Physics 2, 20 (2009)

Trends
Classifying multiferroics: Mechanisms and effects
Daniel Khomskii
II. Physikalisches Institut, Universität zu Köln, Zülpicher Strasse 77, 50937 Köln, Germany
Published March 9, 2009

Subject Areas: Materials Science

Introduction. A bit of history

Electricity and magnetism were combined into one
common discipline in the 19th century, culminating in
the Maxwell equations. But electric and magnetic order-
ing in solids are most often considered separately—and
usually with good reason: the electric charges of elec-
trons and ions are responsible for the charge effects,
whereas electron spins govern magnetic properties
There are, however, cases where these degrees of free-
dom couple strongly. For example, in the new, large
field of spintronics, the effects of spins on the transport
properties of solids (and vice versa) allow the possibil- FIG. 1: Multiferroics combine the properties of ferroelectrics
ity to control one by the other. The finding of a strong and magnets. In the ideal case, the magnetization of a ferro-
coupling of magnetic and electric degrees of freedom in magnet in a magnetic field displays the usual hysteresis (blue),
insulators can be traced back to Pierre Curie, but the real and ferroelectrics have a similar response to an electric field
beginning of this field started in 1959 with a short re- (yellow). If we manage to create multiferroics that are simulta-
mark by Landau and Lifshitz in a volume of their Course neously ferromagnetic and ferroelectric (green), then there is a
of Theoretical Physics[1]: magnetic response to an electric field, or, vice versa, the modi-
fication of polarization by magnetic field. And, in principle we
“Let us point out two more phenomena, which, in principle,
have here the basis for making a 4-state logic state: (P + M+),
could exist. One is piezomagnetism, which consists of linear (+−), (−+), (−−). (Illustration: Alan Stonebraker)
coupling between a magnetic field in a solid and a deformation
(analogous to piezoelectricity). The other is a linear coupling
between magnetic and electric fields in a media, which would
cause, for example, a magnetization proportional to an elec- tricity, the spontaneous ordering of electric dipole mo-
tric field. Both these phenomena could exist for certain classes ments—can coexist in one material in the absence of ex-
of magnetocrystalline symmetry. We will not however dis- ternal electric and magnetic fields. Following Schmid
cuss these phenomena in more detail because it seems that till [4], these materials are now called multiferroics (Fig. 1).
present, presumably, they have not been observed in any sub- (In principle one often includes in this category a third
stance.” type of ordering, spontaneous deformation, which leads
The situation changed soon thereafter, when to ferroelasticity. However, nowadays what most people
Dzyaloshinskii predicted [2], and Astrov observed mean by multiferroic predominantly applies to the co-
[3], this type of coupling, which is now known as the existence of magnetism and ferroelectricity.) Boracites
linear magnetoelectric effect. This was rapidly followed were probably the first known multiferroics [5], and
by the discovery of many other compounds of this soon several others were either found in nature, or syn-
class and by a rather complete classification of possible thesized artificially [6].
symmetry groups allowing for the effect. Essentially, three events majorly revived the study of
A new twist in this problem was the idea that not multiferroics, which had slowed down by the 70s. One
only can strong cross-coupling of responses exist in solids was that theory began to treat the general problem of
(i.e., the appearance of magnetization M in an electric why the coexistence of magnetism and ferroelectricity is
field E, or the inverse effect of electric polarization P so rare [7, 8]. The second and third were experimental
generated by the application of magnetic field H), but achievements that identified two distinct classes of mul-
that there may exist systems in which two types of tiferroics.
ordering—(ferro)magnetism, the spontaneous ordering In 2003, Ramesh’s group successfully grew thin films
of orbital and spin magnetic moments, and ferroelec- of one of the most popular multiferroics, BiFeO3 [9]. The
DOI: 10.1103/Physics.2.20
c 2009 American Physical Society
URL: http://link.aps.org/doi/10.1103/Physics.2.20
 Physics 2, 20 (2009)

