Vot, +MR 21, NUMRRR $ QOToBER, 19 19

. . . 157sica. . .. . 1eery eI.' .~ erromagnetic . Domains

CHARLES KITTKL
Bel/ Telephone Laboratories, Murray Hill, 1Vew Jersey

1. Survey of Domain Theory been obtained confirm the theory in all essential aspects
1.1 Introduction and thus give one conhdence that the more complicated
1.2 The domain assumption
1.3 The origin of domains situations obtaining in polycrystalline materials may
1.4 Coercive force, hysteresis, and reversible permeability be understood, at least in qualitative terms, on the
2. Contributions to the Domain Energy basis of existing theory.
2. 1 Exchange The present paper is undertaken as a comprehensive
2.2 Magnetocrystalline (anisotropy)
review of the physical principles of domain theory and
2.3 Magnetoelastic
2.4 Magnetostatic of the crucial experiments which bear directly on the
3. The Bloch Wall foundations of the subject. The theory of the origin of
3.1 Introductory remarks domains is not treated adequately in the existing text-'
3.2 Estimate of thickness and energy of Sloch wall books on ferromagnetism; for example, none of the
3.3 180' walls in (100) plane of iron
4. Theoretical Domain Structure existing textbooks' discusses the original paper in this
4. 1 Introduction field, written by Landau and Lifshitz in 1935.
4.2 Flux closure domain configurations The organization of the present paper falls into two
5. Experimental Studies of Domains stages. There is erst an introductory survey which
5.1 Powder pattern technique
5.2 Powder pattern results describes the basic physical concepts of domain theory
6. Magnetic Properties of Small Particles in general non-mathematical terms. The ideas are then
6.1 Critical particle size for single domain behavior developed in detail in the subsequent sections of the
6.2 Coercive force of small particles
Permeability and Coercive Force
7. Initial paper.
7.1General remarks
7.2Kersten inclusion theory 1.2 The Domain Assumption
7.3Strain theory
7.4Neel magnetization Quctuation theory The essential aspects of ferromagnetism are illus-
Appendix A: Mathematical expressions for the anisotropy of trated by the implications of the following experimental
cubic crystals. fact
Appendix 8:
Magnetic interaction energy of dipoles on a cubic
It is possible to change the oner all magnetization -of a
lattice.
Appendix C: List of Symbols. suitably prepared ferromagnetic specimen from an initial
value of zero (in the absence of an applied magnetic geld)
1. SURVEY OF DOMAIN THEORY to a saturation value of the order of 1000 gauss, by the
1.1 Introduction application of a geld whose strength may be of the order
' 'N recent years there has been a major growth of the of 0.01 oersteds
Such a magnetization curve is shown in Pig. j.. The
i. body of experimental and theoretical knowledge magnetization is defined as the magnetic moment per
regarding the origin and behavior of ferromagnetic unit volume. The ordinate axis of the curve is, however,
domains. We possess now a general theoretical founda- which in this case is closely equal to 4x
Qux density,
tion for the subject, a foundation which is intellectually
times the magnetization,
satisfying and which has been verified experimentally
in considerable detail in several simple situations in ~ A Russian book has just appeared which gives a brief treat-
ferromagnetic single crystals and, in less detail, in very ment of this subject (Vonsovsky and Shur, 1948). (An alphabetic
fine ferromagnetic powders. The results which have list of all references is given at the end of this article. )
5 41
 CHARLES KI TTE L
x f03 independent elementary magnetic moments. For ex-
20
ample, at room temperature a field of 0.01 oersted will
increase the magnetization of a paramagnetic salt such
as ferrous sulfate (FeS04) by about 10 ' gauss, as
12
compared with 10' gauss in the ferromagnetic specimen.
The small e8ect in the case of the paramagnetic salt is
known to be caused by thermal agitation which acts to
oppose the ordering inQuence of the applied magnetic
field. In the paramagnetic salt electively only one
magnetic moment in 10' is "oriented" by a field of
c/l
0.01 oersted, so that the distribution of magnetic
0.
moment directions remains essentially random. This
X
high degree of chaos is, as we have said, the result of
Cl 4 the predominant role played by thermal agitation in
a system where the electron magnetic moments are
independent, without important mutual interactions.
Pierre gneiss (1907) pointed out that the difficulty
caused by thermal agitation could be largely circum-
vented if one postulated in ferromagnetic materials the
-16 existence of a powerful internal "molecular" field; we
describe this now as a mutual interaction between
-20 electrons which would tend to line up the magnetic
-0.08 ~06 -0.04 -0.02 0 0.02 0.04 0.06 0.08
H IN OERSTEDS moments parallel to one another.
FIG. 1. Magnetization curve of single crystal of silicon iron. The The required magnitude for the Weiss molecular field
8 scale is only approximate. (Williams and Shockley, 1949.) may be estimated readily. At the Curie temperature
T, the thermal energy kT. of an electron spin is of
The fact just referred to contains two significant the same order of magnitude as the interaction energy
observations: p~H J of the Inagnetic moment p~ of an electron acted
on by the effective molecular field H J . .
(a) It is possible in some cases to attain saturation magnetiza-
tion by the application of a very weak magnetic Geld. kT~= p~Hm~, (1.2. 1)
(b) It is possible for the magn. etization gf the same specimen
to be zero in zero (or nearly zero) applied 6eld. so that'
The first observation is remarkable since it is known H y= I4T./14m =10 "10'/10 "=10' oersteds. (1.2.2)
from the study of paramagnetism that the application
This is an exceedingly powerful effective held; it is
of a field of 0.01 oersted has an entirely negligible
about twenty times more intense than any actual
effect on the magnetization of a system of free and
magnetic Geld produced in a laboratory. At tempera-
tures below the Curie temperature the eBect of the
molecular field outweighs the thermal fluctuation energy
and the specimen is accordingly ferromagnetic. Mag-
netic moment orientations in paramagnetic and ferro-
magnetic materials are illustrated schematically by
Fig. 2; the variation of the saturation magnetization of
(o) (e) iron as a function of temperature is plotted in Fig. 3.
PARAMAGNETIC SALT
{OR FERROMAGNET
F ERROM AGNET
AT VERY LOW TEMPERATURE
It is known now that the origin of the molecular
ASOVE CURIE POINT) field lies in the quantum-mechanical exchange force;
another and better known manifestation of this force
is the chemical valence bond, although in the case of
the chemical bond the exchange force usually tends to
make the spins of neighboring electrons anti-parallel,
instead of parallel as in the case of ferromagnetism.
Weiss himself did not make any specific predictions
about the origin of the molecular field, but he did point
(c) {d) out that the ordinary magnetic moment interaction
FERROMAGNET FERRONIAGNET between electrons is much too weak to account for the
AT LOW TEMPERATURE JUST BELOW CURiE POINT molecular field. The magnetic held at one lattice point
FIG. 2. Orientation of magnetic moments of electrons in
paramagnetic and ferromagnetic materials. ' A list of symbols is given in Appendix C.
 P H YS I CAL THEORY OF F E RROMAGNE TI C DOM A I NS

arising from the magnetic moment of an electron at a ).0

~~
neighboring lattice point is of the order of

—=
p, g 10 " 0.8— ~ ~
~ ~
H= = 1000 oersteds, (1.2.3) ~4
r3 (2X 10-')' 0.6—
0'0
&ce
which is smaller than the effective molecular field H ~ 0.4-
by a factor of the order of 10 4. The magnetic moment
interaction by itself would lead to a Curie temperature 0. 2—
in the neighborhood of 0.1'K. The situation is quite
diferent in dielectric materials, as the order of magni- 0 I I I 1 I I l

0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00

tude of electric dipole moments is. about 100 times T
larger than magnetic dipole moments, leading to inter- Tc
action energies 10' larger. It is therefore not surprising FIG. 3. Saturation magnetization of iron as a function of
to find materials which are ferroelectric at room temper- temperature. At room temperature the saturation value is 2.0
percent below that obtaining at zero degrees Kelvin.
ature, as a result of electric dipole interactions.
We have now seen how the existence of the powerful We therefore see that Weiss was able to explain the
Weiss molecular field enables saturation magnetization
principal aspects of ferromagnetism by means of two
to be obtained. How then do we explain statement (b) assumptions: the existence of the molecular field and
above, that it is possible for the magnetization to be the existence of domain structure. Weiss did not justify
zero in zero applied field? It seems at first sight contra- either of these assumptions in terms of atomic forces.
dictory, in view of the 10' oersted molecular field, to The explanation of the molecular field in terms of
'
suppose that a 10 oersted applied field can alter the exchange forces was contributed by Heisenberg in 1926,
magnetic moment of the specimen by an appreciable and the explanation of the origin of domains in terms
amount. of magnetic field energy was given by Landau and
Weiss extricated the theory from this difficulty by Lifshitz in 1935.
making the assumption that actual specimens are
We go on now to consider qualitatively the causes
composed of a number of small regions called domains,
responsible for the formation of domains.
within each of which the local magnetization is satu-
rated; the directions of magnetization of different 1.3 The Origin of Domains
domains need not necessarily be parallel, however. A
schematic arrangement of domains with zero resultant In this section we shall show that domain structure
magnetic moment is shown in Fig. 4a for a single is a natural consequence of the various contributions
crystal. In polycrystalline samples it was imagined by —
to the energy exchange, anisotropy, and magnetic-
early workers that each crystallite might contain a of a ferromagnetic body. But first it is appropriate to
single domain, and that the resultant magnetic moment consider brieQy the experimental evidence for the
could be zero by virtue of a random distribution of existence of domains. We have seen already that the
grain axes, as indicated in Fig. 4b. existence of domains may be inferred from the character
The increase in the value of the resultant magnetic of the magnetization curve itself. But by far the most
moment of the specimen under the action of an applied
magnetic field may be imagined to take place on the
domain theory by two independent processes, as was /

suggested by R. Becker: by an increase in the volume t

of domains which are favorably oriented with respect
to the field at the expense of unfavorably oriented
FIG. 4. Schematic domain
domains; or by rotation of the directions of magnetiza- arrangements for zero resultant to)
tion towards the direction of the field. These two SINGLE CRYSTAL
magnetic moment in a single
methods by which the resultant magnetization may crystal (a) and in a polcrystal-
line specimen (b). The domain
change are shown in Fig. 5. structure of the polycrystalline
It turns out on closer examination that in weak specimen has been drawn for
simplicity as if each crystallite
fields the magnetization changes usually proceed by contained only a single domain;
means of domain boundary displacements, so that the this is not usually the case.
domains change in size. In strong fields the magnetiza-
tion usually changes by means of rotation of the
direction of magnetization. A typical magnetization
curve is shown in Fig. 6, with the regions designated (b)
in which each process is dominant. POLYCRYSTAL
 CHARLES KITTEL
direct and cogent evidence of domain structure is this structure has been given in detail by Williams,
furnished by micro-photographs of domain boundaries Bozorth, and Shockley (1949).
obtained by the technique of magnetic powder patterns. We may understand the origin of domains by con-
This method, applied originally by Bitter (1931), has sidering the structures shown in Fig. 9, each repre-
in the hands of H. J. Williams and his collaborators senting a cross section through a ferromagnetic single
(1947—49) provided ample and convincing proof that crystal. In (a) we have a saturated configuration con-
domains exist in the shapes and sizes expected theoreti- sisting of a single domain; as a consequence of the
cally, and furthermore that they behave under applied magnetic "poles" formed on the surfaces of the crystal
mechanical and magnetic forces as predicted by theory. this configuration will have a high value of the magnetic
The powder pattern method consists in placing a energy (1/Sir) J'IPd V. The magnetic energy for a
drop of a colloidal suspension of finely divided ferro- square cross section will be of the order of '=10' I,
magnetic material, such as magnetite, on the carefully- ergs/cc; here I, denotes the saturation magnetization.
prepared surface of the ferromagnetic crystal under
In (b) the magnetic energy has been reduced by a
factor of roughly one-half as a result of dividing the
study. It is found on observation through a microscope
that the colloid particles in the suspension become crystal into two domains magnetized in opposite direc-
tions. The subdivision process may be carried further
as in (c): with 1V domains it turns out that the magnetic
energy is reduced (because of the reduced spatial exten-
sion of the field) to approximately 1/X of the magnetic
energy of the saturated configuration (a).
The subdivision process may be expected to continue
until the energy required to establish an additional
UNMAGNETIZED
boundary layer or interface, separating two domains
magnetized oppositely, is greater than the reduction in
magnetic field energy consequent on the finer subdivi-
sion. It may be appreciated that a boundary layer does
indeed have a certain amount of energy associated with
it: on opposite sides of the boundary the magnetization
is directed in anti-parallel directions; now since the
FIG. 5. Funda-
mental magneti- exchange forces favor parallel and oppose anti-paralleI.
zation processes. orientations of the magnetization, it will naturally
M AG NET IZ E D BY
require the expenditure of energy to establish a bound-
DOM A IN GROWTH ary layer. In Section 3 we shall calculate this energy
(BOUNDARY. DISPLACEMENT after an examination of the nature of the boundary
layer, and we shall find that the energy is of the order
of 1 erg/cm' of boundary surface. If then we suppose
tentatively that there are N=10' domains per centi-
meter, the total boundary energy in a crystal cube
one cm on each edge will be of the order of 10' ergs
and the magnetic energy will also be of the order of
10' ergs. This situation represents approximately the
equilibrium number of domains for the particular geo-
MAGNET IZED BY
metrical arrarlgemerlt shove.
DOMAIN ROTATION
It is possible to devise domain arrangements such
as (d) for which the magnetic energy is zero. In (d)
strongly concentrated about certain well-defined lines the boundaries of the triangular prism domains (termed
which represent the boundaries between domains "domains of closure" ) near the end faces df the crystal
magnetized in different directions. The reason why ——
make equal angles 45' with the magnetization in
the colloid particles concentrate near these boundaries the rectangular domains and with the magnetization
is that in their vicinity there exist very strong local in the domains of closure: therefore the component of
magnetic fields which attract the magnetic particles. magnetization normal to the boundary is continuous
A photograph of a relatively simple domain structure across the boundary, and no poles are formed anywhere
in iron is shown in Fig. 7, along with the interpretation in the crystal. As there are no poles there is no magnetic
derived from the photograph and from certain auxiliary field associated. with the magnetization, and we may
experiments. A more complex type of domain structure speak of the Qux circuit being completed within the
is shown in Fig. 8; structures with this general "tree" —
crystal thus giving rise to the phrase "domains of
character arise when the crystal surface is slightly closure" for the domains near the surfaces of the
inclined with respect to a cube face; the explanation of crystal which act to complete the Qux circuit.
 PH YSI CAL THEORY OF FERROMAGNETI C DOMAINS

The extent to which the subdivision of the closure

configuration (e) proceeds will depend on the energy
requirements of the domains of closure. It is not
immediately obvious that the optimum closure con-
6guration of type (e) will necessarily have a lower
energy than the optimum butt-end configuration of
type (c), and in fact approximations to both types of
termination are found in diGerent materials.
The energy required to form a domain of closure in
an uniaxial crystal such as cobalt comes principally
from what is called the crystalline anisotropy energy.
The anistropy energy tends to make the magnetization
of a domain line up along certain crystallographic axes.
The axes thus favored are known as preferred axes, or
axes of easy magnetization. Such axes are well-estab-
Frc. 6. Representative magnetization curve, shoveling the dominant
lished experimentally, and it is known that a consider- magnetization processes in the different regions of the curve.
ably larger amount of energy may be required to
saturate a specimen along an arbitrary axis than along 45' lines shown in (a); when part of the crystal is
one of the preferred axes. In cobalt the hexagonal axis magnetized clockwise and part counter-clockwise, then
of the crystal is the only preferred axis, and cobalt is the square-shaped boundary in (b) is formed in addi-
accordingly referred to as uniaxial. In iron, which is tion. Magnetization changes are then found to take
cubic, the preferred axes are the cube edges; in nickel, place by the movement of the square-shaped boundary;
which is also cubic, the preferred axes are the body it was observed that the flux changes for the magnetiza-
diagonals. Fig. 10 shows magnetization curves for Fe, tion curve plotted in Fig. 1 actually corresponded
Xi, and Co, in directions of easy and hard magnetiza- quantitatively to the displacements of domain wall.
tion.
In cobalt, if the basic rectangular domains are
Sise of Domains
magnetized along the easy axis of magnetization, then
the domains of closure wi11 by necessity be magnetized An essential feature of the new viewpoint regarding
in hard directions. In a cubic crystal such as iron it is domains is that the size of domains in a single crystal
possible for both the basic domains and the closure is expected to be largely a function of the size and shape
domains to be magnetized along di8erent easy axes. of the crystal. That is, the "size" of a domain is not a
The energy expenditure in this case arises from mag- fundamental length of physics, but is rather fairly
netostriction: since the closure domains are magnetized sensitive to the actual crystal in hand. The volume of
along different axes from the basic domains they will certain superficial domain structures may be relatively
tend to be elongated by magnetostriction along different small (say 10 ' cc), while the underlying domain
axes, and in order to fit the various domains together structure will usually contain larger domains ( 10 '
in the crystal structure we have to do work against cc, perhaps). The limit to domain size in a properly
elastic forces.
The termination structures revealed by powder s33 +III
patterns are often more complicated than the simple liiiII ~~ Ilg, gIIR
I ',
cases we have discussed. But the underlying principles p
are always the same: domain structure has its origi ie
w» Pj'j,
the possibiLity of Lowering the energy of a system by going
I

from a saturated configuration such as (a) with high i

magnetic energy to a domain configuration, such as (c)
or (e), with a lower energy. The domain structures
found are essentially induced by the presence of the
external surfaces of the crystal, and it is therefore not
unexpected to find that the domain structure may be
altered radically by changes in the crystal surfaces.
A particularly simple type of domain structure is
shown in Fig. 11; this structure has been realized by
Williams and Shockley (1949) with a single crystal of
silicon iron which was cut to the form of a hollow
rectangle with legs accurately parallel to $001j and
L010$ crystal axes. When the crystal is saturated FIG. 7. Simple domain structure in Si-Fe single crystal,
entirely in one sense the d.omain boun. daries are the (Williams, Bozorth, siid Shockley. ) (&(500)
546 CHARLES K I TTE L
the crystallite axes are not necessarily distributed at
random, but a considerable degree of orientation may
exist. For example, in the case of rolled iron tape' the
crystallites have a strong tendency to line up in such a
way that the rolling plane is a (001) plane, and the
rolling direction is a 110j direction.
I

If there is a very high degree of orientation, the

whole polycrystalline specimen may be expected to
behave with respect to domain structure more or less
as a single crystal.
On the other hand, if the crystallites are oriented at
random, so that there is a good chance that the direc-
tions of easy magnetization of adjacent crystallites
make fairly large angles with each other, then each
crystallite may be expected to behave largely as if it
itself were a single crystal isolated from the neighboring
FB'. 8. Complex domain structure in Si-Fe single crystal.
crystallites.
(Williams, Bozorth, and Shockley. ) ()&500) We see then that, in principle, the size of a domain
may be either larger or smaller than the size of a
shaped crystal can be determined to order of magnitude crystal grain in polycrystalline material. It is, however,
mainly by the crystal size itself. quite unusual to have a sufFiciently high degree of
orientation with a single domain to encompass com-
Polycrystalli rIe Materials pletely several grains. Figure 12, which is due to
Williams, shows a case in which domain boundaries
The greater part of this paper is directed specifically
are practically continuous across more than one grain.
towards understanding the domain structure of single
We may also get a certain amount of information
crystals. The magnetic materials used commercially
regarding domain dimensions in polycrystalline speci-
are, however, always polycrystalline; that is, each
mens by means of measurements of the depolarization
specimen is composed of a large number of small
of polarized neutron beams passing through the ferro-
crystallites. In many materials the crystallites are
magnetic specimen.
oriented more or less at random, so that for some
purposes certain properties of the polycrystalline speci- 1.4 Coercive Force, Hysteresis, and
men may be obtained by averaging the corresponding Reversible Permeability
single crystal property over al1. directions. In other
materials, in particular those subjected to cold working, The coercive force4 is perhaps the most sensitive
property of ferromagnetic materials which is subject to
our control, and is one of the most important criteria
in the selection of ferromagnetic materials for practical
N N S S S N S N
applications. The essential diQerence between material
I
I

