Properties of Magnetic Materials with a

Nonuniform Saturation Magnetization.

I. General Theory and Calculation of the
Static Magnetization
Cite as: Journal of Applied Physics 38, 5027 (1967); https://doi.org/10.1063/1.1709271
Submitted: 20 April 1966 • Published Online: 29 June 2004

Ernst Schlömann

ARTICLES YOU MAY BE INTERESTED IN

New Determinations of the Saturation Magnetization of Nickel and Iron

Journal of Applied Physics 39, 669 (1968); https://doi.org/10.1063/1.2163571

Past, present, and future of soft magnetic composites

Applied Physics Reviews 5, 031301 (2018); https://doi.org/10.1063/1.5027045

A review of high magnetic moment thin films for microscale and nanotechnology
applications
Applied Physics Reviews 3, 011301 (2016); https://doi.org/10.1063/1.4941311

Journal of Applied Physics 38, 5027 (1967); https://doi.org/10.1063/1.1709271 38, 5027

© 1967 The American Institute of Physics.

 JOURNAL OF APPLIED PHYSICS VOLUME 38, NUMBER 13 DECEMBER 1967

Properties of Magnetic Materials with a N onuniform ~aturation .Ma~n~tization. I. General

Theory and Calculation of the Static MagnetizatIOn
ERNST SCHLOMANN
Raytheon Research Division, Waltham, Massachusetts
(Received 20 April 1966; in final form 6 July 1967)

A general theory concerning the properties of magnetic ~aterials h~vin~ a nonuniform saturation
magne t·Izat'IOn IS
. de loped
v In ethl's theory
. , the local saturatIOn magnetizatIOn . and the h local exchange
d h
interaction are assumed to vary in an arbitrary. manner througho~t the sample. It}S furt. er assume t ~t a
dc magnetic field is applied which is sufficiently large to magnetIz~ thesample ~ubstantIa~ly to saturatIOn.
The general theo;y is then specialized for a calculation of the static magnetlza~lOn? assummg.that ~he host
material contains inclusions of a material with a different satur~~ion magnetizatIOn. If. all mclus~ons are
. Iy Iarge, the effect of the exchange interaction becomes.neghglble
fair . h and a result
. .prevIOusly
f . denved
t f by
Nee! is obtained. The inclusion of exchange reduces the deViation of t. e magnehzatI~n rOI? I~S sa ura IOn
I 'cularly at low values'· of the internal field. If exchange IS neglected, this deViatIOn becomes
va ue, partl .• . h d . . . . fi't Th
infinitely large at zero internal fie!d. If exchange is mc~uded, how~ver, t e eVlatlOn remams ill e. e
theory is, therefore (for suitable materials), still approximately vahd at remanence.

