding to the various antiferromagnetic
re6ections of MnO, FeO, CoO, and NiO. The points are from (a) the curve (Solid line) for complete quenching,
experimental data; the solid line corresponds to that calculated with the magnetic moment corresponding to 2S and
for spin-only scattering with complete quenching of orbital
moments, the dashed line corresponds to that calculated for the (b) the curve (dashed line) for the free ion, with the
free ion in which the spin and orbital moments add, and the dotted moment corresponding to (25+L) It is t.o be seen that
line corresponds to that calculated for the level splitting in the the data for Mn++ and Ni++ are in good agreement with
crystalline Geld. All of the curves are calculated with the Mn++
form factor and represent a Grst approximation to that which is the expected spin-only curves. This also indicates that
correct for the other ions. the approximation of all of the form factors by the
Mn++ form factor is not grossly in error. For Co++,
mentum is quenched to some degree, and (c) antiferro- and to a lesser extent for Fe++, the experimental data
magnetic coupling leads to a definite spatial orientation are in reasonable agreement with the theoretical curves
of the spin moment for a given atom in the lattice. The based on partial quenching. It is to be noted again that
simplest case is that of Mn++, where the free ion is in the very low value for the innermost (111) Feo re8ec-
the spectroscopic state '5~~2 and therefore displays tion is the result of a diGerent eGect, namely, that q'
none of the diKculties mentioned above. Co++ and vanishes for this re6ection as already discussed. In view
Ni++ are in F-states (L=3) in the free ion (4F9~2 and of the approximations involved, the general agreement
'F4, respectively), and according to Bethe~ an F-level of the observed intensities or cross sections with those
splits in a cubic field into three levels, two triply de- expected appears satisfactory.
generate levels (magnetic) and one nonmagnetic single The aforementioned quenching eGects in Fe++,
level. Van Vleck'4 and Schlapp and Penney" have Co++, and Ni++ compounds appear also in the para-
assigned the single level in Ni++ to be the lowest lying magnetic susceptibility and gyromagnetic ratio mea-
and normally occupied level, and hence, we expect Ni++, surements, and, in fact, the early theory was directed
like Mn++, to show a spin-only magnetic moment. In towards an understanding of these observations. The
other words, the crystalline field has completely
quenched'" the orbital contribution to the magnetic
moment. (Although Ni++ has a spin-only moment, it is a- Fee O~
to be noted that the form factor for scattering may still
display spatial orientation effects. ) As its lowest level X- RAYS
Co++ has one of the triply degenerate magnetic levels,
and therefore, a partial quenching of the orbital moment
is to be expected. Similarly Fe++, 'D4 for the free ion,
has the D-state split by the cubic field into two levels,
one triply degenerate and one doubly degenerate, with
the latter nonmagnetic. Again Van Vleck~ assigns the
triply degenerate level as the normal one in Fe++, and
(Ill@00) (IIO) {Ril& II|OS {RI0~

partial quenching is to be expected.

R
To calculate the expected magnetic moment, ere use 5 40"
the fact that the spins are oriented in the antiferro-
magnetic lattice and this, by virtue of the I 8 force,
orients the remaining orbital moment. Since the L-S ~ RO-
J(

L
~ H. Bethe, Ann. Physik 3, 133 (1929). I
~ J. H. Van Vleck, Phys. Rev. 41, 208 {1932).
R~ R.
ScMapp and %'. G. Penney, Phys. Rev. 42, 666 (1932).
~ This picture is somewhat oversimpliGed for the Ni++ ion IOO IOo ROO 254 304
since recent experiments (see discussion by C. Kittel, Phys. Rev. SCATTERINO ANOI. K
76, 743 (1949)) do suggest some orbital contribution in this ion.
This is not pronounced, however, and, since the magnetic scatter- FIG. 12. Comparison of x-ray and neutron diffraction patterns
ing intensities are so low for this ion, the simple theory was con- for a-Fees hematite. The innermost two reflections are anti-
sidered adequate in initial interpretation. ferromagnetic in origin and are absent in the x-ray case.
 NEUTRON DIFF RACTION 343

effective magnetic moments for these ions as calculated

from the paramagnetic susceptibility show departures
from spin-only values in the fashion described for 40-
crystalline field level splitting, as do also the g-factors.
This is summarized in several tables of Van Vleck2e and
of Stoner. ~ 30-
As mentioned, Tombs and Rooksby" have shown
for all of these simple cubic oxides at temperatures zv
CO

