Magnetism in Solids: Hysteresis

DS Schmool and D Markó, Université de Versailles/Saint-Quentin, Université Paris-Saclay, Versailles, France

r 2018 Elsevier Inc. All rights reserved.

1 Introduction 1
2 Theory of Magnetic Hysteresis 3
2.1 Models of Hysteresis 3
2.1.1 Stoner – Wohlfarth model 4
2.1.2 The Preisach model 5
2.1.3 Jiles – Atherton model 8
2.1.4 Globus–Guyot model 9
2.2 Magnetic Domains, Domain Walls and Hysteresis 12
3 Experimental Methods to Measure Magnetic Hysteresis 14
3.1 B-H Looper 14
3.2 Superconducting Quantum Interference Device Magnetometer 15
3.3 Vibrating Sample Magnetometer 16
3.4 Alternating Gradient Force Magnetometer 17
3.5 Magneto-Optical Kerr Magnetometer 17
4 Soft and Hard Magnetic Materials 18
4.1 Soft Magnetic Materials 18
4.1.1 Iron 19
4.1.2 Silicon steels 19
4.1.3 Ni–Fe alloys 19
4.1.4 Co–Fe alloys 19
4.1.5 Amorphous alloys 20
4.1.6 Soft ferrites 20
4.1.7 Rare-earth iron garnets 21
4.2 Hard Magnetic Materials 21
4.2.1 Martensitic steels 21
4.2.2 Precipitation-hardened alloys 21
4.2.3 Superstructure alloys 21
4.2.4 Ceramic hard ferrites 21
4.2.5 Rare-earth element alloys of SmCo- and NdFeB-type 22
5 Applications and Practical Aspects of Hysteresis 22
5.1 Applications of Soft Magnetic Materials 22
5.1.1 Static applications 22
5.1.2 Low frequency applications 22
5.1.3 High-frequency applications 23
5.2 Applications of Hard Magnetic Materials 23
5.2.1 Static applications in uniform magnetic fields 23
5.2.2 Static applications in non-uniform magnetic fields 23
5.2.3 Dynamic applications 23
References 24

1 Introduction

Magnetic phenomena associated with solids have been known for over two millennia; for example, lodestone (which is better
known as magnetite, Fe3O4) appeared in Greek and Chinese writings many centuries before the Christian era. Though its physical
origin was unknown at the time, magnetism and magnetic materials continued to attract much attention due to their “mysterious”
physical properties. The origin of the name magnet derives from “the stone of Magnesia”, a Greek town and province in Asia Minor.
One of the first true scientific works dedicated to magnetism appeared in 1600 with the publication of De Magnete by William
Gilbert. Research in magnetics in the seventeenth and eighteenth centuries was dominated by the military and in particular by the
British Navy. While much progress was made during this time, it was with the developments concerning the unification of
electricity and magnetism in the 19th century that enormous strides were made. In 1820, Oersted discovered that an electric
current could generate a magnetic field. This was followed by the production of the first electromagnet in 1825 by Sturgeon.

Reference Module in Materials Science and Materials Engineering doi:10.1016/B978-0-12-803581-8.11413-4 1

2 Magnetism in Solids: Hysteresis

Warburg produced the first measurement of the hysteresis loop for iron in 1880. While Faraday, Gauss and Maxwell etc. made
much theoretical progress in the mid to late 1800s, however, it was not until the 20th century that the foundations were laid for a
more comprehensive physical description of magnetic materials. Curie and Weiss developed early models of the phenomenon of
spontaneous magnetisation and its variation with temperature. Indeed the critical temperature for the transition from ordered
(ferromagnetic) state to the disordered (paramagnetic) state is known as the Curie temperature, TC. Pierre Weiss postulated the
existence of magnetic domains in order to explain the magnetisation of a magnetic material from its virgin state with a zero net
magnetisation. The regions between these domains of spontaneous magnetisation are referred to as domain walls and were
described in detail by Néel, Bloch and Landau.
Of the magnetic properties associated with magnetic materials, magnetic hysteresis is especially important. Indeed, it is this
property which allows ferromagnetic (and ferrimagnetic) materials to retain a magnetised state, or remanence, in the absence of an
applied magnetic field. This property alone permits the many applications of magnetic materials, such as the magnetic compass,
electric motors and magnetic recording. Magnetic hysteresis is most commonly visualized with the magnetisation – applied
magnetic field curve, as illustrated in Fig. 1. In this figure we illustrate the main features of the curve, which schematically traces the
magnetisation, M, as a function of the applied magnetic field, Happlied; the saturation magnetisation, MS; the remnant magneti-
sation (at Happlied ¼ 0), MR; the saturation field (the applied field for which the magnetisation is at saturation, M¼ MS); the coercive
field, HC (which is the applied magnetic field necessary to reduce the magnetisation state to zero, M¼ 0). In this representation we
assume that the sample is taken to its saturated state, i.e., the largest value of the magnetisation for the sample. This is referred to as
a major loop. Minor loops occur when the magnetic field is reduced before the sample is taken to the saturated state, or when the
cycle is made between limits that do not include both magnetic saturation points.
Of the magnetic properties encountered in condensed matter physics, diamagnetism and paramagnetism are the most com-
mon. In fact, a vast majority of elements and compounds fall into the categories of paramagnetism and diamagnetism, which
display weak magnetic susceptibilities, as determined from the relationship between the magnetisation, M, induced by an applied
magnetic field, H:
M ¼ χH ð1Þ
For weak applied fields this relation is a good approximation to a linear isotropic homogeneous response for paramagnetic and
diamagnetic materials. In the former the magnetic susceptibility is positive and weak, while for diamagnetic systems it is negative
and weak. A more sophisticated model for paramagnetic materials was postulated by Langevin in 1905, who developed a semi-
classical model for a system of non-interacting spins. Here the magnetisation follows the so-called Langevin function, which can
be expressed by:
1
LðaÞ ¼ coth a  ð2Þ
a
with
mB
a¼
kB T
where m is the magnetic moment of a single ion, B is the applied magnetic induction (The relationship between the magnetic
induction, B, magnetic field strength, H, and the magnetisation, M, is expressed as: B¼ m0(H þ M)), T the temperature and kB the
Boltzmann constant. The magnetisation follows the relation:
M ¼ MS LðaÞ ð3Þ
in which the saturation magnetisation is given by MS ¼ nm, where n is the total number of ions per unit volume in the system.

Fig. 1 Schematic illustration of a magnetic hysteresis loop indicating the magnetisation as a function of the applied static external field.
 Magnetism in Solids: Hysteresis 3

The quantum mechanical approach to this problem allows for the quantisation of the spin of the ions and leads to an
expression in which the Langevin function is replaced by the Brillouin function, which takes the form:

  
2J þ 1 ð2J þ 1Þy 1 y
BJ ðyÞ ¼ coth  coth ð4Þ
2J 2J 2J 2J
with y ¼ gmBJB/kBT, where mB is the Bohr magneton, J is the spin number, which defines the quantisation of the spin. It is noted that
for a non-quantised, classical system J¼ 1 and the Brillouin function reduces to the Langevin function.
One of the properties observed from both the classical and quantum models for paramagnetism is that the magnetisation
reduces to zero for no applied field and there will be no hysteretic behaviour in the system. Indeed for non-interacting spin systems
there will be no hysteresis. It is only when an interaction between spins exists that any cooperative effects are manifest, i.e., via
some form of exchange interaction, and hysteretic effects become observable. There are many forms of interaction between spins,
the nature of which depends on the material in question. For example, in metallic systems, such as Fe and Ni, direct exchange
coupling between neighbouring spins arises from an overlap of electronic wave functions. However, more complex forms of
exchange occur in oxide systems in which the coupling between spins occurs via intervening oxygen atoms typically arising in
antiparallel alignments. This form of indirect coupling is referred to as super-exchange and is common in antiferromagnetic and
ferrimagnetic systems. Only the latter exhibits a spontaneously measurable magnetisation and magnetic hysteresis. Anti-
ferromagnetic systems, such as CoO and MnFe, while being fully magnetically ordered, have no net magnetisation, since oppo-
sitely aligned magnetic sublattices are fully compensated. In ferrimagnetic systems, the differing populations of oppositely aligned
magnetic sublattices mean that a net moment per unit cell is non-zero, or for spins originating from different atomic/ionic species
for the different sublattices, which do not fully cancel. The measured magnetic properties of these materials resemble those of
ferromagnetic systems, including spontaneous magnetisation, magnetic domains and of course hysteretic behaviour. Temperature
dependent magnetisation can be rather different however due to differing temperature dependencies of the magnetic sublattices,
which can also give rise to a compensation temperature at which the global magnetisation disappears. The ordering temperature in
these materials is known as the Néel temperature, TN, in honour of Néel’s work on the development of the theory of ferrites.
Many applications of magnetic materials rely directly on the hysteretic behaviour for their specific functions. For example,
magnetic recording media require a reasonably strong coercive field to provide sufficient stability to store information on magnetic
bits, however this should not be too elevated since the writing process would become problematic. In this short contribution we
aim to provide the reader with a basic introduction to the phenomenon of magnetic hysteresis where we outline the physical
origins and a theoretical description. We also give an outline of the principal experimental methods used to characterise magnetic
hysteresis in solids. Finally we will discuss hard and soft magnetic materials and some applications of magnetic materials that rely
on specific hysteretic behaviour.

2 Theory of Magnetic Hysteresis

There are many models that have been developed over the past century which explain the phenomenon of magnetic hysteresis. In
the following sections we will outline some of the principal elements of the most common of these models, which will illustrate
how hysteresis can arise from the fundamental properties of magnetic materials.

2.1 Models of Hysteresis

Historically there have been many attempts to explain the phenomenon of hysteresis and in general the variation of the mag-
netisation of a ferromagnetic material as a function of the applied magnetic field. In early approaches the explanations can broadly
be categorized into two main groups; the first was based on a frictional type force, while the second considered the strong mutual
interactions between the magnetic moments of individual atoms. The original idea for the existence of a restoring force to
maintain moments in their initial unsaturated state was postulated by Weber (1852), which while possibly providing a
mechanism for the initial magnetisation curve would not explain the remnant state after the removal of a magnetic field.
Wiedemann (1886) considered the frictional resistance to rotation of magnetic moments to account for the effects of hysteresis. It
was Maxwell (1873) who appears to have been the first to consider the effects of mutual interactions between arrays of magnetic
moments to explain the effects of hysteresis. This was further developed by Ewing (1890) who believed that such interactions
between magnetic moments could fully account for the observed magnetic hysteresis in magnetic materials.
As will be seen in the following section, strong interactions between magnetic moments, which maintain them in a ferro-
magnetic state, can indeed give rise to hysteretic behaviour. However, magnetic reversal can be rather abrupt and is strongly
anisotropic, a fact not incorporated into the early attempts at an explanation for hysteresis phenomena. In the frictional force
based models, smoother changes to the magnetisation were envisaged and can be related to the effect of pinning of domain walls
by defect sites within the solid. This produces a force opposing the changes to the magnetisation that would otherwise occur under
the influence of an applied magnetic field.
Progress in the modelling of hysteresis behaviour was hampered by their restriction to specific ranges of applicability. For
example, Cullity (1972) finds only three cases where algebraic expressions were obtained for magnetisation curves. These are for
high field magnetisation curves in single crystals, Williams (1937), in polycrystalline materials, where the law of approach to
4 Magnetism in Solids: Hysteresis

saturation governs the variation of the magnetisation, Chikazumi (1964) and finally for low-field magnetisation curves and
hysteresis loops for polycrystals that display Rayleigh loops, Rayleigh (1887).
In the following we review some of the principal models that have been developed for a more in-depth critique of the basic
principles of the approach to the simulation and modelling of hysteresis phenomena. In each case there exist variants and
modified versions which build on the pioneering ideas of each model. We will mainly consider these basic principles.