I. INTRODUCTION "stray field") prevents complete alignment of the local

magnetization vector with the applied magnetic field in
It has long been recognized that local variations of the vicinity of the inclusion. Hence, the component of
the saturation magnetization can have a profound the magnetization vector parallel to the applied field is
effect upon the properties of magnetic materials. ~ee~l.2 reduced by the presence of the inclusion, the amount of
has investigated the influence of nonmagnetIC lU- reduction depending upon the strength of the applied
clusions upon the manner in which the magnetization magnetic field.
approaches saturation at field strengths much larger It is also evident from Fig. 1 that in a material with
3
than the coercive force. Goodenough has discussed nonuniform saturation magnetization a "resonance"
the effect of magnetic inhomogeneities upon the square- can be excited with the rf magnetic field applied parallel
ness of the B-H loop. The influence of porosity upon to the dc field. This follows from the residual mis-
the width of the ferromagnetic resonance line has been alignment between the local magnetization vector and
discussed by several authors.4-s In practical applica- the dc magnetic field, which is present even at very
tions of ferrites at microwave frequencies, porosity high field strength. The "longitudinal" resonance (rf
broadening is frequently the dominant line broadening and dc fields are parallel to each other) differs in
mechanism. A detailed theory of this effect has recently some important details from the conventional, "trans-
lo
been proposed by Sparks9 and by Seiden and Sparks. verse" resonance (rf and dc fields are perpendicular
A qualitative, intuitive understanding of the mag- to each other). The longitudinal resonance here con-
netic properties of ferromagnetic materials with a non- sidered is a linear (low-power) phenomenon and should
uniform saturation magnetization may be obtained by not be confused with the nonlinear response observed
considering Fig. 1, which shows an inclusion in an with the dc and rf fields aligned parallel ("parallel
otherwise uniform magnetic matrix. If the saturation pumping") .
magnetization of the inclusion differs from that of the Inhomogeneity of the saturation magnetization also
matrix, a divergence of the magnetization vector occurs has a strong effect upon the transverse resonance.
at the surface of the inclusion. The magnetic field From Fig. 1 it is clear that the magnetic field is strongly
generated by this divergence of magnetization (the inhomogeneous in the vicinity of the inclusion. Thus
* Supported by the U.S. Air Force under Contr. No. the magnetization vector, if disturbed from its equilib-
AF33 (615)-1423. rium position, will tend to precess at different rates
1 L Nee!, Compt. Rend. 220, 738 (1945). in different parts of the sample. This implies that the
2 L Neel, J. Phys. Radium 9, 184 (1948).
3 J. B. Goodenough, J. App\. Phys. 36,2342 (1964). • width of the ferromagnetic resonance line is broadened.
4 E. Sch16mann, Proc. Conf. on Magn. and Mag. Matenals, In the absence of inhomogeneity and anisotropy effects
AlEE Spec. Pub!. T-91 (1956) p. 600. . . d I b h " .. "
6 L G. Van Uitert, F. W. Swanekamp, and S. E. Haszko, J. the resonance IS eXCIte on y y t e pOSItive sense
App!. Phys. 30, 363 (1959). of circular polarization, because the free precession of
6 P. E. Seiden, C. F. Kooi, and J. M. Katz, J. App!. Phys. 31, the magnetization vector follows a circular path. In
1291 (1960).
7 P. E. Seiden and J. G. Grunberg, J. App!. Phys. 34, 1696 the presence of inhomogeneity this is no longer true.
(1960). Thus the transverse resonance can, to a small extent,
8 M. Sparks, R. London, and C. Kittel, Phys. Rev. 122, 791
(1961). . also be excited by the negative sense of circular po-
g M. Sparks, Ferromagnetic Relaxation Theory (McGraw-Hill larization.
Book Co., Inc., New York, 1964); M. Sparks, J. App!. Phys. 36, In the present sequence of papers a consistent theory
15?o0p~1§~fJ~n and M. Sparks, Phys. Rev. 137, A1278 (1965). of the various effects attributable to nonuniformity of
5027
 5028 ERNST SCHLOMANN