Ro-
below their antiferromegnetic Curie points a slight I
x'
departure from cubic symmetry. Their precision x-ray
studies have suggested (a) a rhombohedral unit cell IO-
with enclosed angle slightly larger than 60' (if strictly
cubic, the equivalent angle would be exactly 60') I l I I

for MnO and NiO, (b) a rhombohedral cell with an ROO 400 eCXP 800 IOOO

angle slightly smaller than 60' for FeO, and (c) a TEMPERATURE {4K)

tetragonal unit cell with axial ratio about 0.995 for CoO. FIG. 14. Variation of O.-Fe203 magnetic scattering intensity with
Here again the neutron scattering similarity from oxide temperature. The Curie temperature T, is that obtained from
magnetic susceptibility data.
to oxide shows the same grouping as does the x-ray line
splitting. With respect to moment alignment in the STUDIES ON e-F6203
unit cell MnO and NiO show the same behavior with no
crystalline 6eld orbital momentum effects with respect The magnetic properties of a-Fe20~ (hematite) have
been studied by many observers, and widely different
to moment alignment. FeO behaves differently, and the
conclusions have been drawn concerning its magnetic
CoO scattering data indicate a considerable coupling of
structure. At room temperature very pure samples
the orbital momentum to the spin momentum. No
show weak ferromagnetic properties, which disappear
detailed picture of the correlation between the scat-
at temperatures above 675'C. This Curie temperature
tering data and the splitting observations are available is suspiciously close to that of Fe304 (magnetite), and
as yet, but it may be that a detailed investigation will Neep' considers that the observed ferromagnetism is
show such a correlation. parasitic and that the basic magnetic structure is that
of an antiferromagnetic material. Very recently, Morin"
(ti~) (ioo) (i~o) has shown the room temperature ferromagnetism to
disappear sharply at a temperature of — 20'C, and
F620& below this the magnetic susceptibility can be accounted
for on the basis of an antiferromagnetic lattice. Further-
more, Morin finds no crystallographic transition (by
0 G
x-rays) above and below this low temperature trans-
formation.
I Neutron diffraction patterns have been taken for this
C/l
X substance over an extensive range, from 80'K to 1000'K,
LLI
I and these indicate the existence of an antiferromagnetic
SSO'C lattice throughout this temperature region but with
some difference with temperature in structural detail.
0
0 0~ The room temperature pattern is shown in Fig. 12 along
with an x-ray pattern for comparison purposes. All of
the observed neutron reflections can be accounted for on
7IO G
the basis of permitted reQections for the chemical unit
cell; therefore, the magnetic unit cell is suggested to
Og be of the same size as the conventional chemical unit
I cell as determined by x-ray diffraction. This is a
IO' lS' 20 rhombohedral cell of edge length 5.42A and rhombo-
COUNTER ANGLE
hedral angle 55'17', with two Fe203 molecules contained
FIG. 13. Inner structure of 0.'-Fe20g diffraction pattern at several
temperatures. The (111) and (100) reflections are magnetic in in the unit cell. The innermost two refiections (111)and
origin and fall off at the higher temperatures, whereas the (110) (100) in Fig. 12 are the most interesting and informative
is a nuclear reflection and is insensitive to temperature changes.
in the neu, tron pattern. On the assumption that all Fe
"J. H. Van Vleck, Electrk und Magnetic Sgsceptibjlities (Oxford ions are alike in scattering power (justifiable for x-radi-
University Press, London, 1932).
~~ E. C. Stoner,
Mugnetism und Mutter (Methuen and Company, "[Link],Ann. phys. 4, 249 (1949).
Ltd. , London, 1934). "F.J. Morin, Phys. Rev. 78, 819 (1950}.
 SHULL, STRAUSER, A ND WOLLAN

is still a remnance of the magnetic reflections. The

(tI) (OO) (to) (tll) (tOO)(tt0)
patterns gave no indication of a reorganization of the
x I20- structure, only a gradual fall-oft' in intensity.
I In addition to the high temperature studies, the
pattern at 80'K has been studied, and here a reor-
ganization of the pattern appears. Figure 15 compares
a portion of the 300'K and 80'K patterns, and it is seen
that the normally strong (111) magnetic reflection has
now become very weak or absent, whereas the (100)
magnetic reflection has increased in intensity. The
remaining reflections in the complete pattern showed
l
little significant change; and these are not of much use
0 I
20'
I