multiferroic properties of bulk BiFeO3 are fairly weak, materials in which ferroelectricity and magnetism have
but in thin-film form they are greatly enhanced. The sec- different sources and appear largely independently of
ond major experimental development, also in 2003, was one another, though there is some coupling between
the discovery of a novel class of multiferroics, in which them. In these materials, ferroelectricity typically ap-
magnetism and ferroelectricity do not just coexist, but pears at higher temperatures than magnetism, and the
in which magnetism causes ferroelectricity. Tokura and spontaneous polarization P is often rather large (of or-
Kimura discovered this phenomenon in TbMnO3 [10] der 10 − 100 µC/cm2 ). Examples are BiFeO3 (TFE ∼
and Cheong found a similar effect in TbMn2 O5 [11]. 1100K, TN = 643 K, P ∼ 90µC/cm2 ) and YMnO3
Since 2003, these three systems have served as test
grounds (“drosophiles”) for the main physical ideas in (TFE ∼ 914K, TN = 76 K, P ∼ 6µC/cm2 ). The
the field. second group, which we can call type-II multiferroics,
is the relatively recently discovered materials [10, 11],
The realization that these materials have great poten-
in which magnetism causes ferroelectricity, implying a
tial for practical applications has lead to an extremely
strong coupling between the two. However, the po-
rapid development of the field of multiferroics. Appli-
larization in these materials is usually much smaller
cations include the ability to address magnetic mem-
ory electrically (and without currents!), the creation ( 10−2 µC/cm2 ). Many groups are also investigating
of new types of 4-state logic (i.e., with both up and composite multiferroics that consist of known magnets
down polarization and up and down magnetization) and ferroelectrics in the form of multilayers and self-
and magnetoelectric sensors. If there is any doubt about organized nanostructures [16].
the growth of this field, consider that in 2001, Nicola In this article, I will mainly discuss the microscopic
Spaldin organized the first special session on multifer- picture that is at the root of the properties of multifer-
roics at the American Physical Society’s March meet- roics, but for many multiferroics the symmetry consid-
ing. By 2007 there were seven of these sessions, more erations are indispensable, especially for type-II multi-
than double the number of sessions on the previous “hot ferroics. In this field they were mostly developed and
topic”—colossal magnetoresitive manganites. The liter- successfully applied by A. B. Harris, A. Aharony, and
ature now contains special reviews on multiferroics [12– their coworkers, see, for example, Ref. [17], No. 434202.
16], and in particular, a comprehensive discussion of dif- A detailed classification of possible symmetries allow-
ferent aspects of this field can be found in the collection ing for multiferroelectricity and magnetoelectricity is
of short review articles in Ref. [17]. given by H. Schmidt in Ref. [17], No. 434201.
To understand the basic phenomena in the field of
multiferroics, and appreciate the main achievements
and remaining problems in this field, it is necessary Type-I multiferroics
to classify multiferroics by the microscopic mechanism
that determines their properties. The first part of this ar- Type-I multiferroics are “older” and more numerous.
ticle is therefore devoted to this classification. In the sec- These are often good ferroelectrics, and the critical tem-
ond part of the article, we explore theoretical and prac- peratures of the magnetic and ferroelectric transitions
tical ideas that are emerging from the different mecha- can be well above room temperature. Unfortunately,
nisms underlying multiferroic behavior. the coupling between magnetism and ferroelectricity in
these materials is usually rather weak. The materials
challenge for this group of multiferroics is to keep all
their positive features, but enhance this coupling. As
How to combine magnetism and fer- we will see later, the opposite problem exists for type-II
multiferroics.
roelectricity: Different types of multi- One can single out several different subclasses of
ferroics type-I multiferroics, depending on the mechanism of
ferroelectricity in them. We will focus on four of the ma-
The microscopic origin of magnetism is basically the jor subclasses, but there are certainly others.
same in all magnets: it is the presence of localized
electrons, mostly in the partially filled d or f shells of
transition-metal or rare-earth ions, which have a cor- Multiferroic perovskites
responding localized spin, or magnetic moment. Ex-
change interactions between the localized moments lead Probably the best-known ferroelectrics are the per-
to magnetic order. The situation with ferroelectrics ovskites like BaTiO3 or Pb(ZrTi)O3 (PZT). There are
is quite different. There are several different micro- many magnetic materials among perovskites [18], and
scopic sources of ferroelectricity, and accordingly one also many ferroelectrics [19]. But comparing these ex-
can have different types of multiferroics. Generally tensive compilations, each containing more than 100
speaking, there are two groups of multiferroics. The pages of tables, demonstrates that there seems to be mu-
first group, called type-I multiferroics, contains those tual exclusion of magnetism and ferroelectricity in per-

DOI: 10.1103/Physics.2.20
c 2009 American Physical Society
URL: http://link.aps.org/doi/10.1103/Physics.2.20
 Physics 2, 20 (2009)

ovskites: there is practically no overlap of these tables.

