3398 IEEE TRANSACTIONS ON MAGNETICS, VOL. 38, NO.

5, SEPTEMBER 2002

Classics in Magnetics

Summary of Losses in Magnetic Materials

John B. Goodenough

Abstract—Current concepts of the physical origin and mech-

anisms of losses in magnetic materials are reviewed under
three traditional categories: hysteresis losses, eddy-current or
dielectric losses, and residual losses. The phenomenological
Landau–Lifshitz equation is used to quantitatively express the
various manifestations of the residual losses. Possible physical
mechanisms contributing to the phenomenologically introduced
Landau–Lifshitz damping term are qualitatively discussed.

I. INTRODUCTION

A NY DISCUSSION of energy losses in magnetic or

other materials almost necessarily presupposes specific
engineering applications in which these considerations are of
extreme practical importance. Certainly magnetic losses have
historically captured the interest of physicists, metallurgists, or
ceramists because of their practical significance to the electrical Fig. 1. Initial magnetization curve and hysteresis loop for a typical
engineer or applied physicist. Whatever the origin of the ferromagnet.
interest, a systematic pursuit of loss minimization leads directly
to a study of the basic physical and chemical mechanisms rently believed responsible for losses in magnetic materials. The
of matter. The study has, therefore, intrinsic interest for the second step of the problem is neglected. In Section II, the var-
physicist, metallurgist, and chemist as well as for the electrical ious factors that contribute to magnetic hysteresis are discussed.
engineer. In Section III, eddy-current and dielectric losses are consid-
Such a study can be broken down into two levels or steps: ered. The Landau–Lifshitz formalism and its application to do-
in the first, the various engineering parameters are expressed main-wall damping and ferromagnetic resonance are reviewed
in terms of basic, measurable parameters characteristic of mag- in Section IV. In Section V, there is a brief discussion of the cur-
netic materials in general. This requires an understanding of the rent theoretical ideas of what actual physical mechanisms are
basic macroscopic properties of magnetic matter. Since most of responsible for the damping phenomenologically expressed by
these basic parameters are, or must eventually be, expressible in the Landau–Lifshitz equation.
terms of crystalline energy states and wave functions, in terms
of chemical-bonding energies, electron-spin and orbital mag-
netic moments, exchange interactions, and crystal symmetry, II. HYSTERESIS
the second step is to determine the relationship between these A. Origin of Hysteresis Loss
parameters and chemistry or processing procedures of particular Over certain portions the magnetization curve ( versus ,
materials. Only after this second step has been accomplished is where is the magnetic induction through the material and
it possible to intelligently invent new materials or processing is the applied external field) for a torroidal core is irreversible
procedures for optimizing specific applications. as shown in Fig. 1. Because the flux-change mechanisms can be
In this brief review, discussion must be limited to a qualita- irreversible, energy is dissipated in the medium in the form of
tive description of the macroscopic physical mechanisms cur- heat with each flux-reversal cycle. In reversible processes, en-
This work was supported by the Army, Navy, and Air Force under contract ergy is stored in the lattice just as potential energy is stored in
with the Massachusetts Institute of Technology. This paper was originally a spring that is compressed by an external force. In irreversible
published in Conference on Magnetism and Magnetic Materials, Boston, MA, processes, energy is dissipated as heat in the additional degrees
October 16–18, 1956. New York: American Institute of Electrical Engineers,
February 1957, pp. 368–387. For this republication, the original paper was of freedom that are excited in the lattice. The actual physical
edited slightly by the author. processes by which energy is dissipated in the course of a qua-
The author was with the Massachusetts Institute of Technology, Lincoln Lab- sistatic traversal of the hysteresis loop are identical with those
oratory, Lexington, MA 02420-9108 USA. He is now with the Texas Materials
Institute, The University of Texas, Austin, TX 78712 USA. responsible for the dynamic losses. This is because the qua-
Digital Object Identifier 10.1109/TMAG.2002.802741. sistatic hysteresis loop is traversed by successive rapid, irre-
0018-9464/02$17.00 © 2002 IEEE

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on September 15,2021 at 19:35:39 UTC from IEEE Xplore. Restrictions apply.
 GOODENOUGH: SUMMARY OF LOSSES IN MAGNETIC MATERIALS 3399