2.1.1 Stoner – Wohlfarth model

The Stoner – Wohlfarth (SW) model, which dates back to 1948, is a simple attempt to evaluate the hysteretic behaviour of
ferromagnetic systems (Stoner and Wohlfarth, 1948). It has subsequently become a very important first approximation method for
the study of ferromagnetic nanoparticle systems. In the SW model it is assumed that the particle or magnetic body has a single
magnetic domain with uniaxial anisotropy. This assumption is based on the original premise of the model that the magnetisation
remains constant throughout the sample. This has the consequence of maintaining the total exchange energy constant during the
magnetisation reversal process. Such behaviour is termed uniform or coherent rotation and is also appropriate for weakly or non-
interacting magnetic nanoparticle assemblies. The model breaks down for very small nanoparticles, where the surface anisotropy is
different from the bulk and non-collinear spin configurations are predominant.
The magnetic anisotropy of the particle, which we will consider as uniaxial, can be of magnetocrystalline or magnetostatic
origin, since it has no important bearing on the outcome of the model. We use an effective anisotropy for this reason. As such, we
can express the free energy density of the system as:

ESW ¼ Keff sin2 y  m0 MS Hcosðy  fÞ ð5Þ

where angles are defined in Fig. 2. For our definition we use Keff40, so that an energy minimum occurs for y ¼ 0, i.e., with the
magnetisation along the easy axis. The effective anisotropy
 can, for a single small magnetic nanoparticle, be expressed as (Skomski,
2008; Tannous and Gieraltowski, 2008): Keff ¼ K1 þ m0 N>  N8 M2S =2. The constant K1 refers to the uniaxial magnetocrystalline
anisotropy constant which aligns the individual atomic magnetic moments along its easy-axis, the second term originates from the
magnetostatic anisotropy, due to the shape of the particle, where N> and N|| are, respectively, the demagnetising factors per-
pendicular and parallel to the easy-axis. For this example we consider the case where these contributions are aligned to minimise
the energy density along the same axis, as illustrated in Fig. 2. For further discussion of such considerations, see Skomski (2008).
Other configurations are possible as is the possibility of other forms of magnetocrystalline anisotropy. The example we give here is
the simplest model possible and allows us to demonstrate the appearance of magnetic hysteresis.
In Fig. 3 we show the energy landscapes for some applied magnetic field values. The energy minima denote the orientations for
the stable configuration of the magnetisation vector. For zero and low applied (reversal) magnetic fields, there are two
stable minima at y ¼ 0 and y ¼ 7p. As the reversal field increases (HoHK where HK ¼ 2Keff/m0MS is called the anisotropy field), one
of the minima will be deeper than the other; this defines the most stable configuration of the system. We note that the energy
barrier between the global minimum and the local minimum (meta-stable state) reduces. At an applied field of HK the local
minima vanishes, this corresponds to the coercive field of the system. If the magnetisation happens to be “trapped” in this local
minima then at the point where H ¼ HK, it will switch or reverse its direction. Clearly, depending on the sample history, i.e., the
state of the sample when we start applying the magnetic field, the magnetisation remains in its original state or will reverse. This
signifies the origin of hysteresis in its simplest form and arises due to the fact that the spins act as a macrospin state and the
magnetic anisotropy produces a coercive field. This is a consequence of the exchange interaction between spins, which is direc-
tional and leads to magnetocrystalline anisotropy in solids.
The equilibrium orientation of the system is defined by the configuration which gives a minimum energy, indicated as y¼ y*.
The condition for minimising the free energy is given by:
   
∂ESW ∂2 ESW
¼ 0 and 40 ð6Þ
∂y y¼y ∂y2 y¼y

Fig. 2 Schematic diagram for a magnetic single domain with uniaxial anisotropy. Relative orientations of the applied magnetic field, H, and the
magnetisation, M, are shown with respect to the easy axis.
 Magnetism in Solids: Hysteresis 5

Fig. 3 Energy landscape for the case of the (a) aligned and (b) non-aligned Stoner–Wohlfarth model. In the aligned case, the landscape exhibits
two minima at low fields, one at y¼0 (↑) and the other at y¼1801(↓). As the field increases and approaches the coercive field, the local minima
become very weak, and finally disappear for H4HK and only the (↑) minima remain. In this case the position of the minima does not change. For
the case where the applied field is not aligned with the easy axis, as the field increases the positions of the energy minima change and tend
towards the direction of the applied field. As H approaches HK, the shallow local minimum gradually disappears and the system will stabilize in the
deeper energy minimum.

From this we obtain:

½sinycosy  hsinðy  fÞy ¼ y ¼ 0 ð7Þ
and
½cos2y þ hcosðy  fÞy ¼ y  0 ð8Þ
where we have used the following reduced quantities: h¼ H/HK and m ¼ M/MS. The calculations of the magnetisation state are
made by varying the strength and direction of the applied magnetic field, where the magnetisation will respond by aligning in that
field due to the relative strength and orientation of the magnetic anisotropy as well as that of any applied magnetic field. It is
customary to decompose the magnetisation into parallel and perpendicular components, which we express as: m|| ¼ cos(y–j) and
m> ¼ sin(y–j). The variation of these transverse and longitudinal components of the magnetisation is illustrated in Fig. 4 for
different orientations of the applied magnetic field.
In the Stoner–Wohlfarth model the hysteretic magnetic behaviour arises from the magnetic anisotropy, where the coercive field,
HC, is related to the anisotropy field. The model is also used to show the magnetisation reversal or switching behaviour in single
domain particle systems. This is achieved by evaluating the boundaries of the hysteretic behaviour via the first and second
derivatives of the free energy. In fact the entire model is based on the variation of the energy landscape, in which the minima
determine the orientation of the magnetisation, which is affected by the strength and orientation of an applied magnetic field with
respect to those of the magnetic anisotropies, as illustrated in Fig. 3. The switching field dependence can be expressed in terms of
the anisotropy and orientation of the applied magnetic field in the form:
HK
HSW ¼ 3=2
ð9Þ
ðsin2=3 f þ cos2=3 fÞ
This is displayed for the applied fields in the parallel and perpendicular directions. The form of the switching field curve is
commonly known as the Stoner–Wohlfarth astroid. This is illustrated in Fig. 5, where we show the theoretical switching
field curve.
While the Stoner–Wohlfarth model is a good first approximation, it does assume that the magnetisation is always homo-
geneous. This situation is generally favoured only in small magnets, which have the form of an ellipsoid of rotation. This means
that the particle acts as a single macrospin and the exchange energy is a dominant energy in the system. Deviations from this
model can occur for a number of reasons, such as edge and surface anisotropies, which can cause some form of non-collinear
alignment of the magnetisation at the surface, meaning ∇M¼ 0. Defects and other magnetic inhomogeneities will also bring about
deviations from the model and add further contributions to the coercive field, HC, which in the Stoner–Wohlfarth model arise only
from the anisotropy. Non-coherent reversal of the magnetisation can also occur in elongated single domain particles and therefore
is not well described by the SW model. Such processes are usually described by a curling of the magnetic moments in the particle.

2.1.2 The Preisach model

The Preisach model of hysteresis was an early theoretical model of hysteresis dating from 1935. This model can be viewed as a
phenomenological representation of hysteresis and has a number of variants that have been developed to overcome some of the
6 Magnetism in Solids: Hysteresis

Fig. 4 (a) Longitudinal, m||, and (b) transverse, m>, components of the magnetisation. The angles refer to the direction of the applied magnetic
field for f ¼01, 301, 601 and 901. Reproduced from Tannous, C., Gieraltowski, J., 2008. “The Stoner – Wohlfarth model of ferromagnetism”. Eur.
J. Phys. 29, 475–487.

Fig. 5 Switching fields for a single domain particle with uniaxial anisotropy. Due to the shape of the curve, this is popularly known as the
Stoner–Wohlfarth astroid. Inside the astroid, the particle is capable of switching, while outside it is not. Reproduced from Tannous, C.,
Gieraltowski, J., 2008. “The Stoner – Wohlfarth model of ferromagnetism”. Eur. J. Phys. 29, 475–487.

shortcomings of the earlier models. In the following we shall present only the basics. Essentially the Preisach model uses the
hypothesis that the free-energy profile of a magnetic system is characterised by multiple minima and meta-stable states that can be
decomposed into a set of many elementary bi-stable contributions (Benabou et al., 2003). The bi-stable units or hysterons, which
we denote as “ þ ” or “  ”, are characterised by two reduced fields; hC – a local coercive field, and hU – an interaction field, these are
illustrated in Fig. 6(a). A switching field couple, which can be expressed by the plane (ha,hb), where ha ¼ hU þ hC and hb ¼ hU–hC,
characterises a bistable unit and must conform to certain conditions. Representing the saturation magnetisation and saturation
magnetic fields as MS and HSat, respectively, when H4HSat, all bistable units are positive and the magnetisation will be M ¼ MS. At
the other end of the hysteresis loop, if H’HSat, all bistable units will be negative and M¼  MS. The above assumptions lead to
the following general conditions for the couple, (ha,hb):
ha rHSat

hb  HSat
 Magnetism in Solids: Hysteresis 7

Fig. 6 (a) Elementary Preisach loop, with a local reduced coercive field, hC, and an interaction field, hU. (b) The Preisach plane (ha,hb) showing
the triangle defined by the saturation fields.

Fig. 7 The triangle defining the Preisach plane with the separation of the surfaces for the (  ) and ( þ ) states and the boundary between them
as given by the line L(t).

Given that hysteresis is an energetically dissipative phenomenon, we must have haZhb. These three conditions allow us to
define a triangle in the Preisach plane, as illustrated in Fig. 6(b). Each couple for a particular bistable unit must belong to
this plane.
The magnetisation process is determined by the Preisach density function, p(ha,hb) of the material in question. Mayergoyz used
a geometrical interpretation of the magnetisation process using a one-to-one correspondence between an operator ^g a;b and the
points (ha,hb) in the Preisach plane for the half-plane haZhb, (Mayergoyz, 1986a,b). The specific triangle, at any instant in time and
depending on the sample history, can be split into two distinct regions: S þ (t), which consists of points (ha,hb) for which ^ g a;b ¼ 1,
and S  (t), which consists of points (ha,hb) for which ^
g a;b ¼  1. This is schematically illustrated in Fig. 7. The boundary between
the S þ and S regions, defined as L(t), is typically a stepped function whose vertices have ha and hb coordinates coinciding with
local maxima and minima of input at the previous instant of time. The line L(t) commences at the line ha ¼ hb and moves when the
input changes. As such L(t) represents the history of the sample. For example, if the magnetic field is decreased, the horizontal part
of the line L(t) moves downwards, while if the field increases, the vertical part of the line L(t) moves to the right. Using this logic,
the magnetisation state of the sample is evaluated via the following expression (see for example, Hejda and Zelinka, 1990;
Mayergoyz, 2008a, b):
M ¼ MS ∬ha hb pðha ; hb Þ^
g a;b dha dhb ð10Þ
Eq. (10) can be rearranged in the following manner:
M ¼ MSat ð11Þ
Therefore the magnetic state of the system is explicitly related to the broken line L(t). As stated above, this is determined
by the extremal values of input at previous instants of time. Consequently, the past extremal values of input shape the
8 Magnetism in Solids: Hysteresis

Fig. 8 Example of a magnetisation process in which the sample is taken from the virgin state to saturation in the positive sense, then the field is
cycled from HSat to H1, H2 and H3. Reprinted from Benabou, A., Clénet, S., Piriou, F., 2003. “Comparison of Preisach and Jiles-Athertonmodels to
take into account hysteresis phenomenon for finite element analysis”. J. Magn. Magn. Mater. 261, 139–160.