Stray fIeld produced Direc ti on of foeof

by inclusion Here 'Y is the gyromagnetic ratio, h Planck's constant,
Mo the saturation magnetization, and the operators a
and a+ obey the commutation relation
a(r)a+(r') -at(r') aCr) =o(r-r'), (2)
where oCr) is the delta function.
Applied
magnetic
Equations (1) and (2) are derived from the basic
field commutation relations of the angular momentum
vector. We note that they remain valid when the
saturation magnetization Mo and/or the gyromagnetic
ratio 'Yare functions of position. In the following, we
assume 'Y to be constant, but Mo to be variable. For
convenience we define a function if;(r) such that
(3)
Nonmagnetic where (Mo) is the average saturation magnetization.
inclusion
t , Consider now a small, rectangular test volume within
FIG. 1. Stray field and direction of local magnetization in the the sample. Let Mo(r) be the local saturation mag-
vicinity of a nonmagnetic inclusion. netization within this small test volume. If the test
volume is representative of the sample as a whole, it
the saturation magnetization is developed, and com- is permissible to assume that outside the test volume
pared to experimental data where applicable. Local all local properties are periodically repeated and can
(crystalline and/or strain-induced) magnetic anisotropy hence be expanded in a Fourier series.
can produce effects quite similar to those produced by The Fourier transform a(k) of the operators aCr)
the nonuniformity of the saturation magnetization. The are defined by
local magnetic anisotropy is neglected in the present
theory, but the formulation is such that it can readily aCr) = V-1/2 L a(k) exp(ik·r),
Ie
be included. For convenience, a quantum-mechanical
formulation in terms of creation and annihilation
operators has been chosen, even though the problem f
a(k) = V- 1/2 d3ra(r) exp( -ik·r), (4)
is basically a classical one.
In the first paper of this series, a general theory and obey the commutation relation
concerning the properties of magnetic materials having a(k)a+(k') -a+(k')a(k) =ow, (5)
a nonuniform saturation magnetization is developed.
Quantitative predictions of the various effects induced where Ow is the Kronecker symbol and V is the
by nonuniformity of the saturation magnetization can periodicity volume.
be derived from this general theory. The present paper We similarly define Fourier transforms of the mag-
also contains detailed theoretical results concerning the netization vector and of the function if;(r)
static magnetization. A brief comparison of experi- M(r)=L:M(k) exp(ik·r),
mental data with the present theory of the high field
magnetization curve has previously been given. I1 A M(k) = V-If d3rM(r) exp( -ik·r),
more detailed comparison and an extension of the
theory to the high-frequency permeability (both longi- if;(r) = L if;(k) exp(ik· r),
tudinal and transverse) will be described in later
publications. if;(k) = V-Ifd rif;(r) exp( -ik· r).
3
(6)
II.· GENERAL THEORY
Note, that in defining the Fourier transforms M(k)
It is convenient to formulate the problem in terms and if;(k) a slightly different convention has been used
of creation and annihilation operators12 a+(r) and a(r) than in Eq. (4) for defining the Fourier transforms
which are related to the components of the local a(k). Because of its definition (3) the Fourier trans-
magnetization M(r) by the relations form if;(k) obeys the relation
M+=Mx+iMy= (2'YhMo)112a+[1- C'Yh/2Mo)a+aJl/2, (7)
M_=Mx-iMlI = (2'YhMo)1/2[1- ('Yh/2Mo)a+aJ1I2a,
The theory developed iu this paper is intended to be
M.=Mo-'Yha+a. (1) applicable as long as the magnetization vector is ap-
11 E. Schlomannand T. Kohane, J. Appl. Phys.38, 1118 (1967). proximately aligned with the direction of an applied
l2 T. Holstein and H. Primakoff, Phys. Rev. 58, 1098 (1940). de magnetic field (the z axis). It is, therefore, reasonabJe
 MATERIALS WITH NONUNIFORM MAGNETIZATION. I 5029

to approximate the square roots in Eq. (1) by unity. The dipolar energy is, therefore,
The Fourier transforms of the magnetization are thus JCdip=27rV L k-2[k·M( -k) ][k·M(k)]. (14)
given by k

M+(k) = (2'YIi(M o)/V) 1/2L: 1f(k-k')a+( -k'), We now use Eq. (8) to express the energy in terms of
k' the operators a(k) and a+(k). It is apparent that
M_(k) = (2'YIi(Mo)/V) 1/2 L: 1fCk-k')aCk'), terms of zero, first and second order in these operators
k' occur, in addition to higher order terms which may be
Mz(k)=Mo(k)-C'/'h/V) La+(-k+k')a(k'), (8) neglected in the present context. Detailed expressions
k' for the first- and second-order terms are given below.
where Mo(k) is the kth Fourier component of the local The exchange energy in an inhomogeneous (but iso-
saturation magnetization Mo(r). tropic) material can generally be expressed as l3
In the following the spatial average of M.(r) is of
particular interest. According to the third line of JCex = f d3rC(r) [(VMx)2+ CVMy) 2+ (VMz)2], (15)
Eq. (8),
where C(r) is a local exchange constant, which is
(M z )= V-If d rMzCr)
3 proportional to the local exchange integral. Expressed
in terms of the operators a and a+ the exchange energy
=M.(k=O) contains terms of zeto and second order, in addition
to the higher order terms, which are here neglected.
= (Mo)- ('YIi/V) L a+(k)a(k). (9) The zero-order term of the Hamiltonian is of no
k
interest in the present context. The relevant terms of
Consider now the various contributions to the energy. the total Hamiltonian (sum of Zeeman, dipolar, and
If the applied dc magnetic field is Ho, the Zeeman exchange energy) can be expressed as
energy is
JC= L[fka+(k) +H.c.]
k
JCZ eeman = - !M.Hod3r= - V (M.)Ho
+ L{Akk,a+(k)a(k') +tCBkk'a+(k)a+( -k') +H.c.]l
kk'
=[ - V(Mo)+'Y1i L a+(k)a(k)]Ho. (10)
(16)
The dipolar energy is where H.c. represents the Hermitian conjugate of the
~ fM.HdiPd3r,
expression preceding it. The first-order contribution
JCdip= - (11) to the Hamiltonian arises from the dipolar energy. The
coefficientjk are according to Eqs. (14) and (8)
where the dipolar field H dip is the field generated by
the magnetization. In the quasistatic approximation jk= 27r (2')'1i (MoW) 1/2 L 1f(k- k') MoCk') k/k_' /k'2,
k'
Hdip(r) satisfies
(17)
V XHdip=O, where
V·Hdip = -47rv·M. (12) (18)