t50 tso 20 in the structural arguments, because their weak inten-

SQATTERtNG ANGLE sity is not resolved from strong nuclear intensity. From
FIG. 15. Comparison of magnetic scattering in a-Fe20& at observations on the pattern intensity when the sample
295'K and 80'K. The magnet;ic intensities show a considerable was being cooled to low temperatures, it was established
rearrangement in cooling the sample below room temperature.
that the transformation occurred in the general vicinity
ation), there should be no scattered intensity in these of the — 20'C transition, which had been found by
reflections, because the crystal structure factors vanish Morin, although this was not studied in complete detail.
in the summation over the unique atoms of the unit MAGNETIC STRUCTURE OF e-Fe20»
cell. There is pronounced neutron intensity at these
As already mentioned, the neutron diffraction data
positions, however, so that the magnetic orientation,
or scattering power, must dier among the four Fe suggest that the magnetic unit cell is the same size as
atoms of the unit cell. TAsLE III. Comparison between observed and calculated F'
Before discussing the magnetic structure of this values for various magnetic structure models of Q, -Fe203. The
bers are values calculated and observed for the differential
material, it is desirable to present the temperature- scattering cross section F' in absolute uruts of 10M4 cm 2/Fe2 3
0 ~
eo
dependent data which have been taken. The region of molecule.
the diffraction pattern containing the (111), (100), and Calculated for
(110) reflections was studied at various elevated tem- Model (a} Model (b) Observed
peratures up to 730'C. The latter reflection is wholly (I) (I I) (II I) (I) (II} (II I) 300oK 80 K
nuclear in origin and should change only slightly due to {111) 1.25 0 4.3 0.23 0 0.81 4.9 &0.05
the Debye-%aller temperature factor, whereas the (100) 1.40 1.59 0.96 2.32 2.64 1.59 0.91 1.37
other two magnetic reflections should exhibit charac-
teristic Curie temperature behavior. Figure 13 shows the chemical unit cell and, hence, that there are four
typical curves at temperatures of 20', 650', and 710'C, uniquely positioned Fe atoms in the cell, shown as
and it is seen that the nuclear reQection remains essen- ABCD in Fig. 16. There are three possible antiferro-
tially unchanged, whereas the magnetic reflections are magnetic arrangements of relative spin orientation for
sensitive to temperature changes. Figure 14 summarizes these four atoms represented by Model (a)
——, —
—+ — +,
the magnetic intensity data, after correction for the Model (b) ++— and Model (c)
— indicates that atoms A and D have
+ + —.
Here
Debye-%'aller temperature factor as determined in the Model (a) + +
uclear reflection at various temperatures. The inten- parallel orientation (say, directed upward) while 8 and
sity variation is not nearly as pronounced as In thee C both have antiparallel orientation (downward) with
MnO case and, in fact, falls o6 with essentially a linear respect to that of A and D. Of these models, the crystal
dependence on temperature. Even at the Curie tem- structure factors for Model (c) vanish in the (111) and
perature of 675'C, suspected from magnetic data, there (100) reflections, and hence, this model can be immedi-
ately excluded. Models (a) a, nd (b) both predict finite
intensities for these reflections, so it is to be hoped that
the neutron data can show preference for one over the
other. The observed intensities will depend not only
upon the above model but also upon the orientation of
the moments in the model with respect to the dift'racting
planes. For the absolute sense of orientation, the three
most reasonable possibilities which have been con-
sidered are:
I. The moments are directed along the unit cell edges.
F G. 16. Magnetic structure of ar-Fe203 at room temperature. II. The moments are directed along the space diagonal of the
The rhombohedral unit cell containing four unique iron atoms is unit cell and thus are perpendicular to the ferromagnetic (111)
shown. sheets in the lattice.
 PROPERTIES OF (y, e) CROSS SECTIONS
III. The moments are in the (111) sheets and directed towards
for the directional properties of the weak ferromag-
one of the three nearest neighbors in the sheet. netism in single crystals of O.-Fe&O3 observed by Smith"
Crystal structure factors have been calculated for these many years ago, Neel envisages the magnetic moment
various cases using the magnetic scattering amplitude direction in the magnetite inclusions and in the O.-Fe~03
for Fe+++ given by Eq. (6) with S=5/2, there being 5 layers to be within the (111) sheets of the lattice. This
electrons in the 3d-shell which contribute to the mag- is just the conclusion drawn from the neutron scattering
netic moment just as in the Mn++ ion. Table III sum- observations.
marizes these calculations for the (111) and (100) mag- The magnetic lattice for O, -Fe203, shown in Fig. 16,
netic reflections for comparison with the observed exhibits characteristics similar to those of the magnetic
values. All the values are given as diGerential scat- lattice for the simple cubic oxides shown in Fig. 5.
tering cross sections per Fees molecule, and both When additional unit cells to those shown in Fig. 16
calculated and observed values are on an absolute scale. are visualized, it is seen that the structure consists of a
It is seen that the room temperature data suggest series of (111) sheets, within which all moments are
Model (a) with orientation III as being correct, while arrayed ferromagnetically but with antiferromagnetic
at low temperatures the suggested structure is that of coupling between neighboring sheets. Interestingly,
Model (a) with orientation II. Thus, the low tem- there are sheets of oxygen ions between each of the
perature data could be accounted for as simply a reorien- antiferromagnetically coupled sheets of iron ions. The
tation of the magnetic alignment from within the (111) alternating ferromagnetic sheet structure with inter-
sheets to one perpendicular to these sheets when the mediate oxygen planes was just that found in the cubic
temperature is lowered. oxide magnetic structure.
The orientation results for the room temperature %e wish to acknowledge our appreciation to Mr.
lattice 6t in rather well with Neel's picture of the weak, %. C. Koehler and Mr. R. A. Erickson who obtained
parasitic ferromagnetism which is always observed. In and analyzed some of the later period data of this inves-
his picture there exist small, distorted crystallites of tigation, and to Dr. L. C. Biedenharn for many helpful
magnetite intimately mixed with normal o.-Fe203 layers discussions on the theoretical aspects of data inter-
in the layer structure built up along the trigonal axis pretation.
perpendicular to the (111) sheets. In order to account ~ T. Townsend Smith, Phys. Rev. 8, 721 (1916).