versible changes each within a small region of the lattice, the they order parallel to one another. In Néel ferrimagnetic
so-called Barkhausen jumps. These rapid, irreversible changes materials there are two sublattices. The atomic moments order
are produced by relatively strong local fields within the ma- parallel to the other moments of their sublattice, antiparallel to
terial. These small, rapid changes have associated with them the moments of the other sublattice. The material has ferromag-
the same losses that are associated with macroscopic dynamic netic properties because the net moment on one sublattice is
changes. However, it is convenient to isolate the quasistatic from greater than that on the other. As a magnetic material is cooled
the macroscopically dynamic processes as the total effect of the through the Curie temperature, ordered regions of different
varying internal magnetic fields can be known from a measure- orientation are created. These regions are called magnetic
ment of the quasistatic loop. domains. The transition regions between neighboring domains
The energy per unit volume of a magnetic material of magne- are called domain walls. In contrast to the transition regions
tization in an external field is given by . Therefore, between crystallographic domains (grains)—the grain bound-
the sum of the energy stored and the heat dissipated in a mag- aries—domain walls can move through the lattice at lower
netic material whose magnetization changes from to temperatures. There are, therefore, two mechanisms by which
when placed in a magnetic field is flux may change inside a magnetic material: domain rotation
and domain-wall motion. Both processes may be reversible or
irreversible.
In the first of these mechanisms, the magnetic moments
within a domain rotate in unison. In the second, the moments
Reference to Fig. 1 shows that if the loop had been plotted with
within a domain wall rotate as the wall moves to a new position.
instead of for the ordinate scale, the total energy put into
This flux change is characterized by a sequential rotation of
the system on going from to would be the area be-
spins. It is intuitively obvious, therefore, that a domain-rotation
tween the ordinate and the corresponding magnetization curve.
mechanism is intrinsically faster than a domain-wall-motion
Similarly the energy restored to the specimen when the field is
mechanism. Since the basic physical process for flux reversal
removed is the area between the ordinate and the magnetization
is rotation of the atomic moments, the damping process which
curve from to . Therefore, in going around the hys-
limits the speed of rotation is the same for either process. This
teresis loop, energy equivalent to the area of the loop is dissi-
fact is important for the design of magnetic components with
pated as heat in the specimen. This energy dissipation is known
fast flux-reversal characteristics.
as hysteresis loss. This loss is present even if the loop is tra-
2) Energy Considerations: The factor that determines the
versed in a quasistatic manner. Because , this en-
relative importance of domain-wall motion and domain rotation
ergy loss per cycle may be written in cgs units as
in any particular flux change is the free energy of the system.
There are four principal magnetic contributions to the free en-
erg/cm
ergy of a magnetic specimen. The first of these is the magneto-
static energy
Since the area of the hysteresis loop increases with increasing
maximum induction , this energy loss can be plotted as a
function of . From such a relationship, Steinmetz [1] ex-
pressed the energy loss by the empirical equation
which results from the interaction of the atomic magnetic mo-
erg/(cm s). ments with the local internal magnetic field . This local field
varies from point to point within the specimen and is a function
This expression holds fairly well for values of lower than of the externally applied field strength, the saturation moment
the knee of the initial magnetization curve. Values of the Stein- , the shape of the specimen, the crystalline imperfections,
metz coefficient and the constant are often given for sheet and the degree of magnetization.
materials to be used in transformers and in parts of rotating elec- The second and third contributions to the magnetic free
trical machinery. energy result from the interaction between the atomic magnetic
moments and the crystalline lattice. For purposes of conve-
B. Factors Influencing Shape and Area of – Loop
nience, the results of this interaction are expressed in terms
1) Flux-Change Mechanisms: The magnetic flux per unit of two empirically determinable energies, the magnetocrys-
cross-sectional area is defined as the magnetic induction . In talline-anisotropy energy
cgs units . As is indicated in Fig. 1, ap-
proaches a saturation value as is increased to very large
values. The magnitude of is determined by the density and
magnitude of individual atomic moments that contribute to it.
Flux changes resulting from externally applied magnetic fields and the magnetostrictive energy
are normally not the result of variations in the magnitude of the
individual atomic moments; rather they are due to changes in
their individual orientations.
Below a critical temperature, the Curie temperature , the Since the total interaction energy is most stable when the mo-
atomic moments become ordered. In ferromagnetic material ments are oriented along certain crystallographic axes—the axes

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on September 15,2021 at 19:35:39 UTC from IEEE Xplore. Restrictions apply.
 3400 IEEE TRANSACTIONS ON MAGNETICS, VOL. 38, NO. 5, SEPTEMBER 2002

Fig. 3. Theoretical hysteresis loops for uniaxial thin magnetic films.

Fig. 2. Schematic diagram of a 180 domain wall.

The effective anisotropy is determined not only by the magne-
of easy magnetization—the magnetocrystalline-anisotropy en- tocrystalline anisotropy and magnetostriction, it is also deter-
ergy density expresses, in terms of empirically determinable mined by the local fields , which depend upon the orientation
proportionality constants , the energy required to rotate the of the moments with respect to the dimensions of the sample and
magnetic moments away from an axis of easy magnetization to lattice imperfections.
with zero applied mechanical stress. Thus, symmetry consid- 3) Irreversible Rotation: It was pointed out in Section II-A
erations dictate that that only the irreversible processes contribute to hysteresis
losses. In most materials, domain-wall creation and motion
contribute to the irreversible flux changes. However, there are
two geometries that are of practical importance because irre-
versible domain-rotation processes can be made to predominate
where are the direction cosines of the magnetic mo- in materials with these shapes.
ment with respect to the crystalline axes in a cubic crystal, and The first geometry is the fine particle: a magnetic particle
is the angle the moment makes with the unique axis in a lattice that is too small for domain-wall formation to be energetically
with but one axis of easy magnetization. favorable. This geometry is extremely important for permanent
Magnetostriction is the nomenclature applied to the changes magnets. Materials with large, uniaxial, magnetocrystalline
that occur in lattice dimensions as the magnetic-moment orien- anisotropy are formed in, or are ground to, small particle size
tation is changed with respect to the crystallographic axes. If with needle-like shapes, the axis of the needle coinciding with
mechanical stresses or constraints are present, the stable energy the easy axis of magnetization. If the particle is smaller than a
state is influenced by this magneto-mechanical coupling. domain-wall width, flux reversal takes place only by rotation
Finally there is the contribution to the free energy of the against the large shape and crystalline anisotropy fields. High
system that comes from the “magnetic-exchange” interaction, coercivities result.
the interaction between magnetic moments to orient them par- The second geometry is the thin ( 1000 Å) magnetic film.
allel or antiparallel to their neighbors. The physical origin of The purpose of this geometry is to provide a low-coercivity ma-
this interaction resides in the symmetry properties of wave me- terial in which flux reversal takes place by domain rotation. In
chanics. For many purposes, the matrices of atomic moments this geometry, shape anisotropy keeps the magnetic moments
may be considered as classical vectors. In this case it is possible in the plane of the film. If the film is deposited in the presence
to express the exchange energy density as a function propor- of a magnetic field parallel to the plane of the film, uniaxial
tional to an exchange constant [erg/cm] and a term expressing anisotropy develops along the direction of the magnetic field.
the misalignment of neighboring atomic moments with distance Fig. 3 illustrates the hysteresis loops that are characteristic of
of separation. The Curie temperature is an empirical mea- such a film when the driving field is, respectively, perpendic-
sure of the magnitude of . ular and parallel to the unique axis. These characteristics are
In Fig. 2 is shown a diagram of a 180 wall. Through the wall calculated from the anisotropy energy and the
the spins are oriented so as to keep the magnetostatic energy magneto-static energy. In the first case, the flux-reversal process
a minimum. Further, in a domain wall spins are oriented away is completely reversible and there is no hysteresis. This type of
from an easy axis of magnetization and are not parallel to the loop has been experimentally observed. In the second case, there
axis of their neighboring spins. Therefore, the total anisotropy is no rotation until the torque due to the applied field is suf-
energy per unit area of wall is reduced if the wall is narrowed ficient to overcome the anisotropy forces. Since the torque in-
whereas the exchange energy is reduced if the wall is broad- creases faster than the anisotropy restoring forces as the magne-
ened. The total energy per unit area of domain wall is the sum of tization is pulled away from the easy-magnetization direction,
these two contributions and is a minimum when these two con- a sudden irreversible 180 rotation takes place. This provides
tributions are equal. Thus, the energy per unit area of a domain an extremely square hysteresis loop. In practice, domain walls
wall has the form , where is an effective perpendicular to the film are also energetically feasible and do-
anisotropy constant, and the thickness of a wall is . main-wall processes compete with the domain-rotation process.