Thus, its Fourier transform Hdip(k) is given by The second-order term of the Hamiltonian contains
contributions from all three types of energy here con-
H diP (k)=-(47r/k2)k[k·M(k)]. (13) sidered. The coefficients A kk , and Bkk' are given by

Akk'/ 'Yh=HoOkk'+27r (Mo)

"" 1fCk-k2)1f(k2-k), (k2+k2-
1;' ~
Ckz-k.'F)
-2 (k-k')2

+2(Mo) L 1f(k-k2)CCk2-k3)1f(k3 -k') Ck2-k+k')· (k 3 +k-k'), (19)

k2k a

Bkk'/'Y1i=27r(Mo) L 1f(k-k2)1f(k2-k')k2-2/k22. (20)

k2

These expressions contain indeterminate contributions (zero over zero) which arise from the uniform com-
ponent (zeroth Fourier component) of the dipolar field. Since this field is identical with the macroscopic de-
magnetizing field, we can combine it with the applied field and define the sum as the internal magnetic field. This
applies both to the de and the rf fields. In the following, we adopt this convention. The internal (demagnetized)
dc magnetic field is denoted by H, the externally applied magnetic field by Ho. We similarly distinguish between
the internal rf field h and the externally applied rf field h o. If all formulas are expressed in terms of the internal
fields, the indeterminate contributions to Akk' and B kk , must be set equal to zero for consistency.
13 See, e.g., C. Herring and C. Kittel, Phys. Rev. 81,.869 (1951).
5030 ERNST SCHLOMAC-;I'\

The diagonal components of the matrices Akk' and B kk , are of particular interest. They are given by
Ad'Yn==Akkhn=H+27r(Mo) L 11f(k-kz) 121 k2+ 12/kl+2(Mo) L1f(k-kz)C(k2-k3)1f(ka-k)k2'ka, (21)
k2 k2k,

B k hh==Bkkhh=27r(M o) L 11f(k-k
k2
2) 12 kZ_2/k22. (22)

For convenience, we introduce two significant approximations at this stage. The first of these is to assume the
exchange interaction C(r) to be uniform. With this approximation, the double sums in Eqs. (19) and (21) reduce
to single sums. The result of the calculation shows that the exchange energy is of secondary importance in the
present context. Thus, a more elaborate consideration of nonuniform exchange appears unwarranted.
We further assume that the inhomogeneity of the local saturation magnetization Mo(r) is "weak". This implies
that
(23)
k";O

Consistent with these approximations, Eqs. (17), (19), and (22) become
jk= 27r(2'YhMo V) J 12Mo(k) kzL/k2, (17a)
Akk'hh=+27r(Mo)1f(k-k') II k+ \z/k2 +1 k/ \2/k'2_4[(k z -k/)z/(k-k')2]1
+1¥(k-k') (4k· k'-kZ -k'2) krfk', (19a)
Bkk'hh= 27r(Mo)1f( k- k') (k_2/k2+k_'2/k'2) krfk', (20a)
Akhh=H+27r(Mo)(1 k+ 12/k ) +Dk2, 2 (21a)
Bkhh=27r(Mo)L2/k2, (22a)