Why? Whereas for magnetism one needs partially filled
d shells of a transition metal, practically all ferroelec-
tric perovskites contain transition metal ions with an
empty d shell, such as Ti4+ , Ta5+ , W6+ . Ferroelectric-
ity in these systems is caused by the off-center shifts of
the transition metal ion, which forms strong covalent
bonds with one (or three) oxygens, using their empty
d states. And somehow, the presence of real d electrons
in dn configurations of magnetic transition metals sup-
presses this process, preventing ferroelectricity in mag-
netic perovskites. This so called ” d0 vs dn problem” was
one of the first to be studied theoretically at the begin-
ning of the recent revival of multiferroics [7, 8], and al-
though there has been some progress, largely from ab
initio calculations, still in my opinion there is no full so-
lution of this problem [13]. The answer may well lie in
the fact that this mutual exclusion is not a “theorem,”
but rather a matter of numbers: in most cases a mag- FIG. 2: Different microscopic mechanisms found in type-I mul-
netic dn ion is stable in the center of its O6 octahedra, but tiferroics. (a) In “mixed” perovskites with ferroelectrically ac-
there may still be the cases where it is not. One possible tive d0 ions (green circles) and magnetic dn ions (red), shifts of
way around this problem may be making ”mixed” per- d0 ions from the centers of O6 octahedra (yellow plaquettes)
ovskites with d0 and dn ions, see Fig. 2. Unfortunately lead to polarization (green arrows), coexisting with magnetic
the coupling of magnetic and ferroelectric subsystems in order (red arrows). (b) In materials like BiFeO3 and PbVO3 ,
mixed perovskites is rather weak. the ordering of lone pairs (yellow ”lobes”) of Bi3+ and Pb2+
ions (orange), contributes to the polarization (green arrow).
(c) In charge ordered systems, the coexistence of inequivalent
sites with different charges, and inequivalent (long and short)
Ferroelectricity due to lone pairs bonds, leads to ferroelectricity. (d) The “geometric” mecha-
nism of generation of polarization in YMnO3 [24] describes the
In BiFeO3 , and probably in BiMnO3 and PbVO3 , Bi3+ tilting of a rigid MnO5 block with a magnetic Mn remaining at
the center. Because of the tilting, the Y-O bonds form dipoles
and Pb2+ play the major role in the origin of ferroelec-
(green arrows), and there appears two ”up” dipoles per one
tricity. In these ions, there are two outer 6s electrons “down” dipole so that the system becomes ferroelectric (and
that do not participate in chemical bonds. They are multiferroic when Mn spins order at lower temperatures). (Il-
called lone pairs, or sometimes dangling bonds. They lustration: Alan Stonebraker)
have a high polarizability—the condition required for
ferroelectricity in the classical description. More mi-
croscopically one can explain the origin of ferroelectric- more often we meet the situation in which there exist
ity in these compounds by ordering of these lone pairs ions with different charge (often because they are dif-
(with certain admixture of p orbitals) in one direction ferent elements), but on top of that there occurs dimer-
[Fig. 2]. Apparently this is what happens in many Bi− ization. TbMn2 O5 belongs to this class [11], as does the
and Pb−containing ferroelectrics and multiferroics such newly discovered multiferroic Ca3 CoMnO6 [21], see be-
as BiFeO3 , but it also helps to improve the ferroelectric low. Another similar possibility is when the bonds are
properties of Pb(Zrx Ti1− x )O3 . inequivalent because of the structure of the material, the
site-centered charge order appearing on top of that. This
is the case in an organic ferroelectric (TMTTF)2 X[22],
Ferroelectricity due to charge ordering and also in the multiferroic LuFe2 O4 [23].

One more mechanism that can lead to ferroelectricity

and to type-I multiferroicity can be charge ordering, of-
“Geometric” ferroelectricity
ten observed in transition metal compounds, especially
those formally containing transition metal ions with dif-
ferent valence. If, after charge ordering, both sites and Lastly, we consider a case realized in, for example,
bonds turn out to be inequivalent, this can lead to ferro- YMnO3 [24]. Ferroelectricity in YMnO3 has nothing to
electricity [Fig. 2] [20]. do with the magnetic Mn3+ , but is caused by the tilting
A corresponding mechanism can work in systems like of the practically rigid MnO5 block. This tilting occurs
Pr1/2 Ca1/2 MnO3 [20] or in nickelates RNiO3 (see Refs. just to provide closer packing, and as a result the oxygen
[15] and [17], No. 434217) with charge ordering. But ions move closer to the rather small Y ions [Fig. 2].