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on September 15,2021 at 19:35:39 UTC from IEEE Xplore. Restrictions apply.
 GOODENOUGH: SUMMARY OF LOSSES IN MAGNETIC MATERIALS 3401

Fig. 4. Typical B –H loops for materials with (a) H < 0, (b) H > 0.
Although square, low-frequency hysteresis loops have been ob-
served, these can usually be shown to be due to domain walls.
These loops have a coercivity .
In films with sufficiently low anisotropy that
is of the order of 1 Oe, domain rotation processes can be initiated
by a fast-rise-time ( 1 ns) driving-field pulse that is uniform
over the film. Flux-reversal speeds at least compatible with a
rotation process are observed [2]. Attempts are currently being
made to exploit these potentialities for switching devices and
storage elements in high-speed digital computers.
4) Domain-Wall Processes: Except in the special instances Fig. 5. Hysteresis loops of 68 permalloy under tension. Maximum field
strength 5 Oe.
cited above, the principal irreversible mechanism responsible
for the magnetic hysteresis loop is domain-wall motion. There-
fore, in order to understand the usual mechanisms contributing zation despite the varied grain orientations. A typical effect is
to the hysteresis losses, i.e., to the area of the hysteresis loop, it illustrated in Fig. 5. The development of oriented materials has
is necessary to understand the mechanisms that determine do- been extremely important for transformer design, magnetic am-
main-wall creation and motion. plifiers, and magnetic switches.
In order to reverse saturated magnetization, domains of re- In materials with such extremely square loops that the ex-
verse magnetization must be created. These domains will be ternal field for wall creation is greater than that for wall mo-
created in the neighborhood of crystalline imperfections or at tion, the reverse-domain nucleation field may determine the co-
a crystal surface where domain creation can reduce the energy ercivity ; but in most materials this quantity is determined by
associated with large local fields . If the local fields are suffi- domain-wall surface tension and internal fields resulting from
ciently strong to create many domains of reverse magnetization specimen shape or lattice imperfections, the walls stabilizing in
when the external field is still in the direction of the initial sat- positions that minimize the energy associated with them. Pre-
uration field, the remanence value will be low (see Fig. 4). In cipitates, inclusions, cold work, grain size, grain orientation,
some materials the remanence value is practically zero, and the and specimen shape are all important in determining . The
loop has a butterfly shape. In other materials there is little or important intrinsic parameters that enter into the various terms
no reverse-domain creation until the external field is opposed to contributing to the coercivity are and . The
the direction of the initial saturation field. These materials have control of by suitable chemistry and processing techniques
a high remanence. It is often possible to control the shape of the is extremely important in magnetic-material design.
loop by either controlling the shape and lattice imperfections of
the specimen, or controlling the effects of these imperfections. III. EDDY-CURRENT AND DIELECTRIC LOSS
A striking illustration of this is control of the effects of the grain
boundary. Because the crystallographic directions change on A. Flux Changes Throughout a Conductor
passing through a grain boundary, the normal components of the In 1831 Faraday found that whenever the number of tubes of
magnetization vectors in the two grains adjacent to a boundary magnetic flux linking a closed circuit are changed, an induced
are not usually equal at remanence, and grain-boundary mag- current results. The current so induced flows in such a direc-
netic poles exist. Commonly the local fields associated with tion that its magnetic field opposes the change in magnetic flux
these poles are sufficiently great to cause reverse-domain cre- that produces it. Consequently whenever the flux is changed in-
ation at the grain boundary [3]. Three methods have been used to side a ferromagnet, local currents that flow in planes perpendic-
reduce these grain-boundary poles: a) grain orientation by cold ular to the magnetic lines of force are induced within the mate-
rolling to reduce the misalignment of easy-magnetization direc- rial. These local currents are called eddy currents. The associ-
tions from grain to grain; b) a magnetic anneal; or c) application ated fields oppose the domain-wall motion producing the flux
of a tensile stress to produce a common axis of easy magneti- change. In a magnetic field varying with angular frequency

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on September 15,2021 at 19:35:39 UTC from IEEE Xplore. Restrictions apply.
 3402 IEEE TRANSACTIONS ON MAGNETICS, VOL. 38, NO. 5, SEPTEMBER 2002

, the resulting power loss may be expressed in terms of the lag The average power dissipation occurs at . Also
between impressed field and induction change. If both and
vary with time as , such a lag is expressed mathe-
matically by introducing a complex permeability where and are the maximum and minimum values of .
with being the phase angle between Here, , , and
and . Since is a measure of the lag, which in turn is . Therefore
directly related to the power loss, is called the magnetic
loss tangent.
If one assumes a homogeneous permeability throughout the
The discrepancy between these cases is seen to be the factor
material in the calculation of eddy-current losses, the calcu-
. This means that calculations of eddy-cur-
lated losses are always somewhat smaller than the measured
rent losses based on any model between these two extremes will
losses. The discrepancy gets worse the larger the spacing be-
have the same functional dependence on specimen dimensions,
tween domain walls compared with the small dimension of the
frequency, , and resistivity. They will differ in absolute mag-
specimen perpendicular to the lines of magnetic flux. This is to
nitude by less than a factor of 2.
be expected since the permeability is not at all homogeneous
If the power dissipated by eddy currents about the expanding
throughout the material, but is localized to the vicinity of the
cylindrical wall is set equal to 2 , the power
moving domain walls. This discrepancy is known as the eddy-
released by magnetization change in the effective driving field
current anomaly.
, the reciprocal radial velocity of the domain
As an illustration of the order of magnitude of this effect, the
wall is calculated as
special case of eddy currents in a cylindrical rod of length ,
radius , will be considered. In the extreme of flux change due
to domain rotation, the assumption of homogenous permeability
where
is correct. In this approximation, the eddy currents are assumed
to flow tangentially about the rod so that the voltage induced at
any radius is given by
The fields and refer, respectively, to the externally ap-
plied field and a threshold field for domain-wall motion. Since
if the varying enclosed flux is given by . In 2 is the effective driving force per unit area on the wall,
Gaussian units, the power dissipated due to a current is it is apparent that gives the damping per unit area of wall due
therefore to eddy currents. The complete equation of motion for an ex-
panding 180 cylindrical wall would be