where local magnetization vector and the dc field under

D=2(Mo)C(0) (24) equilibrium conditions. According to Eqs. (26) and
(16) the static excitation is determined by
is an average exchange stiffness constant. For k-->k',
Eqs. (19a) and (20a) do not reduce to Eqs. (2la) L[Akk'ao(k') +Bwao*( -k')J=A. (27)
and (22a) , respectively. This arises from the fact k'
that in sums of the type The off-diagonal elements of the matrices Akk' and
Bkk' can be neglected by way of approximation, be-
L o/(k-kz)o/(k2- k')j(k2), cause they represent nondiagonal elements of a small
k2
perturbation. Thus,
two large terms occur when krfk' (those in which k2
equals either k or k'). Only one large term occurs for Akao(k)+Bkao*( -k) =k (28)
k=k'.
By combining this equation with its "adjoint", the
It should be noticed that the off-diagonal elements
complex conjugate equation with k replaced by -k,
of the matrices A kk , and Bkk' are considerably smaller
one obtains
than the diagonal elements, provided that the in-
homogeneity is "weak" in the sense of Eq. (23). Note (Ak2 -1 Bk \2)ao(k) =jkAk-j-k*Bk. (29)
also that Akk' is Hermitian (Akk' = Ak'k*) and that Bkk'
This can be simplified by noting that according to
satisfies
Eqs. (17a) and (22a),
(25)
(30)
III. STATIC MAGNETIZATION Thus,
The equations of motion for the variables a(k) are (31)
generally
The average z component of the static magnetization
a(k) = - (i/h) [a(k), X]
is, therefore, according to Eqs. (9) and (17a)
=-(i/h)aX/aa+(k). (26) . ( I Mo(k)
12 cos 2(h sin21h )
This equation of motion can also be derived on a purely (M z )= (Mo) 1-8r {; (H+Dk2+47r(Mo) sin20k)2 •
classical basis. In this case, the variables a(k) are c
(32)
numbers rather than operators. In the following we
adopt this interpretation. The Hermitian adjoint opera- Here, Ok is the angle between the wavevector k and
tor a+ then becomes the complex conjugate variable a*. the z axis. The term k=O must be excluded from the
Because the Hamiltonian contains terms which are summation becausejo=O.
linear in a(k) a "static excitation" ao(k) occurs, which If the saturation magnetization is isottopically dis-
describes the lack of complete alignment between the tributed, as we shall assume, the coefficients MoCk)
 MATERIALS WITH NONUNIFORM'MAGNETIZATION, I 5031

actually depend only upon the magnitude of knot where the function <I>(k) depends only upon the magni-
upon its direction. The summation in Eq. (32) can, tude, not upon the direction of k, and is normalized
therefore, be carried out by first taking an average such that
over the angular factors and then summing the terms
dependent upon 1 k 1= k.
The problem simplifies considerably if the local
saturation magnetization varies relatively "slowly" From Eqs. (32), (33), and (39), we now obtain
with position. This is the case treated by Neel,1·2
In this case Mo(k) is significantly large only for small k. (Mz)=(Mo)[1-tuF(u)], (41)
In the present context, the variation of Mo(r) is where
called "slow", if the exchange contribution to Ak is
negligible (i.e., Dk2«H) for all k with a significantly F(u) = (V/27r2) ["k 2dk<I>(k)f(U+Dk 2/47r(Mo»)
o
large Mo(k). We show later that this applies for
instance in the case of spherical inclusions, if the
radius of the inclusions is sufficiently large. (42)
For convenience, we introduce the abbreviation
Here,f(u) is given by Eq. (38),
u= 2: 1Mo(k) 12/Mo(0) 2. (33)
kr'O ~=Dk2/47r(Mo), (43)
Here the sum is equal to the mean square deviation and
of the local saturation magnetization from its average wW = (V~1/2/ 47r2) (47r(Mo)/ D) 3/2<I>[ (~47r(Mo)/ D)1/2]
value
(44)
2: 1Mo(k) 12= 2: 1Mo(k) 12_Mo(0)2
kr'O all k is a weight function which is normalized to unity:
= (M02)- (MO)2
(45)
= «~MO)2), . (34)
where the angular brackets denote a spatial average. In the Appendix, the Fourier spectrum <I>(k) and the
The parameter associated weight function w(~) are calculated in detail
(35) for the case of spherical inclusions. If all inclusions are
assumed to have the same saturation magnetization
is a measure of the degree of inhomogeneity. and the same radius Ti, the result is
We also introduce a reduced, dimensionless variable
u proportional to the internal field wW =tt"lJ3/2 2 [(Uc)1/2], (46)
where
u=H/47r(Mo). (36) C = D/T147r(Mo) ( 47)
With this convention from Eq. (32), and J3/2 is the Bessel function of order t.
In Fig. 2, the function cw(~) is plotted vs Uc. For
(Mz)= (Mo)[1-tuf(u)], (37)
Uc»1, the weight function behaves as
where
2 2 w(~)"-'(3hr)cl/2t"3/2 cos2[(Uc)1/2]. (48)
fe u) -/'oldx (u+1-X
x (1-x
2)2
)
If the quasiperiodic variation is disregarded (i.e., the
cos2 term replaced by t) the weight function can for
3 3u+2 . ~»c be approximated by
=- -
2
+ 2(u+1)1/2 smh- (U- 1/2 )
l (38)
wW = (3/27r)Cl/2~-3/2. (49)
and x=cosO. This agrees with Neel's result,1·2
Consider now the effect of the exchange interaction This function is shown in Fig. 2 as a broken line.
upon the static magnetization. It is obvious that this For spherical inclusions of equal radius the weight
interaction tends to keep the local magnetization vectors function w(O depends only upon a single parameter
in adjacent regions of the sample more nearly aligned. (namely c). We shall denote the function F(u) cal-
In order to carry out the calculation the Fourier spec- culated using this particular weight function as F(u, c).
trum 1Mo(k) 12 of the local saturation magnetization In Fig. 3, the function F(u, c) is plotted vs u with
must be determined. As long as the magnetization is c as parameter. The curves were obtained by nu-
isotropically distributed the Fourier spectrum can, in merically integrating Eq. (42) using the actual weight
general, be expressed as function wW defined by Eq. (46) for O<Uc< 100
and the smoothed out, approximate function w(~) de-
fined by Eq. (49) for 100 < Uc< 00. It should be noted
5032 ERNST SCHL()MANN