I
I

I for permanent magnets and material for transformer

I
I
cores lies in the coercive force, which may range from
~'
I&&i
I
I I I I
the value of 600 oersteds in a loudspeaker magnet
&' I

(Alnico V) and 20, 000 in a special high stability magnet

I

I I
I I
I

I
I I

I
(Fe-Pt) to the value of 0.5 in a commercial power
I
I
I
I I g
transformer (Silicon-iron) or 0.004 in a pulse trans-
S S N N N S N S former (Supermalloy). Thus the coercive force may be
varied over a range of 5&(10 .
(a) (b) (c) 3 Yager has given a beautiful demonstration of the orientation
y~/ ~~i anisotropy in rolled iron tape by means of a microwave resonance
Y y Y
I
I I
experiment.
'r
I
4 The coercive force
H, is defined customarily with reference to
I I
the saturation hysteresis cycle (Fig. 13) as the value of the
I I magnetizing field corresponding to the point B=O; that is, it is
I

I
I the reverse field necessary to carry the induction from the satura-
A,
I I
tion value down to zero. In theoretical work, however, it is often
l
more direct to consider the coercive force as the field corresponding
I

A to the point for which the magnetization I —

is zero that is, where
B—H is zero; when used in this sense the coercive force is denoted
(e) by the symbol IH, . The distinction between H. and IH, is sig-
—
nificant only for large values of the coercivity of the same orde&
Fro. 9. The origin of domains. as the saturation magnetization I, .
 P H YS I C A L TH E0 R Y 0F FER R. 0 lVI A G N E T I C 0 0 NI A I N S 547

1600 p oj
[I t
31001
[0001]

~ 1200 1200
(
FIG. 10. Magnetization curves goo]
for single crystals of Fe, Ni, and z 800 800
Co. (Honda and Kaya, 1926; 200
F1010]
Kaya, 1928.) 400 4oo
100
Ni
00 200 400 600 0 200 300 00
100 2000 4000 6000 8000
H IN OERSTKDS

The problem of the theory is to interpret the observed taining a precipitated phase are Inagnetically hard.
values of the coercivity in terms of the physical state While there is little doubt of the general correctness of
of the material and to predict methods by which the our picture there remains the question of the detailed
coercivity may be increased in magnetically hard quantitative correlation of coercivity with specific
materials and decreased in magnetically soft materials. physical factors; the examination of this question we
Several theories have been advanced (Becker, 1932; defer until Section 7.
Kersten, 1938, 1943; Neel, 1946; Stoner and Wohlfarth, The coercive force of one type of magnetically hard
1947—48) and a certain amount of progress has been material may be understood from a quite diferent
made, although the problem is beset with the usual picture; we refer to materials composed of very small
difhculty in explaining any material property which is grains or fine powders where each particle is always
highly structure sensitive, namely, the difhculty in magnetized to saturation as a single domain. The fact
determining quantitatively the relevant physical factors that a suSciently small particle, with diameter less
— such as impurities, lattice imperfections and internal than 10 4 or 10— ' cm, is composed of a single domain is
strains. a result of domain theory which has been confirmed by
When we have understood the coercive force we will experiment. It can be shown that with such very small
be a long way towards understanding the saturation particles the formation of a domain boundary is
hysteresis loss at low frequencies, since the area energetically unfavorable: this is essentially because
enclosed by the hysteresis loop (Fig. 13) is a'pproxi- too large a proportion of the volume of a small particle
mately given by the product of the saturation induction would be contained within the wall the wall thickness —
8, times the coercive force. That is, the energy dissi- being independent of the particle size.
pated on going once around a hysteresis loop is of the If a small particle is constrained to remain as a single
order of B, H„ to within a factor of 2 to 4. We may, domain it will not be possible for magnetization changes
therefore, devote our attention to the single factor H, . and reversal to take place by means of the process of
The coercive force in "magnetically soft" (low II, ) boundary displacement which usually requires rela-
materials may be understood from the following tively weak fields; instead the magnetization of the
picture: The total energy of a given specimen may particle must rotate as a whole (Fig. 16), a process
vary depending on the position of a domain boundary, which may require large fields depending on the ani-
as a result of local variations in internal strains, sotropy energy of the material or the shape of the
impurities, crystallite dimensions, etc. ; the variation is particle: this is because we must rotate the magnetiza-
indicated schematically in I'ig. 14. In the absence of
an applied magnetic field the boundary will be situated I
I I
at some minimum position such as (A) in the figure. I
I
I

In the presence of a field the boundary will be unable I

I
I
I
I
to make a large displacement to the extreme right (D) I
I
I
I

unless the energy is increased by a sufhcient amount to I I

I

8
I

enable the boundary to pass over the point corre- )

I

II& I/
I
I I&
I

sponding to the maximum boundary energy.

I
I I

The increase in energy must be furnished by the I

I
I
I

reorientation of the local magnetization I,

in the applied I
I
I
I
I

field H, and the value of H which suKces to reverse I

I
I
I
I
about one-half of the magnetization of the specimen I
I I
I

will be the coercive field H. .

I
I
I y

this picture of the coercive process /

Qualitatively
explains the fact that the coercive force diminishes as (b)
the (precipitated) impurity content decreases (Fig. 15)
FIG. 11. Simple domain structures in single crystal of iron in
and also as internal strains are removed through form of rectangular loop, with legs parallel to [001$ and [010j
annealing; it also explains why it is that alloys con- axes.
 CHARLES KI TTF L

II IR2
I ~ P

..., +~N[~pi, n

PlRt~~
J8 '-

= ~iIiill8@Q

11661
'"' Illa'
I...&2:"'"
Frc. 12. Domain structure in
silicon iron grain-
6% 6'IWk 26
' ~2 "2l~akF~ ~
II ~ a] I
polycrystalline
oriented (Williams). ()& 500)
PQ

16 ~
111 ~ m

$ lI'
P II.
'~ I.III
&I I

6 .611
0 IFVSii ---- I

. sai~
I I ggml —
P —
qII Il pP"

, 122&2
R'
II 2
—~
~

611.,
12~ '
.
I
--
~K
'%II
. I

@I ! 'I" g gP:I KRKh I' JR6~IQIRB2~2', :QNI 122265P

-"I l'thha I Rl»-
III I 2 ~ ,121

'%P)62, ~ :.
I,. ;=;-'l:44 aSS W
~ -12'"
'1
. "62'
III' I ll1 I

!12. "I

tion over the energy hump corresponding to a direction If the small particles possess an elongated shape we
of hard magnetization. may have a high coercivity because of the anisotropy
The coercive force of fine iron particles is expected of the energy in the demagnetizing field, even if the
theoretically to be about 250 oersteds on the basis of crystalline anisotropy energy is low. That is, the
rotation opposed by the crystalline anisotropy energy, magnetization tends to line up along the long axis of
and this is of the order of the value reported by several the specimen, and a strong field may have to be applied
observers. Similarly the high coercivities of the com- to turn the magnetization through the short axis. This
pounds MnBi (IH, &12,000) snd FePt (zII, =20,000) appears to be the explanation of the high coercive force
seem to be in line with the rotation concept, with of the alloy FeCo in fine powder form; the alloy is
anisotropy energy as the tactor opposing rotation. known to have a low anisotropy energy from single
crystal measurements, so that the anisotropy energy
alone cannot explain the observations, but the shape
—Bs e6ect must be invoked.

Reversible Permeability
The extent of the range of field strength over which
the permeability is reversible is determined by the
distance through which a domain boundary may move
without passing over a peak in the curve of wall energy
vs. distance; with reference to Fig. 14, one such region
of reversible permeability is the region CAB when —
the domain boundary leaves this region it moves
irreversibly to the extreme right or extreme left of the
figure.
The reversible permeability is determined by the
irregularities of the curve of boundary energy vs.
displacement, and thus is determined by essentially
the same physical conditions as the coercive force. A
comparison of the initial permeabilities p, ~ and the
coercive force II, for a wide range of magnetic ma-
terials is shown in Fig. 17. It is seen that there is a
very close correlation, materials with high coercivities
Pro. 13. Definition of the coercive force. having low permeabilities, and vice versa.
 PHYSICAL THEORY OF FERROMAGNETI C DOMAINS
Barhhagsem Eeet
Since many physicists have encountered the concept z
of domain structure only in elementary textbooks, in V
Ld

connection with discussions of the Barkhausen eGect, LI)

it may be well at this point to correct the widespread 0

notion of the connection between the size of the V
LLj
Barkhausen discontinuities and the size of domains. z
LILI

Recent experiments of Williams and Shockley (1949)

have shown quite clearly that there is usually no direct 0
I-
connection here, but that the Barkhausen discontinui-
ties correspond to irregular fluctuations in the motion
COORDINATE:
of a domain boundary (Bloch wall) under the influence NORMAL TO DOMAIN BOUNDARY

of an applied magnetic field, rather than to a complete FrG. 14. Variation in energy of specimen as a function
domain reversal. Very pronounced and continued of the position of the boundary.
Barkhausen noise has been observed attending the
motion of a single domain wall. This discovery frees and free. energy; the di8erence is important in ferro-
the subject from the difFiculties which had been raised magnetism only near the Curie point.
by the earlier interpretation; for example, the apparent The most important equations for energy density
"domain volume" of 10 ' to 10 ' cc indicated by the derived in the various parts of this section are summa-
Barkhausen eGect has no direct connection with actual rized here, for a cubic crystal:
domain volumes, which may be very much greater.
Exchange: f-= J~'Ev v' (2. 1.5)
Perroelectri c Domains
Certain dielectric crystals such as rochelle salt and Anisotropy: f&= E&(aPn2'+u2 n32+032~P) (2.2.4)
barium titanate show ferroelectric behavior — that is,
regions of spontaneous electric polarization are found, Magnetoelastic: f .= —,'XT sin28 (2.3.22)
analogous to the regions of spontaneous magnetization
in ferromagnetic crystals.
Magnetic: f „=— 2H I (for self-energy) (2.4.2)
In some crystals of barium titanate arrangements of
ferroelectric domains have been found (Matthias and
Here J
is the exchange energy integral; y is the angle
between the directions of neighboring spins S; Eq is
von Hippel, 1948), which at 6rst suggest that the
the anisotropy energy constant; cx&, o.2, n3 are the direc-
formation of ferroelectric domains is governed by
tion cosines of the magnetization vector referred to the
electrostatic energy, just as the formation of magnetic
crystal axes; X is the isotropic. magnetostriction, and 8
domains is governed by magnetostatic energy. How-
is the angle between the tension T and the magnetiza-
ever, the electrostriction in barium titanate is 10—,
tion.
which is considerably larger than the values of the
magnetostriction ( 10 ') in ferromagnetic materials.
It is therefore likely that the electrostriction will play
an important role in ferroelectric domains, along with
considerations of closure of electrical flux. There is also
the possibility of charge neutralization by means of
free charges in the atmosphere. V)
O
The saturation polarization in BaTi03 is approxi- 4J
I-
V)
ec
mately 50,000 esu. LLI

2. CONTRIBUTIONS TO THE DOMAIN ENERGY

The purpose of this section is to derive quantitative
expressions for the several types of energy which enter
specifically into the theory of domain structure. The
energies with which we are most particularly concerned

static energy, f „.
and the symbols by which we denote the corresponding
energy density are: exchange energy, f, ; anisotropy
energy, f&,. magnetoelastic energy, f,
; and magneto-
All of these terms are necessary;
commonly the omission of any one would cause pro-
found modihcations in the nature of the domain
structure. We neglect the distinction between energy
0
0 40 60
PERCEl4T COPPER BY WEIGHT

FIG. 15. Effect of precipitated copper on the coercive force of iron

(Kussmann and Scharnow, 1929).
550 CHARLES K I TT E L

200 or more at room temperature, while their electrical

resistivities are of the order of 10' to 10' ohm-cm. By
comparison the resistivity of iron is about 10 ' ohm-cm.
SY
H=P ES H4 Hc f
In semi-conducting magnetic materials, such as the
ferrites, we are inclined to believe that both the 4s
FIG. 16. Magnetization changes in very small particles occur by
rotation of the total magnetic moment of the particle. electrons and the ferromagnetic 3d electrons are more
or less fixed to individual atoms, rather than wandering
2. 1 Exchange Energy widely throughout the crystal. It appears probable
that in more metallic materials the 4s electrons con-
Ke shall start the treatment of exchange energy with tribute largely to the conductivity, while the 3d elec-
a preliminary discussion of the present picture of the trons are largely localized as before. The results of
origin of ferromagnetism in relation to the electronic neutron polarization experiments also indicate that the
structure of magnetic materials. electrons with uncompensated spins are well-localized
It is believed that nearly all of the magnetic moment in the crystal.
of ferromagnetic substances arises from electron spin The principal difficulty with the model of 3d electrons
motion, rather than from orbital motion of the electron fixed on individual atoms is that it does not immedi-
around a nucleus. This conclusion is indicated by ately explain the non-integral values of the magneton
results of measurements of the magnetomechanical number obtaining in most substances. ' The alternative
ratio. The magnetomechanical ratio is the ratio of model of "collective" electron ferromagnetism empha-
magnetic moment to angular momentum, and the ratio sizes the band concept according to which the 3d
is expected theoretically to be equal to e/mc for spin electrons may wander more or less freely throughout
motion, and e/2mc for orbital motion. The experi- the entire crystal, and on' this picture the non-integral
mental observations, as summarized by Barnett (1944), values of r4rt find an easy explanation (Table I).
are close to e/mc, with small but probably significant Neither the atomic nor collective models taken
deviations; from the deviations we may tentatively separately in their simplest form will give a complete
suppose that the orbital motion electively contributes and consistent description of all the numerous phe-
something like 10 percent and the spin motion 90 nomena associated with ferromagnetism. We shall in
percent of the saturation magnetization. This conclu- this paper employ the atomic model exclusively, not
sion is supported by the results of microwave resonance because of a firm belief in its universal applicability,
experiments (Kittel, 1949a). For most purposes the but because on this model one may treat in a simple
orbital contribution is neglected. way the quantity of most direct interest to domain
We next ask how many electron spins per atom —
theory the exchange energy due to gradual variations
participate in the magnetization. We may obtain the in spin direction, such as occur within the boundary
effective number of Bohr magnetons per magnetic wall between domains. The good agreement which
atom from the relation obtains between the theoretical and the experimental
values of the surface energy of the boundary wall is,
saturation magnetization
in fact, a significant achievement of the atomic model.
Rgff
(Bohr magneton) It is, however, possible to give a quantitative treatment
X (number of magnetic at, oms per unit volume) of the boundary on the band picture, and this is accom-
plished in a forthcoming paper by Herring and Kittel.
Some typical values are given in Table I.
A further question is that of the relationship between Exchange ErMrgy on the Atomic (Heitler
the electrons responsible for ferromagnetism and the London;Heisenberg) Model
electrons responsible for the electrical conductivity of
the metal: are the same or diferent electrons responsible We shall work then with a model on which at each
for the two processes' It appears likely that the lattice point of the crystal there is situated an atom
with total spin quantum number S, where 25 is equal
electrons responsible for the ferromagnetism make only
a minor contribution to the electrical conductivity. In to the number of unpaired electron spins in the atom
and is an integer.
the iron group of the periodic table the conduction
electrons are believed to come largely from the 4s shell,
The relevant result of the quantum-mechanical
treatment' of the many-electron problem may be
while the ferromagnetic electrons are in the 3d shell of
summarized by saying that there is a term of electro-
each constituent atom; the 3d shell lies nearer to the
static origin, which does not enter on strictly classical
nucleus and is more tightly bound than the 4s shell.
dynamics, in the energy of interaction between neigh-
In this connection it may be noted that materials are boring atoms, and this term tends to orient the electron
known which are strongly ferromagnetic, yet are very
poor electrical conductors. Compounds such as manga- ~ Van Vleck
(1945) has pointed out that a slight variant of the
Heisenberg model will account for non-integral values of the
nese ferrite MnO Fe203 and nickel ferrite NiO Fe203 average magneton number.
have values of saturation magnetization of the order of 6
See, for example, the reviews by Van Vleck (1945, 1947).
 PH YSI CAL THEORY OF F E RROMAGN ETI C DOMA I NS
spins of the atoms either parallel or antiparallel to each TABLE I. Calculation of effective number n.yy of Bohr mag-
—
other according to the algebraic sign of a certain
netons per magnetic atom; and data on saturation magnetization
and Curie points.
energy integral J, known as the exchange integral,
which enters the problem. The usual convention is to Ferro-
define J in such a way that when it is positive the Saturation
magnetization I& Density
magnetic
Curie
temp.
energy for parallel orientation of two spins is lower Substance Room temp. O'K &~ ~~ (0 K) gr/cc 'K
than the energy for antiparallel orientation by an Fe 1707 1752 2.221 7.86 1043
amount 2J, for spin 2. Co 1400 1446 1.716 8.8 1388
In this review we are taking for granted the existence Ni 485 510 0.606 8.85 631
Gd 1090 1980 7.10 7.83 289
of an exchange interaction with the specified properties, MnBi 600 675 3.52 9.0 670*
as it is not possible in this space to develop the necessary Cu2MnAl 430 (540) 3.0 1 72++ 600
background of quantum theory. We shall, however,
give here a method whereby the exchange integral J + Extrapolated;
**Density
limited by apparent phase transformation
refers to manganese atoms alone.
at 620'K.
may be related approximately to the Weiss molecular
field H, „j. Let us imagine that the electron spin of a
given atom and its s nearest neighbors are oriented in
fundamental result of quantum theory and is the
starting point for our calculations of the exchange
the same direction. The exchange energy of the selected energy of configurations with varying spin directions.
atom is then equal to — 2sJS'. The molecular field is The equation is derived in many places, and we may
essentially defined so that the energy of interaction refer again to the review by Van Vleck (1945).
—2Sp~H ~ of the magnetic moment of the atom with For many purposes it turns out that we may think
the molecular field is equal to the exchange energy: of the spin matrices as approximately related to classical
vectors, and in this sense we rewrite Eq. (2.1.3), as
2s JS' = 2Sp~Hmg, (2. 1.1)
so that
II sSJ/IJ, ~. — (2. 1.2)
..
z = —+2J,,S' cosy;,
i)j
(2.1.4)
r
The effective coupling between spins caused by the where y;; is the angle between the directions of the spin
exchange effect is equivalent to a potential energy of angular momentum vectors, understood in a classical
the form sense; m„
is now the exchange energy. The circum-
stances and the extent to which the use of Eq. (2.1.4)
Vg= —2J@S; S, , (2. 1.3)
is justifiable will be discussed in a separate publication;
where J,; is the exchange integral connecting atoms i we may summarize the results by saying that the
and j,and S; is the spin angular momentum of atom i, semiclassical approximation is good when neighboring
measured in multiples of h/2s. . This equation is a very spins make only small angles with each other, as

105
SUPE RMALLOY
1040 ALLOY
~ SENDUST

MUMETAL
104
~ 78 PERMALLOY
~ HYDROGEN TREATED IRON

HI PER NIK
O
i' HYPERSIL
&- 10
I-
~ S IL ICON
FIG. 17. Correlation be- d) IRON
tween the initial permea- 4J
bility and coercive force of i& IRON
a wide range of magnetic 4J NICKEL
materials. ~ 102 0