where is the domain-wall mass per unit area. The problem as

stated above has neglected acceleration and therefore assumed
where is the volume of the cylinder, is the resis- . It has also lumped the work done against elastic forces
tivity, and is the velocity of light. The average power dissipated per unit area, , and the work done against domain-wall sur-
is then face tension into an effective threshold field . These approxi-
mations are valid provided large, irreversible wall motions occur
and the inertial term is much smaller than the damping term. The
The general form of this expression is typical for all such eddy- justification for the latter condition is given in Section IV, where
current calculations except that is replaced by the small spec- another contribution to , the relaxation contribution , is dis-
imen dimension perpendicular to the lines of magnetic flux. cussed. The former condition arises because the elastic forces
In the other extreme case, the flux change may be considered dominate only while a wall is restricted to reversible motions
to be due to the motion of a single domain wall of radius within a potential-energy well about some lattice imperfection.
concentric with the cylinder. In this case the eddy currents are It should be noted in passing that although is usually the
confined to the region and the induced voltage at dominant damping factor in metallic ferromagnetics, is the
is [4] dominant damping factor where laminations of less than 25 m
thickness or high-resistivity ferrites are used.

B. Skin Effect
The above discussion of eddy currents assumes that the ex-
where is the saturation induction. Again, the ternal field penetrates the entire ferromagnet so that flux change
power dissipated is is occurring in the interior as well as the exterior portions of the
specimen. This assumption is valid for the thin sheets or wires
that are customarily used in transformers or metallic-tape cores
that are designed to reduce eddy-current effects. In bulk metals,
however, the assumption of a homogeneous permeability is

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on September 15,2021 at 19:35:39 UTC from IEEE Xplore. Restrictions apply.
 GOODENOUGH: SUMMARY OF LOSSES IN MAGNETIC MATERIALS 3403

completely inadequate. In these materials, the eddy currents

shield the interior from the external driving fields so that the
amplitude of the flux change decreases and the phase angle
increases with penetration depth.
A calculation of the penetration depth follows directly from
the Maxwell equations for sinusoidally varying fields:

Fig. 6. Equivalent circuit for calculation of effective dielectric constants in

ferrites.

decreased rapidly to a very low value at about 2 MHz. This fre-

These equations may be combined in the usual way to give the quency is much lower than would be expected from previously
wave equation measured losses in ferrites.
It was pointed out above that the wavelength of an electro-
magnetic wave within a material is inversely proportional to
. If the effective dielectric constant is large, the wave-
where and are, respectively, the absolute values of the length may be small enough to be comparable with the dimen-
permittivity and the permeability . sions of the sample. A dimensional resonance can then occur
The ohmic and displacement currents are seen to both appear in in which standing waves tend to be set up within the material
the permittivity. Here, and are the total electric loss angle and a greater amount of energy is transferred from the electro-
and magnetic loss angle, respectively. If a plane electromagnetic magnetic wave to the magnetic lattice. This resonance of en-
wave propagating in the direction with magnetic field im- ergy transfer is typically reflected in a resonance peak for ,
pinges normally upon the plane surface of an isotropic medium, the imaginary part of the effective permeability, and a comple-
then mentary dispersion in . The resonance is similarly reflected
in the effective dielectric constant. Two methods have success-
fully been used to shift the resonance frequency to higher values:
where the propagation constant is and is 1) the use of blocks of smaller dimension and 2) the application
known as the skin depth. The power dissipated by eddy currents of a dc field to reduce . The elimination of this effect is im-
is equivalent to that lost by total effective eddy currents in a dc portant in studies of ferromagnetic resonance.
resistance of a skin of thickness . At , the penetrating field is The ferrites investigated by Brockman, Dowling, and Steneck
attenuated to of its value at the surface. If this solution had a certain content of ferrous ions. If ferric and ferrous ions are
is substituted into the wave equation, the skin depth is given randomly distributed on the same type of lattice site, electrical
by conductivity is enhanced because of the ease of electron transfer
between Fe and Fe . To understand better the origin of the
high dielectric constants, Koops [6] investigated nickel–zinc fer-
rites prepared under different oxidizing conditions to vary the
conductivity from 10 to 10 ( cm) . It was found that
For metals the contributions of the dielectric constants are a high dielectric constant (as high as 10 ) occurs with a higher
negligible compared to those of the conductivity , and conductivity. Further dispersions in the dielectric constant and
. If magnetic losses are negligible resistivity were observed at high frequencies: above 9300 MHz
, the usual relationship for skin depth and wavelength the dielectric constants of all ferrites decrease from the 12 to 15
follow: range to below 10. The high values of dielectric constant and
resistivity also decrease with increasing electric field strength.
The physical origin of these effects is somewhat obscure. An
equivalent circuit that can account for the dispersions and also
In metals of high conductivity, is extremely small. In ferrites, has a physical interpretation is shown in Fig. 6. It is supposed
however, may be more than 10 times smaller. These mate- that a polycrystalline, sintered ferrite consists of large regions
rials are therefore useful at microwave frequencies. However, with relatively large conductivity ( and of Fig. 6) sepa-
magnetic losses are not negligible in those instances. Control rated by thin layers (the grain boundaries) of a relatively poor
of the magnetic losses in these materials is an area of current conductor ( ).
concern in many laboratories. These losses are discussed in
Section IV. IV. RELAXATION PROCESSES
C. Cavity Resonance A. Phenomenological Equation
During the design of a massive ferrite core composed of fitted If the thermal equilibrium of a magnetic system is suddenly
brick-shaped pieces with cross-sectional area 2.5 1.25 cm to altered by some external force, the reestablishment of a thermal
be used in the lower megacycle region, Brockman, Dowling, and equilibrium of this system is governed by relaxation processes.
Steneck [5] discovered that the initial permeability of the bricks Aside from an inertial moment, the interactions that modify the