DOI: 10.1103/Physics.2.20
c 2009 American Physical Society
URL: http://link.aps.org/doi/10.1103/Physics.2.20
 Physics 2, 20 (2009)

Type-II multiferroics: Magnetic multiferroics

The biggest excitement nowadays is caused by the

discovery of a novel class of multiferroics in which fer-
roelectricity exists only in a magnetically ordered state
and is caused by a particular type of magnetism [10, 11].
For example, in TbMnO3 magnetic ordering appears at
TN1 = 41 K, and at a lower temperature, TN2 = 28 K,
the magnetic structure changes. It is only in the low-
temperature phase that a nonzero electric polarization
appears. Similar behavior occurs in TbMn2 O5 .
The first paper to study TbMnO3 showed that a
magnetic field can strongly influence the electric po-
larization: e.g., in TbMnO3 the polarization rotates (or
FIG. 3: Different types of spin structures relevant for type-II
“flops”) by 90 degrees when a critical magnetic field is
multiferroics. (a) Sinusoidal spin density wave, in which spins
applied along a certain direction [10]. In TbMn2 O5 [11] point along one direction but vary in magnitude. This struc-
the influence of an external field is even stronger: the ture is centrosymmetic and consequently not ferroelectric. (b)
polarization changes sign with field, and a field alternat- The cycloidal spiral with the wave vector Q = Q x and spins
ing between +1.5 and −1.5 Tesla leads to corresponding rotating in the (x,z)-plane. It is in this case where one finds
oscillations in the polarization. Since the discovery of nonzero polarization, Pz 6= 0. (c) In a so-called “proper screw”
these materials, a number of other type-II multiferroics the spins rotate in a plane perpendicular to Q. Here the inver-
with strong magnetoelectric coupling have been discov- sion symmetry is broken, but most often it does not produce
ered and studied. polarization, although in certain cases it might [27]. (Illustra-
From the point of view of the mechanism of multifer- tion: Alan Stonebraker)
roic behavior, one can divide type-II multiferroics into
two groups: those in which ferroelectricity is caused
by a particular type of magnetic spiral and those in
which ferroelectricity appears even for collinear mag- Magnetic frustration is the source of spiral magnetic
netic structures. ordering in an insulators. Therefore, these type-II multi-
ferroics are usually found in frustrated systems—a good
example of how the study of magnetic frustration, itself
Spiral type-II multiferroics a hot topic, has lead to new findings with potentially
very important applications.
Most of the type-II multiferroics known to date be- From these ideas and Eq. (1), one can understand the
long to this subgroup. Ferroelectricity appears in con- role of a magnetic field. The effect of an external field
junction with a spiraling magnetic phase, mostly of the H on a cycloid is analogous to the well-known spin-flop
cycloid type. This is the case in TbMnO3 , Ni3 V2 O6 , and transition in an antiferromagnet: in a magnetic field, the
MnWO4 . In TbMnO3 below TN1 = 41 K the magnetic sublattice moments prefer to lie in a plane perpendicu-
structure is a sinusoidal spin-density wave, where all lar to the field so they cant in the direction of the field
spins point in one direction, but the size of the local mo- with less cost in exchange energy between neighboring
ment varies periodically in space [Fig. 3]. (Note, this is spins. Consequently, when a magnetic field forces the
actually a sort of antiferomagnetic phase because the to- plane of a magnetic cycloid to flop by 90 degrees, the
tal moment for the magnet is zero.) Below TN2 = 28 K, polarization P also flops: according to Eq. (1), P lies in
the Mn spins order so that the tip of the spins sweep out the plane of spins but perpendicular to Q.
a cycloid [Fig. 3].
Katsura, Nagaosa, and Balatsky [25], using a mi- One word of caution about Eq. (1): Its derivation
croscopic approach, and Mostovoy [26], using a phe- assumes a particular crystal symmetry, such as simple
nomenological approach, showed that in a cycloidal spi- cubic or tetragonal. In these cases, Eq. (1) is indeed
ral a polarization, P, appears, that is given by: valid, and, for example, the polarization of a proper
screw indicated in Fig. 3, in which spins rotate in the
P ∼ rij × [Si × S j ] ∼ [Q × e] (1) plane perpendicular to the wave vector of a spiral Q,
would be zero. However, as recently shown by Arima
where rij is the vector connecting neighboring spins [27], this may not be the case for other symmetries, in
Si and S j , Q is the wave vector describing the spiral, which case there may appear a nonzero polarization
and e ∼ [Si × S j ] is the spin rotation axis [Fig. 3]. The even for a proper screw spiral. Experimentally, this
microscopic mechanism of this polarization is connected was observed in RbFe(MoO4 )3 [28] and more recently in
with the spin-orbit interaction (see Refs. [25] and [17], other layered triangular systems such as CuFeO2 and
No. 434208). ACrO2 ( A = Cu, Ag, Li, Na)[29].