interface and thus leave a mark on the sample history, i.e., the nonlocal memory. These input values can be erased in what is
referred to as the wiping-out property: Each local maximum wipes out the vertices whose ha coordinates are below this
maximum and each local minimum wipes out the vertices whose hb coordinates are above this minimum. This can be
incorporated into the memory vector h, which includes extrema, Hi, with coordinates that adhere to the following condi-
tions (Benabou et al., 2003):
H0 ¼ 0; h ¼ fH0 ; H1 ; H2 ; H3 ; …; Hn g ð12Þ
for i ¼ 1,…,n–1 and di ¼ Hi–Hi–1. Where we have:
di :diþ1 o0
jdiþ1 jojdi j ð13Þ

In Eq. (12) Hn, the last component, is the current value of the magnetic field. From these relations the memory vector is
constructed. An example is illustrated in Fig. 8 for the magnetic state defined by the memory vector h¼ {0, þ HSat,H1,H2,H3}.
A further property of the model, known as the congruency property, states that all hysteresis loops (major or minor) corre-
sponding to the same extremal values of input are congruent. This means that given a set of input variations within the same range,
the output increments to the magnetisation must be the same.
The missing element is the knowledge of the Preisach density function. From this it is possible to determine the form of
the hysteresis loop and more generally the variation M(H) from any starting position within the hysteresis loop. This can be
determined experimentally, as outlined by Mayergoyz (1986a,b), see also Hejda and Zelinka (1990). In general there is a
good agreement between experiment and theory and modelling using the Preisach formalism provides a good insight into the
magnetisation process. However, critics of the model state that the Preisach model, while being widely applicable on the
macroscopic scale, does not always give a physical picture and only really allows a modelling of irreversible processes.
Reversible processes must be introduced arbitrarily (Jiles, 2002). The relation between microstructure and magnetic prop-
erties is an essential aspect of understanding the magnetisation process. Often this can be investigated using computational
micromagnetics, as illustrated by Dupré et al. (1999, 2002). A more generalized Preisach model includes an extension to the
classical approach and describes the vector nature of the magnetisation process (Mayergoyz and Friedman, 1988; Hejda and
Zelinka, 1990; Bertotti, 1991).

2.1.3 Jiles – Atherton model

The Jiles – Atherton or J – A model in its original presentation (Jiles and Atherton, 1986) aims at expressing the relation between
the magnetisation of a ferromagnetic material as a function of the applied magnetic field. As a starting point, the model is based on
the anhysteretic behaviour, Man(H), which can be expressed in the form of a modified Langevin function:

  
He a
Man ðHÞ ¼ MSat coth  ð14Þ
a He
where He ¼ H þ aM is the effective field experienced by the magnetic moments within the ferromagnetic domains. H is the applied
magnetic field and a is a mean field parameter analogous to the Weiss mean field. The parameter a has the dimensions of magnetic
field and characterises the shape of the anhysteretic magnetisation. Typically this expression represents the effects of moment
rotation within magnetic domains but does not account for losses due to domain wall motions. Jiles and Atherton (1986) note
that for elevated values of the mean field constant, a, which translates as strong interactions between magnetic moment, hysteretic
behaviour is observed, however, these values are unrealistic in practical terms.
Central to the J – A model is the consideration of the domain wall displacement through pinning sites, such as impurities and
defects, where energy will be dissipated. The effects of both rigid, planar walls and flexible walls gives rise to irreversible and
reversible magnetisation processes. Reversible processes can be thought of as the effect of wall bending between anchor points due
to pinning sites. Irreversible processes are more associated with rigid wall motion. Energy losses due to hysteresis arise from the
 Magnetism in Solids: Hysteresis 9

irreversible changes, which are proportional to the average pinning energy:

〈ϵp 〉
〈ϵpin 〉 ¼ ð1  cosWÞ ð15Þ
2
where 〈Ap〉 denotes the pinning energy of a site for 1801 domain walls and W is the relative orientation of the magnetisation on
either side of
R Mthe domain wall. The pinning energy can be related to the change of magnetisation via (Jiles and Atherton, 1986):
Epin ðMÞ ¼ k 0 dM, where k¼ n〈Ap〉/2m, here n denotes the average density of the pinning sites in the solid and m the magnetic
moment per unit volume. Assuming a uniform distribution of pinning sites, where each is considered as having the mean pinning
energy, Rthe work done against pinning by the applied field will be proportional to the change in magnetisation. The magnetisation
R
energy MdBe, with Be ¼ m0He, will Rbe the difference in energy between that obtained in the ideal lossless case, i.e., Man(He)dBe,
minus the loss due to hysteresis, k dM:
Z Z Z  
dM
MdBe ¼ Man ðHe ÞdBe  k dBe ð16Þ
dBe
From which it is a simple matter to obtain:
 
dM
M ¼ Man ðHe Þ  δk ð17Þ
dBe
The parameter δ has a value of þ 1 for dH/dt 4 0 and  1 for dH/dt o 0, ensuring that the pinning opposes changes to the
magnetisation. A consideration of the magnetic energy under the assumption of a uniform distribution of pinning sites leads to
conclusion that the magnetisation energy will be the difference between the energy which would be obtained in the anhysteretic
case minus the energy due to the losses induced by domain wall motion. It is then possible to obtain the differential susceptibility
for the irreversible magnetisation, Mirr, in the form:
dMirr Man  Mirr
¼ ð18Þ
dH k=δm0  aðMan  Mirr Þ
The reversible contribution to the magnetisation process is linked with the anhysteretic and irreversible magnetisation via the
linear relation:
Mrev ¼ cðMan  Mirr Þ ð19Þ
where the reversibility coefficient, c, determines the proportion of the magnetisation that is reversible and can take values between
zero and unity. Given that the total magnetisation can be expressed as the sum of reversible and irreversible components;
M¼ Mrev þ Mirr, and taking into account Eq. (19), we can express the magnetisation as (1 þ c)M¼ Mirr þ cMan and the total
differential susceptibility takes the form:
dM 1 Man  Mirr c dMan
¼ þ ð20Þ
dH ð1 þ cÞ k=δm0  aðMan  Mirr Þ ð1 þ cÞ dH
This equation can then be used to derive the magnetisation as a function of the magnetic field. An alternative form of this
equation can be expressed as (Tumanski, 2011):
 
dM Man  Mirr dMan dMirr
¼ þc  ð21Þ
dH k=δm0  aðMan  Mirr Þ dH dH
The J–A model has a number of parameters that require adjusting to obtain the correct shape for the hysteresis loop, these
include k,c,a and a, which in general can be determined from the experimental measurement.
Optimisation of the fitting parameters can be computed and once correctly determined provide the model with excellent
agreement to experimentally determined M(H) measurements. An example is illustrated in Fig. 9. The importance of the model is
that the fitting parameters have a specific physical meaning and that it provides a useful tool for discussing the physical
mechanisms of the magnetisation process.
It is worth noting that under certain conditions (Pasquale et al., 1998), the Preisach and the J–A models are equivalent. In
particular, if the Preisach distribution function decreases exponentially with hc, according to the coercive force of the Preisach
formalism, then the rate of change of the magnetisation with the applied magnetic field in the two models will be equivalent.
Modified forms of the Jiles–Atherton model have been developed; for example, Hamimid et al. (2010) introduce the magneti-
sation, M, in place of the irreversible magnetisation, Mirr, into the effective magnetic field, He. The authors demonstrate that this
allows a better fitting of their experimental magnetic hysteresis loops.

2.1.4 Globus–Guyot model

The original model ascribed to Globus considers the initial magnetisation of soft magnetic materials, which is due to the motion
of a domain wall, as opposed to the domain rotation in hard magnetic materials at high magnetic fields. Domain wall motion can
occur via the wall bowing mechanism or wall displacement. The bowing of domain walls is considered in an analogous manner to
the elastic deformation of a membrane, which is fixed, or pinned, by a grain boundary. The domain wall is subject to the force
created by an applied magnetic field which produces domain growth of those domains with a favourable direction of magneti-
sation with respect to the orientation of the applied magnetic field. Such motion, since it is only due to bowing, is reversible. This
10 Magnetism in Solids: Hysteresis

Fig. 9 (a) Experimental magnetic hysteresis loops for a sample of Fe-C 0.06 wt. %. (b) Theoretical magnetic hysteresis loops for the J–A model
with MSat ¼1.6  106A/m; a ¼1100 A/m; a ¼1.6  103 and k¼400 as determined from Eq. (18). The reversible portions were determined using
Eq. (20). Reprinted from Jiles, D.C., Atherton, D.L., 1986. “Theory of ferromagnetic hysteresis”. J. Magn. Magn. Mater. 61, 48–60.

means that if the magnetic field is reduced the bowing will correspondingly reduce, with the domain wall following the same path
that it took with increasing field. Should the magnetic field be sufficient to cause the depinning of the domain wall from the
pinning site, then domain displacement will occur and the process is irreversible, with energy being dissipated and lost to the
domain wall. Initial work on pinning mechanisms was proposed by Becker and Döring (1939) and Kersten (1943).
As stated above, the initial model (Globus, 1962) considered the reversible portion of the motion of domain walls in
ferrimagnetic materials, discussing the relation between the initial susceptibility and the domain wall continuity, which is
dependent on sample crystallinity (grain size) as well as the relative orientations of the crystalline axis with respect to the domain
wall direction. Globus and Duplex (1969) also considered the effect of temperature, with wall continuity reducing with tem-
perature. Experimental results were well explained using the model, which shows that the diameter of the wall inside the grains
controls the susceptibility of the material (Globus and Duplex, 1970). The authors also indicate that wall motion dominates
domain rotation, which are dependent on anisotropy but not grain size, Dm. Furthermore, the model predicts a magnetisation that
is proportional to the applied field and the grain size. Beyond a critical value of the applied magnetic field, HCp1/Dm, the wall
becomes detached from the pinning site and moves. This leads to a reduction in the size of the domain wall area (Globus and
Guyot, 1972). For fields above this critical value, the position of the domain wall is determined by a balance between forces on the
wall produced by the applied field, which are proportional to d2, and the viscous forces proportional to d, where d is the wall
diameter (for a spherical grain) and is proportional to 1/H (Globus et al., 1971). Globally the model presents a schematic
hysteresis loop as illustrated in Fig. 10, where a domain wall is considered under the differing applied fields and directions and
responds accordingly. Further considerations for the demagnetising field produce a tilting of the loop.
The concept of domain wall bulging and displacement are in general accord with the Jiles–Atherton model, though the latter
consider that this is too restricted and that domain wall pinning also occurs on other inhomogeneities within the grains, such as tangles
of dislocations, regions of inhomogeneous strain and other precipitates or non-magnetic inclusions (Jiles and Atherton, 1986).
The model was extended to include dissipative processes resulting from the translation of domain walls at more elevated values
of applied field (Globus and Guyot, 1972; Guyot and Globus, 1973). For a polycrystalline material, the global anisotropy is
considered to be composed of the magneto-crystalline contribution, K1, as well as a magnetoelastic term, such that K ¼ K1 þ lSsn,
where sn denotes the natural stresses within the material and depend on the sample composition. This extension of the model also
considers the energy loss mechanism during wall displacement, which contributes to the irreversible mechanisms of the M(H)
loop. A simple model is presented of the displacement of the domain wall in a spherical grain, of radius R and saturation
magnetisation MS. In the remnant state, where the domain wall is flat, i.e., a circular membrane, the magnetisation will take a
value given by:
dV
MR ¼ 2MS ð22Þ
V
where dV is the volume due to the wall displacement and V is the volume of the grain. With respect to Fig. 11, we can envisage the
remnant magnetisation as the difference of the magnetisations of the two domains separated by a domain wall, of negligible
thickness. The problem is reduced to a question of geometry and allows the remnant magnetisation to be expressed as
MR ¼ MS(V↑  V↓). However, V↑ ¼ V  V↓, from which, using the volume formulae; V¼ 4pR3/3 and V↓ ¼ ph2(3R  h)/3, with
 Magnetism in Solids: Hysteresis 11

Fig. 10 The Globus–Guyot model for a spherical grain showing the roles of domain wall bulging and displacement. Hatched areas denote the
magnetisation due to the bulging of the domain wall. Reproduced from Globus, A., Guyot, M., 1972. “Wall displacement and bulging in
magnetisationmechanisms of the hysteresis loop”. Phys. Stat. Solidi (b) 52, 427–431.