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on September 15,2021 at 19:35:39 UTC from IEEE Xplore. Restrictions apply.
 3404 IEEE TRANSACTIONS ON MAGNETICS, VOL. 38, NO. 5, SEPTEMBER 2002

immediate response of the atomic magnetic moments to an im- The factor 2 on the right-hand side of the equation is valid only
pressed force can only be interactions of these atomic moments for 180 walls. Any average value must be between 2 and 1, the
with themselves or with the lattice. Since the atomic moment is factor for 90 walls.
associated with the electron “spin” of the atom, these processes Whatever the equation used, both and can be expressed
may be described as either “spin–spin” relaxations or “spin–lat- in terms of measurable physical constants through the phe-
tice” relaxations, respectively. nomenological Landau–Lifshitz equation. To do this, use is
The precise interaction mechanisms that are responsible made of a concept introduced by Becker [10] in 1951. Becker
for the magnetic-relaxation processes are not completely showed that domain-wall velocity can be translated directly
understood. A brief summary of present theory is given in into the precessional velocity of the atomic moments inside
Section V. Although a quantitative explanation of the relaxation the moving wall. The precessional velocity can be imagined
effects that are observed must await a better understanding of due to a field which is perpendicular to the wall. The total
the detailed interaction mechanisms, the qualitative features of field responsible for wall motion is, therefore, ,
these effects may be described by a phenomenological equation where is the component of the applied field parallel to the
of motion of the magnetization subject to a torque imposed wall. The precessional velocity of the spins can then be shown
by an external field : from (1) to be
(1) (3)
where is the magnetomechanical ratio . This where
equation follows directly from equating the rate of change of an-
gular momentum to the torque reduced by a frictional
term that is directed to oppose the direction of motion. The ar- and the proportionality constant follows from the
bitrary parameter must have dimensions s and is therefore Landau–Lifshitz equation with the displacement variable
called the relaxation frequency. This equation, first proposed by or . The kinetic energy of the wall is the difference in the
Landau and Lifshitz [7] in 1935, does not attempt to differen- energy of a domain wall in motion and at rest
tiate between spin–spin and spin–lattice processes. Since sev-
eral physical processes may simultaneously contribute to the
relaxation frequency , it is extremely dangerous to assign a
measured value of to any actual physical process. The integral can be shown proportional to
Other phenomenological equations have been suggested. through use of (3) and a calculation of the domain-wall energy
One such formalism, adopted by Bloembergen [8] from Bloch’s per unit area . It follows that the mass per unit area of wall
[9] original description of nuclear magnetic resonance, attempts is , where is the exchange parameter
to separate the spin–spin and spin–lattice interactions by the defined in Section II. In typical materials, where
introduction of two relaxation frequencies, and . g/cm , the inertial term in the equation of wall motion is much
However it, with all other phenomenological equations, is smaller than the damping term and may be neglected. Only in
subject to the same major limitation: it cannot give any details high-resistivity ferrites driven at frequencies above 10 MHz
of the physics of the relaxation processes themselves. Since the does the inertial term have a measurable effect.
Landau–Lifshitz formalism can be easily adapted to describe From the Landau–Lifshitz equation it also follows that the
both domain-wall and domain-rotation processes, it is used in power dissipated by the motion of the wall is
this descriptive summary.

B. Domain-Wall Damping
1) Flux-Reversal: As was pointed out in Section III, the
equation of motion of a cylindrical 180 domain wall may be
where is the wall area.
expressed as
If inertia is neglected and the elastic terms are lumped into an
effective threshold field for irreversible wall motion, this power
(2) dissipation can be equated to 2 , the power sup-
plied by the effective applied field if is the
where the damping coefficient is composed of a volume of the growing domain bounded by the moving wall. It
relaxation contribution as well as the eddy-current contribu- follows that for walls experiencing large, irreversible motions
tion . If large, irreversible flux reversals take place, an equa-
tion for cylindrical walls is probably most appropriate as it de- (4)
scribes nucleation and/or growth of many ellipsoidal domains of where
reverse magnetization with large eccentricity. If initial-perme-
ability measurements are made, the domain-wall configuration
is more complicated and a plane-wall equation is a sufficiently
This relaxation damping is extremely important for pulse-cir-
satisfactory representation:
cuit applications. It not only limits the speed of flux reversal,
(2 ) it also contributes a residual energy dissipation per unit volume

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on September 15,2021 at 19:35:39 UTC from IEEE Xplore. Restrictions apply.
 GOODENOUGH: SUMMARY OF LOSSES IN MAGNETIC MATERIALS 3405

a relaxation phenomenon is observed. Ideally the half-width of

the resonant peak is . Because many walls contribute to the
resonance, however, the resonant linewidth is inhomogeneously
broadened to a considerable extent.
Rado and his co-workers [11] have observed a resonance
at 50 MHz in the initial permeability of a Ferramic A
(45–14–3–2–1 percent Fe–Mg–Mn–Ca–Zn ferrite, remainder
presumably oxygen) that appears attributable to domain-wall
motion. Observations of initial permeability by other workers
on other materials give no unambiguous domain-wall reso-
nance. Interpretation of the measurements is difficult because
there is a domain-rotation contribution to the initial perme-
ability that also has a resonance. The relative contributions
Fig. 7. Switching characteristics of metals and ferrites. of the two effects are not always easy to separate. They vary
markedly with composition and processing. An experimental
per half cycle , where is the largest distance any wall technique for determining the relative contributions of the two
moves in the flux-reversal time . In coincident-current memory effects has recently been worked out by Grimes [12].
elements, driving fields are limited to . But in a flux-re- 3) Amplitude-Permeability Dispersion: The domain-wall
versal process involving irreversible domain-wall motion contributions to the initial permeability discussed above are
due to reversible wall motions. Using a maximum driving field
Oe, Went and Wijn [13] observed a dispersion of the
and amplitude permeability in manganese-ferrous
ferrite and nickel–zinc ferrite. This dispersion can be satisfac-
torily accounted for by the damping of the irreversible wall
Typical plots illustrating this last equation are shown in Fig. 7. motions.
The contribution to from alone is large enough that if small
values of are desired ( s), must be large C. Ferromagnetic (Rotational) Resonance
enough to produce considerable hysteresis loss. Heating due to 1) Resonance Frequency: The electrons whose spins con-
hysteresis and relaxation losses becomes the major factor lim- tribute to the net magnetic moment of the magnetic atoms also
iting ultimate repetition frequencies for ferrite switching and possess an angular momentum. If a torque is exerted on the mag-
memory components. netic moment by an external field , the moment begins to pre-
2) Domain-Wall Resonance: The frequency variation of the cess about the direction of the field in much the same manner as
domain-wall contribution to the initial permeability a spinning top precesses about the vertical under the influence
also follows directly from the equation of motion of a do- of the gravitational field. The natural frequency for this rotation
main wall. If the driving field for the plane-wall equation is is the well-known Larmor frequency
, the form of the equation is immediately seen to
be equivalent to that for forced harmonic motion or a harmon-
ically driven LCR circuit. If is an average domain size, the where is the spectroscopic splitting factor and is called the
domain-wall magnetic susceptibility is given by , gyromagnetic ratio. (For a discussion of the difference between
where . The steady-state for from (2 ) then the magneto-mechanical ratio and the gyromagnetic ratio ,
gives see [14] and [15].) This frequency condition is derived directly
from the simple equation of motion (damping neglected)
(5)
If an atomic moment is aligned in a strong static field, a pre-
cessional motion is induced. If a small ac field is superimposed
at right angles, the moment experiences a torque that increases
its angle with the field. As the frequency of the ac field goes
through the Larmor frequency, resonance occurs and energy is
absorbed. In the calculation of this resonant field, care must be
taken to include in demagnetizing and effective anisotropy
with resonance frequency , where fields as well as the applied static field [16].
is the static value of the susceptibility due One interesting contribution to this effective field occurs
to domain-wall displacements, , and . It is when the ac field is parallel to a 180 wall [17]. In this case
to be noted that the elastic coefficient can be associated with the spins in the two adjacent domains precess in such a phase
a measurement of , whereas and are related to mea- relation that they have simultaneous components toward or
surable constants as previously indicated. If is so small that away from their common boundary (see Fig. 8). The internal
, the resonance frequency becomes imaginary and demagnetizing fields that result contribute a term to which