DOI: 10.1103/Physics.2.20
c 2009 American Physical Society
URL: http://link.aps.org/doi/10.1103/Physics.2.20
 Physics 2, 20 (2009)

Type-II multiferroics with collinear magnetic structures cally shown in Fig. 4. The wall shown in Fig. 4 is called
a Bloch domain wall; in it the magnetization at the wall
The second group of magnetically driven ferro- rotates in the plane of the wall, perpendicular to the di-
electrics are those in which ferroelectricity appears in rection Q from one domain to the other. In the other
collinear magnetic structures—that is all magnetic mo- case, Fig. 4, which is called a Néel domain wall, spins
ments aligned along a particular axis—without the nec- rotate in the plane containing Q. Comparing Fig. 4 and
essary involvement of the spin-orbit interaction. Polar- Fig. 3, one immediately sees that one can consider the
ization can appear in these materials as a consequence of Néel domain wall as a part of a cycloidal spiral. Accord-
exchange striction because the magnetic coupling varies ing to the expressions in Eq. (1) we then should expect
with the atomic positions. The simplest example, found that in such a domain wall the electric polarization will
in Ca3 CoMnO6 [21], coincides with that shown in Fig. form, localized in the wall and directed as shown in Fig.
2. Ca3 CoMnO6 consists of one-dimensional chains of al- 4.
ternating Co2+ and Mn4+ ions. At high temperature the If now we put a system with such a Néel domain wall
distances between the ions along the chain are the same, in the gradient of an electric field, the wall and the elec-
the chain has inversion symmetry, and polarization is tric dipole associated with it will be either pulled toward
absent. Magnetic ordering, however, breaks inversion the region of stronger field or pushed out of it, depend-
symmetry: the spins form a ↑↑↓↓ type magnetic struc- ing on the orientation of a dipole. This effect was re-
ture. Due to an exchange striction the distortion of ferro cently suggested by Dzyaloshinskii [35], but even before
and antiferro bonds (↑↑ and ↑↓) is different, and we end his prediction it was actually discovered and studied ex-
up in the situation of Fig. 2: the material becomes fer- perimentally by a group in Moscow [36]. They did a
roelectric. Theoretical calculations [30] confirm this pic- conceptually very simple experiment: using films of the
ture. well-known insulating ferromagnet with a high transi-
In the case of Ca3 CoMnO6 , exchange striction arises tion temperature, iron garnet (BiLu)3 (FeGa)5 O12 , they
from transition metal ions with a different valence (Co2+ brought a sharpened copper wire to the surface and ap-
and Mn4+ ). One can, however, get the same effect even plied a voltage pulse. Under the influence of the result-
for identical magnetic ions, when one takes into account ing inhomogeneous electric field, the domain walls in
that the exchange in transition metal oxides usually oc- the film, which they visualized magneto-optically, start
curs via intermediate oxygens and depends on both the to move (or rather bend). They concluded that this hap-
distance between the metal ions and the metal-oxygen- pens because of the mechanism described above, and
metal bond angle. In RMnO3 perovskites where R is the authors could even estimate the resulting velocity
a small rare earth, the Mn magnetic order in the basal of a domain wall. This observation, besides confirming
plane is of the type ↑↑↓↓. As proposed by Sergienko, our general understanding, may be potentially very im-
Sen, and Dagotto [31], exchange striction in this case portant, as it could prove to be a means of controlling
can cause the oxygen ions to shift perpendicular to the domain structure, and consequently magnetic memory,
Mn-Mn bonds, which produces a polarization along the by applying a voltage rather than electric currents.
direction of the shift. This effect was observed in [32], Similarly, local electric polarization should appear in
although its value was much less than estimated theo- magnetic vortices, which are also widely studied nowa-
retically [31, 33]. days as a possible route to nonvolatile magnetic mem-
One more mechanism giving ferroelectricity even in ory [37]. To the best of my knowledge nobody yet tried
a collinear magnet is the “electronic” ferroelectricity in to influence such vortices in magnetic insulators by an
frustrated magnets [34]: the polarization of a triangle electric field (without currents).
of spins is proportional to the spin correlation function One more effect can be proposed. Consider an insulat-
S1 (S2 + S3 ) − 2S2 S3 (the numbers denote positions on ing ferromagnet that contains a spin wave (a magnon).
vertices of the triangle), and is nonzero if the latter is The real microscopic picture of a spin wave is the one
nonzero. shown in Fig. 5: the magnon is a spin deviation from
the average magnetization M, with each spin precessing
around M in time and with the phase shifts along the
propagation direction. But from Fig. 5 we see that the
Some spin-offs of the multiferroic instantaneous picture (the snapshot) of a spin wave is a
cone with constant M and with the perpendicular com-
story ponent forming a cycloid. According to Eq. (1), there
should appear in such a spin wave an electric polariza-
The lessons we learned by studying multiferroics, es- tion perpendicular to the magnetization and to the prop-
pecially those of type-II, are very helpful to predict or agation vector. Actually such a polarization was dis-
explain a number of miscellaneous phenomena, some covered for a static conic magnetic structure in CoCr2 O4
of which can be very useful for potential applications. (Ref. [17], No. 434204). Here, if we create a spin wave in
Consider for example the usual domain wall in a ferro- the form of a wave packet, this wave packet will carry
magnet. There are two types of such walls, schemati- with it both magnetization ∆M and an electric dipole