Fig. 11 Remnant state magnetisation, MR, according to the Globus model, where a single flat domain wall in a spherical grain is displaced from
the centre by a distance a.

h¼ R  a, it is possible to express the remnant magnetisation in the form:


MS a a 3
MR ¼ 3  ð23Þ
2 R R
The pinning force on the domain wall is expressed as F ¼ 2prf, where f is the pinning force per unit length. F is considered
to be a frictional force acting to resist the motion of the domain wall under the action of an applied field. The work done on
the wall during a cycle (considered as the movement of the wall between positions a0 and  a0) is given as (Guyot and
Globus, 1973, 1977):
Z a0 Z a0
Wpin ¼ 4 Fda ¼ 8pf rda ð24Þ
0 0

As the wall displaces there will be a variation of its total energy due to the variation of the wall area, DS. This leads to an energy
loss during the cycle, which is expressed in the form:
 
Wloss ¼ 4ngDS ¼ 4npg R2  ro2 ¼ 4npga20 ð25Þ
where g denotes the wall energy per unit area and n is a constant of proportionality. The total loss energy can then be expressed as:
 Z a0 
W ¼ Wpin þ Wloss ¼ 4p 2f rda þ nga20 ð26Þ
0
12 Magnetism in Solids: Hysteresis

Under the conditions of weak magnetisation the energy per unit volume can then be expressed in the simplified form:
 
3a0  a0 4 MR ng MR
E ¼ 2 2f þ ng ¼ fþ ð27Þ
R R R MS 3 MS
where we have used the substitution: a0/R¼ 2MR/3MS. The model expressed by Eq. (27) is in good agreement with experimental
pffiffiffiffiffiffiffiffiffi
results. Experimental results can then be used to determine the domain wall energy, g ¼ 2 AK1 and f, (Guyot and Globus, 1973,
1977). If the anisotropy constant, K1, is known, then knowledge of the domain wall energy permits the evaluation of the exchange
constant, A.

2.2 Magnetic Domains, Domain Walls and Hysteresis

In the above section we have outlined some of the principal ingredients for the theoretical understanding of magnetic hysteresis. In
this section we will briefly discuss the relation between the hysteresis phenomenon and the existence of magnetic domains,
including domain walls. This will complete some of the elements that were presented in the previous section. This will naturally
take into account some of the energy considerations that we have outlined above.
A magnetic domain wall is a region of transition in which the atomic magnetic moments vary smoothly in orientation
from one magnetic domain to a neighbouring domain. When the magnetic moments of the neighbouring domains are anti-
parallel, the domain wall is referred to as a 1801 wall. There are two principal types of domain wall; the Bloch wall is
characterised by the variation of the orientation of the magnetic moment in a direction perpendicular to the direction of
transition or in the plane of the wall; the Néel wall, on the other hand, concerns a rotation of the moment orientation in the
direction of transition or perpendicular to the domain wall. In either case the domain wall represents an additional energy
contribution due to the excess energy associated with neighbouring magnetic moments which are non-collinear and hence
have higher exchange energies and also a contribution due to the fact that the moments are not aligned with the easy axis as
defined by the magnetocrystalline energy. The energy per unit surface area in say the y–z plane, where the transition direction
is parallel to the x-axis, can be expressed as:
Z 1
gW ¼ ½uexch ðxÞ þ uAn ðxÞdx ð28Þ
1
where the first and seconds are respectively the exchange and magnetic anisotropy energies. The wall energy expressed in Eq.
(28) requires the explicit form of the contributions for the exchange and anisotropy terms, which will depend on the type of
sample and the type of domain wall. By way of example we consider the case for a 1801 Bloch wall for a sample with uniaxial
anisotropy in zero applied magnetic field. In this case we can express the exchange energy in the form (Cullity, 1972):
uexch ¼  2JSi  Sj ¼  2JSi Sj cosW ð29Þ
where J is the exchange integral, Si and Sj are the spin angular momenta for neighbouring atoms i and j, respectively and W
denotes the angle between these neighbouring spins. Allowing for the case where the spin momenta are equivalent,
Si ¼ Sj ¼ S, and using the series expansion for cosW¼ 1  W2/2 þ W4/24  …, we can express the exchange energy as:
uexch ¼ JS2 W2  2JS2 ð30Þ
where we have limited the range of the series taking into account that W is very small. Since the second term has no
dependence on angle, we can neglect it, as it will be that same for regions within a domain as within a wall. It is the first term
which expresses the excess energy due to the existence of the domain wall itself. For a wall thickness of N atoms, which are
separated by a distance a, we can express the excess exchange energy per unit area of the wall:
N
gexch ¼ JS2 W2 ð31Þ
a2
Since we are considering a 1801 domain wall we can write: W ¼ p/N, which yields:
JS2 p2
gexch ¼ ð32Þ
Na2
We now consider the anisotropy contribution to the energy from Eq. (28). In the simplest case we can expect that the excess
energy will be related to the anisotropy constant K times the wall volume, leading to the anisotropy term:
gan ¼ KNa ð33Þ
Summing these contributions and writing the wall width as δ¼ Na, we obtain:
JS2 p2
g¼ þ Kδ ð34Þ
δa
Minimising the wall energy with respect to the wall width allows us to obtain the relation for the wall width in terms of the
exchange stiffness constant; A ¼ JS2p2/a and the anisotropy constant:
rffiffiffiffi
A
δ¼ ð35Þ
K
 Magnetism in Solids: Hysteresis 13

pffiffiffiffiffiffiffiffiffi
A more refined calculation of the domain wall width gives a value which is somewhat larger: δ ¼ p A=K , Skomski (2008).
Substituting the result from Eq. (35) into (34) then yields a total domain wall energy of:
pffiffiffiffiffiffiffi
g ¼ 2 AK ð36Þ
The more refined approach yields a value which is twice this value, Skomski (2008). The competing energy terms illustrate that
the magnetocrystalline anisotropy favours narrow domain walls, since this would align more spins along the easy axis directions
and reduce the excess anisotropy energy. However, narrow walls mean that within the domain walls the magnetisation gradients
would be large and unfavourable for the exchange energy, which is reduced when the angle between neighbouring spins is small. It
is therefore the ratio of these quantities that determines the extent of the transition region between magnetic domains. For
example, the domain wall thickness for Fe is 64 nm with a wall energy of 4.1 mJm2 (Coey, 2009).
The phenomenon of hysteresis is intimately linked with the microscopic processes occurring in a magnetic body and concerns
the manner in which the magnetisation is reversed due to the application of an external magnetic field. This relates not only to the
direction in which the field is applied but also to the specific magnetic state of the sample, which depends on the magnetic history
and the particular configuration /distribution of the magnetic spins. Also of importance will be the purity and crystallinity of the
sample itself. Impure samples and grain boundaries will have a significant influence on the manner in which domain walls are
nucleated and how they propagate with the motion of domain walls. We have already touched on some of these issues in the
preceding sections.
There are several modes of magnetisation reversal, which include coherent rotation, curling and multi-domain reversal. This
latter is associated with hysteresis losses and is most common for magnetic bodies with dimensions greater than the single domain
particles. The critical size can easily be calculated from energy considerations and for a spherical body of radius Rsd can be
expressed as, (Hubert and Schäfer, 1998):
pffiffiffiffiffiffiffi
AK
Rsd E36 ð37Þ
m0 M2s
with values ranging from a few nm in soft magnets to around a micron in hard materials, Skomski (2008). The critical moment of any
reversal process is the nucleation, which for the multi-domain mechanism corresponds to the creation of a reverse domain typically
due to a defect at the surface of the sample. Once nucleation has taken place, the motion of the domain wall under the action of a
reverse applied field requires very little energy and leads to the expansion of the reversal domain unless pinning of the domain wall
occurs. Domain sizes and shapes can be rather difficult to predict, and will depend on a number of factors, such as the exchange
energy, anisotropy, sample dimensions, sample shape and defects. Micromagnetic simulations have been quite successful in many
cases for small magnetic structures, however, larger structures are more difficult to predict (Hubert and Schäfer, 1998).
In general domain walls tend to be flat or planar, due to energy minimisation of magnetostatic and exchange energies.
However, non-planar walls can exist due to pinning of domain walls at inclusions, voids and internal stresses in the sample. Strong
pinning tends to occur on planar defects, especially those with dimensions comparable to the domain wall width, δ. Large planar
defects can be very effective pinning centres since they will tend to reduce the wall area and hence reduce the total energy or excess
energy of the system. Weaker pinning can be expected for smaller defects. Typically a sample will contain a number of defects of
varying sizes and will be distributed throughout the body of the magnetic entity. Indeed, pinning is the principal source of
hysteresis and coercivity in inhomogeneous magnets. We can think of pinning as the domain wall being trapped in a region with
low domain wall energy, and the coercivity is determined by the “depinning” field, which is the magnetic field necessary to push
the wall over the pinning energy barrier.
A simple model of wall pinning considers a planar wall, with position x, in which the energy of the wall depends on position.
The coercivity is evaluated from the total wall energy, which can be expressed in the form, (Skomski, 2008):
EW ðxÞ ¼ ZðxÞ  2m0 MS Hx ð38Þ
where Z(x) depends on the microstructure and is essentially proportional to the local wall potential and the second term is a
Zeeman energy. For a given applied field, H, the wall position can be deduced from the minimisation of the energy, EW(x). The
field can be expressed from this as: H ¼ 2m 1MS dZ
dx : The derivative dZ/dx is a function of position and will have a maximum
0

in the vicinity of a defect. Increasing the field at this point will push the domain wall over the barrier and lead to its depinning. The
depinning field can thus be expressed as:
 
1 dZ
HP ¼ ð39Þ
2m0 MS dx max
The passage of the domain wall through a defect is typically envisaged as the progress of the wall bulging from the defect
position until the field is sufficient to depin it, resulting in a jump of the wall to a new position and returning to its planar
state. This mechanism of domain wall pinning was first presented by Kersten (1943), see also work by Gaunt (1983, 1986).
These wall jumps are associated with the discontinuities in the magnetisation, known as Barkhausen jumps. These jumps
represent the irreversible movements of the domain wall and hence the magnetisation, leading to coercivity and the phe-
nomenon of hysteresis. Indeed the jumps can be heard as Barkhausen noise with clicks and illustrate the dissipation of
energy in the system. This has lead to the study of magneto-acoustic emission in ferromagnetic and ferrimagnetic materials.
In addition to the pinning mechanism, Guyot and Cagan (1991) also suggest that magnetic hysteresis is more fully described
14 Magnetism in Solids: Hysteresis

Fig. 12 Schematic illustration of the initial magnetisation curve and demagnetisation in the first and second quadrants.

by the addition of domain wall creation and annihilation, which was shown to be a dissipative process (Guyot and Globus,
1973, 1977). Guyot and Cagan (1991) suggest that the physical mechanism of magnetic energy conversion into heat arises
from the acoustic emission process.
We can summarise the magnetisation process from the virgin state to saturation and back to zero field and then to the first reverse
coercive field. This allows an overall view of the principal mechanisms at play. This magnetic path can be schematically represented as
illustrated in Fig. 12. In the first stages, magnetisation increases as domain walls move such that those domains that are favourable
with the direction of the applied field grow. Here the wall can be pinned by defects and can bulge or bow. However, since this process
is reversible, any reduction of the applied field results in reduction of the magnetisation along the same path back to the virgin state. If
the field is further increased, Barkhausen jumps can occur as domain walls over come the energy barrier created by pinning sites and
the magnetisation increases with the increase of the favourable domain orientation. This region is irreversible since the erratic jumps
dissipate some of the energy and a reduction of the field from this point will result in a reduction of the magnetisation along a
different path and will eventually lead to some remnant magnetisation at zero field. As the magnetic field is further increased, the non-
favourable domains are eliminated from the sample and we enter the reversible coherent rotation phase, also referred to as the
approach to saturation, where the magnetisation rotates towards full alignment with the applied magnetic field. This leads to full
saturation of the sample. When the field is reduced, reverse domains can nucleate and begin to propagate as the sample returns to a
multidomain state. At zero field, the remnant magnetisation is attained, where there is still an overall magnetisation in the direction of
the previously applied field. Applying a magnetic field in the reverse direction will increase the domains in the reverse direction until a
state of zero magnetisation is reached at the coercive field. This microscopic state of the sample is very different from the virgin state,
which can now only be re-established by heating the sample to a temperature above the Curie point and then cooling in zero field.