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on September 15,2021 at 19:35:39 UTC from IEEE Xplore. Restrictions apply.
 3406 IEEE TRANSACTIONS ON MAGNETICS, VOL. 38, NO. 5, SEPTEMBER 2002

important. In most practical cases this problem can be formu-

lated in terms of the phenomenological Landau–Lifshitz equa-
tion (1). If a damping term is added to the simple equation of
angular motion (5), the resonant frequency is shifted slightly
away from the Larmor frequency for a free atom, and there
is a natural linewidth of the resonance absorption. This shift
and linewidth can be expressed [18], [19] in terms of the re-
laxation frequency .
If the static field is , the procedure is to assume a time de-
pendence of the and components of and in
the Landau–Lifshitz equation. An expression for
and, therefore, and , follows. Since the demagnetizing
fields and the effective anisotropy fields enter into these expres-
sions through , the effective field inside the material, there
Fig. 8. Magnetic poles induced on 180 domain walls by an ac field
(a) perpendicular and (b) parallel to the walls.
is no shape-independent plot for and . For spherical sam-
ples or samples of infinite size, the resonant frequency is

may be as high as . This effect is important, of course,

only if the static applied field is so small that domain walls are
where
present in the material.
If a small ac signal of increasing frequency is applied to a
demagnetized specimen, a resonance in the rotational contribu-
tion to the imaginary part of the initial permeability may and the half-amplitude linewidth is
be observed. This resonance is the result of a precession of
the atomic moments about the directions of the internal mag- -
netic field. In the experiments of Rado on Ferramic , the res- Although the actual resonance line measured on polycrystalline
onances in and were resolved. This is not in general the samples is broadened by the variations of effective anisotropy
case. Experimentally the resonance in initial permeability is not from grain to grain and about impurity centers, resonance
sharp. Instead of a single resonance peak, the losses extend over linewidths of well-annealed, spherical single crystals do give
a frequency range. The additional losses may be attributed to a measure both of the magnetocrystalline anisotropy constants
the effects of the magnetic-domain structure that is present in and the Landau–Lifshitz relaxation frequency . As was
the demagnetized material. The domain structure in a demagne- pointed out earlier, the latter value is of little fundamental
tized specimen is nearly always extremely complex. This means significance so long as it is not a measure of any specifically
that there are contributions to the local internal fields from do- known physical mechanism.
main-wall structure which may vary from 0 to . With such
a large variation in effective internal field strength from region to V. THEORETICAL CONSIDERATIONS
region in the material, there is a correspondingly wide variation
in local resonant frequency, and the resonance is smeared out. Since the linewidth problem, or the damping phenomenolog-
These effects are important in low-frequency microwave de- ically expressed by the Landau–Lifshitz equation, has no com-
vices. In the microwave application, the frequency is kept con- plete theoretical explanation, it has recently been the object of
stant and the static field strength is varied to produce resonance. considerable research. In a few specific cases it has been pos-
With small applied fields, domain walls exist in the material. sible to identify a definite mechanism that contributes at least in
The demagnetizing fields associated with the poles induced at part to the observed linewidth. These mechanisms are all either
the domain walls by the ac field may be sufficient to cause res- spin–lattice or spin–spin interactions.
onance in some regions of the material even though the applied
field is low. This effect gives rise to large losses at low ap- A. Spin–Lattice Interactions
plied fields that persist until the applied field is large enough The spin–lattice interactions include those between the
to remove most of the walls. These are usually referred to as system of atomic moments and the vibrations of the crystal lat-
low-field losses. In low-frequency ( 100 MHz) devices, the res- tice and those between the atomic moments and the conduction
onance field strength is relatively small. In these devices the electrons.
walls may not be removed before the influence of resonance 1) Interactions of Spin Wave With Phonons: A calculation
in the applied plus anisotropy fields alone is felt. This problem of the spin–lattice relaxation time characterizing the exchange
makes the construction of low-frequency devices difficult. Cur- of energy between the system of spin waves and the system of
rent ferrites for low-frequency applications are being designed lattice vibrations has been made with the methods of quantum
with a small value of in order to reduce these effects. field theory. These calculations start from the standard phe-
2) Resonance Linewidth: In most microwave devices using nomenological, two-constant equation for the magnetostrictive
magnetic insulators to give components with a nonreciprocal contribution to the free energy, the magnetostriction repre-
character, the width of the spin-resonance absorption line is senting the magnetoelastic interaction of the magnetization