I-
Z ~ 5 lo TUNGSTEN
STEEL
RE MALLOY
10
~ VICALLOY I
AGNI CO 2
ALNICO I
&i VECTOL ITE
F ERRO —PLATINUM ~
10 10 10 1 10 10 10 10
COERC I VE FORCE, H C, IN OERSTEDS
 CHARLEs KITTEL
within a domain boundary wall. In domain theory we involved in the statistical theories are reviewed by
are interested in the exchange energy only for con- Van Vleck (1945, 1947). We summarize the results of
Ggurations in which neighboring spins make small two methods here. P. R. Weiss (1948) finds by an
angles with each other, extension of the Bethe-Peierls method the results:
If we suppose that only interactions between nearest
neighbors are important for the exchange energy and
I=0.54kT, (simple cubic; spin i~) (2.1.12)
that these interactions are equal, then I= 0.34k T, (body-centered cubic; spin )
—',
(2.1.13a)
w~~= 2IS Q cos(po. 7=0.15kT, (body-centered cubic; spin 1) (2.1.13b)
i& j'
For lion,
If neighboring spins make a small angle q;;«1 with J = (0.15)(1043)k—160k, (2.1.14)
each other, then we have the important result
taking the spin as one.
Another method, used by Lifshitz' (1944), is to relate
and the exchange energy between each pair of spins is the experimental value of the constant C in the Bloch
law
Am;; —JS'y'. (2.1.6) I = Io(1 —CTl) (2.1.15)
It is often convenient to express Eq. (2.1.5) in for the temperature dependence of the saturation
another form. I et us suppose that the direction cosines magnetization at low temperatures, to the effective
'.
of the spin at lattice point r; are n, *, o.p, o., Now the
;, '
direction cosines 0. a;&, o., at the neighboring lattice
exchange integral by means of the equation (Bloch,
1931; Moiler, 1933)
point r; may be expanded in a Taylor series:
C = (0.0587/2S) (k/2SJ)'* (2.1.16)
n ;*=n, *+. (x,; (8/», ;)+y g(B/By;, )+z,, (B/Bs,;) )n, ' for a body-centered cubic structure and spin S. Now
from the measurements of Fallot (1936) we have for
iron C=3.5X10 ' so that
—

J= 205k,
On summing over nearest neighbors on a body-centered
for S=1, in fair agreement with Eq. (2. 1.14) obtained
cubic lattice with lattice constant a,
by a quite different method.
Aw„~—2JS'a,'Q (n "Pn. ) (2. 1.8) We shaB use the value J=205k resulting from the
Bloch theory since the spin wave picture is most closely
allied with the behavior of a domain boundary wall
which may be manipulated into the form
and actually (as can be shown) leads to a value of A
&~ = 2&S'~'PL(~n ')'+ (~n ")'+ (|7n ')q (2. 1.9) independent of assumptions regarding the eBect of
interactions from other than nearest neighbors.
Sy use of the above value the constant A in Eq.
by using the fact that (2.1.11) becomes (with S= 1):
(2.1.10)
2JS' 410k
The exchange energy density becomes (taking account = 2.0&(10 ' ergs/cm. (2.1.18)
of the two atoms per unit cell and taking care not to 86X10
count interactions twice):
The measurements of Fallot (1936) indicate that for
f., = A L(V'n, )'+ (V'n, )'+ (7'ns)'], (2.1.11)
4 percent (by weight) Si-Fe one should use A 1.7 —
&(10 ' ergs/cm. This composition corresponds approxi-
where A =2JS'/a.
mately to that used by H. J. Williams in his domain
We have now obtained two useful forms for the pattern studies; it is used rather than pure iron because
exchange energy; namely Eqs. (2.1.6) and (2.1.11). crystals of
pure iron are considerably more dificult to
The next problem is to relate the value of the exchange
pl o duce.
integral Jwhich enters these equations to some experi-
mental quantity which is strongly dependent on J, 2.2 Anisotropy Energy
such as the Curie temperature or the variation of the
saturation magnetization with temperature. Here we The anisotropy energy or, as it is sometimes called,
run head on into the fact that the accurate determina- the magnetocrystalline energy of a ferromagnetic
J
tion of from thermal data presupposes the existence crystal acts in such a way that the magnetization
of an accurate statistical theory of ferromagnetism, tends to be directed along certain definite crystallo-
and this we do not possess at present. ~ Several numerical errors in the work of Lifshitz are corrected
The mathematical details of the type of calculations in the present treatment,
 PHYSICAL THEORY OF FERROMAGNETIC DOMAINS 553

graphic axes which, accordingly, are called Chrections of X IO~

ap
easy megeeHsatioe; while the directions along which it is
most dif5. cult to magnetize the crystal are called hard 30
directions. It is found experimentally to require the I-
expenditure of a certain, and often considerable, amount 20
(
of energy to magnetize a crystal to saturation in a hard Z IP
direction, referred to the lower energy required to 0U
saturate along a direction of easy magnetization. The &-
p
excess energy required in the hard direction is the 0
I- — P
ariisotropy ertergy 0 I

As an example of anisotropy energy we may consider Z -20

the case of cobalt, which is a hexagonal crystal. It is
found that the direction of the hexagonal axis is the
-30
30 40 50 60 70 80 90 IOO
direction of easy magnetization (at room temperature), PER CENT NIC KEL
while all directions in the basal plane, normal to the FIG. 18. Anisotropy constants for face-centered Fe-Ni alloys
axis, are hard directions. The magnetization curves of at room temperature. Note the region of low anisotropy near
a single crystal of cobalt are shown in Iig. 10. The 70 percent nickel.
energy represented by the magnetization curve in the
hard direction is given by J'IICI per unit volume and the restrictions imposed by cubic symmetry. For ex-
amounts to an excess energy of about 5&(10~ ergs/cc ample, the expression for the anisotropy energy must
for the curve shown. be an even power of each n;, and it must be invariant
under interchanges of the o.; among themselves. The
Matherrtatical Represemtatiort of the Anisotropy Energy lowest order combination satisfying the symmetry
requirements is nP+n2i+nP, but this is identically
We now consider the problem of determining the equal to unity and does not describe anisotropy effects.
work done in magnetizing a cobalt crystal in a direction The next combination is of the fourth degree: 0, ~'n2'
making an angle 0 with the hexagonal axis. It is natural +nPnP+o. Po!ii, which is sometimes written 'in the
to expect that the anisotropy energy density fJr may equivalent form ni'+o. i+o.q. The equivalence' of the
be represented by a series expansion of the form two forms follows from the relation
fx PE„'sini~——
g, (2.2. 1) 1 = (Ap+ ct!i +o! ) = (11 +o!i +As
g

+2(&Paa +o's &a +&a &1 ) &

where odd powers of sin8 have not been included since, whence
by symmetry, the —
0 direction is equivalent magneti- —(&s +o'a
cally and crystallographically to the +0 direction; the
o'p&p+&u &i +&a &p= 2 g +&i ).
E„' are constants, independent of 0. In cobalt it is The next term is of the sixth degree: n. 'n2'a. 3'. This
actually found that a very good representation of the is as far as one usually needs to go in fitting the observa-
experimental observations is given by the erst two tions, so that for iron
terms:
f~ E~' sin'0+Ei' —
sin'tt, (2.2.2)
fx= Ei(cxq (xi +o!i cia +(xi txP)+Ego!PQ!i (xi (2.2.4)

where at room temperature where, at room temperature,

E~' = 4.1)& 10' ergs/cc Ei' — E~ —4.2&&10' ergs/cc; Ei 1.5&&10' ergs/cc. —
(—
2.2.5)
1.0)& 10' ergs/c—
c. (2.2.3)
Other ways of writing the E& term are given in Ap-
It is found unnecessary to include any terms which
pendix A.
distinguish among directions in the basal plane, so
that the specifically hexagonal nature of the crystal
Physical Origin of the Artisotropy Energy
does not emerge, but only the uniaxial character, in the
anisotropy energy. Before going more deeply into the experimental data
Iron is a cubic crystal, and the magnetization curves on anisotropy energy, we wish to consider the origin of
(Fig. 10) show that the cube edges [100], [010) and the anisotropy energy in terms of interatomic inter-
[001J are the directions of easy magnetization, while actions. Anisotropy energy in effect relates the direction
the body diagonals ([111] and equivalent axes) are of magnetization to the crystal axes. We may at the
hard directions. The excess work done in magnetizing outset point out three very important difhculties in
along [111)is about 1.4X10' ergs/cc at room temper- the way of understanding the origin of anisotropy:
ature. (a) the exchange energy by itself does not lead to any
In attempting to represent the anisotropy energy of anisotropy, regardless of the actual geometrical ani-
iron in an arbitrary direction with direction cosines sotropy of the crystal structure; (b) the magnetic
a~, 0.2, n3 referred to the cube edges, we are guided by moment interaction leads to only very small values of
 CHARLES KIT TEL
600, 000, 800, 000
OBAND K
500,000 X + WANDB
O
NICKEL
IRON

400,000 600, 000

4J
Z
O
V
300,000
I-
Z
400,000
V)
200,000
0z
O

Q.
IO0, OOO 0
I-
0
V)
200, 000
0 z
-400 -200 0 200 400 600 800 I 0OO
TEMPERATURE IN DEGREES C

0
0 100 200 300 400
TEMPERATURE IN DEGRE ES K

X I06
I2

8XI06 MrI Bj.

8
LIJ
V COBALT
ytl) 6
Z
V
K 6
V
z dl
~4
K
0
L4
X
4J
'~~a~~ K2
06. 2
K
I
O
Vl
X
0 0-

-2 -2
"200 100 0 IOO 200 300 400 0 40 80 I20 160 200 240 280 320
TEMPERATURE IN C TEMPERATURE IN DEGREES KELVIN

Fn. 19. Temperature dependence of the principal anisotropy constants of (a) iron, (b) nickel, (c) cobalt, (d) MnBi.

the anisotropy constants, much smaller than are spontaneously gives only very small values of the
observed; (c) the anisotropy constants are found to be anisotropy —
about 10 ' of the observed values in iron
very sensitive to temperature changes, and it is not and nickel. In uniaxial crystals the dipole-dipole inter-
unusual for even the sign of the constants to change action may contribute a small amount to the ani-
between low and high temperatures. sotropy energy, but the eGect is not usually significant.
First with regard to the exchange energy: the ex- The anisotropy energy of the uniaxial crystal MnBi is
change energy operator refers only to the angle between of the order of 10' ergs/cc, while the dipole-dipole
various spins: I,
energy is only of the order of 2, or less than 10' ergs/cc.
II= —2JQS,"S; (2.2.6) Furthermore, the change in sign of the anisotropy
constant in the system of face-centered alloys of iron
and not at all to the angle between the spin and the and nickel, as shown by Fig. 18, is not intelligible in
crystal structure; that is, we may rotate the whole terms of magnetic dipole interactions.
spin system by any angle with respect to the crystal The temperature dependence of the principal ani-
structure without changing the exchange energy of the sotropy constants of iron, nickel, cobalt, and the
system. compound MnBi is shown in Fig. 19; it is seen that
The ordinary magnetic moment interaction between the variations are quite erratic and discourage any
electrons leads to too small values of the anisotropy. simple interpretation.
In Appendix 8 we prove that the magnetic dipole The origin of the anisotropy energy is believed at
moment interaction gives zero anisotropy for an un- present (Sommerfeld and Bethe, 1933; Brooks, 1940;
strained cubic lattice; allowing a cubic lattice to deform Van Vleck, 1947) to be the result of the combined
 PH YSI CAL THEORY OF FERROMAGNETI C DOMAI NS 555

sects of spin-orbit interaction and the partial quench- of high permeability. For example, the permalloy and
ing of the orbital angular momentum (by inhomo- allied series of Fe-Mi alloys containing about 78 percent
geneous crystalline electric fields and by orbital ex- Ni have exceptionally high permeabilities; in the case
change interaction with neighboring atoms). In other of supermalloy the maximum permeability is of the
language, the magnetization of the crystal "sees" the order of a million. Near 75 Ni —25 Fe the magneto-
crystal lattice through the agency of the orbital motion striction also is very low: this is another requirement
of the electrons; the spin interacts with the orbital for high permeability. The exceptional properties of
motion by means of the spin-orbit coupling and the the Permalloys are attributable to the fact that both
orbital motion in turn interacts with the crystal struc- the anisotropy and magnetostriction are small for the
ture by means of the electrostatic fields and overlapping particular alloy composition.
wave functions associated with neighboring atoms in
the lattice. The theory as developed along these lines is 2.3 Magnetoelastic Energy
quite complicated, even when rather drastic approxi- The magnetoelastic energy is that part of the energy
mations are made. An excellent review of the present
of a crystal which arises from the interaction between
theoretical position is given by Van Vleck, 1947. the magnetization and the mechanical strain of the
One result which may be anticipated from the theory lattice. The magnetoelastic energy is defined to be
of anisotropy is that the magnetic anisotropy energy zero for an unstrained lattice.
will tend to be large for crystals where the lattice of
The close physical relationship which exists between
magnetic ions is of low symmetry, while the anisotropy the anisotropy and magnetostriction constants is not
energy is expected on the whole to be low for crystal revealed clearly in the standard discussions of the
lattices of high symmetry. This tentative rule is sug- in ferromagnetics. It is of the
energy relationships
gested by the fact that the anisotropy energy enters primary importance to recognize that there mill be mo
only in a higher order approximation in a cubic crystal tznear rnagnetostrzction if the anisotropy energy is inde
than in a uniaxial crystal. The observational data pendent of the state of strain of the crystal. Magneto-
support our hypothesis: the anisotropy energy of the striction occurs because the anisotropy energy depends
cubic crystals Fe and Ni is of the order of 10' ergs/cc, on the strain in such a way that the stable state of the
while it is of the order of 10' ergs/cc for the hexagonal
crystal is deformed with respect to a cubic lattice.
crystals Co and MnBi and probably also of the order That is, a crystal will deform spontaneously if to do
of 10' ergs/cc for the ordered state of the alloy FePt in so will lower the (anisotropy) energy. We proceed
which the Fe atoms by themselves form- a trigonal
now to examine the nature of the interaction between
lattice. There is considerable interest in high anisotropy magnetization and strain in cubic crystals, and to
energy in connection with the development of perma- derive energetic relationships in terms of the experi-
nent magnet materials with high coercivity.
5X105
Experirnenta/ Data on Anisotropy Energy
We review here brieAy a selection of the experimental 4'4
material on anisotropy energy; for further data and a
description of measurement methods the useful sum-
mary article by Bozorth (1937) may be consulted.
—
Iron-nickel alloys. The anisotropy constant E& at
room temperature for the face-centered (y-phase) sys-
tem Fe-Ni are plotted in Fig. 18. For pure Ni, E~ = — 3.4 V
X10' ergs/cc and E2=5.0X10' ergs/cc; for pure Fe,
Ez —4.2X10z ergs/cc and Ez 1.5X10z ergs/cc-— . V
—
Iron-cobalt alloys. Results at room temperature for
CL'

z
0

the body-centered (n-phase) of the Fe-Co system are

plotted in Fig. 20. For pure Co, w ich is hexagonal,
E,'=4. 1X10' ergs/cc and Ez' 1.0 10' ergs/cc. —
Manganese coznpozznds —
At room. temperature Guil-
laud (1943) gives Ez'=8. 9X10' ergs/cc and Ez' 2.
X 10' ergs/cc for MnBi; and Ez' 1.8X 10' ergs/cc an—
Ez'=0. 8X 10' ergs/cc for MnzSb.
The low values of the anisotropy energy for certain
— d
7— «4
—

alloys, such as those having compositions near 75 -5

20 40 80
percent Ni and 25 percent Fe, and near 40 percent Co PERCENT COBALT
and 60 percent Fe, are of considerable importance in
Fn. 20. Anisotropy constants for body-centered
connection v ith the development of magnetic materials Fe-Co alloys at room temperature.
 CHARLES K I TTE L
mental magnetostriction constants. We give a simpli6ed magnetization vector:
treatment of the standard discussions of Seeker and
Akulov. f E—
(nPa0 +n0 u0 +u0 uP)
+B~(a Pe„+u0'e»+n Pe„)
The elastic erIergy density in a cubic crystal is given
by (Love, 1944, p. 160)
+B0(n&n2e»+usage„, +n0nge, )
+0cyy(eg +e» +eg )
.
f, = 2c»(e„'+e»'+ e»')+ 0 c44(e»'+ e„,'+ e„') + 0 c44(e,„'+e„,'+ e„')
+ c~0(e»e„+e„e„+
t

+cq0(e»e„+e„e„+e, e») (2.3.1) e»e»). (2.3.6)

The equilibrium configuration of the crystal, that is,
where the cg are elastic moduli and the e;; are strains, as
the stable state of strain, when magnetized in the
de6ned by Love, p. 38. For iron (Kimura and Ohno,
1934)
direction a may be found by minimizing with respect f
to the e;, . The solutions for the e;; may finally be
c»=2.41X10"ergs/cc; interpreted in terms of the usual saturation magneto-
1.46 X 10" ergs/cc; striction constants X~pp and ) ~~~.
——
c&0 (2.3.2)
c44 = 1.12 X 10 ergs/cc. Furthermore, the solutions e;; of the minimal equa-
tions depend on the direction cosines in such a way
For nickel (Bozorth et at , 1949. ) that the energy in the equilibrium configuration may
be expressed in the form
c» —2.50 X 10" ergs/cc;
c~0= 1.60X10" ergs/cc; f= (E+EE)(aPuP+n0'n0'+n0'nP) (2.3.7)
c44= 1.185 X 10" ergs/cc. where DE is independent of 0. but is simply related to
the elastic moduli and the magnetoelastic coupling
The amisotropy erjergy density in an Nrsstrailed cubic
constants.
crystal is of the form (Section 2.2)

fx —E(nPaP+n0'n0'+ n0'aP) (2.3.3)

The Minima/ Eqmatioes
The 6rst problem is to determine the values of the
to the first order. Here E is a constant
independent of
e;; which minimize f
in Eq. (2.3.6):
the direction of saturation magnetization in the crystal;
ca~, u~, u3 are the direction cosines of the direction of Bf/Be„=B&nP+ C»e, +C&0(e»+e„)= 0,
magnetization referred to the cubic axes. For iron, Bf/Be» = Bla2 + c»e»+c10(e„+e„) = 0,
K=4.2X10' ergs/cc. Bf/Be» BqnP+ —+ c~0(e„+e») = 0, (2.3.8)
c»e„—
To express the dependence of the anisotropy energy Bf/Be, „= B0ugn0+c44e,
Bf/Be„=B0aqn0+ C44e„=0,
„=0,
on the strain we expand the energy in a Taylor's series
in the strains: Bf/Be„,= B0n0n, + c44e„,= 0

The solutions are

frc= (fx) 0+2(Bfx/Beg) 0e', + (2.3.4)
eii B1LC10 ap(c»+2cy0)7/
L(c„—
c„)(cgg+2cg0)], (2.3.9)
Here (fx)0 must satisfy cubic symmetry, but the terms
(Bfx/Be;, )0e;; may have lower symmetry as these terms e,;= —B0n,n;/c44 (iW j). (2.3.10)
refer to the deformed lattice.
Relati orl, to MogrIetostri cti orI, Coestumts
Considering only the lowest order terms dependent
on orientation, we may take from symmetry considera- The conventional first order magnetostriction equa-
tions tion which is frequently used in the analysis of experi-
mental measurements (on cubic crystals) is (Becker
,= B,np;
Bfx/Be. Bfx/Be» = B]n0
„= and Derring, 1939)
Bfx/Be„=Bqn0, Bflc/Be.
Bfx/Be„,= B0usn„. Bfx/Be. .=
B2n jn2 ',
B0n~n0,

where 8& and 82 are constants which may in principle

be calculated knowing the details of the interactions in
(2.3.5)
—
bl
= 0~100(a1'PP+n0'P0'+u0'l4'
+3Xggg( n, ,nPP 0+u, aP
0)—
0A+u,uP, P,), (2.3.11)
the solid. The B's will be called mageetoetostic coupling where a= (a~, n0, n0) is a unit vector in the direction of
coestaets. For iron, as we shall see later, S~ — —
2.9)&10' the magnetization, g= (p~, p0, p0) is a unit vector in the
—
ergs/cc; B0 3.2X10' ergs/cc. direction in which 8/ is measured; X]pp and ) ~~~ are the
The result of combining the foregoing equations is saturation values of the longitudinal magnetostriction
given by the following expression for the part of the in the directions L1007 and [1117 respectively. We
total energy density which depends on the strain and wish now to relate the magnetostriction constants happ
on the crystallographic direction of the saturation and )~~~ to what we have called the magnetoelastic
 PH YSI CAL THEORY OF FERROMAGNETI C DOMAINS

coupling constants, 81 and 82, which have more existence of magnetostriction would have no direct
fundamental signi6cance. effect on the result of the anisotropy measurement.
Now in terms of the strains we have readily However, in practice one actually measures the ani-
sotropy at contest stress, so that the lattice is allowed
st to deform under the action of magnetoelastic forces.
=—
Ze P P'' (2.3.12)
We consider the general expression (2.3.6) for the
energy, eliminate the strains through Eqs. (2.3.9) and
since by definition of the strains (noting particularly (2.3.10), and express the 8's in terms of the )1's, and
the definition of the shear strains used by Love): obtain finally