3 Experimental Methods to Measure Magnetic Hysteresis

The objective of the present section is to discuss the underlying concepts behind the most common and most popular experimental
methods used to characterise magnetic materials by means of their hysteresis curves. Magnetometry techniques for determining
hysteresis loops and magnetisation as function of an applied magnetic field can be classified as closed-circuit or open-circuit
measurements. In open-circuit measurements, the sample is not part of a complete magnetic circuit and due to demagnetisation
effects, the externally applied field H0 is different from the internal field H of the sample. The relationship between both quantities is
given by H¼ H0  NM, where N is a demagnetising factor which depends on the shape of the sample. Open-circuit methods can
further be divided into two categories, force- and induction-based, where either the force acting on a magnetised sample in a
magnetic field gradient is measured or the change of magnetic flux in a circuit is sensed while the sample is moved, respectively. More
specific techniques are based on optical effects such as the magneto-optical Kerr effect (MOKE). In closed-circuit techniques, the
sample is part of a complete magnetic circuit and in the obtained B(H) or M(H) loops, the applied field H0 and the internal field of
the sample H are identical. A common example is the hysteresisgraph used to characterise permanent magnets. The principal open-
circuit techniques used nowadays include the vibrating sample magnetometer (VSM) and the alternating gradient force magnet-
ometer (AGFM). When ultimate sensitivity is paramount, a superconducting quantum interference device (SQUID) magnetometer is
the method of choice. For fast measurements of the B(H) dependence without the need to know the absolute value of M, either a B-H
looper or a MOKE magnetometer can be used. In the following, we will briefly discuss these five open-circuit techniques.

3.1 B-H Looper

An ac hysteresis loop tracer, or B–H looper, typically consists of a pair of Helmholtz coils, capable of generating ac magnetic fields
of up to 150 mT at frequencies in the Hz to lower kHz range. The sample is placed in the centre of a pair of pick-up coils, which are
 Magnetism in Solids: Hysteresis 15

connected such that the voltages induced by the changing flux due to the applied ac field are cancelled out, whereas the voltages
induced by the changing magnetic flux, F, of the sample are enhanced. According to Faraday’s law of induction, the voltage Vi
induced in the pick-up coils is given by:
∂F ∂B
Vi ¼ N ¼ NA ð40Þ
∂t ∂t
where N is the number of turns of the pick-up coils and A is the cross-sectional area. The time integral of the above equation yields
the value of the magnetic flux density B. The value of the magnetic field H can be measured either by a Hall probe attached to one
of the pick-up coils inside the Helmholtz coils or from the voltage across a resistor connected in series to the Helmholtz coil.
B-H loopers are cheap and simple and are typically used for quickly characterising soft magnetic films by measuring their B(H)
loops, but due to the small fields generated by the Helmholtz coils, their application is limited to thin films with small coercivity.
Moreover, relatively large sample volumes are required due to the low sensitivity, and absolute magnetic moment measurements
require difficult calibration.

3.2 Superconducting Quantum Interference Device Magnetometer

SQUIDs are the most sensitive devices available today for detecting the magnetic flux F or any quantity that can be transformed
into it. Their operation is based on two physical phenomena: flux quantisation in a closed superconducting loop in units of the
flux quantum F0 ¼ h/2e ¼ 2.07  1015 Tm2, and Josephson tunneling. The key part of any SQUID sensor, which is depicted in
Fig. 13(a) and (b), is a Josephson junction which separates two superconductors by means of a “weak link.” The most important
type of barrier used nowadays is a Josephson tunnel junction made from an insulator such as AlOX. If the junction is biased with a
constant current Ib, the operational point of its I–V curve lies midway between superconducting and normal behaviour. Since the
current–voltage characteristic of a Josephson junction exhibits a hysteresis itself, an additional shunt resistor is placed across the
junction to dampen its dynamics (Fagaly, 2006). When magnetic flux is inductively coupled into the SQUID loop, screening
currents are generated which effectively increase or decrease the critical current Ic of the junction, that is given by the first Josephson
equation as Ic ¼ Ic0 sinj with j ¼ F1–F2 being the phase difference across the junction (Josephson, 1962). Ic0 is the maximum
critical current of the junction and depends on the thickness of the insulating barrier, the junction area as well as the operating
temperature. When biased with Ib4Ic0, the phase difference across the junction changes over time and the voltage drop across the
junction will vary. This is described by the second Josephson equation (Josephson, 1962):
∂j 2e 2p
¼  VDC ¼  VDC ð41Þ
∂t ℏ F0
where ħ¼ 1.055  1034 Js is the reduced Planck’s constant. As shown in Fig. 13(b), an increase or decrease of external magnetic
flux leads to periodic changes of the voltage VDC with the corresponding period being that of the flux quantum F0.
Different types of SQUIDs are available that differ in the manner of biasing the Josephson junction as well as in the number of
junctions itself (Fagaly, 2006). In dc SQUIDs, two junctions are biased by a dc current, whereas in rf SQUIDs, there is only one
Josephson junction which is biased by a rf current. Due to their generally superior noise performance, nearly all commercial
SQUIDs are dc SQUIDs. For use as a magnetometer, a magnetic thin film or wire wound pick-up coil can be connected to the input
coil of a current sensor SQUID. Other applications include the use as magnetic gradiometers, current sensors and voltmeters, as
well as amplifiers (Schmelz and Stolz, 2017). While an unparalleled sensitivity of 10–10 Am2 or better can be attained, the
measurements are slow and time-consuming since they involve both data extraction and changing the field of the superconducting
magnet from one measuring point to the next during a hysteresis cycle.

Fig. 13 (a) dc SQUID with two weak links (Josephson junctions) and (b) oscillations of the voltage across a Josephson junction as a function of
the magnetic flux.
16 Magnetism in Solids: Hysteresis

3.3 Vibrating Sample Magnetometer

The VSM, developed by Foner (1959) and schematically drawn in Fig. 14, is one of the most commonly used instruments in
academic and industrial research to measure the magnetic moment not only of soft and hard magnetic materials, but also of
weakly magnetic or paramagnetic samples. Its operating principle is based on attaching the sample to a vibrating rod located in the
centre of a single or double set of static pick-up coils. The exact coil arrangement and direction of sample vibration relative to the
external field depend on the type of magnetic field source used. If a vertical field is applied, as with a superconducting solenoid, a
pair of Helmholtz coils together with longitudinal sample vibration is employed. When a horizontal field is applied, either with an
electromagnet or by permanent magnet sources such as, e.g., Halbach cylinders (Halbach, 1980), an assembly consisting of, e.g., a
four-coil configuration known as Mallinson’s set (Mallinson, 1966) in conjunction with perpendicular sample vibration is used. In
either case, upper and lower coils are oppositely wound such that the signals induced in them by the oscillating sample are of
additive nature, whereas any signal resulting from fluctuations of the external field is minimised or ideally cancelled. The voltage Vi
generated in the pick-up coils is proportional to the magnetisation of the sample and depends on the orientation of the magnetic
moment relative to the coils. For a magnetic moment m directed along the (x-)axis of a coil pair and moving with a velocity x_ the
induced voltage Vi is given by (Fiorillo, 2010):

d 
Vi ðx; t Þ ¼ mkðxÞ ðxÞ  x_ ¼ mgx ðxÞ  x_ ð42Þ
dx

Here, gx(x)¼ dkx(x)/dx represents the distinct spatial sensitivity distribution of the coil arrangement, and thus the dependence of
the VSM signal on the sample position (Zieba and Foner, 1982). To obtain time dependence of the signal, the above expression
must be integrated over the volume of the sample for given amplitude and frequency of oscillations. Ideally, sensing coil
arrangements are used that exhibit a saddle point in the symmetry centre in whose vicinity the sensitivity is independent of the

Fig. 14 Schematic drawing of a VSM with an electromagnet field source. The voltage induced in the pick-up coils is amplified using a lock-in
amplifier. Its reference signal is used to control both amplitude and frequency stability of the oscillations of the vibrating rod. Reproduced from
Fiorillo, F., 2010. “Measurements of magnetic materials”. Metrologia 47, S114–142.
 Magnetism in Solids: Hysteresis 17

pffiffiffi
sample position. For example, for thin coils spaced apart by a distance of d ¼ 3 a, with a being the coil radius, the sensitivity
function is not maximal, but very flat at the centre of the coil arrangement, and the linearity of the field derivative is enhanced.
The measurement of the magnetic moment of the sample can be performed by either varying the applied magnetic field
continuously with time or by generating the magnetic field in steps. The latter offers an improved signal-to-noise ratio if the signal
is averaged over a reasonably long integration time of the order of 1 s or longer. Typical lower and upper detection limits for the
magnetic moment in commercially available VSMs are of the order of 10–9 Am2 and 1 Am2, respectively. The vibration frequency
generally ranges from 10 Hz up to several 100 Hz, with the vibration amplitude varying between 0.1 and several tens of
millimeters (Graham, 2000). Frequency stability and oscillation amplitude are controlled by a lock-in amplifier-based feedback
loop. The volume of the sample is generally chosen to be a few cubic millimeters such that it does not protude from the region of
uniform sensitivity determined by the sensing coil configuration. While any kind of sample shape, even irregular ones, can be
tested, its influence on the measured hysteresis curve has to be properly taken into account by calculating the corresponding
demagnetisation factors. The ideal sample geometry is that of a sphere, as it not only ensures uniform magnetisation of the sample,
but also allows for precise and easy retrieval of the actual hysteresis curve after correction for the demagnetising field Nd ¼ 1/3.
Magnetic thin films can conveniently be approximated by the disc geometry with the corresponding demagnetisation factors
being similar to those of an oblate ellipsoid having the same axes length. In any case, the sample has to be centred prior to a
measurement in order to determine the saddle point of the sensitivity function.
When samples with very small volumes such as thin films are measured, it is typically necessary to subtract the diamagnetic
contribution from the sample holder or substrate, which manifests itself in a negative slope of the entire hysteresis curve. Properly
calibrated, VSM measures the absolute value of the magnetic moment of the sample from which, when divided by the sample
volume, the saturation magnetisation can be determined. However, since an absolute calibration by numerically calculating the
sensitivity function g is difficult, it is much more common to calibrate instead by comparing to the well-known magnetic moment
of a standard sample such as a Ni sphere measured at a high field value. As an additional benefit, this type of calibration offers a
correction from the image effect, which shows up as an apparent decrease in magnetisation at high fields due to saturation of the
soft magnetic pole shoes of the electromagnet.