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on September 15,2021 at 19:35:39 UTC from IEEE Xplore. Restrictions apply.
 GOODENOUGH: SUMMARY OF LOSSES IN MAGNETIC MATERIALS 3407

with the lattice. The direct interaction between the spin system, Yager, Galt, and Merritt [21], working with single crystals of
in a mode of uniform precession, and the lattice is found to ferrous-nickel ferrite, have measured the resonance linewidth
be very small, leading to a relaxation time 10 s. The and determined as a function of temperature. They also
observed relaxation times are 10 s. This discrepancy may determined an activation energy in good agreement with that for
be accounted for by spin–orbit, orbit–lattice interactions. electrons found from dc resistivity measurements.
2) Spin–Lattice Interactions via Conduction Electrons in Although these experiments indicate that there can be a con-
Semiconductors: Wijn and van der Heide [20], working with siderable contribution to the relaxation frequency due to an
iron-rich ferrites, have measured the initial permeability and interaction between the atomic moments and the conduction
determined the magnetic-loss factor as a electrons in ferrites containing both ferric and ferrous ions, it
function of temperature for various frequencies. They observed is not yet clear just what it is that relaxes when the electrons re-
a maximum that shifted to higher temperatures with higher distribute themselves. This effect is reduced, of course, by the
frequencies. Such a relationship is characteristic of a relaxation elimination of the ferrous ions.
phenomenon with relaxation time 3) Spin–Lattice Interactions via Conduction Electrons in
Metals: If an electromagnetic field penetrates a ferromagnetic
metal, the phase angle between the electromagnetic field and
where the heat of activation is approximated by an activation the induced RF component of the magnetization vector varies
energy and is the Boltzmann gas constant. Except in the with depth of penetration. This variation is due to the eddy
immediate vicinity of the Curie temperature, the entropy will not currents that tend to concentrate at the surface of the material
contribute significantly to the heat of activation. The relaxation within a skin depth . If the skin depth is
phenomenon may be expressed by the basic equation small (experimentally obtained by choosing a material with
large ), the variation of the orientation of with penetration
can be made relatively large so that there exists a measurable ex-
where is the relaxation time for with constant , is the change torque tending to orient neighboring spins parallel. This
relaxation time for with constant , and is the relaxed exchange torque is capable of modifying the motion of and,
permeability. These definitions follow at once if, respectively, therefore, the nature of the ferromagnetic resonance. Ament and
and are substituted into the above Rado [22] have calculated the linewidth to be expected from
equation. If a periodic field is impressed on a magnetic material this interaction by adding to in the usual Landau–Lifshitz
such that , , the above equation an exchange field , where
equation gives is the usual exchange parameter discussed in Section II. In the
special case where this effect is large compared to the damping
introduced through the relaxation frequency , the agreement
between experiment and theory is satisfactory.
For small , and
B. Spin–Spin Interactions
In nonmetallic materials of high resistivity (ferrites with low
ferrous content, for example), there are no conduction electrons
(6) to dissipate energy. Since the spin-phonon interactions are too
weak to give the observed linewidths of 50 Oe (relaxation
where times 10 s), there must be at least one other source of line
broadening. A possible source is semicovalent exchange, which
involves transfer of two electrons from an oxygen, one each, to
and the d-state manifolds of the interacting spins.
1) Dipole–Dipole Interactions: The ordinary exchange in-
teraction that aligns neighboring spins parallel (or antiparallel)
It follows that is a maximum if . Therefore, at to one another cannot cause any broadening of the resonance
a given temperature , there will be a frequency for which line since the Weiss molecular field is always par-
is a maximum and allel to the magnetization so that the torque must
vanish. However, there are dipole–dipole interactions between
similar spins situated at the sites of a regular lattice that can
The activation energy can therefore be determined from the give rise to an intrinsic resonance linewidth. These calculations
slope of the line versus , where is the temperature [23] predict a strongly temperature-dependent linewidth that
at which is a maximum for a given . goes to zero at the absolute zero of temperature. Well below the
Wijn and van der Heide measured activation energies Curie point the experimental linewidths tend to increase slightly
varying from 0.41 eV for a nickel–zinc ferrite with resistivity as the temperature is lowered, suggesting strongly that a finite
10 cm. These activation energies were found to linewidth remains even at absolute zero. Attempts to amend the
correspond to the activation energy determined from dc original calculations [24] to give a finite linewidth at absolute
resistivity measurements on the same zero have been challenged [25]. A possible answer [23] to this
materials. dilemma lies in the effect of magnetic inhomogeneities that gen-

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on September 15,2021 at 19:35:39 UTC from IEEE Xplore. Restrictions apply.
 3408 IEEE TRANSACTIONS ON MAGNETICS, VOL. 38, NO. 5, SEPTEMBER 2002