.
x'= (1+e. )x+-'2e y+-', e. s
y = 2egox+ (1+cog)y+2e2 z
. (2.3.13) where
I' = (E+&E)(C21'4222+422'423'+C33'al') (2.3.17)

.
s' = —e2, x+-2e„,y+ (1+e„)
s, hE= (9/4)[(cll —C12)Xioo' —2C44X111']. (2.3.18)
whence
For AE= — 5X102
7. ergs/cc, so that here E/E
—10iron
Il
b(P) =2l bl=2P+e;;P, P; '. For nickel we have AE— 2X10' ergs/cc, whence
from which Eq. (2.3.12) follows immediately. d, E/E — 10 '.
On substituting the values of the e;; given by Eqs.
(2.3.9) and (2.3.10) we have IsotroPi c Magn etostricti on

+1 It is often assumed for the sake of simplicity that

= F111= X;this is the case of "isotropic magneto-
C11 C12
y y &a z X1pp
"
striction. For Xi it is customary to take ) as — 34
—
X10 ' and, for Fe, X= 7)&10 ', although the isotropic
82
c44
(~~— .
Po+~~*P*P.+~2~.PuP. )
oP—
assumption does not give in either case a very good
representation of the experimental results.
Equation (2.3.11) reduces to

(cl1+ 2C12) (cll C12)

(2.3.15) —
61
= p[(C31P1+422P2+a3P3)' 1/3]—
, (2.3.19)

which may be written in the form of Eq. (2.3.11) if we

set —
81
= 32X[cos28 —1/3], (2.3.20)
2 8'
~100
3 C11 — C12 where 0 is the angle between the magnetization and the
(2.3.16) direction in which 8/ is measured. It is seen that this
182 expression contains no reference to the crystal axes and
~111 is therefore isotropic.
3 C44
It is of considerable importance to calculate the
and drop a residual term which is constant with respect contribution to the anisotropy energy caused by a
to 42 and g. We have thus succeeded in relating the uniform tensile stress T applied to the specimen. The
magnetostriction constants X1pp and )111 to the strain stress components relative to the crystal axes of a
gradient of the anisotropy energy and to the elastic tension with direction cosines y1, 72, y3 are I';y, = Ty, ~l„
constants of the crystal. giving
For iron, ) 100=19.5X10 ' and X111=—18.8X10 ' as 2 44'7172 e* 2 [ 1171 + 12(72 +73 )]
the result of experiment; using these values we calculate
81 —2.9X 102 ergs/cc, — 82= 6.4X10' ergs/cc. For where the s's are elastic coefficients. The energy terms
nickel, Amp= — 46)&10 ' and X111=— 25X10 '; we have in Eq. (2.3.4) dependent on strain become
81—6.2X10' ergs/cc, 82=9.0X10' ergs/cc. The values f~g= 81T[($11 $12)(421 Vl +C32 V2 +C33 73 )]
of the X's are from Becker and Doring (1939, pp. (2.3.2 1)
277—280).
82T$44(cKlc32 rl|
2+422cE3r2'r3+c2341'Iro+1) ~

If now we suppose ) 100=)111='A, we have, from

Contribf4tion to AnisotroPy Energy Eq. (2.3.16),
82 (Cll C12) 281C44,
We now wish to show that the presence of magneto-
striction gives rise to an apparent contribution to the so that, using well-known relations between the c's and
anisotropy energy of the crystal. If experimental s's, and the relation
determinations of the anisotropy energy could be
carried out at coestarIt lattice dimensions that is, on a — cos8= (ulyl+u2y2+c23y3),
crystal held at constant strain by clamping then the — where 8 is the angle between the magnetization and the
 C HARLES K I TTE L

Self En-ergy of a Permanent Magnet

When the Geld against which the work is being done
.
FIG. 2 j. Model for the
is not external, but is due to the magnetization itself,
calculation of the magnetic we have the usual factor of one-half coming in which
6eld energy of coplanar occurs in all self-energy problems. It enters because we
tL

rr
strips of alternate sign. —
must not count dipoles twice the expression (2.4.6)
I when applied to self-energy effectively counts each
dipole once as a source of field and once as a magnet
in the field. The correct result is

(2.4. 7)
tension, we also have
Etti psoidal Speci men
f ,= ~'AT sin'8. (2.3.22)
Terms independent of 8 have been dropped. We shall If the specimen is in the form of an ellipsoid and is
require this expression in Section 6.2. magnetized along one of the principal axes, the self-field
is given by
2.4 Magnetostatic Energy H= —EI, (2.4.8)

We shall not give a detailed discussion of the subject where X is the demagnetizing factor. Values of are E
of magnetic energy, for this would be a rather long tabulated in useful form by Osborn (1945). Then
story, but we shall discuss several particular situations Eq. (2.4.7) becomes
of direct interest to domain theory. A more general f ,
g= 'iVI2. (2.4.9)
treatment may be found in papers by Guggenheim This relation is very useful.
(1936) and Fokker (1939).
Here we are concerned with: Distribution of Poles on a P/ane
(1) The energy of interaction of a permanent magnet We consider the magnetic field energy of coplanar
with a uniform external magnetic field: strips of alternate sign (Fig. 21). Let the plane of the
f„.
, = —I H, (2.4. 1)
strips be the (x, y) plane with the y axis parallel to the
axis of the strips; the width of a single strip is D and
per unit volume. the pole strength per unit area of strip is The mag- I.
(2) The self-energy of a permanent magnet in its netic energy per unit area is, from Eq. (2.4. 7),
own field is
f„.
, = —')I(-, H, (2.4.2) o = —(1/2) H Idk. (2.4. 10)
per unit volume, which for an ellipsoid may be written
Now the vertical or s-component of the magnetic field
(2.4.3)
directly below the plane of the strips is given by the
per unit volume, where T is the demagnetizing factor; I'ourier expansion of a square-wave of amplitude —2ml.
and for parallel strips of poles of alternating sign, The approximate solution of the Laplace equation is
o-,g = 0.8525I,'D, (2.4.4) H. = %2irIl (4/s. ) sinkx e+~'

per unit area, where D is the width of a strip. +terms in odd multiples of k] (2.4.11)
(3) The eRect of finite anisotropy energy on Eq. where k= (~/D). For the present the harmonic terms
(2.4.4), which holds only for the case of infinite ani- in the expansion will be neglected. Then
sotropy energy (completely "frozen-in" spins):
o .,=l 2/(1+tr*)](0.8525I'D) t *=1+27rI2/E (2.4.5) =4P t e"*ds(lsinkxl).
for glancing angles of the magnetization vector with
respect to the surface. The mean value of sinkxl isl
2/m. , so that

Interaction of Permanent Magnet with External Field o = 8I2D/n-', (2.4. 12)

It is a familiar result that the energy of interaction which is the contribution from the first order term
of a permanent magnet dipole p with an external field alone. The complete expression including the effect of
H is given by the quantity — p H. The same result the harmonic terms is obtained by multiplying by
holds for a rigid assembly of dipoles, so that the Qn ', where the sum is over odd integers and is
magnetic energy density may be written as approximately equal to 1.0517. We have finally
., = —I H. (2.4.6) , = 0.8525I,'D. (2.4. 13)
 PHYSICAL THEORY OF FERROMAGNETI C DOMAI AS 559

We can easily treat along similar lines the energy of in a medium of permeability (p, 1, p), subject to the
an arbitrary periodic distribution of poles on a plane. boundary condition that the diGerence in H, at the
Let p(x, y) be the surface pole density, which is sup- surface shall be equal to &4xI, sin8. Suppose that the
posed periodic in the fundamental area which is a solution for the potential problem for p 1 is y(x, s).
rectangle of sides 2~L, and 2mL, . Then p may be For the actual problem we take the potential as
expanded in a double Fourier series Ap(x, ns) for s)0 and as Ay(x, ps) for s(0.
Then the
equality of the surface charge in the two problems leads
to
p(x, y) =P PC „exp[i(mg+nq)), (2.4. 14)
An(8q/Bs), ~+pAP(8qr/Bs), 0
= 2(8q/Bs), =+, (2.4.20)
where $= /xL„&=y/L„, and
2w 2m A = 2/(n+Py). (2.4.21)
C„„=
~2J
I
' ' "'
J 0 p(,
g)e "' (2
'~m&+. ~id dg
' 415)
' ' The condition that div E=O gives

The surface energy density turns out to be

~**+P'v **=0, (2.4.22)
which is satisfied for P=1; similarly n=1. Thus
0 ., =~QQC C E (2.4. 16) A=2/(1+@). (2.4.23)
where Since the potential is proportional to A, the magnetic
-
pm't'
I,-= — — energy is also proportional to A.
—
(L)I+IEL)
(2.4. 17)
I I

. We consider the proper value to use for p that is,

the effective permeability for small displacements about
In this way we find that the energy of a checkerboard an easy axis. For both cubic and uniaxial crystals
array is (with K) 0) we have
a g = 0.53I2D (2.4. 18)
where D is the side of each minor square. where q is the angle (supposed to be small) between
For a circle of one polarity imbedded in a square of the magnetization vector and the easy axis. If a mag-
opposite polarity, netic 6eld is applied perpendicular to the easy axis,
0. = 0.374I2D, (2.4.19)
fmag= —&I.q (2.4.24)
where the fundamental area is a square of side D in
which there is imbedded a circle of radius D/(2m)&. The total energy E.p' —III,p is a minimum for

The p~-correct oe
2Ep —III, = 0, (2.4.25)
so that
The question of the energy of a pole distribution on y=HI, /2E. (2.4.26)
a plane surface is not quite as simple as it may appear
at first sight. The complication arises from the fact
Now the magnetization parallel to II is l, q, so that the
susceptibility is
that the spins are not really "frozen-in" along the
directions of easy magnetization, but can deviate from y=I, p/H=I 2/2E, (2.4.27)
these directions under the inhuence of the Geld caused
by the pole distribution. Only for extremely high values
of the anisotropy energy can the spins be considered as
frozen-in along the easy directions.
The corrections to the magnetostatic energy which
must be made on this account will depend on the
particular situation. Various situations have been dis-
cussed by Lifshitz, Keel, and Shockley. We follow the
treatment of Shockley (1948). We consider the case of
parallel strips of poles of alternating sign &I, sin0,
where the easy axis makes a small angle 8 with the
surface of the crystal, as shown in Fig. 22. The response
of the magnetization to a magnetic held may be
described by three permeabilities p, p, and p, Here
p„— 1, since the magnetization in the x-direction cannot
be increased appreciably; and p —p„by
symmetry.
The problem is first to solve for the Geld distribution FIG. 22. Model for p* calculation.
 CHARLES K»TE"

l t

N N
t

I t

N t
N
, f~I'~ ~)
P/
p~

'I
I
~i~
xf &I Ill I &
6t ~p'
) a ~g
g ~
g
/'

/' //
I

I !
t

I I

FlG. 2 3 The plough w

wa&&.

whence (Fq (2, 1.6))

&+=1+(2~ I' &). ' (2.4.28) ~„=JS'q' (3.1 .1)
""-" .
~

;„,
The notatio P
effective p
'
'. y
cab jlity aused by an so op
p* are 46 for Fe an d 3 6 fpi Cp
' Vl-"«
the
-,l.
)or the ex

-dS-
r between
h each other;
h"p",
.ld""d
t o making a
e exchang~
t m measure
hng, h.
LOCH ~PLL
THE gL
3 1 ggtj. od&ctog" Re~ ~s angle change
. the ex ge energy between each pa'
The ter
erm h wall" denptes th transitionlayer atoms is
'
d acent doma»ins m g ed in differ- 'LV ex =ps'(v w'&)'. (3.1.2
h (1932)' ho
he total exch»g
n e energ f the line «& ] atpms is
further deve p
layer were made b y Landau an gs'po'/&. .
(3..13)
l, ,fs itz (1944 Qee l (1944a);. an Herring an d Kittel lf the tota ange o a
(t t e submitt ed correspon din g tP a rev a netization
The essentiaja]. idea the g]och wal is that e entire
. on passing throug h the wall, th en the exc an
.
change in sP' s ln directio between do ma lns mag netized
. diferent dlrirections goes «
of a line o& aatoms t h ugh a wa ll y00 atpms lin thickness
. o le not cur in one
occu pntin- is of the prder pf T 1pp, as comp d wl'th k~ f
us jump ac o atomic p lane. R ther the wall onn] y pne ato — -layer thic
j ckness ~ ~

change of direction w ill take place l way Since the etc hange
a energy pf a wa ll is inverse y
~ ~

. .
over maan y atomic p anes (F'g 23 ' The reas n fpr the prppprtip 'naltote
h. thickness q 3 13)) the wa Kq,
di,
o

dua] nature p the change 's is the fact that for a given m&gi ht sp~~~d pu t until it fjlled a siza l ep rppprtlon
' '
total c"ange spln tion the excha n g e energy is the crystal, w ere it not fpr o the restrai in g egect of th e
'
lpwer when the change is distribut' d pver many sp ins anisptropy energy, w hich ic acts to lm the wi th of t
'mit
than when the change pccurs abrup tl y'
.
translsition lay . The splns con tained within the wa]l
beh» o' be understoo eel
ex ression re largely dlrec ted away f«m
are m the ax pf ea, sy magne-t-
 PHVSICAL THEORY OF FERROMAGNETIC DOMAINS
ization, so that there is a certain amount of anisotropy The total wall energy per unit area is
energy associated with the wall. The amount of ani-
sotropy energy will be roughly proportional to the
0;„=2m[JE. S'/a]'* (3.2.7)
thickness of the wall, since the thickness is a measure which in iron is of the order of magnitude
of the total volume directed away from the axes of
easy magnetization.
o=(k.T,E/a j»= [10 "10'/10 '$»
The actual thickness and energy of the transition
= 1 erg/cm'.
layer is the result of a balance between the competing It turns out that the exchange and anisotropy energy
claims of exchange energy and anisotropy energy, the contributions are equal to each other.
former tending to increase the thickness and the latter In the above estimate we have rather arbitrarily
tending to decrease the thickness. supposed that the total change in spin direction is.
shared equally by each of the E atoms on a line through
3.2 Estimate of Thickness and Enexgy
the wall; we have also used a very rough estimate of the
of Bloch Wa11
anisotropy energy of the spin system within the wall.
We proceed to make a rough order of magnitude These approximations are dispensed with in the more
estimate of the thickness and energy of a Bloch wall, exact and rigorous treatment which follows.
deferring' until later detailed treatments of specific
situations and comparison with experimental results. 3.3 180 Degree Walls in (100) Plane of Iron
Let us consider a wall parallel to the cube face of a We consider now in detail the important case of a
simple cubic lattice and separating domains magnetized wall parallel to a (001) plane of iron, separating domains
in opposite directions, as in Fig. 24. We wish to deter-
magnetized in opposite directions. The directions of
mine the thickness of the wall in terms of the number domain magnetization may be taken as [100) and
.E of atomic planes contained within the wall, and we
[100j, as shown in Fig. 24. Equivalent solutions for
wish also to determine the energy per unit surface the properties of the wall in this case were given by
area, 0-„. I.ifshitz (1944) and Neel (1944a).
The energy per unit surface area may be represented We suppose that the rotation of the spin direction on
to a good approximation as the sum of contributions passing through the wall is such that the spiv directions
from exchange and anisotropy energies: lie in the plane of the wall; this is the result for this
&w = &ex+ &anis. (3.2. 1) specific case of the more general requirement that the
normal component of the magnetization remain con-
The exchange energy is given approximately by Eq. stant through the wall, so that no poles are formed.
(3.1.3) for each line of atoms through the wall and The absence of poles is decided by considerations of
normal to the plane of the wall. There are 1/a' such minimum magnetostatic energy: it may be noted that
lines per unit area, where u is the lattice constant; the magnetostatic energy of a double layer IOOOA in
whence
„= 7r'JS'/Na'. (3.2.2)
thickness with surface pole density &I, per unit area
0

The anisotropy energy is of the order of the anisotropy

1S
0 „= 2+I,'d = (10) (10') (10 ') = 100 ergs/cm' (3.3.1)
constant times the volume, or which greatly exceeds the wall energy 0. =1 erg/cm'
0 ~~ jg ~ E'lVa (3 2 3) estimated above on the tacit assumption that the
so that change of spin direction takes place in such a way that
0. = (~'JS'/Na')+KNa (3.2.4)
which is a minimum with respect to E when

= 0= —(7r'J S'/N'a')+Ea, (3.2.5)

ol
N = $x' JS'/Ea')'. (3.2.6)
We have then the result that the thickness of the
wall measured in atomic separations is approximately
equal to the square root of the ratio of the exchange
integral to the anisotropy energy per unit cell. For
order of magnitude, in iron, [10
ohio]
N= [kT./Ea'3»= [10 '3/10'10 "j»
=300 lattice constants FIG. 24. j.80' wall parallel to the cube face of a cubic crystal
and separating domains magnetized in opposite directions parallel
= 1000A. to a cube edge.
562 CHARLES K ITTEL
0
~ 90 reduces to

8 since
f„=(d8/ds)',
A (3.3.3)
2
o
n~= cos9, n2= sine, 0.
n3 ——
2 80
0
I- The wall energy per unit area is then
O
4J

G
2
"I 3o' [E sin'8 cos'8+A (d8/ds)'$ds, (3.3.4)
2

I- which may be written more compactly as

CO

lal
o
Z
0
'-3 -2 -I 0 I 2 3 4
NORMALI2ED COORDINATE PERPENDICULAR TO PLANE OF WALL [g(8)+A8' jds (3.3.5)
VA/ K

FIG. 25. Variation of spin direction within a 90 Bloch wall, by setting g(8) = E sin'8 cos'8 and 8'= d8/ds.
as from L100] to [010] in the (001) plane. In Eq. (3.3.5) 8 is to be determined as a function of
s to minimize the integral on the right. %e require
the normal component of the magnetization is constant therefore that the variation 80. be identically zero for
on passing through the wall. any small variation 88:
The calculation of the properties of the wall will
6rst be carried through neglecting magnetoelastic [g'(8) 88+ 2A8'd88/ds jds = 0. (3.3.6)
energy; the eGect of the magnetoelastic energy associ-
ated with the wall will be considered subsequently.
Let 8 be the angle between the spin direction and the On integrating by parts, noting that
x-axis. The anisotropy energy density in the x, y plane 8'(d/ds) 88= (d/ds) (8'88) —88(d8'/d. ),
is then, according to Eq. (2.2.4),
and that 8'88 vanishes at both limits, we have
= E(n 'n '+nrsn '+nssnss) = E sin 8 cos'8 (3.3.2)
fry

since ns —0. The exchange energy density (Eq. (2.1.11)) 8o „= [g'(8) 2A (d8'/ds) —
j88ds= 0; (3.3.7)
f., = A[(Vn, )'y(Vn, )'+(Vn, )sj
this equation can be identically satisfied over the range
of s only if
g'(8) —2A (d8'/dz) = 0. (3.3.8)
This is the Euler equation of the problem.
Z//A/K On multiplying by 8' and integrating over s between
—~ and z, we 6nd
60
g(8) = A (d8/ds)', (3.3.9)
since 0'=0 at s= —~.
This relation shows that at
every point of the wall the local anisotropy energy
density g(8) is equal to the local exchange energy
90 90 density A(d8/ds)', it follows that in directions of high
anisotropy energy neighboring spins make larger angles
with each other than in directions of low anisotropy
energy.
From Eq. (3.3.9) we have
120

ds=+A (3.3.10)
(g(8))'
so that Eq. (3.3.5) becomes
~82

-8 ~.= 2+A (g(8)) Id8

Fzc. 26. Polar plot of variation of spin direction with a 180
Sloch wall in iron; s is the coordinate normal to the plane of the = 2(EA) I ~
sin8 cos8~ d8 (3.3.11)
wall; the plane of the wall is a (100) plane. 40
 PHYSICAL THEORY OF FERROMAGNETIC DOMAINS 563