3.4 Alternating Gradient Force Magnetometer

An AGFM can be thought of as sort of an inverted VSM. Its original idea was developed by Zijlstra (1970), whereas first complete
instruments were described by Reeves (1972), Roos et al. (1980), and Flanders (1988). In Fig. 15, the side view of an AGFM is
shown. A magnetic sample attached to one end of a cantilevered quartz or glass fibre is magnetised by a dc field of varying
amplitude and simultaneously subjected to a small ac field gradient. The alternating field gradient, whose maximum is of the order
of several hundreds of mTm1 at the location of the sample, exerts an alternating force on the sample, that is proportional to both
the amplitude of the field gradient and the magnetic moment of the sample. For example, in a vertical gradient AGFM, a sample
with its magnetic moment directed along the x-axis would be subjected to an alternating force FZ ¼ m0m(∂Hx/∂Z) in z-direction. The
resulting deflection of the rod is converted into a voltage by means of a piezoelectric bimorph mounted to the other end of the rod.
By operating at or close to the mechanical resonance frequency of the cantilever, typically between 100 and 1000 Hz, the
corresponding output signal is greatly enhanced (Graham, 2000).
Compared to a VSM, the large Q factor of the mechanically resonant system directly translates into a gain in sensitivity by a
factor of 50–100 (Graham, 2000). However, maximum measurable sample size, mass and moment are reduced which largely
compensates this advantage. Factoring in mechanical and acoustic noise from the environment, a sensitivity of 10–10 Am2 is more
realistic. When measuring soft magnetic materials, the changing field gradient can lead to errors in the measurements of rema-
nence, coercivity, and magnetisation decay.

3.5 Magneto-Optical Kerr Magnetometer

Optical magnetometers measure the magnetisation by means of magneto-optical effects such as the Faraday (1846) and Kerr
(1877) effect and are especially useful in the case of thin films whose signals might be too weak for inductive or force-based
methods. What is exploited for optical magnetometry are their respective rotational effects, both of which can be interpreted as
circular magneto-optical birefringence causing a magnetisation-dependent polarisation rotation of light upon reflection from or
transmission through a magnetic sample. The Faraday effect occurs in optically transparent samples, whereas the Kerr effect applies
to any metallic or otherwise light-absorbing magnetic material with a sufficiently smooth surface. The latter is also much weaker
than the transmission effect since the light only interacts with the magnetisation of a relatively thin surface layer whose thickness is
given by the penetration depth of the light. Both effects are linear in magnetisation and consequently, the corresponding rotation
is inverted when the magnetisation direction is reversed.
The physical origin of both effects is the interaction of light as an electromagnetic wave with the magnetisation-dependent
properties of a sample. Linearly polarised light impinging upon a reflective magnetic surface will induce electrons to oscillate
parallel to its plane of polarisation, which is the plane of the electric field, E. A fraction of the incident light beam is regularly
reflected, N, i.e., it maintains its original polarisation. At the same time, the Lorentz force induces of small component of
vibrational motion perpendicular to both the magnetisation and the primary motion. This secondary motion, which is
18 Magnetism in Solids: Hysteresis

Fig. 15 Side view of a vertical gradient AGFM setup. A vertical ac field gradient is generated by means of a suitable coil arrangement.
The vibration of the sample (magnetic moment pointing along the x-axis) along the z-axis is transmitted to the cantilevered piezoelectric
bimorph using a glass or quartz fibre mounted to the tip of the bimorph. Adapted from Fiorillo, F., 2010. “Measurements of magnetic
materials”. Metrologia 47, S114–142.

proportional to the Lorentz velocity vLor ¼  m  E, generates an additional magnetic contribution to the amplitude of the reflected
light: the secondary Faraday amplitude F for transmission and the secondary Kerr amplitude K for reflection. The superposition of
N with F or K then leads to magnetisation-dependent polarisation rotations.
Depending on the relative orientation of magnetisation, plane of incident light, and polarisation direction of illuminating light,
three geometries can be defined: polar, longitudinal, and transverse (Hubert and Schäfer, 1998). In case of the polar effect, which is
depicted in Fig. 16(a), the magnetisation is perpendicular to the sample surface. If the magnetisation lies both in-plane, i.e.,
parallel to the sample surface, and along the plane of incidence, the longitudinal effects are obtained as shown in Fig. 16(b). For
transverse orientation, illustrated in Fig. 16(c), the magnetisation is in-plane and perpendicular to the plane of incidence, which
only leads to a magneto-optical effect in reflection.
The measuring principle of a typical MOKE magnetometer is the following. A laser, e.g., a He–Ne model with l ¼ 633 nm, or
monochromatic light emitted by a high-intensity diode is passed through a polariser (generally a Glan–Thompson prism) set to a pre-
defined polarisation. The light then passes through an objective lens and is focused onto the region of interest of the sample. Due to the
interaction between the polarised light and the domain structure of the sample, the polarisation plane of the reflected signal is rotated, with
the magnitude of the rotation being proportional to the local magnetisation. Next, the signal passes through a birefringent photoelastic
modulator (PEM) to improve the signal-to-noise ratio. In addition, the PEM also introduces a phase-retardation (e.g., l/4) allowing to
separate Kerr rotation and ellipticity components of the MOKE signal. The modulated signal is used as a reference signal for a lock-in
amplifier. The beam is finally transmitted through an analyser and focussed by a second objective lens onto a photosensitive diode, which
provides the input signal for the lock-in amplifier. By plotting the relative Kerr intensity as a function of the applied magnetic field, the
magnetisation process of thin films is characterised.
The advantages of MOKE magnetometry include high temporal and spatial resolution, low cost, as well as the possibility to
perform optical measurements during preparation or treatment of a sample inside a vacuum chamber. In contrast to the other
techniques discussed in this section, MOKE is a local probe as the magnetisation is only probed at the position of the laser,
offering the possibility of spacial magnetisation mapping by scanning over the sample surface.

4 Soft and Hard Magnetic Materials

Based on their specific hysteretic behaviour, ferromagnetic materials can be classified into two distinct categories, soft magnetic
and hard magnetic materials, which we will discuss in the present section.

4.1 Soft Magnetic Materials

Soft magnetic materials exhibit low coercivities (typically o1 kAm–1), high permeabilities, low remanence. and narrow hysteresis
loops (see Fig. 17). They can be easily magnetised by low magnetic fields owing to their high permeabilities, but only retain a
fraction of their magnetisation when the magnetic field is switched off. As a result of their low coercivities, they can be demag-
netised by applying small magnetic reversal fields. When designing a soft magnetic material for a particular application, a trade-off
 Magnetism in Solids: Hysteresis 19

Fig. 16 Schematic representation of the different geometries of the magneto-optical Kerr effect. Adapted from Hubert, A., Schäfer, R., 1998.
Magnetic Domains: The Analysis of Magneticmicrostructures, Springer.

between high saturation magnetisation, maximum permeability, minimum coercivity, low magnetostriction, low ac losses, low
temperature dependence and a high thermal stability of its magnetic properties, and cost has to be made. While some of these
properties are strongly sensitive to grain size and trace impurities, others are influenced by the material structure. For example, the
addition of trace amounts of C, N, or S is known to increase hysteretic losses, whereas alloying with Fe, Co, Ni has to opposite
effect (Cardarelli, 2000).
Soft magnetic materials include Fe in various degrees of purity (e.g., Armco and carbonyl iron), low-carbon steels and silicon
steels, Fe–Al–Si alloys (e.g., Alfer, Alperm, and Sendust), Fe–B alloys (e.g., Finmet and Metglas), Fe–Co alloys (e.g., Hyperco and
Permendur), Fe-Cr alloys (e.g., Ferrochrome), Fe–Ni alloys (e.g., Mumetal, Permalloy, Supermalloy, Invar, Perminvar, and
Hypernik), spinel ferrites (e.g., Ni–Zn, Mn–Fe, and Ni–Co), as well as synthetic garnets (e.g., YIG) (Cardarelli, 2000). In Fig. 18,
the ranges of coercivity and saturation magnetisation of selected soft magnetic materials are shown.

4.1.1 Iron
Pure iron has an excellent soft magnetic properties, but suffers from the fact that even small amounts of impurities cause them to
significantly deteriorate. In addition, it has a high conductivity which leads to large eddy-current losses and practically prohibits
the use of pure iron for ac applications. These problems have been solved by progress in steel metallurgy over the years, which have
lead to a steady increase in purity of low-carbon steels made from powdered iron carbonyl (Tumanski, 2011). Despite the
deteriorated magnetic properties due to the presence of C, its low price and good mechanical properties make low-carbon electrical
steel an acceptable material for many cheap and basic applications and devices.

4.1.2 Silicon steels

Drastically better electrical performance as well as improved mechanical and magnetic properties were achieved by the invention
of silicon steel in 1900. Depending on the exact amount of Si in the composition, either nonoriented, isotropic alloys or grain-
oriented, anisotropic alloys, which are typically rolled into thin sheets, can be produced. Today, these two types of materials
account for roughly 40% and 25%, respectively, of the global market for soft magnetic materials (Coey, 2009).

4.1.3 Ni–Fe alloys

Ni–Fe alloys provide the most versatile system of soft magnetic materials available today. Depending on the nickel content, which
varies in the range from 30 to 80 wt%, either permeability, saturation magnetisation, or electrical resistance are maximised for
high, medium, or low Ni concentrations, respectively. Both anisotropy and magnetoelasticity constants are almost zero around a
nickel concentration of 80%. Though it is not possible to obtain zero value of both parameters simultaneously in binary Ni–Fe
alloys, the addition of small amounts of Mo and Cu makes it possible to obtain a nonmagnetostrictive material with zero
anisotropy (Warlimont, 1997). This has permitted to design alloys which exhibit ideal soft magnetic properties called Permalloys
in order to indicate that high permeability is their outstanding feature. Annealing treatments combined with field-induced uniaxial
anisotropy are used to control and adjust the magnetisation process to specific applications.

4.1.4 Co–Fe alloys

Co–Fe alloys are significantly less versatile and much more expensive due to the costly cobalt. Crystalline anisotropy and
magnetostriction have minima at different Co contents, allowing compositions to be optimised for different applications. For
example, Permendur, Fe50Co50, has an unrivaled saturation magnetisation of 2.45 T which is the highest value of all magnetic
materials at room temperature, whereas NiFeCo alloys called Perminvar exhibit constant permeability across a wide field range.
20 Magnetism in Solids: Hysteresis

Fig. 17 Schematic hysteresis loops of a soft magnetic (dashed curve) and a hard magnetic (solid curve) material.

Fig. 18 Ranges of coercivity and saturation magnetisation of commercially available magnetic materials as an example of products offered by
Vacuumschmelze. Reproduced from Tumanski, S., 2011. Handbook of Magnetic Measurements. Boca Raton: CRC Press.

Adding 2% vanadium improves machinability, delays the structural ordering transformation, and drastically increases electrical
resistivity, thus reducing eddy current losses, without sacrificing the magnetic properties.

4.1.5 Amorphous alloys

Amorphous alloys, also known as metallic glasses, are compositions close or equal to the form T80M20, where T¼ Co, Fe, Ni and
M¼ B, Si. There are two major groups of amorphous alloys: Fe-rich glasses of FeSiB-type with high saturation magnetisation and
Curie temperature, and Co-rich alloys of CoFeMnSiB type with zero magnetostriction and very high permeability. Metallic glasses
are produced by melt-spinning thin tapes or ribbons which together with their very high resistivity allows them to compete with
ferrites in the medium frequency range up to 100 kHz. Fully or partially crystalline materials, e.g., Finmet, obtained by annealing
amorphous alloys, exhibit excellent soft magnetic properties that can easily compete with the best permalloys.

4.1.6 Soft ferrites

Soft ferrites are a very important group of soft magnetic materials thanks to their low raw material and production cost as well as
ease of fabrication into a wide variety of shapes. Due to their ferrimagnetic nature, saturation magnetisation and permeability are
mediocre compared to metallic ferromagnets. However, their insulating character makes them suitable for high-frequency
applications since eddy-current losses are strongly suppressed. Structurally, ferrites are iron spinels, and the two major types of
 Magnetism in Solids: Hysteresis 21

high permeability ferrites are Mn-Zn and Ni-Zn ferrites, which can be operated at frequencies up to around 1 MHz and from 1 to
300 MHz or even more, respectively (Coey, 2009). The latter have a lower saturation magnetisation, but much higher resistivity.