appraximatian,
o.
(M.)= (Mo)(l-a/H) , (SO)
where in the limit c---c>O
ewl(lo t i'[l t Iyt 1]2
yo (Ie 1
J
(51)
The caefficient a is frequently referred to as the mag-
004 netic hardness caefficient.
0.03
IV. DISCUSSION
002

001 In this paper, a general Hamiltanian has been derived

which describes the properties af magnetic materials
10 20 30 40 50 - 60 70 80 90 KlO 110 120
y"(lc having a nonuniform saturatian magnetization (Sec. II).
The general theary was applied to. a calculatian of the
FIG. 2. Normalized weight function w(~) [Eq. (46)] applicable static magnetization at fairly high dc fieldstrength
in the <;ase of spherical inclusions of equal radius. The broken line
is explained in the text. (Sec. III). The result agrees in an impartant limiting
case (slawly varying saturatian magnetizatian) with
that previausly derived by Neel,1,2 Neel's formulation
that f( u) becames lagarithmically infinite as u---c>O. Far af the problem is different but entirely equivalent;
C~O, hawever, F(u, c) remains finite as u---c>O. This substantially the same approximatians must be adopted
implies that with the inclusian af exchange into. the
calculation, the theory remains approximately valid 3.0,.----.--,----,-----.----,
dawn to. zero internal magnetic field strengths, pravided
that of(O, c)«1. \
In Fig. 4 the function F(O, c) is platted vs c. It can
readily be shawn that for c«l this functian behaves 2.0.
as ! log (co/c) , which is shawn in Fig. 4 as a braken
line. Co has been determined graphically with the
F(O,c)
result c~2.6Xl(F-2.
In Fig. 5 the functian F(u, c) is platted vs 1/u 1.0
with c as parameter. It has been previausly painted
aut by NeeF,2 that (far c=O) feu) =F(u, 0) varies
co=2.6x 10.-2
approximately linearly with 1/u far!< 1/u< 10. Figure
OL--~IO--~4-~IO--~3-JIO--2~~~~-I-=~I.0
4 shaws that this is also. true for C~O. Thus, to. this c

FIG. 4. Deviation of remanent magnetization from saturation

value (in reduced units) vs c=D/r;247r(Mo).