$0

IV V
0 )0-4 Z
co N C/)
LL V)
LLJ

&. tA
z
~ t9
O
x
~
Z2 0$ 10-&
LLJ
O

001 10 6
102 10 104 105 10 10
ANISQTROPY ENERGY DENSITY IN FRGS/CM
FIG. 27. Approximate dependence of wall energy and wall thickness on crystalline anisotropy energy;
exchange integral approximately as for iron. (180' wall)

which gives the result rather rough experimental value deduced by Williams,
Bozorth and Shockley (1949) based on domain pattern
o =2(EA)'*.
measurements.
Value of the Wall Energy War Tbzcumess
Now for iron A = 2.0X 10 ' ergs/cm from Eq. The thickness of the wall may be found with the
(2.1.18), while E~ —4.2X10' ergs/cc, so that help of Eq. (3.3.10):
o (Fe) =1.8 ergs/cm'. (3.3.13) ds= (A/E)-'* (3.3.14)
This is an important and basic result. Several remarks sin0 cos0
or
may be made regarding the probable reliability of this s—so
—(A/E) l log(tan8/tan8O). (3.3.15)
value for the energy of a 180 degree transition layer in
an (001) plane. Taking s'o=0 at OO=45', we have
(1) The anisotropy constant E2 is omitted because s= (A/E)'*
the associated energy term E2n&'n2'o. &' is zero in an log tan0. (3.3.16)
(001) plane. This equation is plotted in Fig. 25 for the range 0(0
(2) Neel (1944) arrives at o = 1.4 ergs/crn' for the (s. /2. The s coordinate is plotted as a multiple of the
same situation, as a result of using an estimate of the fundamental length (A/E)' which is equal to 2.3X10 '
exchange interaction constant A obtained in a different cm, or 230A, in iron. Of the 90 degrees change shown
manner (from the Curie point, using a Weiss field); on the vertical scale, 70 degrees takes place for As
any estimate of A must be regarded as somewhat — 3.5(A/E)'*, or (ds)~o =800A=280 lattice constants.
tentative, but we would like to suggest that the value The angle between successive spins is of the order of
deduced from the experimental coefFicient in the Bloch 4 degree.
T' law should be used for wall energy calculations, as However, the expression 3.3.16 does not lead to a
the physical situation in a spin wave is quite similar to convergent answer for even the approximate thickness
that in a Bloch wall. of a 180' wall; this is because the wall tends to split
(3) It will be shown below that consideration of up into two 90' walls, separated by an in6nite distance.
magnetoelastic energy has only a negligible effect on The difFiculty is somewhat 6ctitious and is removed by
the numerical value (Eq. (3.3.13)) of the wall energy in the effect of magnetostriction. A large domain can be
iron, although magnetostriction does have a profound formed at 90' between two antiparallel domains only
effect on the thickness of the wall. at the expense of a considerable magnetoelastic energy:
(4) For 3.85 percent SiFe we have approximately since each domain is elongated in the direction of its
A=1.7X10 ' ergs/cm; E=2.8X10' ergs/cc; so that magnetization the 90' domain does not freely jibe
o„=1. 4 ergs/cm'. This value is compatible with the across its boundaries with the antiparallel domains, and
 CHARLES KI TTEL
E:ASY AXIS Mathematically, the effect of magnetostriction wi11
be shown to be equivalent to adding to the anisotropy
energy density (Eq. (3.3.2)) the term
(a) f. (9/4) (crt —C12))ttpp2 sin'8, (3.3.17)
where X100 is the saturation value of the longitudinal
EASY AXIS
magnetostriction in the L100j direction, and c11, c12 are
the moduli of elasticity. The proof is given here.
As a consequence of magnetostriction the part of the
crystal magnetized in the &x direction is strained by
(Eq. (2.3.9))
C11+C12
e = —81 (3.3.18)
d«ooA (cl1 C12) (cl1+ 2C12)

(c)
e»= e» —81 (3.3.19)
(Cl1 C12) (Cll+ 2C12)
Fro. 28. Stable saturated magnetic configurations. (a) Long
ellipsoid parallel to easy axis; (b) hollow rectangle with each side Here 8& is a magnetoelastic coupling constant and is
parallel to easy axis; (c} very Gne particle.
related to ) 100 by Kq. j2.3.16jj:
t',

a system of stresses must be set up to make the bound-

aries jibe. As a result of this factor the 90' region ~100 (3.3.20)
between the two 90' walls will vanish, and the two 3 (crt —C12)
90' walls will coalesce into a single 180' wall.
Now the material in the domain wall is effectively
Egect of cVageetostnction held at the deformation of the surrounding domains.
The excess magnetoelastic energy of the wall material
Both Neel and Lifshitz have included the term in referred to the domains as a basis is, from Eq. (2.3.6),
magnetoelastic energy in the solution of the wall
problem. We follow the treatment of Lifshitz here. ' ~f„.=a, f(~, —1)e„+~2'e„„]
= Bt sin'8(e„„—e„) (3.3.21)
or, substituting for 81, e» and e„,
Af, = (9/4) (c11—C12))trpp' sin'8. (3.3.22)
EASY The wall thickness is derived as before from Eq.
AXES
(a)
(3.3.10), where we must substitute
g(8) = E sin'8 cos28+ (9/4) (c11—c12))ttpp' sin'8, (3.3.23)
whence

(3.3.24)
(2/E) '*
(sin'8 cos'8+P sin'8)'
N S N S N S
P= (9/4)(c11 —C12))ttpp /E. (3.3.25)
For iron, P=2X10 ',so that the effect of mag22etoehastic
EASY
AXIS
ie irorI, is negligible.
energy on the mull energy
The solution of Eq. (3.3.24) is

(1+Pp &

sinhs(E(1+P)/2) = —
*
S S S ctn8 (3.3.26)
t. P )
I
N N N I

FIG. 29. Examples of single crystal shapes which favor domain

structure: (a) Hollow rectangular specimen with legs parallel to where s is taken from the midpoint of the wall. In the
L110j and equivalent axes; (b) Crystal with rectangular cross
section with easy axis as shown. neighborhood of the midpoint
Noel's treatment is somewhat in error since he fails to recognize
the distinction between anisotropy energy at constant lattice P (3.3.27)
dimensions and at constant stress {see Eq. (2.3.17) above); this
leads him to associate with magnetoelastic energy a term which
d(./(a/E) &]
is automatically included in the experimental value of the ani-
sotropy energy. for P((1, which shows again that for I'=0 a 90'
 PHYSICAL THEORY OF FE RROMAGNETI C DOMAI NS

domain may exist between the two halves of the wall.

The angle variation is plotted in Fig. 26 for P= 2)& 10
as for iron.
In Fig. 27 we give the approximate dependence of wall
energy and wall thickness on the crystalline anisotropy
energy. The exchange energy constant is taken as for
iron: A=2X10 ' ergs/cm.

4. THEORETICAL DOMAIN STRUCTURES

4. 1 Introduction IX
4J
z
The size and shape of domains are not constant LLI

attributes of bulk ferromagnetic material but are func-

tions of the dimensions and the orientation of the
boundary surfaces of the crystals as well as of the state
of strain and of the magnetic field in tensity. The
essential reason why domains exist is that their forma-
tion in general reduces the magnetic field energy which DOMAIN WIDTH, D

would be associated with a saturated magnetic con- FzG. 30. Energy of domain structure as function
Gguration. of domain width D.
The incentive that a given specimen has to form
domains will then depend on the magnitude of the shown in (b) for a uniaxial crystal in the form of a
demagnetizing eBects associated with a single domain rectangular cylinder.
or saturated conGguration in the specimen. For example, The area of Bloch wall is L/D per unit area of the
all of the arrangements shown in Fig. 28 are stable. crystal surface, looked at from above, The wall energy
In (a) we have a long fine prolate spheroid with a is then
direction of easy magnetization parallel to the shape
axis; for sufficiently large values of the axial ratio the per unit, surface area, where o- is the surface energy
saturated single domain configuration will have a lower density of a Bloch wall. The magnetic field energy
total energy than an arrangement with domain struc- associated with the parallel charged strips normal to
ture. That is, the decrease in magnetic energy accom- the paper is, according to Eq. (2.4.4),
panying domain structure can be in this case less than
the energy required to set up the requisite transition wm, g
—1.7I,'D
layers (Bloch walls) between domains.
per unit surface area when both top and bottom surfaces
In (b) we have a structure with four domains. There are taken into account; here I,
is the saturation magnet-
is no magnetic Geld energy here, since there are no poles
anywhere (except for a relatively trivial pole strength
ization. The variation of ~~ and x,
w D is „with
plotted in Fig. 30. For large D the magnetic field energy
on the edges of the Bloch walls). The normal component
is dominant, and for small D the wall energy is domi-
of the magnetization is continuous across the diagonal
nant.
Bloch walls shown as dotted lines, and this is the The total energy per unit surface area is
condition that there be no poles on the walls. A crystal
with the domain structure discussed here has been ~ = ~wall+ ~magq (4 1 3)
produced by Williams and Shockley (1949). ol
In (c) we have a very fine particle, with a diameter w = o„L/D+1.7I,'D, (4.1.4)
of the order of (for iron) 100A or less. Here a single
domain structure is stable because the amount of 0 +—
exchange energy required to set up a non-magnetic
configuration will exceed the magnetic energy of the
single domain configuration. This inequality is only
valid for very small dimensions. There is considerable
experimental evidence for the stability of single domain EASY
AX!.S 1(
structure in small particles, as will be discussed in a
later section.
The crystal shapes shown in Fig. 29 are, on the other
hand, favorable to domain structures along the lines
indicated, provided always that the crystals are of Xl-V-'V-~~
macroscopic dimensions. We shall consider now the
optimum slab thickness D for a domain structure as FzG. 31. Flu& closure domain configuration in uniazial crystal.
566 CHARLES KITTEL

~I~PV'IWP&lfr~ll II'/ lQ/!

I ~ 5IISWI M ~»
l
IlilIm II~~
)ieieeeeeeee I ~ ~
I

FIG. 32(h). Details of flux closure of "domain of closure" Q.

Landau and Lifshitz (1935) for a uniaxial crystal, is

shown in Fig. 31. A similar structure (Fig. 32(a)) has
been observed by H. J. Williams on an iron crystal, in
unpublished work. Iron is, however, cubic.
The "flux circuit" is completed entirely within the
crystal by means of the small triangular prisms on the
upper and lower surfaces of the crystal. As shown in
Fig. 32(b) these domains of closure" transfer the
FiG. 32(a). Retouched photograph of domains of closure magnetization lux from the upward-directed domains
in Si-Fe crystal (Williams). to the downward-directed domains without any poles
being formed. The absence of poles is the result of the
which is a minimum with respect to the domain width continuity across the prism sides of the normal compo-
D when nent of magnetization. We proceed to calculate the
—(o„L/D')+1.7I '=0, optimum value of the domain width and the corre-
Bw/BD= (4.1.5) sponding energy density. The wall energy per unit area
or
of the crystal surface is approximately
D= [o.„L/(1.7I e)]'. (4. 1.6)
wwatt= &uL/D. (4.2. 1)
Therefore, the width is of the order of
The magnetic energy is zero, but the anisotropy energy
D=[(2)(1)/(1.7)(1.7X10')']'*=10 ' cm (4. 1.7) is not zero (if the crystal is uniaxial). The volume
for a crystal with X=1 cm. The domain width D contained within the domains of closure is oriented in a
should not be confused with the transition layer thick- direction of hard magnetization and involves an energy
ness 6 which is of the order of 10 ' cm. IC per unit volume, where E is the anisotropy constant.
The energy of the domain structure is Per unit area of crystal surface on one side the volume
w = 2[1.7I 'oL]'*- (4. 1.8)
in the domains of closure on both sides is D/2, so that

„,
w, , = ED/2. (4.2.2)
which is of the order of
The wall energy tends to increase the domain width,
w= 2[(1 7) (1.7X10')'(2) (1)]'=7 X 10' ergs/cm'. (4.1.9) while the anisotropy energy tends to decrease the width.
Since the crystal was taken with L=1 cm, the energy The total energy is
per unit volume is of the order of 7X10' ergs/cc, w= (o L/D)+(ED/2) (4.2.3)
whereas the magnetic field energy density obtaining
with a saturated single domain structure would be of per unit area, and this is a minimum with respect to
the order of I,'=10' ergs/cc. This shows that, quali- the domain width D when
tatively, the formation of domairts has reduced the energy
(Bw/BD) = (o„L/D')+(K/2) =—
0. . (4.2.4)
of the system by a very considerable amount
If we had taken L= j.0 ' cm, as in a thin film, then The condition for the minimum is then:
the energy density of the domain structure would also
have been of the order of 10' ergs/cc, which is of the D= [2o„L/E]t (4.2.5)
same order as the magnetic energy density of the
saturated film. It becomes apparent that size plays an
important role in domain structure. FIG. 33. EGect of magnetostric-
tion on domains of closure. Dotted
curve shows on exaggerated scale
4.2 Flux Closure Domain Con6gurations the volume which would be occu-
pied by the material in a domain
It is possible to devise a domain arrangement for the of closure if the constraint exerted
rectangular slab just discussed which will have no by the rest of the crystal were
removed.
magnetic poles. Such an arrangement, first treated by
 PHYSICAL THEORY OF FERROMAGNETIC DOMAINS 567

and the corresponding energy per unit area is

w = [2rr„I.E]'* (4.2.6)
FIG. 34. Domain branching near
The energy per unit volume is the crystal surface, as discussed by
Lifshitz. This structure would be
fsetlLee s [20 to%/~ 3 (4.2.7) expected when the anisotropy is
high, and related structures have
If we arbitrarily substitute the approximate values been observed.
of the constants for iron, and take the length I. as 1 cm,
we have
D= [(2)(2)(1)/4X10s]''=3X10 ' cm (4.2.8)
Numerically, for iron, with I.= 1 cm,
f=[(2)(2)(4X10s)/(1)]'=1.3X10s ergs/cc. (4.2.9) D—[(4) (2) (1)/(10') j'=0. 1 cm (4.2. 13)
and
For these values of the various constants the energy
of the fiux closure con6guration Fig. 31 is somewhat fq —[(2)(10')/(1)]'*=50 ergs/cc. (4.2. 14)
lower than that of the simple slab con6guration Fig. General Case
29(b), but with increasing values of the ratio IC/I s
the domains of closure gradually open up and for The nature of the domain structure near the crystal
E/I, '))1 the simple slab arrangement prevails. surfaces may actually be expected to be more compli-
cated than the simple situations we have discussed.
Cubic Crystals with Positive Anisotropy Constant For example, I ifshitz has shown that the arrangement
shown in Fig. 34 has under certain conditions a lower
In cubic crystals with the anisotropy constant E)
0 energy than the triangular prism arrangement. The
the directions of easy magnetization are the cube edges, complete variational problem to find the nature of the
so that in this case the direction of magnetization in domain structure which minimizes the total energy has
the domains of closure can be in an easy direction at not been solved; instead we assume certain types of
the same time that the magnetization in the basic domains on the basis of physical insight, and then
slab-like domains (Fig. 31) is in another easy direction. minimize the energy with respect to one or more
It would appear then that the domain width will be adjustable parameters.
free to increase almost without limit until the whole In general it is probably a fair approximation to
specimen consists of four domains. Usually, however, suppose that the total energy consists of two parts:
this condition is not expected to occur because of the one part being the energy of the Bloch walls, m «, and
eGect of magnetostriction.
There will be in general a magnetostrictive energy
associated with the domains of closure. This results
from the tendency of domains to change slightly in
length in the direction of magnetization, so that
domains magnetized along different lines will- not fit
together smoothly except by the expenditure of elastic
energy in forcing the material to 6t together. The
domains of closure, for example, may be regarded as
being squeezed to join on to the basic domains, as
shown in exaggerated form in Fig. 33.
In the case of a domain of closure the strain imposed
by the basic domain structure is of the order of the
longitudinal magnetostriction constant happ as may be
understood by referring back to Section 2.3. The elastic
energy density of the closure domains is then (approxi-
mately)
fel 2cll)tloo i (4.2. 10)
which is of the order of 500 ergs/cc in iron.
For the situation in this type of cubic crystal we
have as the domain width
D= [40wL/cllX100 j (4.2. 11)
and as the energy density of the domain structure
FlG. 35. Powder pattern on hexagonal face of cobalt
fdo [&wert%os IL] (4.2. 12) single crystal (Williams).
 CHARLES KI TTEL
PHOTOGRAPHIC pattern, such as shown in Fig. 8, whereas the under-
30 KV PLATE
HEXAGONAL FACE lying basic structure in single crystals is believed to be
ELECTRONS
fairly simple in most cases (Fig. 7).

FIG. 36(a). Arrangement for detecting surface magnetic „.

The width of the basic structure is determined by
the surface energy density g(D) of the superficial struc-
ture and the wall energy o. The nature and scale of

„,
6elds by electron scattering.
the superficial structure depends on the relative values
of magnetic, anisotropy and magnetoelastic energies
the other part, m, &„f, t, the local energy associated at the crystal surface.
with the outer surfaces of the crystal. We may suppose In cobalt the anisotropy energy is dominant and as
that for slab-like domains a result the Aux closure is incomplete. A typical domain
~mali=& L/D (4.2. 15) pattern obtained on a hexagonal face of a cobalt crystal
is shown in Fig. 35. L. H. Germer (1942) has confirmed
where I. is same characteristic average length, while by electron beam methods the existence of strong local
w, „~,„y„,g(D)—
i
— (4.2. 16) magnetic fields of the order of 10,000 oersteds just at
the hexagonal face of a cobalt crystal. 30 kilovolt
is some function of the slab width. The nature of the electrons scattered from a hexagonal face produce on a
dependence is determined by the structure of the photographic plate an exceedingly complex pattern
surface domains. The minimum energy occurs for the (Fig. 36) made up of curves or arcs interlaced in elabo-
slab width Do for which rate fashion. Iron and nickel crystals show no similar
phenomenon, and it may be inferred that the magnetic
(& I-/Dos)+g'(Do) =O. (4.2. 17) fields are very weak, as would be expected from domain
It is important to appreciate the fact that the arrangements in which the Aux circuit is closed inside
domain structures of most crystals fall naturally into the crystal. The results for cobalt, of course, suggest
two classifications, a superficial domain structure which that the Qux circuit is not closed inside the crystal, in
is connected with the local details of Aux closure on the this case, and this conclusion is in agreement with
crystal surfaces, and the basic domain structure running theoretical expectation.
through most of the volume of the sample. The super- In the extreme case of very high anisotropy energy
ficial structure very often assumes a highly complicated one may expect the domains to be slab-like over their

xiii

~ ~ I
" sss I IR6IBSl
goal, %! 5!Il+

Nil
II ~

, ,BllPlf)'„

;
K,

Fro. 36(b). Results obtained in scattering of electrons from the hexagonal face of a cobalt single crystal (Germer, 1942).
 PH YS I CAL THEORY OF F E RROMAGN ET I C DOMAINS 569

entire extent, and there will then be no special super-

ficial domains. In the other extreme, where the ani-
sotropy and the magnetostriction are both zero, all
semblance of a discrete domain structure disappears,
and the requirements of Aux closure are over-riding,
except when signi6cant changes in direction of magnet-
ization on a scale of 10 ' cm is required —
in which case
the exchange energy must also be considered. Two
possible domain structures for this case are shown in
Fig. 37. The domains have swollen so that each domain
occupies a large part of the volume of the crystal.
This limiting case may be expected to set in when the
thickness of a Bloch wall becomes comparable with the
crystal dimensions. For a thickness of 10 ' cm, the
(a)
anisotropy energy must be of the order of only one
erg/cc. It is barely possible that we are approaching Fxo. 37, Theoretical possibilities for domain structures in the
limit of zero anisotropy energy.
such a situation in supermalloy in thin tape form
(Bozorth, 1948).
%le have not discussed in this section the question of
investigation. The liquid contains a colloidal suspension
domain structure in the presence of applied magnetic of a fine ferromagnetic powder, which is usually
fields. The calculations proceed along the same general magnetite with a particle size of the order of one micron.
lines as above, and for details the reader may consult
We are now going to show that a great local concen-
the papers by Neel (1944c) and Kholodenko (1947). tration of particles in the colloid 61m is expected as a
Hates and Neale (1949) have confirmed experimentally result of variations in magnetic Geld at the edge of a
the predicted dependence of domain width on applied Bloch wall,
magnetic Geld in the [011$ direction in silicon iron. The magnetic particles are acted on by a force when
in the presence of an inhomogeneous magnetic field,
S. EXPERIMENTAL STUDIES OF DOMAINS and are drawn to the points on the surface of the sample
where the magnetic 6eld intensity is highest. Actually,
In the preceding sections of this review the general it appears as if the particles are in thermal equilibrium,
theoretical basis of the structure of ferromagnetic so that the density distribution of particles in the liquid
domains has been described in considerable detail. We
now discuss some aspects of the experimental basis of
the theory.
There are a number of diGerent sources of experi-
mental information concerning domain structure. By 32
far the most important source, with regard to funda-
mental understanding, is that of magnetic powder 28
patterns. The powder pattern technique was introduced
by Bitter, Elmore and others, and finds its highest
in the definitive papers of Williams, 24
development c~
Bozorth, and Shockley (1949) and Williams and Shock- V)

ley (1949). The principle of the method is analyzed in 0 20

Section 5.1, and a selection from the experimental I-
O
results is discussed in Section 5.2. For further details Z
16
the reader is referred to the papers of Williams and
his collaborators. cA
2.'
Other experiments on domain structure include: CI
t2

(a) Scattering of electron beams (Germer, 1942; Marton, 1948);

(b) Depolarization of beams of polarized neutrons (Surgy,
Hughes, and %allace, 1948);
(c) Sixtus and Tonks (1933) domains in stressed wires;
(d) Dependence of magnetostriction and magnetoresistance on
applied field and stress (Bozorth, 1946).