PHYSICAL REVIEW VOLUME 83, NUM BER 2 JULY 15, 1951

Properties of (q, n) Cross Sections*

L. MARsHALL
INstitute for Nuclear Studfes, University
of Chicago, Chicago, Illinois
(Received March 12, 1951)

The mean energies of photons producing the reactions Cu™(y,n)Cu~, Zn~(y, e)Zn~, and C'~(y, e)C"
have been found by measuring their absorption coefBcients for many values of Z. Monitor and detector were
made of the element investigated. The absorbers were Be, C, Al, Ti, Fe, Ni, Cu, Zn, Se, Mo, and Sn. Less
than 1 percent of the radiation striking the detector originated in the absorber. In this geometry, (1/Z)
Xabsorption cross section was a linear function of Z. The Compton cross section was separated from the
pair cross section by considering its difFerent Z-dependence. The mean energy was evaluated in two ways:
(1) from the observed pair cross section and the Bethe-Heitler formula, and (2) from the observed Compton
cross section and the Klein-Nishina formula. A correction was applied for the deviation of pair cross section
from the value given by the Born approximation. With this correction, the values of the mean energy were
found to be as follows:
from Compton cross section from pair production
Zn~ 16.6 Mev 20.0 Mev
Cue' 17.3 21.0
Cls 23 32
The integrated (y, e) cross sections for Cr and Zn~ have been found to be 0.086' 10 4 Mev-cm~ and
0.77&10 Mev-cm~, respectively.

PART I or more neutrons, protons, heavier charged particles,

HEN high energy gamma-rays fall on a nucleus, gamma-rays. In all nuclear species which have been
~
a reaction is usually observed. that corresponds examined so far the nuclear absorption seems to rise
~

to absorption of the gamma-ray with emission of one to a maximum and then fall oft again with increasing
* This work was supported by the joint program of the QNR energy of the photon. As yet none of the nuclear species
and ABC. examined up to 100 Mev sho~s a second region in which