erally occur in the experimental materials. A microscopically in- [14] J. H. Van Vleck, Phys. Rev., vol. 78, p. 266, 1950.
homogeneous distribution of two kinds of magnetic ions on the [15] C. Kittel, Phys. Rev., vol. 76, p. 743, 1949.
[16] C. Kittel, Phys. Rev., vol. 73, p. 155, 1948.
octahedral sites of a ferrospinel, for example, leads to a variation [17] D. Polder and J. Smit, Rev. Mod. Phys., vol. 25, p. 89, 1953.
of the dipolar forces between neighboring ions. This variation [18] W. A. Yager, J. K. Galt, F. R. Merritt, and E. A. Wood, Phys. Rev., vol.
predicts a finite linewidth at the absolute zero of temperature. 80, p. 744, 1950.
[19] C. Kittel and E. Abrahams, Rev. Mod. Phys., vol. 25, p. 233, 1953.
2) Interactions of Spin Waves With Spin Waves: Recent [20] H. P. J. Wijn and H. van der Heide, Rev. Mod. Phys., vol. 25, p. 98, 1953.
studies of ferromagnetic resonance at high levels of signal [21] W. A. Yager, J. K. Galt, and F. R. Merritt, Phys. Rev., vol. 99, p. 1203,
power by Anderson and Suhl [25] caused these workers to 1955.
[22] W. S. Ament and G. T. Rado, Phys. Rev., vol. 97, p. 1558, 1955.
suspect that the spin-wave spectrum should be modified when [23] F. Keffer, Phys. Rev., vol. 88, p. 686, 1952.
considering a finite magnetic body. Clogston et al. [26] have [24] N. Bloembergen and S. Wang, Phys. Rev., vol. 93, p. 72, 1954.
recently shown that in a finite medium, there exists a degen- [25] P. W. Anderson and H. Suhl, Phys. Rev., vol. 100, p. 1789, 1955.
[26] A. M. Clogston, H. Suhl, L. R. Walker, and P. W. Anderson, Phys. Rev.,
eracy of the spin wave with many spin waves of large vol. 101, p. 903, 1956.
value. The spin–spin interactions due to the interactions of [27] C. Kittel, “Physical theory of ferromagnetic domains,” Rev. Mod. Phys.,
the degenerate spin waves can give rise to a linewidth of the vol. 21, pp. 541–583, 1949.
[28] J. Smit and H. P. J. Wijn, “Physical properties of ferrites,” Adv. Electron.
observed order of magnitude since the spin waves of higher Electron Phys., vol. 6, pp. 69–136, 1954.
value can transfer considerable energy to the lattice. [29] E. Abrahams, “Relaxation processes in ferromagnetism,” Adv. Electron.
Physically this degeneracy arises from the fact that the effec- Electron Phys., vol. 6, pp. 47–68, 1954.
tive field contains demagnetizing-field contributions of the
form , where and are demagnetizing fac-
tors corresponding, respectively, to the static-field direction and John B. Goodenough was born in Jena, Germany,
on July 25, 1922. He received the A.B. degree in
to all directions perpendicular to the static field. These contri- mathematics from Yale University, New Haven, CT,
butions depress the spin-wave spectrum from that for the infi- in 1943, and the M.S. and Ph.D. degrees in physics
nite medium in which corresponding to lies at the from the University of Chicago, Chicago, IL, in
1951 and 1952, respectively. Prior to graduate study,
very bottom. In an infinite medium, . In a fi- during World War II, he served in the U.S. Air Force
nite medium, remains roughly the same for all spin waves as a meteorologist.
whereas is rapidly reduced for spin waves of large except at He came to the Massachusetts Institute of Tech-
nology’s Lincoln Laboratory in 1952. His initial as-
the surface of the sample, neighboring regions of opposite po- signment was with a small group charged with the
larity being induced on the perpendicular surfaces by the spin development of the square-loop ferrite cores to be used in the coincident-cur-
waves with . Consequently the short-wavelength spec- rent magnetic memory of the early digital computer. Following the successful
completion of that project, he turned to studies of cooperative orbital ordering
trum (large ) is depressed from the infinite-medium case by a on transition-metal atoms in solids, the interatomic spin–spin exchange interac-
greater amount than the spin wave for , and the degen- tions, and the complex long-range magnetic ordering resulting from competitive
eracy follows. interatomic exchange interactions. These studies were summarized in his book,
Magnetism and the Chemical Bond, published in 1963. In the 1960s, he con-
In brief, line broadening occurs because of inhomogeneous centrated on the transition from localized to itinerant electronic behavior. That
internal fields. work was the subject of his book, Les Oxydes des Métaux de Transition, pub-
For important reviews not specifically referred to in the text, lished in 1973 as a translation of his review article, “Metallic Oxides.” He is the
author or coauthor of more than 500 papers and over 70 book chapters. In 1976,
see [27]–[29] he became Professor and Head of the Inorganic Chemistry Laboratory, Univer-
sity of Oxford, Oxford, U.K. While he was at Oxford, he developed the layered
Li CoO and spinel Li Mn O cathodes for secondary Li-ion batteries
REFERENCES used in cell telephones and laptop computers. In 1986, he took up the Virginia H.
[1] C. P. Steinmetz, Trans. Amer. Inst. Electr. Eng., vol. 9, p. 3, 1892. Cockrell Centennial Chair of Engineering at The University of Texas, Austin.
[2] N. Menyuk and J. B. Goodenough, J. Appl. Phys., vol. 26, p. 8, 1955. The discovery of high-critical-temperature superconductivity by Bednorz and
[3] J. B. Goodenough, Phys. Rev., vol. 95, p. 917, 1954. Müller brought him back to his study of the transition from localized to itin-
[4] H. J. Williams, W. Shockley, and C. Kittel, Phys. Rev., vol. 80, p. 1090, erant electronic behavior in transition-metal oxides. He has since demonstrated,
1950. with Jianshi Zhou, that the first-order character of this transition leads to phase
[5] F. G. Brockman, P. H. Dowling, and W. G. Steneck, Phys. Rev., vol. 77, and bond-length fluctuations that give rise to a variety of unusual physical prop-
p. 85, 1950. erties. He also continued with the development of energy materials. These in-
[6] C. G. Koops, Phys. Rev., vol. 83, p. 121, 1951. clude the olivine Li FePO cathode material under commercial development
[7] L. Landau and E. Lifshitz, Physik Z. Sowjetunion, vol. 8, p. 153, 1935. for Li-ion batteries as well as solid oxide fuel cells based on Sr- and Mg-doped
[8] N. Bloembergen, Phys. Rev., vol. 78, p. 572, 1950. LaGaO as the electrolyte. The arrival of his 80th birthday has not diminished
[9] F. Bloch, Phys. Rev., vol. 70, p. 460, 1946. his enthusiasm over the design and discovery of new properties and uses of the
[10] R. Becker, J. Phys. Radium, vol. 12, p. 332, 1951. transition-metal oxides.
[11] G. T. Rado, R. W. Wright, W. H. Emerson, and A. Terris, Phys. Rev., vol. Prof. Goodenough is a member of the National Academy of Engineering and
88, p. 909, 1952. a Fellow of the American Physical Society. He is a member of the American
[12] D. M. Grimes, “Reversible susceptibility in ferromagnets,” Electronic Association for the Advancement of Science, the American Chemical Society,
Defense Group, Dept. Electrical Engineering, Univ. Michigan, Tech. L’Academie des Sciences de L’Institut de France, the U.K. Royal Society of
Rep. 64, Apr. 1956. See also D. M. Grimes and D. W. Martin, Phys. Chemistry, and the Physical Society of Japan. He is a foreign associate of the In-
Rev., vol. 96, p. 889, 1954. dian Academy of Science. He was awarded the 2001 Japan Prize by the Science
[13] J. J. Went and H. P. J. Wijn, Phys. Rev., vol. 82, p. 269, 1951; Physica, and Technology Foundation of Japan for the discovery of environmentally be-
vol. 17, p. 976, 1951. nign electrode materials for high-energy-density rechargeable lithium batteries.

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on September 15,2021 at 19:35:39 UTC from IEEE Xplore. Restrictions apply.