5.1 Magnetic Powder Pattern Technique 0

3 4
The magnetic powder pattern technique consists in x = &H/yT
the application of a thin liquid layer to the carefully Fzo. 38. Plot of particle density function (sinhx)/x vs, parameter
prepared surface of the ferromagnetic sample under x= pH/kT, where p is the magnetic moment of a single particle.
570 CHARLES KI TTE L
particle makes with the field H. Here we assume that
I the particle has a permanent magnetic moment; also,
?
we neglect particle-particle interactions. On averaging
I p(H, 8) over all angles 8, we get the mean value
l'i
1 t"
I
p(H) = — ~l exp~
&pH cos8i
— ~2z- sin8d8,
kT )
= (sinh(iiH/k T)/(liH/k T)). (5.1.2)
MAGNETIZATION NOR MAL TO SCRATCH
A plot of the function p(H) = (sinhx)/x, where
x=iiH/kT, is given in I'ig. 38. It is seen that the
particle density rises sharply when x is of the order of
3 or greater; that is, when pIJ)3k'l. Taking, as a
rough estimate, I, for the particle as 400 and the
~
~ ~

~
~

~
particle volume as 10 "
cc, we must have a field
variation
3(1.4X10 ")(3X10')
H) (3k T/ii) = =3X10 4 oersteds
MAGNETIZATION PARALLEL TO SCRATCH (4X 10s)(10-»)
FIG. 39. Determination of the magnetization axis by in order that an appreciable density variation may
means of the scratch technique. "
occur. For a particle volume of 10 cc, corresponding
approximately to a diameter of 0.1 micron, the magnetic
is given by the Boltzmann distribution function. That
field variation must exceed 0.3 oersteds.
is, the particle density p(H) at a point where the field
Now there are field variations at the surface of a
intensity is H is related to the density p(0) where the
ferromagnetic crystal greatly exceeding 0.0003 or 0.3
field is zero by the relation
oersteds. On a crystal surface with a Aux closure
p(H, 8) = p(0) exp(liH cos8/kT), (5.1.1) domain configuration, as in Fig. 31, there will be local
fields along the lines where the plane of the Bloch walls
where 8 is the angle the magnetic moment p of the intersect the surface. As the change in spin direction
within a Bloch wall occurs in the plane of the wall, it is
evident that if we cut oG the wall, as we must do at the
crystal surface, then the spins in the wall will have a
component of magnetization normal to the crystal
surface. ' That is, there will be a line of either or 5 S
poles wherever a Bloch wall comes out on the surfaces
of the specimen. These lines of free poles produce a
magnetic field which is quite adequate for the purpose
of forming a dense line of colloid particles. As a very
crude estimate, we may consider the perpendicular
intersection of a wall with the surface of a crystal of
iron to produce a pole density I, on a strip about 10 '
cm wide, corresponding to a linear pole density
g= (1700) (10 ') =0.02 gauss/cm.
At a distance of one micron from the wall this produces
a field H= 2g(r = 2(0.02)/10 '=400 oersteds.
This estimate shows that the field intensity due to
the edge of a Bloch wall is several orders of magnitude
greater than that needed to form a dense line of colloid
particles. Actually our estimate must be corrected for
the p*-eGect discussed in Section 2.4. That is, we must
allow for the fact that the line of poles rests on perme-
able material, with permeability p*=1+2rrI.s/E. The
Fzo. 40. Drawing of colloid pattern on a (100) surface with a solution to the boundary-value problem of the field due
set of scratches parallePto the L010) axis. The magnetization is to a line of poles on a uniform permeable half-plane
parallel to L010$';.in~regions where the scratches do not appear,
and is parallel to L001j where the scratches are visible, 'gee Fig. 23,
 PHYSICAL THEORY OI FERROMAGNETI C DOMAINS
20, 000

12.000

8, 000

4, 000

FIG. 41. Magnetization

0
vs. displacement of the 180'
Bloch wall and patterns 4, 000
showing the Sloch wall
in three different positions
(Williams and Shockley, 8,000
1949).
12,000

16,000

20, 000
0 20 40 60 80 )00
DISTANCE IN MM ON FILM

shows that the 6eld in the air is reduced with respect 5.Z. 1. 5cratch TechrIique
to the free space field by the factor 2/(1+p~), which is
In the absence of a magnetic field the direction of
equal to 0.047 for iron. The corrected field intensity
magnetization of the domains will be along the axes of
at one micron distance is about 20 oersteds, which is
still ample for the formation of a dense line of colloid easy magnetization of the crystal. In an iron crystal
there are three axes of easy magne tization, the
particles, as is actually observed.
The theory of line formation has also (Kittel, 1949b) three being mutually perpendicular and along the cube
been developed for the case of colloid particles which edges.
are not permanent magnets, but which are permeable
In interpreting a powder pattern we wish to know
the direction of magnetization in all parts of the
spheres, as is probably more often the case.
Particle concentration will also occur, and often to a pattern. We may determine the axis of the magnetiza-
tion in some important cases by what is known as the
great degree, in cases where there is not a Aux-closure
scratch technique; and the setose is then determined by
domain configuration, but where domains themselves
observing the expansion or contraction of the domain
come to an abrupt end on the crystal surface. This is
when acted on by an external magnetic 6eld parallel to
expected to occur in materials of high anisotropy energy,
and is in fact commonly found in cobalt and on a (111) the domain axis.
plane of iron.
The scratch technique may be understood by refer-
We have shown that large local concentrations of ence to an (001) surface of an iron crystal. The preferred
colloid particles occur on the faces of ferromagnetic axes are parallel to [100] and [010]. If we make a
specimens. These concentrations may be observed under fine scratch on the surface parallel to [100], we shall
an optical microscope, when the specimen is suitably get colloid deposition on the scratch from domains
illuminated. parallel to [010], but the scratch will not collect
colloid particles from domains parallel to [100]. That
5.2 Selection of Results of Powder Pattern Work is, colloid collects when the magnetization crosses the
We discuss now some of the results which have scratch, but not when the magnetization is parallel to
emerged from powder pattern studies. We shall treat the scratch. The situation is made clear by Fig. 39.
these three topics: It is the leakage Qux that attracts the colloid particles,
1. Determination of direction of magnetization by scratch and there is no leakage when the magnetization is
technique (Williams, 1947) parallel to the scratch.
2. Correlation between magnetization changes and boundary A drawing of a domain pattern demonstrating the
displacement {Williams and Shocl~ley, 1949)
3. Dagger blade domains around cavities and crystal imper- effect is shown in Fig. 40, which was obtained by
&ections (Williams, 1947). Williams (1947).
 'sa Ij I te ;:--~l]E;-
«y-r ~

mgj

'
-' '"
Im Its
:.y
':;:?~
.
:,
'%LQ4'j'''""''"""""i~ywiiij'j" i'i%'.
""'::««» "'ll?iiPii ~i SISs

„.
IK

. %!F ?l?

«'?" "-'.

jill Sg ,, ;,c, j:;:.

e:goal
II 8%
~ jS Sl:.: 8 Ihk'ak'it I
?sih't
rrs LmlIkk
ss
4:,:::.Hill «
I@I?, „. sr?sJ
= "" Ran' i
.-:. II ~ I P. ~
a",.A, II NII"t'?""

I I!
I I

,.c;8

J I SII II '.
I IQg ?H «?. s
«
t" . j~::;:::::, :::-:':::::::::.
:::;. :.
'::::3:.
'
5" ~~"'~""'-':.:"~:;:::::::::,
-.,
:::::, )P!. ','««ll ,":::,
im'

159
' -'
'-ltL-'~~ .p'A
, ~

l, ~m
b
~
~ I
I le '
l'
''ll

«l

J
i':::i?rJ' II-.m': . &.
l' '?
?u?' ~ o. I I +
r«. :;. I ~
l:,.

l?«

~r?I?'y. J
' 'i ' l l
?l'«: ~
4 t.'

IIJ

jj ss r I .'««?5'
8 IISI' ~ l
' s
'.

. ~ «?
iud 8'l~l K:-'::=:-:::=-.
"-':::&l

.9 I rj
~::::':::::.
:k::~&, ~.
M I Sl I554cm

FIG. 42. Domain structure around cavities: the structure in the left-hand drawing was predicted by Neel (1944b) on energy
considerations; the predicted pattern was found experimentally by Williams (1947).

5.Z. Z, Correlation betzeem 3Eagnetisatioe Charges aed form around the cavity, with a reduction in the total
Boundary Di sP/acemeut energy. The Keel structure has the effect of distributing
the poles along the curved portions of the boundaries.
In Fig. 11 we have given a sketch of the simple As the domain elongates the magnetostatic energy
domain structure found by Williams and Shockley decreases and the wall energy increases, and that
(1949) in a single crystal of iron (containing a small structure occurs for which the sum of the two energies
amount of silicon). Reversal of the magnetization is a minimum.
circuit shown in Fig. 11 (a) proceeds by the formation Williams (1947) has found domain patterns of the
and movement across the crystal of the Bloch wall expected type around holes in a silicon iron crystal.
shown in (b) in the form of a rectangle. The rectangle Measurement of the ratio of length to width of the
expands or contracts in size according to the relative Weel domains offers a possible experimental method for
amounts of Aux carried clockwise or counterclockwise. determining 0. , the surface energy density of a Bloch
If the wall shown goes straight through the crystal, wall. Values estimated in this way are of the correct
then we would expect to find a linear relationship order of magnitude.
between the change of magnetization and the displace-
ment of the Bloch wall relative to the side of the crystal. 6. MAGNETIC PROPERTIES OF
This result is verified by the measurements shown in SMALL PARTICLES
Fig. 41. This experiment must be considered as one of If the specimen is cooled from above the Curie point
the most fundamental experiments in the field of in field-free space, it is a matter of common observation
ferromagnetism, as it shows that there is a direct that the demagnetized state is the stable state in large
proportionality between wall displacement and net Aux. ferromagnetic crystals. In the demagnetized state
the domains are oriented so that the magnetic Aux
5.Z.3. Domain Structure Around Cavities
circuit lies almost entirely within the specimen, and the
Noel (1944b) pointed out that the presence of a overall magnetic moment of the specimen is approxi-
cavity in the magnetic material included in a domain mately zero.
will give rise to a considerable amount of magnetostatic As the dimensions are diminished, the relative contri-
energy because of the formation of E and S poles on butions of the various energy terms to the total domain
opposite faces of the cavity. He suggested that a local energy are changed, and surface energies become more
domain structure such as that shown in Fig. 42 would important than volume energies. The surface of the
 PH YSI CAL THEORY OF F ERROMAGNETI C DOMA I NS 573

transition layer (Bloch wall) between domains is a

surface energy, while the energy in the magnetic field
(magnetostatic self-energy) is a volume energy. When
very small dimensions are reached, there will be a point
at which it is favorable energetically to do away with (a3
the domain walls, so that the whole specimen becomes
one domain and acts as a permanent magnet. This was FiG. 43. Types of simple domain arrangements in a small
sphere: (a) Applies for low anisotropy, (b) for high anisotropy
first predicted by Frenkel and Dorfman (1930), al- in a cubic crystal, and (c) for high anisotropy in a uniaxial crystal.
though in their note the surface energy of the boundary
between domains is over-estimated by a factor of the radius R, the energy is
order of 50, leading to much too high values of the
critical dimensions for single domain behavior. Im- w= fV = (1/2)(4n/3)'R'I ' (6.1.2)
proved and corrected calculations were first published and is approximately 24)&10' ergs for X=1 cm and
by Kittel (1946); followed by Neel (1947a) and Stoner 24)&10 "ergs for 8=10 ' cm.
and Wohlfarth (1947; 1948). We now must consider the energy of simple arrange-
Experimental evidence of the intrinsic permanent ments of domains. If the anisotropy is low it appears
magnetization of small ferromagnetic particles was reasonable to take the circular configuration shown in
demonstrated first by Elmore (1938, 1941), although Fig. 43(a) as the most favorable Aux-closure configura-
the eAect had been suspected earlier, as for example by tion, whereas if the anisotropy is very high we may
Antik and Kubyschkina (1934), on the basis of evidence expect a situation as shown in (b) for a cubic crystal
regarding the magnitude of the coercive force. The gen- and as shown in (c) for a uniaxial crystaL These three
eral agreement of experiment with theory in the small cases will now be treated individually.
particle area is not as good as might be desired, but in
many cases it seems pretty clear that the theory is on 6.1.1 I.ow Amsotropy
the right track. Complications due to agglomeration in L

actual materials have not been discussed adequately. The energy here is largely in the form of exchange
The single domain behavior of small particles prom- energy. Consider the spins on a circular ring of radius
ises to be of great practical importance in connection r. There are 2~r/u spins on the ring, where u is the
with materials for permanent magnets. This is because length of a side of the unit cell. The total change in
of the very high values of the coercive force which it is angle in going around once is 2x, so that the angle y
possible to obtain with single domains. between successive spins is
Similarly, single domain behavior is expected in very q =u/r, (6.1.3)
thin films (Kittel, 1946), and experimental evidence for and
this is suggested by the recent work of Origo and Pizzo (1/2) J(u/r)2(2irr/u) = vr Ju/r,
w„;„~= (6.1.4)
(1948) on the Barkhausen effect in thin films of Fe, Ni,
and Co. They find that the Barkhausen eGect disap- by Eq. (2.1.5), with S=1. Now consider the sphere
made up of circular cylinders (Fig. 44), each one unit
pears when the thickness of the films is reduced to
10 ' cm, in good agreement with the theoretical value
cell in thickness. The number of rings in a cylinder is
(2/u)(R' — r')«, so that
of the critical thickness for single domain behavior.
w i — r2)«/r, —
2' J(R— 2
(6.1.5)
6.1 Critical Particle Size for Single
Domain Behavior
We shall consider a small spherical ferromagnetic
w, „i„„, = (2m J/u)
~B
[(R' —r')«/rjdr
particle of radius R; it is supposed for the sake of =(2%rJR/u)DB(2R/u) —1j (6.1.6)
concreteness that the particle is a single crystal. We
or, per unit volume,
are interested first of all in the critical particle size for
which the energy of the single domain configuration is
lower than the energy of configurations in which there
.
f, = (3/2) (J/uR') )le(2R/u) —1]. (6.1.7)
This energy density depends on the size of the sphere.
is a domain structure tending towards Qux closure.
The energy density of the saturated single domain
con6guration is in the form of magnetostatic energy,
and is equal to, from Eq. (2.4.3),

f= 'IHIP =2~I '/3- (6.1.1)

Fn. 44. Decomposition of sphere into
circular cylindrical shells.
r

for a sphere. The numerical value is approximately

6X10' ergs/cc in the case of iron. For a sphere of
 CHARLES KITTEL
CRYSTAL possible for the critical radius to exceed appreciably
AX IS
the wall thickness. When this condition is satisfied the
critical radius may be calculated by use of the model
Fxc. 45. Geometrical situation for shown in Fig. 43(b). The energy here is essentially in
the calculation of the coercive force
resulting from magnetocrystalline ani- the form of wall energy and is equal to
sotropy, for the case of a uniaxial
crystal. ~uiall 2&m~+ ~ (6.1.9)
For an iron
particle with J'=10 'cm we have m, ll
—14X10 " ergs; but the value of the critical radius is
about 0.7X10 cm, since
Using the value I/a=2X10 ' ergs/cm for iron from R, = (9/4~) (o. /I, ') (6.1.10)
Eq. (2.1.18), we have
as follows from equating Eq. (6.1.9) with Eq. (6.1.2).
(a) 8=1 cm This value for the critical radius is less than the wall
m„=1,3X10 4
ergs thickness so that the calculation is not applicable. If
(b) 2=10 ' cm the anisotropy energy were increased by a factor of 10
m. ~=23X10 "ergs or more, then the calculation would be applicable.

6.1.3 High Anisotropy, IIniaxial Crystal

f.~= 0.8X 10' ergs/cc.
For the model shown in Fig. 43(c) the energy balance
The values of m may be compared with those calcu-
is roughly given by
lated following Eq. (6.1.2) for the saturated configura-
tion. It is seen that the Aux-closure configuration has (1/2) (1/2) (4n/3) 'E.'I 2+ mrE, 'o
by far the lower energy when the radius is 1 cm, but = (1/2) (4s/3)'R'I ' (6.1.11)
when the radius is 10 ' cm the saturated configuration which leads to
has the lower energy. R.= 9o/4irI, '. . (6.1.12)
Thus for sufficiently small particles the saturated
This is identical with Eq. (6.1.10). For MnBi we may
configuration has a lower energy than the Qux-closure
estimate
configuration. The critical radius E, is given by
L20/(600)'j 4X 10—' cm,
(1/2)(47r/3)'Ec'I'= (ir JR,/o) Lln(2R, /u) —1j, (6.1.8) '
while the wall thickness is of the order of 8=2&10
and is approximately inversely proportional to the cm, so that in this case the basic assumption b/R, ((1
saturation magnetization. The value of the critical of the calculation is valid.
radius for iron is ~10 ' cm. Similar calculations for fine wires and thin films, as
An expression practically equivalent to Eq. (6.1.8) well as for fine particles, were given by Kittel (1946).
was first given by Neel (1947a).
6.2 Coercive Force of Small Particles
6.1.2 High Anisotropy, Cubic Crystal We have just seen that the formation of domain
In the preceding calculation the anisotropy energy boundaries is energetically unfavorable in the case of
associated with the Qux-closure configuration was neg- suf5ciently small particles. Now in the absence of
lected in comparison with the exchange energy; this is domain boundaries changes of magnetization cannot
justifiable if the critical radius is appreciably smaller proceed by the "easy" process of boundary displace-
than the thickness of a Sloch wall in the material. ment (Fig. 5) but must rather proceed exclusively by
This follows because the anisotropy and exchange the "hard" process of rotation of the total magnetic
energies are equal in a Bloch wall, but the exchange moment of the particle (Fig. 16).
energy will be dominant if the change in spin direction When we have ruled out the possibility of boundary
is constrained to take place in a distance less than the movement we may obtain large increases in the coercive
force by means of increasing the effective anisotropy of
wall thickness.
But if the anisotropy energy is high it may be —
the specimen that is, by making it as difficult as
possible to rotate the magnetization of the domain as a
ThsLE Il. Maximum coercive forces of small particles due to
various causes. (Complete orientation assumed; packing effects whole. This was pointed out Grst by Kittel in 1946.
neglected. ) T=2X10" dynes/cm'. In order to reverse the direction of magnetization in a
small particle it is necessary that the magnetic energy
Anisotropy Shape Internal strain acquired by the particle in the external magnetic Geld
Expression 2E/I, 2m-I, AT/I. be greater than the internal energy tending to prevent
Fe 500 10,700 600 rotation of the domain direction. The effective internal
Co 6000 8,800 600
Ni 135 3,150 4000 anisotropy will be high if the magnetocrystalline ani-
sotropy energy is high (Kittel, 1946), or if the shape of
 PH YSI CAL THEORY OF FERROMAGNETI C DOMAINS

the particle is very elongated (Neel, 1947b), or if a 12,000

severe strainis applied in an anisotropic manner (Stoner 10,000
and Wohlfarth, 1947). The first of these causes is most
LLJ

probably responsible for the highest observed values of O 8PPO

the coercive force, namely, 12,000 oersteds in MnBi x
and 20,000 oersteds in FePt. 6, ooo
We discuss now the coercive force due to each of 0
4, ooo
these three causes separately. It is assumed in all cases
that only single domains occur. We first treat the case 0 2, 000
of a single particle, and later consider the modifications
0
introduced when a collection of particles are packed 0 2 4 6 8 10 12 14 16 I8 20
together densely. Table II compares the values of the AXIAL RATIO

coercive force of fine particles of Fe, Co, and Ni, as I'IG. 46. Coercive force in an iron particle due to anisotropy of
calculated for various mechanisms. shape, as a function of the axial ratio of the particle.