DOI: 10.1103/Physics.2.20
c 2009 American Physical Society
URL: http://link.aps.org/doi/10.1103/Physics.2.20
 Physics 2, 20 (2009)

FIG. 5: How polarization emerges in a spin wave (magnon).

(a) The classical picture of a spin wave in a ferromagnet: the
spin (red arrow) precesses about a fixed axis (blue). The devia-
tion is measured by the black arrows. (b) According to Eq. (1),
as a spin-wave packet propagates along Q, it will also carry an
electric dipole moment. (Illustration: Alan Stonebraker)

local polarization P normal to the surface that turns a

collinear magnetic structure into a cycloidal spiral.
Microscopically the mechanism for this effect is anti-
symmetric, or Dzyaloshinskii-Moriya, exchange of the
FIG. 4: Domain walls in a ferromagnet. (a) Bloch walls where type D · [S1 × S2 ]. This was indeed proposed in the
the spins rotate in a plane perpendicular to Q are in most cases original publications [38, 40] as an explanation of the ob-
not ferroelectric. (b) The spin rotation across a Néel wall is served effect. The use of the knowledge acquired in the
similar to the cycloidal spiral in Fig. 3(c), and, by Eq. (1), it will study of multiferroics makes this effect very simple and
develop a spontaneous polarization P localized at the domain transparent. One has an impression that it should be
wall. (c) If now one puts such a wall in the gradient of an present and should be taken into account in all magnetic
electric field, it will be pulled in or pushed out of the region thin films, although the strength of it would of course
of stronger field, depending on the sign of its dipole moment depend on the particular situation.
[36]. (Illustration: Alan Stonebraker)

moment d that moves with it [Fig. 5]. What next?

Another spin-off from the physics discussed above
concerns the magnetic structure of magnetic films on The field of multiferroics, becides promising impor-
different substrates. A beautiful experiment, carried out tant applications, contains rich and interesting physics.
recently in Germany [38], demonstrated that in an atom- It is now a very active and fruitful field. A lot is already
ically thin layer of manganese on tungsten, the Mn does known, many basic principles are established. Thus,
not form a simple antiferromagnetic structure, as one what is next?
would expect, but a long-periodic cycloidal spiral. And, One can think of several promising directions, some
most surprisingly, all such spirals have the same sense of of which have already been pursued. First, the search
spin rotation. Even earlier the observation was made by for new multiferroics is still proceeding at a rapid pace,
the same group that the Néel domain walls in the atom- and one should expect that many new examples and
ically thin layers of iron on tungsten also have a unique even new classes will be discovered. Each new multifer-
sense of spin rotation [39]. roic is a challenge in itself, every substance has its own
At first glance this is a very unexpected phenomenon. material-specific aspects, which one has to explore.
However, it becomes immediately clear if we remember Then, the understanding of the basic physics, together
the physics described by Eq. (1). According to this re- with the progress of ab initio calculations, opens a way
lation, a cycloidal magnetization in a type-II multifer- to create “multiferroics by design.” Examples of this ap-
roic creates a polarization, but we should also expect proach can be found in [41].
an inverse effect: a nonzero polarization would create From the point of view of the basic physics, we
a cycloidal spiral. This is apparently what happens in should mention the idea of toroidal moments—the third
thin films of manganese and iron on tungsten studied set of moments of electromagnetic field, besides the
in [38, 39] (and it is also the origin of a magnetic spi- well-known electric and magnetic moments. Three
ral in BiFeO3 ). Close to the surface there exists a drop basic ferroic properties—ferroelasticity (characterized
of the potential (i.e., the work function), and a dipole by the strain ε ij ), ferroelectricity (polarization P) and
layer forms at the surface, which produces an electric (ferro)magnetism (magnetic moments M) have different
field perpendicular to the surface. This field creates a transformation properties with respect to spatial inver-