4.1.7 Rare-earth iron garnets

Another group of ferrites are iron garnets containing trivalent rare earth ions such as, e.g., yttrium iron garnet (YIG). This is a
perfect insulating crystal with minimal anisotropy as well as very narrow ferromagnetic resonance linewidth, making it an excellent
material for applications in the GHz range.

4.2 Hard Magnetic Materials

Hard magnetic materials have high coercivities (typically 410 kAm–1), low permeabilities, high remanence, high maximum
energy products, and wide hysteresis loops (see Fig. 17). Owing to their ability to retain a high residual magnetisation even after a
high magnetic field has been switched off, they are also referred to as permanent magnets. The ranges of coercivity and saturation
magnetisation of selected hard magnetic materials are depicted in Fig. 18. From both a historic and commercial point of view, hard
magnetic materials can be classified into five major categories (Cardarelli, 2000), whose general properties and preparation
methods will briefly be described in the following.

4.2.1 Martensitic steels

Magnetic steels were originally made by rapid quenching alloys of iron with up to 2 wt% C, known as high-carbon steel, at
temperatures around 8001C. The addition of carbon reduces the saturation of iron, but coercivity and energy product are vastly
improved by means of internal stresses and inhomogeneities connected with a lattice transformation. The resulting steels were
mechanically very hard, which is where the term hard magnetic originates from. In the early 20th century, additional alloying
elements such as tungsten and later chromium were added in order to further increase the coercivity to values between 3 and
6 kAm1, eventually leading to the discovery of stainless steel. In 1917, cobalt was introduced as an alloying element by the
Japanese which allowed to reach coercivities of up to 25 kAm1 and energy products of up to 8 kJm3 (Cardarelli, 2000). Cobalt
steels including the addition of Mo or V, known as Remalloys or Vicalloys, respectively, had commercial use at that time, but
quickly became obsolete with the discovery of Alnico in the 1930s.

4.2.2 Precipitation-hardened alloys

In the early 1930s, the Japanese experimented with carbon-free Fe alloys containing aluminum, nickel, and later also cobalt, which
lead to significant improvements over the magnet steels used earlier. Alnico, an acronym for aluminum–nickel–cobalt, combined
high remanence of up to 65 kAm1, energy products of up to 80 kJm3, low temperature expansion coefficient, high working
temperature of 5501C, and reasonable cost, making it the material of choice at that time (Cardarelli, 2000). The improved
magnetic properties of these alloys are the result of a precipitation phase introduced through tempering and annealing in high
magnetic fields performed prior to the final hardening treatment. Shape anisotropy, determined by size, shape, and orientation of
the nanoparticles formed during heat treatment, supplemented in many cases by an increase in crystalline anisotropy, is
responsible for their high coercivity. Changing the production process of Alnico magnets from casting to sintering not only reduces
their surface roughness as well as their considerably high hardness and brittleness, but also makes it possible to produce very small
magnets, even with complicated shapes.

4.2.3 Superstructure alloys

In the two-component alloys platinum–cobalt, platinum–iron, and platinum nickel, hardening occurs by a similar precipitation
mechanism, but with an additional superstructure formation, i.e., the precipitated phase is ordered, while the underlying matrix is
unordered. A Pt–Co alloy containing approximately equal numbers of atoms of each element, attains a coercivity of B500 kAm1,
which is only surpassed by cobalt-rare earth magnets, as well as an energy product of B75 kJm3, a value only exceeded by the best
Alnico magnets available at the time of its discovery in the 1950s (Goldmann, 1999). These alloys are characterised by excellent
mechanical and corrosion resistant properties. Though their magnetic properties are superior to most other permanent magnet
materials with the exception of Nd–Fe–B, they were replaced by this very same material because of the expensive price of platinum.

4.2.4 Ceramic hard ferrites

Hard hexagonal ferrites such BaO  6Fe2O3, SrO  6Fe2O3, or Pb  6Fe2O3, were developed since the 1950s and represent the most
widely used permanent magnet material today due to their extremely low costs and ease of production. These magnets are made
by a multi-step powder metallurgy process, which in the absence of a magnetic field produces isotropic magnets, whereas the
presence of a strong magnetic field aligns the easy c axis to produce anisotropic magnets with almost doubled remanence, but
unchanged coercivity. In addition, the final heat treatment can be managed such that coercivity is raised at the expense of
remanence, and vice-versa, leading to remanence- or coercivity-biased anisotropic products (Heck, 1974). While their manu-
facturing process allows them to be molded into many complex shapes, the resulting ceramic magnets are inherently brittle and
frequently contain imperfections. The coercivity of hard ferrites can be as high as 250 kAm1, but due to their low remanence, the
corresponding energy product is only of the order of 40 kJm3 (Coey, 2009).
22 Magnetism in Solids: Hysteresis

4.2.5 Rare-earth element alloys of SmCo- and NdFeB-type

The next major milestone in the technology of permanent magnetic materials was the development of alloys combining rare-earths
and the transition metals Ni, Fe, and Co in the late 1960s. The two best known and most important compositions of this family
are SmCo5 and Sm2Co17. The 2–17 material has superior magnetic properties, but contains more of the costly samarium and is
difficult to produce consistently, which is why only the 1–5 material is of commercial significance (Goldmann, 1999). With values
of up to 160 kJm3 for SmCo5 and 240 kJm3 for a Sm2Co17-based alloy, respectively, they held the record for the highest energy
product for many years (Cardarelli, 2000). Both SmCo-types exhibit coercivities in excess of 1000 kAm1, and have excellent
corrosion resistance, thermal stability up to 300 1C, as well as high resistance to demagnetising influences. Their fabrication
process involves powder metallurgy, with sintered magnets having superior magnetic properties over their bonded counterparts.
The most recent development and yet another significant step up from SmCo magnets was the discovery of Nd2Fe14B magnets.
It was made both simultaneously and independently by Sumitomo Special Metals in Japan and General Motors in the USA in
1984, and mainly driven by the limited availability and volatile price fluctuations of Co which were problems for the SmCo
magnet market (Goldmann, 1999). The Japanese production method was based on a powder metallurgy process, which involved
casting an ingot, breaking it up into powder, pressing, sintering, and heat treating. The resulting NdFeB magnets have energy
products of up to 450 kJm3, making them the most powerful commercialised permanent magnets available today (Coey, 2009).
High-coercivity grades of NdFeB are much more prone to oxidation than any other magnet alloy due to the highly reactive nature
of their Nd-rich phase, which is why usually a protective nickel coating is applied in the final production steps if the material is to
be exposed to humidity or chemically aggressive substances. In the USA, melt-spinning was used to produce fine wires or ribbons
of alloy, followed by powdering and pressing. While this process was much simpler, the magnetic and microstructural properties
obtained by it are less good as they are highly sensitive to the quench rate of the ribbons. As NdFeB magnets posses excellent
magnetic properties, come in a wide range of grades, sizes, and shapes, are relatively inexpensive, and their raw materials are
available in abundance, they can be used to replace traditional permanent magnets such as Alnicos or even SmCo by higher
performing, more efficient, more compact, and more cost-effective designs and devices. Despite their high energy product, NdFeB
magnets only have a relatively low Curie temperature which limits their maximum application temperature to 200 1C. However,
their temperature characteristics can be improved by the addition of Co and Dy at the expense of increased production costs.

5 Applications and Practical Aspects of Hysteresis

As discussed in the previous section, soft and hard magnetic materials posses vastly different magnetic properties which are
exploited in a large number of applications that fall into two categories: ac and dc. In this section, we will give a general overview
of their main applications classified by their operating frequency range (Coey, 2009).

5.1 Applications of Soft Magnetic Materials

For dc or static applications (o1Hz) such as in electromagnets, relays, and shielding, soft iron, Co–Fe (Permendur), and Ni–Fe
(Permalloy) are the preferred materials. In the low frequency range from 1 Hz to 1 kHz, silicon steel, Permalloy, Finmet, and
magnetic glasses are the materials of choice for use in transformers, generators, and motors. For audio frequencies (100 Hz to
100 kHz), Permalloy foils, Finmet, amorphous alloys, Fe–Si–Al powder (Sendust) as well as Mn–Zn ferrite are used for inductors,
transformers in switched mode power supplies, and TV flyback transformers. The radio frequency range between 0.1 and 1 GHz
exploits Mn–Zn and Ni–Zn ferrites for inductors and antenna rods. Lastly, in the microwave range beyond 1 GHz, YIG and Li
ferrite are the materials of choice for use in microwave isolators, circulators, phase shifters, and filters.

5.1.1 Static applications

DC applications of soft magnetic materials such as in electromagnets require optimum flux guidance and concentration, which is
why materials such as pure soft iron or Fe50Co50 combining highest saturation magnetisation with both very low remanence and
coercivity are used. Passive magnetic shielding prevents an external weak dc or low-frequency ac magnetic field from penetrating
into an internal space, where it could interfer or even prevent proper device or circuitry operation. Shielding materials either deflect
magnetic flux by guiding it around the sensitive volume to be protected, or contain magnetic flux in direct vicinity of the device
generating it. DC shields are typically made of Permalloy or Mumetal, which are both insensitive to shock and strain because of
their vanishing anisotropy and magnetostriction. Formable and flexible shielding can be made from Finmet sheets as well as
woven Co-rich metallic glass ribbons (Coey, 2009).

5.1.2 Low frequency applications

Low frequency applications consume the largest amount of magnetic material produced and typically involve power generation
and distribution at mains frequencies of 50 or 60 Hz as well as of 400 or 800 Hz in the case of military and avionics applications,
respectively. Soft iron cores are generally used to generate and guide flux in inductors and generators. Eddy-current losses are
reduced by employing thin laminations of electrical steel, such as in the case of induction motors or high-resistivity materials.
Isotropic sheets of silicon steel are suited for generators and motors, where the flux direction varies continually during operation.
 Magnetism in Solids: Hysteresis 23

In transformers, where the flux axis is fixed, it is advantageous to use grain-oriented sheet steel having an easy axis to reduce losses.
Magnetic amplifiers and fluxgate magnetometers rely on saturable soft cores and are typically made from textured Permalloy.
Transducers, linear actuators, and sonar make use of a Tb-Dy-Fe alloy named Terfenol, which possess the highest magnetostriction
of all known alloys (Coey, 2009). Cost is still the main consideration for almost all of these applications, but higher operating
efficiency and thus lower energy consumption become more and more important.

5.1.3 High-frequency applications

High frequency applications rely to a very large degree on the use of soft ferrites, which are used in chokes, inductors such as
loading coils, wideband pulse transformers, narrowband resonant filters, antennae, high-frequency transformers for switch-mode
power supplies, and broadband amplifiers.
At very high frequencies, such as the in microwave range from 300 MHz to 300 GHz, electrical energy is transmitted by means
of electromagnetic waves travelling in waveguides instead of currents flowing in wires and through circuits. Devices operating in
this frequency range are generally based on YIG and exploit the nonreciprocal interaction between a ferromagnetic medium and an
electromagnetic field by either resonantly absorbing microwaves or producing the microwave Faraday effect as a result of circular
birefringence far from resonance. The differences between the permeabilities of oppositely rotating, circularly polarised micro-
waves are exploited in different ways (Coey, 2009). Resonant isolators absorb microwaves propagating in one direction, but not
the other, when the ferrite material is magnetised to its resonance condition. Phase shifters exploit the very same difference, but
above and below resonance. Circulators with either three or four ports transmit signals from one port to the next one, while the
other ports remain isolated. Faraday rotators change the polarisation plane of microwaves.

5.2 Applications of Hard Magnetic Materials

In general, the principal applications of hard magnetic materials are magnetic recording media and permanent magnets. They are
classified according to the nature of their magnetic flux distribution which can either be spatially uniform or nonuniform, as well
as static or time-varying (Coey, 2009). The physical effects exploited in their applications include force, torque, induced elec-
tromotive force, magnetoresistance, or Zeeman splitting.