6.2. 1 Coercive Force Resulting from Magnetocrystalline usually oriented at random. In this case Neel (1947a)
Anisotropy has shown that for cubic crystallites oriented at random
the coercive force for the average hysteresis loop is
The anisotropy energy density in a uniaxial crystal
may be written, to the first order, (H, )~„=0.
64K/I, (6.2. 10)
fx=Ki sli18, (6.2. 1) for E&0. This is about 160 oersteds for iron. If we
assume the same factor for cobalt, which is uniaxial,
according to Eq. (2.2.2); 8 is the angle between the we get (H, )~„=2500oersteds.
crystal axis and the direction of magnetization. The
magnetic energy density is, for an applied magnetic 6.Z. Z Coercive Force Resulting from Anisotropy
field Ho parallel to the crystal axis, of Particie Shape
f„.
, =HpI, cos8 (6.2.2) We suppose that the particle is in the shape of a
prolate spheroid. We restrict the discussion to the case
where the choice of sign was made to correspond to a
field direction opposite to the projection of magnetiza-
tion on the crystal axis (Fig. 45).
The total energy is
f=Ki'sin'8+HpI, cos8 (6.2.3)
and is a minimum with respect to 0 when e= 0 0-
Bf/88=0=2Ki'sin8cos8 —HpI, sin8 (6.2.4)
or
Hp=(2Ki'/I, ) cos8 (6.2.5)
which is a maximum (and therefore equal to the
coercive force) when 8=0. Thus,
H, =2Ki'/I, . (6.2.6) e =90' 0-
This result also applies to the case of cubic crystals
if the field is applied along a $001$ direction. For a
small 8 the anisotropy energy is
fK=K18 (6.2.7)
according to Eq. (A.3). The total energy is
f=Ki8'+HpI, cos8, (6.2.8) e =45' 0-
which is a minimum with respect to 8 when
8f/88=0=2Ki8 HpI sin8— 0
which, in the limit 0=0, gives H
H, =2Ki/I, (6.2.9) 2rrIS
Pro. 47. Magnetization curves for elongated particles, for the
applied field parallel, perpendicular, and at 45' to the long axis
In actual samples of powder materials the grains are of the particle.
576 CHARLES Kl r rEL
I. O of a long circular cylinder; here
THFORETICAL
CURVE
H, = 2prI, = 8,/2. (6.2. 13)
08 Values of the theoretical maximum H, for several
materials on the basis of the shape eAect are listed
below. These values do not consider interaction eGects,
06
9u ~ = EXPERIMENTAL POINTS FROM which are discussed later.
x GUILLAUD S DATA ON Mn Bi. USING
Hc = 20 000 dp = 9.0 MICRON S
~
Matenal Max. H, (oersteds)
U
0.4 Fe 10)700
Co 8, 800
Ni. 3, 150
0.2
The nature of the magnetization curves when the
shape e8ect is predominant is shown in Fig. 47, where
0
curves are given for the applied field parallel, perpen-
0 10 '12
dicular and at 45' to the long axis of the particle.
dp d
Stoner and Wohlfarth (1948) have shown that for
FIG. 48. Comparison of Guillaud's data on MnBi with theo- particle axes distributed in random orientations, the
retical lower limit to coercive force, H„as
function of particle average coercive force is given by
diameter d.
(H, )g, = 0.48(1V(—cVp) I, (6.2. 14)
where the applied magnetic field Hp is parallel to the
so that the above estimates should be reduced by
long axis of the spheroid, but opposite to the original
about one-half where the orientation is random.
direction of the magnetization.
Neel (1947b) has considered the case when the axial
Let N p denote the demagnetizing factor of the prolate ratio c/a of the prolate spheroid is nearly unity:
spheroid in the direction of the major axis, and let N~
denote the demagnetizing factor in any direction at c/8= 1+s.
right angles to the long axis. YVe have Np&N~, . for a
long circular cylinder Np=0 and N& —2m,. for a sphere
Here we have, by manipulating the analytical expres-
sions for the demagnetizing factors,
Ão= 1Vc=47r/3.
The energy is cVg —lIt'p=8v e/5 (6.2. 15)
f„„= (1/2)I, '(Xp cos'8+Nq sin'8)+HpI. cos8 (6.2. 11) so that, from Eq. (6.2. 13),
where the first term is the energy of the demagnetizing H, 87reI/5— (6.2. 16)
field and the last term is the magnetic energy due to
the applied field Hp. The energy is a minimum with
If the directions of the major axis are distributed at
random, Neel finds by graphical calculations that the
respect to 0 when
average coercive force is reduced by the factor 0.48,
Bf/88=0=I, '(X~ Np) cos8 sin8 —HpI, sin8. so that

0=0 and
(H, )g„—
0.48(8v./5) eI, (6.2.17)
The coercive force corresponds to is equal to
which gives (H.)g, = 4100e for iron.
H. = P~ —&o)I' (6.2. 12)
6.Z.3 Coercive Force Resulting from Longitudinal Stress
Values of the
coercive force calculated for iron
(I, =1700) using this equation are plotted in Fig. 46 The magnetoelastic energy density is (for isotropic
as a function of the axial ratio of the prolate spheroid. magnetostriction), according to Eq. (2.3.22),
The coercive force is a maximum for the limiting case f,= ssXT sin'8, (6.2. 18)
where X is a magnetostriction constant and T is the
applied stress. The total energy density in an applied
field Hp parallel to the stress is

I „p
i[

~
f= sPXT sin'8+HI, cos8, (6.2. 19)
which is a minimum with respect to 0 when
fz, I'
Bf/88=0=3XT cos8 sin8 —HI, sin8,
or, for 0=0,
H, = 3XT/I„ (6.2.20)
FH;. 49. Initial stage of magnetization reversal by formation of
Bloch transition wall of width 8. as was given by Stoner and Wohnarth. Where the
 PHYSICAL THEORY OF FERROMAGNETIC DOMAINS

stress is due to internal strains a reasonable upper limit 10

to the stress is taken by these authors to be 200 kg/mm',

or 2X10" dynes/cm'. The maximum coercive forces
for iron, nickel, and cobalt are then 600, 4000, and 600,
respectively. V
z
x 104
6.ZA DePenderice of the Coercive Force ori, Particle Size
X

It
~A

is an experimental fact that the coercive force 0

-U
increases rather gradually as the particle size is de- Z

creased. The variation of coercive force is probably

IX
more gradual than can be accounted for by the disper- 4J
ILI
X 10
sion in the dimensions of the particles in any sample. O
The experimental results of Guillaud (1943) on fine
powders of the compound MnBi are shown in Fig. 48;
it is seen that there is a considerable variation in
coercive force H, over the range of diameters from 3 to 2
7/ IS
100 microns. 10
10"7 10 6 10 10 4
A rough theoretical explanation of this variation has SURFACE ENERGY OF WALL / (SATURATION MAGNETIZATION) IN CM

been given by Kittel (1948), using a model (Fig. 49) FIG. 50. Approximate upper limit to particle diameter for
which applies to spherical particles of high anisotropy manifestations of single domain behavior.
material.
We consider a small sphere magnetized to saturation; is given by
when the field H, is applied it is supposed that a
D= do/2 = 12o/Ia . '
(6.2. 23)
domain wall forms as shown in Fig. 49. In this position
the energy balance is given approximately by the This relation is plotted in Fig. 50; the coordinates
equation calculated for Fe, Co, Xi, and MnBi have been plotted
=H, I, (m'p 8/8)+iI as short lines, so as to suggest the approximate nature
o(~p /4) U(5/2d). (6.2.21)
of the calculation. The particle diameter calculated
Here 0- is the surface energy density of a Bloch wall, from Eq. (6.2.23) may be considered as an approximate
p is the wall diameter, and 8 is the wall thickness; U is upper limit for pronounced manifestations of single
the volume of the sphere. The term on the left approxi- domain behavior. The relation assumes a high aniso-
mates the energy of formation of the wall; the first tropy energy, so that it is not really applicable to nickel
term on the right approximates the magnetic energy of and iron.
the wall material in the applied field H„while the
second term on the right is a crude estimate on dimen- 1000
sional grounds of the change in the magnetic self-energy
PRESSED PONDER
of the sphere. 70 Fe-30 C0
Using the geometrical relation p'=48d, we may write
Eq. (6.2.21) in the form 800

H, /H;=1 —(d/d, ),
where H, "=2o/8I, and do
— 4o/I
2. 2
Now, from th. e 0
600
theory of the Bloch wall, o (EkT,/u) and &

(kT, /Eu)', so that H, "

2E/I„ in agreement with x
0
~a

the value obtained from domain rotation. For MnBi, O

we calculate H, "=40,000 and dp 7&&10 ' cm. The 0IX 400
estimated value of dp is uncertain by an order of magni- LLJ

tude, because of the crudeness of the estimate of the 0

change in self-energy of the sphere. The theoretical 0u
curve plotted in Fig. 48 was fitted to the data using 200
H, "=20,000 and dp —9&10 ' cm.
The present theory may be regarded as setting an
approximate looser limit to the coercive force as a
—
function of particle size a lower limit because it is not
0
0 0.25 0.50
PACKING FACTOR
0.75 &. 00
clear without very detailed calculation that the situa-
tion pictured in Fig. 49 actually corresponds to the FIG. 51. Effect of density of packing on the coercive force of
maximum energy barrier for wall formation. 70 Fe 30 Co 6ne powder magnets, according to Weil (1947).
The packing factor is the fraction of the volume of the magnet
On this model the particle diameter for H, =H, "/2 which is 6lled by magnetic material.
 CHARLES K ITTEL
$00- 6.Z. 6
v 9o Dependence of the Coercive Force on Temperature
SHAPE EFFECT
O
O When the coercive force of a fine powder is deter-
I
0 eo mined by magnetocrystalline anisotropy the tempera-
CI
LJ
ture variation of the coercivity will be determined by
K
LsJ
the temperature variation of K/I„when the coercive
60 force is determined by shape anisotropy, the tempera-
v ture variation of the coercivity will be determined by
x
0 the temperature variation of the saturation magnetiza-
'Q 40 tion. In Fig. 52 we compare the results of calculations
on these two types of temperature variation for the
case of fine iron powders.
20 The results of experimental measurements on fine
4J nickel particles by Weil and Marfoure (1947) are shown
lK
in Fig. 53. The increase of coercive force with decreasing
0
-200 200 400 600 800 1000 temperature is very marked, but is not as rapid as
TEMPERATURE IN DEGREES C
would be expected if the coercive force were due entirely
Fzc. 52. Theoretical temperature dependence of the coercive to crystalline anisotropy.
force of fine iron powders for two different models. Curie point is
at 770'C; V. INITIAL PERMEABILITY AND COERCIVE FORCE

6.Z. 5 Dependence of the Coercive Force on the

7. 1 General Remarks
Density of Paching The initial permeability and the coercive force are
structure sensitive properties of magnetic materials;
The mutual interaction of the magnetic moments of
the particles in a fine powder matrix causes in some
cases a reduction of the coercive force. For situations Is
V////////////////////////////////////////////////////////r //'/'//A
where the effect occurs it is greater the denser the
packing. Figure 51 shows results obtained by Weil
(1947) with fine powders of an Fe-Co alloy.
The eGect is expected to be most important for
elongated particles, whose coercivity is determined
principally by the shape eGect discussed in Section FiG. 54. Displacement of domain boundary by an applied field.
6.2.2. For a square lattice of infinite fine circular
cylinders Shockley and Kittel have shown in unpub- that is, they may change in value by large amounts as
lished work that the coercive force is given by a result of small changes in the metallurgical treatment
and chemical composition of the material. By contrast,
II= (1 —Pp)2mI„ (6.2. 24)
properties which are not normally structure sensitive
where P is a coefficient which is equal to 1.1 for cubic are the density and the saturation magnetization. In
packing and to 1.0 for hexagonal close packing; p is the general our knowledge of the physics of structure
packing factor. The calculation is somewhat too lengthy sensitive properties is not very detailed. The reason for
to be given here. this is that the physical dimensions of the impurity
aggregates or strain centers (or whatever is the cause
200 of the structure sensitivity) are quite small, often of
IA
O
:
—- — FINE POWDER
SOUD
the order of one micron or less, so that we find it
LLJ dificult to get reliable information as to the actual
u) IQO
physical state of the material. Further, it is dificult to
0
LLJ

perform controlled experiments with a single foreign

Z
aggregate or strain center on the required small scale.
IOO
In practice the very existence in bulk material of a
v
LL' non-infinite initial permeability and non-zero coercive
0
force suggests that the specimen is imperfect and non-
so
v' uniform. In an ideal specimen the boundary wall
LL.

0 separating two domains magnetized in opposite direc-

v
0 tions should be easily displaced (Fig. 7.1.1) by the
» 200 - IOO 0 IOO 200 300 400
application of an exceedingly small applied field H.
TENIPFRATVRK IN PMRK4$ CFNTIGRADE
The essential physical problem of the coercive force
FIG. 53. Experimental variation of the coercive force of nickel
with temperature: comparison of nickel in bulk and in tine"powder may be reduced to the problem of determining the
form (Weil and Marfoure, 1947), threshold magnetic geld IIs necessary for translational
 PHYSICAL THEORY OF F ERROMAGNETI C DOMAINS
motion of the transition layer or boundary wall separating
domains magnetized in opposite directions. What the
theory must do is to provide a mechanism whereby the
energy of the specimen will vary in a more-or-less
irregular way with the position of the Bloch walls. If
the energy varies irregularly there will then be positions
of minimum energy, and the walls will naturally take
up these positions. The walls may be displaced from
these positions by the application of a magnetic field,
which in e8ect exerts a pressure on a wall tending to
displace it so as to increase the magnetization in the
direction of the field. The initial permeability is a
measure of the intrinsic restoring force on the wall for
small displacements, while the coercive force is a
measure of the magnitude of the maximum restoring Frc. 55. Model for calculation of coercive force on
force on the wall. The coercive force tells us the field Kersten impurity center theory.
strength needed to carry a wall from one potential
energy valley to another potential energy valley, past It appears probable that experimental work now
the highest intervening energy hump. going on will in the near future give us a better physical
I. et us, formally, associate all energy changes attend- picture of the actual mechanisms at work, thereby
ant on the wall motion with changes in energy of the making possible theories better grounded in experi-
wall itself. Suppose that wall energy per unit area o. mental fact than are any of the present theories of
must increase by 60;, in order that the wall may coercive force.
traverse a distance d, x. The required energy is provided
7.2 Kersten Inclusion Theory
by the reversal of the magnetic moment I,hx of the
volume affected, in a magnetic field H' just sufhcient As an example of the nature of the calculations
to induce the desired displacement. Then involved we give here a short estimate of the eGect
„,
2H'I, kx = Ao. (7.1.1)
where the term on the left is the decrease in magnetic
of non-magnetic inclusions on the wall energy. The
'

coercive force on this model arises because the wall in

a position where it intersects a number of inclusions
energy of the system resulting from a change in direction will have a smaller area, and therefore a lower energy,
of the magnetic moment I,
dx from an orientation than in a position where no inclusions are intercepted.
anti-parallel to H' to an orientation parallel to H'. We use a highly simplified model in which the in-
The magnetic energy is converted into surface energy clusions are pictured (Fig. 55) as spheres of diameter
of the boundary wall. The effective pressure exerted by d arranged in a cubic lattice with lattice constant s.
the field is 2H'I, . When the boundary intersects a sphere the wall energy
The threshold field Ho for the displacement of a is lowered by an amount corresponding to the energy
boundary over the length of a domain will be deter- in the wall area which is effectively removed or "short-
mined by the largest local obstacle encountered in circuited" by the presence of the inclusion.
passage, so that
Po (1/2I. ) (do /dx—)„.
—, (7.1.2)
100

This gives the order of magnitude of the coercive force.