DOI: 10.1103/Physics.2.20
c 2009 American Physical Society
URL: http://link.aps.org/doi/10.1103/Physics.2.20
 Physics 2, 20 (2009)

sion I and time reversal T: strain is I and T even, P is materials have a bright future.
I odd and T even, M is I even and T odd. There exists
yet the fourth possibility: a quantity which is both I −
and T −odd. It can of course be simply a coexistence of
polarization and magnetism, which we have in all mul-
tiferroics. But there exists also a special quantity with References
such symmetry, which is called a toroidal moment (or
sometimes anapole moment). The details one can find in [1] L. D. Landau and E. M. Lifshitz, Electrodynamics of continuous me-
Ref. [17], No. 434203; in the simplest case it can be de- dia (Fizmatgiz, Moscow, 1959).
fined as [2] I. E. Dzyaloshinskii, Sov. Phys. JETP 10, 628 (1959).
[3] D. N. Astrov, Sov. Phys. JETP 11, 708 (1609).
T = σri × Si (2) [4] H. Schmid, Ferroelectrics, 162, 317 (1994).
[5] E. Asher, H. Rieder, H. Schmid, and H. Stossel, J. Appl. Phys. 37,
1404 (1966).
The relation between toroidal moments and multifer- [6] G. A. Smolenskii and I. E. Chupis, Sov. Phys. Usp. 25, 475 (1982).
roicity is not straightforward. Both are I − and T −odd, [7] N. A. Hill, J. Phys. Chem. B 104, 6694 (2000).
but it does not mean that all multiferroics have hTi 6= 0 [8] D. I. Khomskii, Bull. Am. Phys. Soc. C 21.002 (2001).
or, vice versa, all systems with nonzero toroidal mo- [9] J. Wang et al., Science 299, 1719 (2003).
ments are multiferroics. What one can show is that [10] T. Kimura et al., Nature 426, 55 (2003).
[11] N. Hur et al., Nature 429, 392 (2004).
toroidal moments are directly related to the antisymmet- [12] M. Fiebig, J. Phys. D Appl. Phys. 38, R123 (2005).
ric magnetoelectric effect. As toroidal moments have [13] D. I. Khomskii, J. Magn. Magn. Mater. 306, 1 (2006).
now been observed in several systems, we can expect [14] W. Ehrenstein, N. Mazur, and J. Scott, Nature 442, 759 (2006).
to see more emerge from these ideas [42, 43]. [15] S. W. Cheong and M. V. Mostovoy, Nature Mater. 6, 13 (2007).
One more big direction in the study of multiferroics, [16] R. Ramesh and N. A Spaldin, Nature Mater. 6, 21 (2007); Ref. [17],
No. 434220.
which I did not touch upon above, is the investigation of [17] Special issue, J. Phys. Condens. Matter 20, 434201–434220 (2008).
their dynamical properties and elementary excitations. [18] J. B. Goodenough and J. M. Longo, Magnetic and Other Properties
From the intricate coupling between spin and charge de- of Oxides and Related Compounds, Landolt-Börnstein, Numerical
grees of freedom in multiferroics, one can expect quite data and Functional Relations in Science and Technology, New
Series Vol. III. 4 (Springer, Berlin, 1970).
nontrivial features of such excitations, such as the pos-
[19] T. Mitsui et al., Ferroelectrics and Related Substances, Landolt-
sibility to excite magnons with an electric field. And in- Börnstein, Numerical data and Functional Relations in Science
deed, these effects were already discovered experimen- and Technology, New Series Vol. 16 (1) (Springer, Berlin, 1981).
tally, giving rise to the term “electromagnon” (see Refs. [20] D. V. Efremov, J. van den Brink, and D. I. Khomskii, Nature Mater.
[44] and [17], No. 434210). In general, the investigation 3, 853 (2004).
[21] Y. J. Choi et al., Phys. Rev. Lett. 100, 047601 (2008).
of dynamical effects in multiferroics will certainly be ex- [22] P. Monceaux, F. Ya. Nad, and S. Brazovskii, Phys. Rev. Lett. 86,
panded in future. 4080 (2001).
And of course I left out the very big, and from a prac- [23] N. Ikeda et al., J. Phys. Soc. Japan. 69, 1526 (2000); Nature 436, 1136
tical point of view probably the most promising direc- (2005); Ref. [17], No. 434218.
tion: the fabrication and study of “artificial” composite [24] B. B.van Aken et al., Nature Mater. 3, 164 (2004).
[25] H. Katsura, N. Nagaosa, and A. V. Balatsky, Phys. Rev. Lett. 95,
multiferroics. For example, magnetic and ferroelectric 057205 (2005).
materials can be combined in the form of multilayers [26] M. V. Mostovoy, Phys. Rev. Lett. 96, 067601 (2006).
or self-organized nanostructures. The coupling between [27] T. Arima, J. Phys. Soc. Jpn. 76, 073702 (2007).
magnetic and ferroelectric subsystems in this case usu- [28] M. Kenzelmann et al., Phys. Rev. Lett. 98, 267205 (2007).
ally occurs via strain (although in principle it is not the [29] T. Kimura, J. C.Lashley, and A. P. Ramirez, Phys. Rev. B 73, 220401
(2006); S. Seki et al. Phys. Rev. Lett. 101, 067204 (2008); K. Kimura
only coupling channel), i.e., one tries to use magnetic et al., Phys. Rev. B 78, 140401 (2008).
systems with strong magnetostriction and ferroelectrics [30] Hua Wu et al., Phys. Rev. Lett. 102, 026404 (2009).
with large piezoelectric coefficients. The study of these [31] I. A. Sergienko, C. Sen, and E. Dagotto, Phys. Rev. Lett. 97, 227204
objects is now an extremely active field and has already (2006).
[32] B. Lorenz, Y. Q. Wang, and C. W. Chu, Phys. Rev. B 76, 104405
produced some spectacular achievements (such as the (2007).
creation of tiny magnetic sensors with the sensitivity ex- [33] S. Picozzi et al., Phys. Rev. B 74, 094402 (2006); Ref. [17], No.
ceeding that of even superconducting quantum interfer- 434208; A. Malashevich and D. Vanderbilt, Phys. Rev. Lett. 101,
ence devices (SQUIDs) [45] Some of the results in this 037210 (2008).
field are presented appear in reviews [14, 16, 46] and [34] L. N. Bulaevskii et al., Phys. Rev. B 78, 024402 (2008).
[35] I. E. Dzyaloshinskii, Europhys. Lett. 83, 67001 (2008).
Ref. [17], No. 434220. [36] A. S. Logginov et al., JETP Lett. 86, 115 (2007); Appl. Phys. Lett. 93,
Summarizing, I can say that the field of multiferroics 182510 (2008).
and magnetoelectrics as a part of a recent trend to cre- [37] A. Wachowiak et al., Science 298, 577 (2002).
ate and use multifunctional materials, is already an ex- [38] M. Bode et al., Nature 447, 190 (2007); P. Ferriani et al., Phys. Rev.
tremely active field. And one can be sure that there is Lett. 101, 027201 (2008).
[39] A. Kubetska et al., Phys. Rev. Lett. 88, 057201 (2002); E. Y. Vedme-
still much more to come, and the development of this denko et al., Phys. Rev. Lett. 92, 077207 (2004).
field will go still very actively both in the direction of [40] M. Heide, G. Bihlmayer, and S. Bluegel, Phys. Rev. B 78, 140403
basic physics and in applications, where, it seems, these (2008).