in arder to derive explicit results. The present theary

goes beyond the Neel theoryl,2 insofar as it is not
2.
restricted to the limiting case of a slowly varying
saturatian magnetization. It also indicates clearly the
relationship of the various effects induced by the non-
1.5 uniformity of the saturation magnetization, and can
serve as a basis for detailed calculation of those effects.
It shauld be remembered that the theory of the high
field magnetization curve described in this paper (in
1.0. Sec. III) does not take the influence of local crystalline
and/or strain-induced anisotropy into account. The
lacal anisatropy induces a deviatian af the magnetiza-
tian from the saturation value, which may be cam-
parable in magnitude to. that induced by the presence
af inclusians and/or pares. 14- I6 The anisotrapy-induced
contributian has a different field dependence than the
0.0 01 0.2 0.3 0.4 0.5 0.6 0.7 o.s 0.9 1.0 inclusian-induced contribution, varying approximately
It R. Becker and W. Doring, Ferromagnetismus (Springer-
FIG. 3. Deviation of magnetization from saturation value (in Verlag, Berlin, 1939), pp. 167-176.
reduced units) vs internal magnetic field (in reduced units) for u; T. Holstein and H. Primakoff, Phys. Rev. 59, 388 (1941).
various values of c= D/r;'4r (Mo). (r;=radius of the inclusions.) 18 L. Neel,]. Phys. Radium 9, 193 (1948).
 MAT E R I A L S WIT H NON U N I FOR ~M MAG NET I Z A T ION. I 5033

as 1/H2. The two contributions can, therefore, be

distinguished experimentally without major difficulty.
The theory of the magnetization curve described
in the present paper repeatedly uses the assumption
that the misalignment of the local magnetization vector
is small. The theory can, therefore, be expected to be
valid as long as the magnetic field is reasonably large.
As the magnetic field is reduced, the approximations
invoked in the development of the theory becomes less
and less valid. It is difficult to estimate on purely
theoretical grounds at what fieldstrength the theory
ceases to be applicable. For complete alignment, the
strength of the demagnetizing field generated at the
surface of a nonmagnetic inclusion is of the order of
41r(Mo) in the immediate vicinity of the surface. One
may, therefore, be tempted to conclude that the validity
of the theory is restricted to internal magnetic field FIG. 5. Deviation of magnetization from saturation value (in
reduced units) vs inverse of internal magnetic field (in reduced
strengths that are large compared to 41r(Mo). This units) for various values of c= D/r;241r (Mo).
argument is misleading, however, because any local
misalignment produces a volume divergence of the coming being the neglect of local anisotropy fields.
magnetization vector and hence an additional magnetic One should, therefore, not expect that all magnetic
field, which largely c ncels the demagnetizing field materials will behave according to this model. There
arising from the surface of the inclusion. In the present appears to be a sizeable class of important magnetic
theory the magnetic field generated by the volume materials, however, which are adequately described
divergence of the magnetization is rigorously taken by this relatively simple model. If the magnetization
into account, subject only to the condition that the curve could be calculated- without introduction of any
resulting misalignment be small. The misalignment also further approximation, it would show all the qualita-
reduces the strength of the magnetic field arising from tive features observed in actual materials (i.e., hys-
the surface of the inclusion, because the projection of teresis, coercive field, etc.). By suitably extending the
the local magnetization vector upon the surface normal theory described in the present paper, it may eventually
is reduced. Furthermore, even though the misalign- be possible to calculate the complete magnetization
ment of the local magnetization vector may be fairly curve.
large in the immediate vicinity of the inclusion, it
rapidly decreases with increasing distance from the
inclusion. As long as the volume, over which the small- ACKNOWLEDGMENTS
misalignment assumption is violated, is reasonably The author is extremely grateful to Dr. T. Kohane
small the results of the theory should still be sub- for discussions concerning his experimental results, to
stantially correct. Dr. R. 1. Joseph for a critical review of the manuscript
The magnetization curves measured on porous poly- and to Mrs. C. King and H. Graichen for their help
crystalline YIG agree quite well with the theory with the numerical calculations.
described in the present paper not only at high field-
strengths but down to the remanence pointY It is,
APPENDIX: FOURIER TRANSFORM OF THE
therefore, useful to see how the remanent magnetiza-
LOCAL SATURATION MAGNETIZATION IN A
tion Mr as calculated from the present theory depends
MATERIAL CONTAINING SPHERICAL
upon the various parameters of our model. According
INCLUSIONS
to Eqs. (41), (42), and Fig. 4 the remanent magnetiza-
tion should be given by Assume that spherical inclusions n (n= 1, 2, 3, .0.)
Mr= (Mo)[l-iO' log(co/c)] of radii an and containing a material with a saturation
magnetization Mn are located at the positions R"
= (Mo)[1-!0' log(r;/rio)], (52) within the periodicity volume. Thus, the local satura-
where it has been assumed that c«co (i.e., r;»r,'O). tion magnetization Mo(r) equals the saturation mag-
The parameter c is related to the radius of the inclusions netization of the host material M. unless I r-R" I <an.
ri by Eq. (47). According to Fig. 4 c~2.6XIo-2, hence, The Fourier transform of Mo(r) is
rio=6.2 (D/41r(M o»)1/2. For YIG ri~lO-5 cm.
The physical model upon which the present calcula-
tion is based is that of a magnetic continuum with a
Mo(k) = V-l! d3rMo(r) exp( -ikor)