The problem is now reduced to the estimation of
(do /dx) ...
There are three principal types of mechanisms which
0/)
O
4J
V)
30

P'
K
have been considered in connection with the theory of LLI
10
coercive force. F. Bloch first suggested that inhomo- z ~ 0.3(HC)
geneous internal strains might play a role in determining zV 0 g
the resistance to boundary motion, and this idea was
3
developed by Kondorsky (1937) and Kersten (1938).
The theory of the effect of aggregates or inclusions of
foreign atoms was developed by Kersten (1943). An
1
important criticism and extension of both of these 0.3 3 10 30 100
PERCENT EXCESS COPPER
theories has been put forward by Neel (1946), who
stresses the part played by the demagnetizing energy FIG. 56. Coercive force of iron with heterogeneous inclusions of
precipitated copper, as a function of the excess of copper over
resulting from magnetization variations produced by and above the solubility limit 0.5 percent Cu at 600'C (after
both internal strains and inclusions. Kersten).
 CHARLES KI TTEL
and the half-width 8 of the boundary wall, which is of
the order of o /E, where E is the anisotropy energy.
(a) Then
H, = (K'/I, ) (8/d) n*. (7.2.6)
This is derived on the supposition that the inclusion
diameter d is much greater than the wall thickness 6.
We may, to a rough approximation, suppose that the
maximum value of II, will occur when b=d, so that

I
Hp
(b)
Kersten gives
(H, ) „= (K/I, )n*. (7.2. 7)

j
SIBLEE

j
x
(H, ) ..=2.5(E/I, )ni; (7.2.8)
these values are plotted in Fig. 56 and are compared
-i I R REVERSIBLE with the experimental values of H, for iron with
REVER-
precipitated copper.
FIG. 57. Wall displacement in presence of a sinusoidal
stress distribution (Kersten). Initial Permeability orI, the Iecllsioe Model
The initial susceptibility xo is given by
Consideration of Fig. 55 shows that the wall energy
is given by gp = (dI/dx)/(dH/dx) . (7.2.9)
s' —m L(d'/4)
—x'j The magnetization change AI associated with the
o(x) =op boundary shift Ax between oppositely magnetized do-
$2
mains is given by
for j x~ &d/2. Taking the derivative, dZ = 2I,Phx/s (7.2.10)
(do/dx) = 2ops x/ss (7.2.2) where P=s/h; here ls is the average thickness of a
so that domain. From Eqs. (7.1.1) and (7.2.2),
(do/dx) .. =oped/s'. H = (1/2I, ) (do/dx) = 7r(o./I, ) (x/s'), (7.2. 11)
By combining Eqs. (7.1.2) and (7.2.3), the coercive = pr(o/I, ) (dx/s') = 27r(K/I, ) (h/s')dx
force is found to be dH (7.2. 12)
~ ~

H. = ( /2)( -/I. )(d/ ') (7.2.4) Xp

—(pI ss/27rK6). (7.2. 13)
Introduce now the inclusion packing fraction On taking account of 90' as well as 180' walls,
n = xd'/6s' (7.2.5) iup =47rxp —1 6PI,'d/k8n'. (7.2. 14)
for po))1, d))8.
7.3 Strain Theory
In the presence of a stress T the surface energy
density of a Bloch wall can be shown, by the application
V)
O
LJ
of the reasoning of Section 2.3 to the subject matter of
Section 3.3, to be approximately equal to
K
=2[A(E+XT)j'
4J
G o. (7.3.1)
z
x
O where we neglect numerical factors of the order of
unity. Here X is the saturation magnetostriction. The
only new idea in this equation is the explicit introduc-
tion of strain anisotropy as represented by the XT term.
Suppose that T varies with position in the following
way:
10 20 30 T= Tp+hT sin2~x/l, (7.3.2)
T' IN KGJ'MIH2
as shown in Fig. 57. Then
FIG. 58. Coercive force P.
of nickel as a function of the average
internal stress T;: {a) Recrystallized wire, {b) hard drawn wire
{Kersten, 1938). The T; are estimated through independent mag- do„/dx=7 [a/(K+ZT)]
= (2mAAT/l)LA/(K+AT)]'
'(dT/dx)-
netic measurements. cos2mx/l (7.3.3)
 PHYSICAL THEORY OF FERROMAGNETIC DOMAINS 58i
which assumes the maximum value

(do/dx) .,= (2m. XET/l) [A/(E+) T)]'. (7.3.4) S S

/

This gives approximately

lt
~s
(7.3.5) /
/
/
N N N
/
NN &S
where 6 is the thickness of the Bloch wall and / is the /
/

scale of the stress variation. s

From Eqs. (7.1.2) and (7.3.5),
(7.3.6) (a)
Fro. 59. Figure illustrating the dependence of magnetic energy of
Experimental results showing the dependence of II, on an inclusion on the position of a boundary wall.
the magnitude of the internal stress AT are given in
Fig. 58. By a sophisticated development of this basic idea in
This sketchy calculation should be adequate to get the 1946 paper, Xeel obtains
across the idea that the coercive force depends on the
magnitude of the stress variation as a consequence of Iron: H, = 2. 1v+360v' oersteds, (7.4.3)
the dependence of the wall energy on stress. We might Nickel: H, = 330v+ 97v' oersteds. (7.4.4)
expect this mechanism to be of importance in material
Here L' is the fraction of the volume occupied by
with high magnetostriction such as nickel. For further
details it is suggested that the reader consult the inclusions and v is the fraction of the volume subjected
original papers of Kondorsky (1937) and Kersten to irregular internal stresses of magnitude 30 Kg/mm'.
(1938). The work of Williams and Shockley on an iron
crystal with simple domain structure suggests that the
7.4 Neel Magnetization Fluctuation Theory interaction of the principal domain boundaries with the
dagger blade domains (Section 5.2.3) around crystal
Neel (1944b, 1946) has suggested that the inclusion imperfections plays an important part in determining
and strain theory calculations are seriously in error in the coercive force of this particular educated crystal. It
several respects. First, the implicit assumption of a appears likely that further work of this character will
regular distribution of irregularities, as for example the give a more satisfying physical basis to our under-
assumption of inclusions on a cubic lattice, greatly standing of the mechanism of coercive force.
overvalues the coercive force for actual materials where
the perturbations might be supposed to be more-or-less ACKNOWLEDGMENTS
random. Second, the assumption of rigid domain walls
also acts to give too high values of the coercive force.
I wish to thank those who have offered suggestions
on the manuscript: in particular R. M. Bozorth, K. K.
When the appropriate corrections for these factors are
Darrow, J. K. Gait, A. N. Holden, L. Tisza, and J. H.
made the maximum coercive force yielded by the
Van Vleck. The photographs of domain patterns have
theory, according to Xeel, is of the order of one oersted,
been generously made available by H. J. Williams. J. G.
which is much too small to account for the coercivity
Walker has been of constant help in organizing the
of many magnetic materials.
6gures, and in this and other ways has shown to how
Neel points out that the magnetic energy associated
high a value the multiplication factor of a technical
with inclusions or stress variations may be considerably
assistant can rise; Mrs. E. M. Sparks has been of similar
greater than the changes in wall energy associated with
service in organizing the manuscript.
the same structure. Consider for example the two
positions of the wall shown in Fig. 59. In the position
APPENDIX A
(a) the magnetic energy of the inclusion' in the form
of a sphere of radius g is, from Eq. (2.4.3), Mathematical Expressions for the Anisotropy
of Cubic Crystals
tv. = (1/2) (4v./3) (4v a'/3) I.', (7.4.1)
The expression (Eq. (2.2.4))
whereas the magnetic energy in position (b) is fK = +1(~12~22+~12~32+22~32) (A. 1)
xg= 0.46m, (7.4.2) in the direction cosines n1, n2, n3 of the magnetization referred to
the cubic axes becomes
according to a potential theory calculation given in f~= E1 sin20 cos20= 4'E1 sin20 (A. 2)
Neel's 1944 paper. The difference in energy between in the (001) plane, where n3=0.
m and m~ is a measure of the field which must be For small a2, n3'.
applied in order to make the wall move from position f~~E.102 (A. s)
(a) to position (b). where e is the polar angle between n and the L100$ axis.
 CHA RI. ES K I TTE L

If we write Eq. (A. 1) in terms of a polar angle g and an azi- for a face-centered cubic lattice, where
muthal angle q, then (Neel, 1944)
3 ' 1 5l4
f~ = Xi(sin'g —' sin4g —8 sin
—, g cos4q) (A.4) S=-z,
(Pym'+n')"' (P+m'+n')'f'
2
(B 7)
when the polar axis is a cube side; and
For b.c.c., S =0.4; while for f.c.c., S = 0.6. We have
/ E(-; ='3+ 't '3 —3
3 i '3 3g (A 3)
which is body-centered,
Calculated: B3= —0.7X10' ergs/cc;
then for iron,

B2=0.5X10' ergs/cc
when the polar axis is a body diagonal; finally, Observed: B3= —2.9X10' ergs/cc; B2 10.' ergs/cc
—
6 4X—
fx 'E= 3$(1
—, —4 sin'8+4 sin48)+ (6 sin'8 —4 sin48)sinsq and for nickel, which is face-centered,
—3 sin48 sin422$ (A.6)
Calculated: B3 = —0.04X 10" ergs/cc; B2 = 0.03 X 10' ergs/cc
when the polar axis is a face diagonal. The azimuthal angle q is Observed: B3= 6.2 X10" ergs/cc; B2=9.0 X10' ergs/cc.
referred to a cube edge.
The comparison of calculated with observed values demonstrates
APPENDIX B the inadequacy of dipole-dipole interactions.

Magnetic Interaction Energy of Dipoles on APPENDIX C

a Cubic Lattice List of Symbols
We propose to prove now the well-known fact that the magnetic
moment interaction between dipoles on an infinite unstrained B = Magnetic induction = H+4m I.
cubic lattice gives no anisotropy energy. The magnetic interaction
H= Magnetic field intensity.
H y= Effective molecular field.
energy between any two dipoles of the array is
I
= Magnetization = Magnetic moment/volume.
p" p 3(p' r';)(p; r';) = Saturation magnetization.
I,
r rs Io= Saturation magnetization at O'K.
so that the total dipole interaction energy density of a saturated
E= Anisotropy energy density constant.
simple cubic lattice is
J
= Exchange integral.
T, = Curie temperature.
1 3(l~, +m, +~,)2 k=Boltzmann constant=1. 38X10 ergs/degree abso-"
(12+2322+ N2) 3/2 (12+3/32+ N2) 3/2 lute.
rr= (nr, n2, n2) = (n2, n„,32,) = direction cosines of magnet-
where N is the number of dipoles per unit volume, n is the direc- ization vector, referred to the crystal axes.
tion of the magnetization, and r=(z, y, s)=(la, ma, Na) is the
xiii, &ioo= Saturation magnetostriction constants.
vector from the dipole at the origin to the dipole at the lattice
c», c», c44= Elastic moduli of cubic crystal.
point (l, m, n). The first term in the sum does not contain any e;;= Components of strain.
reference to the direction of magnetization and therefore cannot = energy per unit volume= energy density.
f
give anisotropy. The second term may be rewritten, using
nP=1 —n22 —nP,
f„= exchange energy density.
as f3333 anisotropy energy density.

—3 Z' P+n —P)+o.P(n' —P)

'(m' f,„=
= magnetoelastic energy density.
(P+m'+n')@'
(B.3) f magnetic energy density.
f.
3= elastic energy density.
where we have omitted across terms in lm, ln, etc. , since these Bi, B2= magnetoelastic coupling constants.
obviously sum to zero. Now N= number of atoms per unit volume.
u = lattice constant.
l2
= Z' m2
= Z' n2
" (B.4) 0„=Bloch wall energy per unit area.
$~„(P+m+n') ~
$~ {P+m'+n ) 5~'
) (P+m'+n') q
= angle
between directions of adjacent spins.
by symmetry, so that the terms in n2, n3 are identically zero, n, fy= effective number of Bohr magnetons per atom.
A = exchange energy constant.
which leaves only a term independent of the direction of magnet-
ization. There is therefore no anisotropy for a simple cubic /3/3 —
"
Bohr magneton= 0.927 X 10 ergs/oersted.
lattice; the same result applies to body-centered and face- T= stress; absolute temperature.
centered cubic lattices. It may further be seen that =0. f g= linear pole density.
S= spin quantum number.
If, however, we allow the lattice to deform spontaneously the
magnetic dipole interaction will give rise to magnetostriction and x = susceptibility per unit volume I/H.
thus, by an argument which we have given earlier, to an apparent p = permeability = 1+4rx.
and small contribution to the anisotropy. Becker (1930) has p*= effective permeability of a single domain for fields at
shown that the magnetostriction resulting from dipole-dipole right angles to the direction of domain magnetization;
interactions is only about one-fifth of the observed magneto- see Section 2.4.
striction in iron, while in nickel the magnetostriction calculated
on this basis does not agree with the observed value even in BIBLIOGRAPHY
algebraic sign. We, therefore, see that the magnetic dipole inter-
action from saturation magnetization on a regular lattice explains (Different papers by the same author in the same year
are distinguished by letters after the year, as in 1944b.)
neither the observed anisotropy nor the observed magnetostric-
tion. I. Antik and T. Kubyschkina, Uber die Hysteresisverluste in
On Becker's calculation the magnetoelastic coupling constants Ferromagnetika (Amalgamen), Wiss. Ber. Moskauer
Qussigen
(Eq. (2.3.5)) are equal to Staatsuniversitat 2, 143-149 (1934).
Bi = —6SI,'; B2 = 4SI,' (3.5)
I. F. Bates and F. E. Neale, "A quantitative examination of
"
recent ideas on domain structures, Physica 15, 220—224 (1949).
S. J. Barnett, "New researches on magnetization by rotation and
for a body-centered cubic lattice, and
Bi= —3SI ' B2=2SI ' (8.6)
the gyromagnetic ratios of ferromagnetic substances,
Am. Acad. Sci. 75, 109-129 (1944).
Proc. "
 PHYSI CAL THEORY OF FERROMAGNF TI C DOMAINS
R. Seeker, Zur Theoric der Magnetisierungskurve, Zeits. f. Physik E. Kondorsky, On the nature of the coercive force and irreversible
62, 253-269 (1930). changes in magnetization, Physik. Zeit. Sow. 11, 597—620 (1937).
R. Becker, Elastische Spannungen und magnetische Eigen- A. Kussmann and B. Scharnow, Zeits. f. Physik 54, 1—15 (1929}.
schaften, Physik. Zeits. 33, 905-913 (1932}. Uber die Koerzitivkraft. I. Teil. Koerzitivkraft und mechan-
R. Becker and W. Doring, Ferromagnetismus (Verlag. Julius ische Harte.
Springer, Berlin; reprinted J. W. Edwards, Ann Arbor, 1939). L. Landau and E. Lifshitz, On the theory of the dispersion of
F. Bitter, On inhomogeneities in the magnetization of ferro- magnetic permeability in ferromagnetic bodies, Physik. Zeits.
magnetic materials, Phys. Rev. 38, 1903-1905 (1931). Sowjetunion 8, 153-169 {1935).
F. Bloch, Zur Theorize des Ferromagnetismus, Zeits. f. Physik 61, E. Lifshitz, On the magnetic structure of iron, J. Phys. U. S.S.R.
206-219 (1930). 8, 337-346 (1944).
F. Bloch, Zur Theoric des Austauschproblems und der Remanen- L. Marton, Ferromagnetic domain observation, Phys. Rev. 73,
zerscheinung der Ferromagnetika, Zeits. f. 'Physik 74, 295—335 1475 {1948).
(1932). B. Matthias and A. von Hippel, Domain structure and dielectric
R. M. Bozorth, Directional ferromagnetic properties of metals, response of barium titanate single crystals, Phys. Rev. 73,
J. App. Phys. 8, 575—588 (1937). 1378-1384 (1948).
R. M. Bozorth, Magneto-resistance and domain theory of iron- C. Moiler, Zur Theoric der Austauschproblems und des Ferro-
nickel alloys, Phys. Rev. 70, 923 (1946}. magnetismus bei tiefen Temperaturen, Zeits. f. Physik. 82,
R. M. Bozorth, On magnetic remanence, Zeits. f. Physik 124, 559-567 (1933).
519-527 (1948). L. Noel, Quelques propri6tes des parois de domaines 616mentaires
Bozorth, Mason, McSkimin, and Walker, "Elastic constants and Cahiers de Physique 25, 1—20 (1944a}.
internal loss of single nickel crystals, " Phys. Rev. 75, 1954
ferromagnbtiques,
L. Neel, Effet des cavites et des inclusions sur le champ coercitif,
(1949). Cahiers de Physique 25, 21-44 (1944b).
H. Brooks, "Ferromagnetic anisotropy and the itinerant electron L. Neel, Les lois de 1 aimantation et de la subdivision en domaines
model, " Phys. Rev. 58, 909-918 (1940). elbmentaires d'un monocristal de fer, J. de phys. et rad. 5,
Surgy, Hughes, and Wallace, Double transmission and depolar- 241-251, 265-276 (1944c).
ization of neutrons, Phys. Rev. 74, 1207 (1948). L. Neel, Bases d'une nouvelle theoric generale du champ coercitif,
A. Drigo and M. Pizzo, Particolari aspetti della magnetizzazione Annales Univ. Grenoble 22, 299— 343 (1946).
di sottili pellicole ferromagnetiche, Il Nuovo Cimento 5, 1—11 L. Neel, Proprietes d'un ferromagnetique cubique en grains fins,
(1948). Comptes rendus (Paris) 224, 1488—1490 (1947a).
W. C. Elmore, The magnetization of ferromagnetic colloids, L. Neel, Le champ coercitIf d'une poudre ferromagnetique cubique
Phys. Rev. 54, 1092-1095 (1938). a grains anisotropes, Comptes Rendus (Paris) 224, 1550-1551
W. C. Elmore, Theory of the optical and magnetic properties of {1947b).
ferromagnetic suspensions, Phys. Rev. 60, 593—596 (1941). L. Neel, Theoric de 1'anisotropie a aimants traites a chaud dans
M. Fallot, Ferromagndtisme des alliages de fer, Annales' de un champ magnetique, Comptes Rendus (Paris) 225, 109-111
Physique 6, 305— 387 (1936). (1947c}.
A. D. Fokker, Remark on the fundamental relations of thermo- J. A. Osborn, Demagnetizing factors of the general ellipsoid,
magnetics, Physica 6, 791-796 {1939). Phys. Rev. 67, 351— 357 (1945).
J. Frenkel and J. Dorfman, Spontaneous and induced magnetiza- W. Shockley, Energy calculations fox domains, Phys. Rev. 73,
tion in ferromagnetic bodies, Nature 126, 274—275 {1930). 1246 (1948).
L. H. Germer, Stray magnetic fields from cobalt, Phys. Rev. 62, K. J. Sixtus and L. Tonks, Propagation of large Barkhausen
295 (1942). discontinuities, IV, Regions of reversed magnetization, Phys.
E. A. Guggenheim, On magnetic and electrostatic energy, Proc. Rev. 43, 931—940 (1933).
Roy. Soc. A155, 49-70 {1936). A. Sommerfeld and H. Bethe, Elektronentheorie der Metalle,
C. Guillaud, Ferromagnetisme des alliages binaires de manganese,
Thesis, Strasbourg, 1943. Handbuch der Physik 24.2, 333— 622 (1933).
C. Herring and C. Kittel, Energy of a Bloch wall on the band E. C. Storier and E. P. %ohlfarth, Interpretation of high co-
picture, Phys. Rev. (to be submitted). ercivity in ferromagnetic materials, Nature 160, 650 (1947).
H. Honda and S. Kaya, Magnetization of single crystals of iron, E. C. Stoner and E. P. Wohlfarth, A mechanism of magnetic
Sci. Rep. Tohoku Univ. 1S, 721—753 (1926). hysteresis in heterogeneous alloys, Phil. Trans. A240, 599— 644
S. Kaya, On the magnetization of single crystals of nickel, Sci. (1948).
Rep. Tohoku Univ. 17, 639—663 (1928a). J. H. Van Vleck, A survey of the theory of ferromagnetism, Rev.
S. Kaya, On the magnetization of single crystals of cobalt, Sci. Mod. Phys. 17, 27-47 {1945).
Rep. Tohoku Univ. 17, 1157—1177 {1928b). J. H. Van Vleck, Quelques aspects de la theoric du magnetisme,
M. Kersten, I'robleme der techmschen Magnetisierungskurve, Zur Annales de 1'Institut Henri Poincare 10, 57—190 (1947).
Deutung der Koerzitivkraft, edited by R. Becker (Verlag. S. V. Vonsovsky and J. S. Shur, Ferromagnetism (in Russian),
Julius Springer, Berlin; reprinted J. W. Edwards, Ann Arbor, (OGEZ, Moscow, 1948).
1938), pp. 42-72. L. Weil, Variation du champ coercitif en fonction de la density
M. Kersten, Grundlagen earner Theoric der ferromagnetischen de poudres ferromagnetiques agglomerees, Comptes Rendus
Hystorese siad dor Koorsv'tvvfsruft (S. Hirzel, Leipzig; reprinted (Paris) 225, 229-230 (1947).
J. W. Edwards, Ann Arbor, 1943). L. Weil and S. Marfoure, Variation thermique du champ coercitif
L. Kholodenko, On the domain structure of ferromagnetic bodies du nickel agglomere, J. de phys. et rad. 8, 358-361 (1947).
in the presence of a magnetic field (in Russian), J. Exp. Theo.
Phys. USSR 17, 698—707 (1947).
P. %eiss, L'hypothese du champ molbculaire et la propribt5
ferromagnetique, J. de Phys. 6, 661— 690 (1907).
R. Kimura and K. Ohno, On the elastic constants of single
P. R. %eiss, The application of the Bethe-Peierls method to
crystals of iron, Sci. Rep. Tohoku Univ. 23, 359-364 (1934).
C. Kittel, Theory of the structure of ferromagnetic domains in ferromagnetism, Phys. Rev. 74, 1493-1504 (1948).
films and small particles, Phys. Rev. 70, 965—97l. (1946). H. J. Williams, Direction of domain magnetization in powder
C. Kittel, Domain theory and the dependence of the coercive patterns, Phys. Rev. 71, 646 (1947).
force of Qne ferromagnetic powders on particle size, Phys. Rev. Williams, Bozorth, and Shockley, Magnetic domain patterns on
73, 810 (1948). single crystals of silicon iron, Phys. Rev. 75, 155-178 (1949).
C. Kittel, Gyromagnetic ratios and splitting factors of ferro- H. J. Williams and W. Shockley, A simple domain structure in
magnetic substances, Phys. Rev. 76, 743 (1949a). an iron crystal show'ing a direct correlation with the magnetiza-
C. Kittel, Theory of the formation of powder patterns on ferro- tion, Phys. Rev. 75, 178-183 (1949).
magnetic crystals, Phys. Rev. 76, 1527 (1949b). W. A. Yager, Phys. Rev. (to be submitted).