5.2.1 Static applications in uniform magnetic fields

Though the field generated by a magnetic dipole is nonuniform and anisotropic, it is possible to achieve a uniform field by
arranging pieces of magnetic material magnetised along different directions such that the sum of their individual contributions
results in a uniform field over a limited volume in space. For example, in a cylindrical magnet with a hollow bore, a uniform field
can be generated in the central region. An open square cylinder or a design with flat cuboid magnets surrounded by a soft iron
yoke is used in nuclear magnet resonance and magnetic resonance imaging equipment, accelerator magnets, and magnetic
refrigeration devices. Halbach cylinders, made from wedge shaped magnets whose magnetisation varies around a common axis,
produce fields of the order of 1 T used for magnetic annealing to set the exchange bias in spintronic devices.

5.2.2 Static applications in non-uniform magnetic fields

Non-uniform, transverse multipole fields and uniform field gradients can be generated by modified cylindrical magnet structures.
For example, external multipole magnets, which are variants of a normal Halbach cylinder and produce a multipole field outside
the cylinder while the field in their interior is zero, are used as rotors of electric motors based on permanent magnets. The pattern
of an unrolled Halbach cylinder gives a single-sided magnet with a strong field on one side, but none on the other side, which is
used for fridge magnets and other holding magnets based on bonded ferrites.
Highly brilliant, polarised synchrotron radiation is generated by passing a high-energy electron beam through insertion devices
such as wigglers and undulators, whose sinusoidal transverse field of wavelength l stimulates the electrons to radiate with a
frequency c/l Klystrons and magnetrons employ permanent magnets to generate microwave radiation from electrons.
Magnetic separators use rare-earth permanent magnets to generate complex flux patterns with huge spatial fluctuations (102 to
105 Tm1) for separating ferrous and nonferrous materials based on their different magnetic susceptibilities. There are three basic
configurations: pulleys, drums, and suspended magnets, which are used in either high-gradient or open-gradient magnetic
separators. Magnetic separation is used routinely in, e.g., the mining industry for the beneficiation of iron ores such as magnetite
from slag and slurries, the recycling and scrap processing industries for the recovery of potentially salable iron and steel from
municipal solid waste, and water purification systems for eliminating heavy metals from waste water.

5.2.3 Dynamic applications

Permanent magnet-based, variable flux sources such as a double Halbach-cylinder, a four-rod mangle, or an external Halbach-
cylinder with a soft iron sheath are compact and do not require cooling, making them ideally suited for compact benchtop
measurement systems such as VSMs.
When linear and rotary motion need to be combined, but no mechanical contact point is allowed, gears or bearings based on
axially magnetised permanent magnets can be used that are well suited to high-speed rotary suspensions in turbopumps or
flywheels used for energy-storage.
24 Magnetism in Solids: Hysteresis

Magnetic sensors detect a varying magnetic field in an airgap using Hall effect or magnetoresistance probes which deliver a field
dependent voltage from which, e.g., speed and position can be determined in a noncontact fashion.
Actuators, electromechanical devices having a limited linear or angular displacement range, use permanent magnets in three
different configurations. Moving coil actuators are used in loudspeakers as well as ear- and headphones, with low-end designs
employing flat ferrite ring magnets, mid-range designs using cylindrical Alnico or rare-earth magnets, and high-end models
utilizing Nd2Fe14B magnets. Voice-coil and flat coil actuators are used for positioning of read and write heads in computer hard-
disc drives, requiring Nd–Fe–B magnets with the highest energy product available. Moving-iron actuators are designs that are used
for print hammers in dot-matrix printers as well as in reed switches.
Motor designs have evolved remarkably since the availability of ferrite and rare-earth magnets as well as polymer-bonded
magnets in any desired shape. Small dc motors are typically found in domestic appliances such as washing machines and
consumer electronics such a compact disc drives. DC servo motors power tools and robots, whereas stepper motors are used for
precision control clocks, watches, as well as linear motion and rotation stages.

References

Becker, R., Döring, W., 1939. Ferromagnetismus. Spinger.

Benabou, A., Clénet, S., Piriou, F., 2003. Comparison of Preisach and Jiles-Atherton models to take into account hysteresis phenomenon for finite element analysis. J. Magn.
Magn. Mater. 261, 139–160.
Bertotti, G., 1991. Generalized Preisach model for the description of hysteresis and eddy current effects in metallic ferromagnetic materials. J. Appl. Phys. 69, 4608.
Cardarelli, F., 2000. Materials Handbook: A Concise Desktop Reference. Springer.
Chikazumi, S., 1964. Physics of Magnetism. New York: Wiley.
Coey, J.M.D., 2009. Magnetism and Magnetic Materials. Cambridge: Cambridge University Press.
Cullity, B.D., 1972. Introduction to Magnetic Materials. New York: Addison-Wesley.
Dupré, L.R., Bán, G., von Rauch, M., Melkebeek, J.A.A., 1999. Relation between the microstructural properties of electrical steels and the Preisach modelling. J. Magn. Magn.
Mater. 195, 233–249.
Dupré, L., Sablik, M.J., van Keer, R., Melkebeek, J., 2002. Modelling of microstructural effects on magnetic hysteresis properties. J. Phys. D Appl. Phys. 35, 2086–2090.
Ewing, J.A., 1890. Phil. Mag. 5, 205.
Fagaly, R.L., 2006. Superconducting quantum interference device instruments and applications. Rev. Sci. Instrum. 77, 101101.
Faraday, M., 1846. On the magnetisation of light and the illumination of magnetic lines of force. Phil. Trans. R. Soc. Lond. 136, 1.
Fiorillo, F., 2010. Measurements of magnetic materials. Metrologia 47, S114–S142.
Flanders, P.J., 1988. An alternating-gradient magnetometer. J. Appl. Phys. 63, 3940–3945.
Foner, S., 1959. Versatile and sensitive vibrating‐sample magnetometer. Rev. Sci. Instrum. 30, 548.
Gaunt, P., 1983. Ferromagnetic domain wall pinning by a random array of inhomogeneities. Phil. Mag. B 48, 261–276.
Gaunt, P., 1986. Magnetic viscosity and thermal activation energy. J. Appl. Phys. 56, 2728–2731.
Globus, A., 1962. Influence des dimensions des parois sur la permeabilité initiale. Compt. Rend. Acad. Sci. 255, 1709–1711.
Globus, A., Duplex, P., 1969. Initial susceptibility of ferrimagnetic materials and topography of domain walls. Phys. Stat. Sol. 31, 765–774.
Globus, A., Duplex, P., 1970. Size of bloch wall and parameters of the magnetic susceptibility in ferrimagnetic spinels and garnets. Phys. Stat. Solidi (a) 3, 53–61.
Globus, A., Duplex, P., Guyot, M., 1971. Determination of initial magnetisation curve from crystallite size and effective anisotropy field. IEEE Trans. Magn. 7, 617–622.
Globus, A., Guyot, M., 1972. Wall displacement and bulging in magnetisation mechanisms of the hysteresis loop. Phys. Stat. Solidi (b) 52, 427–431.
Goldmann, A., 1999. Handbook of Modern Ferromagnetic Materials. Springer.
Graham, C.D., 2000. High-sensitivity magnetisation measurements. J. Mater. Sci. Technol. 16, 100.
Guyot, M., Globus, A., 1973. Determination of the domain wall energy from hysteresis loops in YIG. Phys. Stat. Solidi (b) 59, 447–454.
Guyot, M., Globus, A., 1977. Determination of the domain wall energy and the exchange constant from hysteresis in ferrimagnetic polycrystals. J. Phys. 38.C1-157–C1-162.
Guyot, M., Cagan, V., 1991. The acoustic emission along the hysteresis loop of various ferro and ferrimagnets. J. Magn. Magn. Mater. 101, 256–262.
Halbach, K., 1980. Design of permanent multipole magnets with oriented rare earth cobalt material. Nucl. Instr. Meth. 169, 1–10.
Hamimid, M., Feliachi, M., Mimoune, S.M., 2010. Modified Jiles-Atherton model and parameters identification using false position method. Phys. B 405, 1947–1950.
Heck, C., 1974. Magnetic Materials and Their Applications. Elsevier.
Hejda, P., Zelinka, T., 1990. Generalized preisach model of hysteresis – Theory and experiment. Czech. J. Phys. 40, 57–68.
Hubert, A., Schäfer, R., 1998. Magnetic Domains: The Analysis of Magnetic Microstructures. Springer.
Jiles, D.C., 2002. Hysteresis models: Non-linear magnetism on length scales from the atomistic to the macroscopic. J. Magn. Magn. Mater. 242–245, 116–124.
Jiles, D.C., Atherton, D.L., 1986. Theory of ferromagnetic hysteresis. J. Magn. Magn. Mater. 61, 48–60.
Josephson, B.D., 1962. Possible new effects in superconductive tunneling. Phys. Lett. 1, 251–253.
Kerr, J.D., 1877. On rotation of the plane of polarisation by reflection from the pole of a magnet. Phil. Mag. 3, 321–343.
Kersten, M., 1943. Zur Theorie der ferromagnetischen Hysterese und der Anfangs-permeabilität. Z. Phys. 44, 63–77.
Mallinson, J., 1966. Magnetometer coils and reciprocity. J. Appl. Phys. 37, 2514–2515.
Mayergoyz, I.D., 1986a. Mathematical models of hysteresis. Phys. Rev. Lett. 56, 1515–1521.
Mayergoyz, I.D., 1986b. Mathematical models of hysteresis. IEEE Trans. Magn. 22, 603–608.
Mayergoyz, I.D., Friedman, G., 1988. Generalized Preisach model of hysteresis. IEEE Trans. Magn. 24, 212–217.
Maxwell, J.C., 1873. Electricity and Magnetism. Oxford: Oxford University Press.
Pasquale, M., Basso, V., Bertotti, G., Jiles, D.C., 1998. Domain wall motion in random potential and hysteresis modelling. J. Appl. Phys. 83, 6497–6499.
Rayleigh, L., 1887. Phil. Mag. 23, 225.
Reeves, R., 1972. An alternating force magnetometer. J. Phys. E Sci. Instrum. 5, 547.
Roos, W., Hempel, K.A., Voight, C., Dederieks, H., Schippan, R., 1980. High sensitivity vibrating reed magnetometer. Rev. Sci. Instrum. 51, 612.
Schmelz, M., Stolz, R., 2017. Superconducting quantum interference device (SQUID) magnetometers. Smart sensors, measurement and instrumentation. In: Grosz, A., et al.
(Eds.), High Sensitivity Magnetometers 19. Springer International Publishing Switzerland.
Skomski, R., 2008. Simple Models of Magnetism. Oxford: Oxford University Press.
Stoner, E.C., Wohlfarth, P., 1948. A mechanism of magnetic hysteresis in heterogeneous alloys. Phil. Trans. Roy. Soc. A240, 599–642.
Tannous, C., Gieraltowski, J., 2008. The stoner – Wohlfarth model of ferromagnetism. Eur. J. Phys. 29, 475–487.
 Magnetism in Solids: Hysteresis 25

Tumanski, S., 2011. Handbook of Magnetic Measurements. Boca Raton: CRC Press.
Warlimont, H., 1997. Soft magnetic materials. In: Hadjipanayis, G.C. (Ed.), Magnetic Hysteresis in Novel Magnetic Materials. Dordrecht: Springer.
Weber, W., 1852. Abhandlungen der Kg. Sächs Gesellschaft des Wissens. 572.
Wiedemann, G., 1886. Phil. Mag. 52.
Williams, H.J., 1937. Phys. Rev. 52, 747.
Zieba, A., Foner, S., 1982. Detection coil, sensitivity function, and sample geometry effects for vibrating sample magnetometers. Rev. Sci. Instrum. 53, 1344.
Zijlstra, H., 1970. A vibrating reed magnetometer for microscopic particles. Rev. Sci. Instrum. 41, 1241.