locally variable saturation magnetization. In some re- =Mso(k) - L(M.-M")I,,, (Al)

spects the model is oversimplified, the major short- n
S034 ERNST SCHLOMA:-.rN

where In is l/V times the integral of exp( -ik.r) small the transition from Eg. (A7) to Eg. (AS) in-
taken over the nth inclusion. volves a finite error, because the inclusions cannot inde-
pendently assume all possible locations in the periodicity
f
In= (l/V) cPr exp( -ik·r) volume.
In order to estimate the magnitude of this error,
consider the special case in which all Mn are equal to Mi.
According to Eq. (34) L.kfo II Mo(k) 12 must generally
equal «IJ.MO)2)=(Mo2)_(Mo)2. It will be shown that
this is only approximately true for Eg. (AS). The dis-
x i:1d (cosO) exp( -ikp cosO) crepancy is a measure of the error involved in the
transition from Eq. (A7) to (AS).
=exp( -ik·Rn) (vn/V) g(kan). (A2) For the case under consideration
Here (Mo2)=M/(1-p) +Mlp (A9)
(A3) and hence
is the volume of the nth inclusion, and (A1O)
g(y) =3y-3 (siny-y cosy) Using well-known integral relations between Bessel
functions,17 the integration over the k dependent factors
= 3 (!1l') J 12y- 3/2J 3/ 2(y) , (A4) in Eg. (AS) can be carried out
where J a/2 is the Bessel function of order l
It should be noticed that g(O) = 1 and hence, assum-
ing that all Mn are equal to M i ( = saturation magnetiza-
tion of the inclusion), Thus, from Eqs. (AS), (A6) , and (All)
Mo(O) =M.(l-p)+MiP, (AS) L.I Mo(k) IAV2 = (M.-MYp. (AI2)
k
where
p= L. Vn/V (A6) Comparison of Eqs. (AlO) and (A12) shows that for
n
p«l the two results agree very closely, so that Eg.
is the ratio of the total volume of the inclusions to the (AS) appears well justified.
sample volume. We now define a normalized spectral function <I>(k)
Consider now I Mo(k) 12 for k~O. According to according to Eqs. (39) and (AS).
Eqs. (AI) and (A2)
<I>(k) = V-I L. vr,zg2(kan )/'E Vn • (AB)
I Mork) 12 = V-2 L. (M.-Mn) (M.-M n,) n n
nn' It can readily be shown that <I>(k) satisfies the normal-
X exp[ik· (Rn-Rn,)]vnvn'g(kan)g(kan')' (A7) ization condition (40) exactly. Thus, the slight error
made in the transition from Eq. (A7) to Eq. (A8) is
Since the location of the inclusions is not actually
largely compensated if we use this convention.
known, it is reasonable to take an average of I MoCk) 12
If all inclusions are equal in size (in addition to being
over all possible locations of the inclusions and to
spherical), the spectral function becomes
assume that this average value is representative of the
actual sample under consideration. If the total volume <I>(k) = (41l'rN3V)g2(kri)
of the inclusions is small each inclusion can be located
with equal probability at any point within the perio- = (61l'2/Vk3)[Ja/2(kri)]2, (A14)
dicity volume. Under these conditions the average of
where r i is the radius of each inclusion. The corre-
exp[ik· (Rn-R n,)] vanishes unless n=n'. Equation
sponding weight function wW as defined by Eg. (44)
(A7) thus becomes is given by Eg. (46).
I MoCk) IA,2= V-2 L.(M.-Mn)2Vn2g2 (ka n). (AS)
n
17 See, e.g., Handbook of Mathematical Functions, U.S. Dept.
If the volume of the inclusions is not infinitesimally of Commerce, Nat!. Bur. Std. Appl. Math. Ser. 55, Eq. (11.4.6)