DOI: 10.1103/Physics.2.20
c 2009 American Physical Society
URL: http://link.aps.org/doi/10.1103/Physics.2.20
 Physics 2, 20 (2009)

[41] C. Fennie, Phys. Rev. Lett. 100, 167203 (2008); K. T. Delaney, M. and 434210.
Mostovoy, and N. A. Spaldin, arXiv:0810.0552 (2008). [45] J. Zhai et al., Appl. Phys. Lett. 88, 062510 (2006).
[42] A. M. Kadomtseva et al., JETP Lett. 79, 571 (2004). [46] C.-W. Nan et al., J. Appl. Phys. 103, 031101 (2008).
[43] B. B. van Aken et al., Nature 449, 702 (2007).
[44] A. Pimenov et al., Nature Phys. 2, 97 (2006); Ref. [17], Nos. 434209

About the Author

Daniel Khomskii
Daniel Khomskii graduated from Moscow University in 1962. Starting in 1965, he worked
in the Theoretical Department of the Lebedev Physical Institute of the Russian Academy
of Science in Moscow. There, he defended his Ph.D. in 1969. In 1980, he obtained a second
doctoral degree–the Russian equivalent to the German Habilitation or a professorship in
the US. From 1992 to 2003, he was a Professor at Groningen University in the Netherlands
and since 2003, he has been a guest Professor in Köln (Cologne University) in Germany.
His main research interests are the theory of systems with strongly correlated electrons,
metal-insulator transitions, magnetism, orbital ordering (“Kugel-Khomskii model”) and
superconductivity. He is a recently elected Fellow of the American Physical Society and
has published roughly 300 papers over the course of his career.

DOI: 10.1103/Physics.2.20
c 2009 American Physical Society
URL: http://link.aps.org/doi/10.1103/Physics.2.20