Introduction on Multiferroic Materials

Xiaotian Zhang([email protected])
Instructor: Elbio Dagotto
Class: Solid State 2, 2010, Spring semester
Department of Physics and Astronomy
The University of Tennessee, Knoxville, TN 37996, USA
(Dated: Feb.16th.2010)

Abstract
A brief introduction on the concept of Multiferroic Materials is present with a simple review on
the recent research discoveries related to this topic. In this presentation, we will first explain what
is so called Multiferroic Materials and then set some popular ones as examples. What’s more, we
are going to explore the important physical properties of these materials and have a look at the
phase diagrams.

1
I. THE CONCEPT OF MULTIFERROICS

A. History

Electric and magnetic fields and related phenomenas have been considered to be im-
portant since Maxwell, and they were clearly shown to be unified in the special theory of
relativity. However, people used to think that ferroelectricity and magnetism are exclusive
to each other and can only have weak coupling when they coexist.
The term multiferroic was first used by H. Schmid in 19944. His original definition
referred to multiferroics as single phase materials which simultaneously possess two or more
primary ferroic properties among the 4 in the following subsection. Today people extend
the definition to include other long-range orders, such as antiferromagnetic.In this extended
definition , we can say that this research field was born in the 60’s even the 50’s. From then
people started to do related studies under the name magnetoelectric.
From the figure it is clear that there is a burst in the number of papers at the beginning
of this centry. The direct reason is that in 2003, people discovered large ferroelectric polar-
ization in the epitaxially grown BiF eO3 6 thin film and strong magnetoelectric coupling in
T bMnO3 3 and T bMn2 O5 8 . Especially in the so called frusstrated multiferroics, the electric
dipole moments are very sensitive to the applied magnetic fields, although its magnitude is
still much smaller than typical ferroelectrics. These new discoveries broke the traditional
thinking about multiferroics and opened a new gate for industry: if charge can be controled
by external magnetic fields and spins can be controled by applied voltage, new devices with
variaties of functions may be designed.5

FIG. 1: The number of papers using the keywords M agnetoelectric and M ultif erroic from
1950 to 2008, this picture is taken from: http : //en.wikipedia.org

2
 B. Ferroics

There are four types of ferroics:

F erroelectric materials possess a spontaneous polarization that is stable and can be

switched hysteretically by an applied electric field; antiferroelectric materials possess
ordered dipole moments that cancel each other completely within each crystallographic
unit cell.

F erromagnetic materials possess a spontaneous magnetization that is stable and can

be switched hysteretically by an applied magnetic field; antiferromagnetic materials possess
ordered magnetic moments that cancel each other completely within each magnetic unit cell.

F erroelastic materials display a spontaneous deformation that is stable and can be

switched hysteretically by an applied stress.

F errotoroidic materials possess a stable and spontaneous order parameter that is taken
to be the curl of a magnetization or polarization. By analogy with the above examples, it
is anticipated that this order parameter may be switchable. Ferrotoroidic materials have
evaded unambiguous observation.

C. Order Parameter Couplings

The order parameter couplings are described by:

Magnetoelectric describes the influence of a magnetic (electric) field on the polarization

(magnetization) of a material.

P iezoelectricity describes a change in strain as a linear function of applied electric field,

or a change in polarization as a linear function of applied stress.

3
 P iezomagnetism describes a change in strain as a linear function of
applied magnetic field, or a change in magnetization as a linear function of applied stress.

Electrostriction describes a change in strain as a quadratic function of applied electric

field.

We will focus on the ferromagnetic and ferroelectric because they are of most physical
concern.

D. Examples of Multiferroics

Typical multiferroics belong to the group of the perovskite (F ig.2) transition metal oxides,
and include rare-earth manganites and ferrites such as T bMnO3 , Y Mn2 O5 and LuF e2 O4 .
Other examples are the bismuth compounds BiF eO3 and BiMnO3 , and non-oxides such as
BaNiF4 and spinel chalcogenides like ZnCr2 Se4 .

FIG. 2: Perovskite structure

4
II. PROPERTIES OF MULTIFERROICS

A. Types and Possible Mechanisms

1.Charge ordering is one way leads to ferroelectricity. When the ions are also magnetic,
multiferroic is formed. A famous example is LuF e2 O4 9 , the ordering of F e2+ and F e3+
provides ferroelectricity. Meanwhile, site-centered and bond-centered charge ordering in
Rx Ca1−x MnO3 10 can also provide ferroelectricity.
2.Geometrical multiferroic is another type. The mechanism of ferroelectricity is consid-

FIG. 3: Left:site-centered charge ordering. Mid:bond-centered charge ordering.

Right:ferroelectricity by Charge Ordering. This picture is taken from ref.10.

ered as nonlinear couplings between different lattice distortions, suggested by first principle
calculations11 . A proper example of this compound would be hexagonal RMnO3 , in which
the coupling of different phonon modes creates ferroelectricity. Geometrical reasons also
produce magnetical induced ferroelectricity.
3.In some systems, Heisenberg spin-spin coupling with ferromagnetic for nearest neighbour
and antiferromagnetic for next nearest neighbour will provide spiral spin phase. And this
phase may come with ferroelectricity via the coupling P~ = γ × (M ~ · ∇M ~ −M ~∇·M ~ )12

(obtained from thermodynamics). The microscopic mechanism maybe explained as the rel-
ativistic correction of the Heisenberg term, which is called Dzyaloshinskii-Moriya interaction
D~ij · S~i × S~j . Moreover, in certain systems Ising type spin-spin (also ferromagnetic for near-
est neighbour and antiferromagnetic for next nearest neighbour) interaction will lead to the
↑↑↓↓ ground state, which will also cause a lattice distortion.
Let us discuss more about the DM interaction here. First of all, what is the origin of it? The

5
 FIG. 4: An illustration of geometrical frustration. The spin on the third site can be in any
direction, but then the interaction between this site and the other 2 will be different. As a
consequence, the third site will move closer to one site and away from another, which will break
the symmetry and induce ferroelectricity. This picture is gained from: http : //en.wikipedia.org

1
answer is the relativistic correction to the Hamiltonian up to the order of c2
. As well known,
the Heisenberg super exchange interaction is due to the second order virtual hopping of the
electrons, and the spin of every electron is conserved during the hoppings. However, in rela-
tivistic quantum mechanics, the z component of spin is no longer a good quantum number:
it may flip once during the second order virtual hopping (It may also flip twice, of course.
1
But that will be at the order of c4
and can be neglected here). These virtual hoppings with
one time spin flip will lead to the DM interaction in the same way as non-relativistic virtual
hoppings lead to the Heisenberg interaction.
Secondly, how does the DM interaction lead to ferroelectricity? Intuitively, people can argue
that the relativistic correction can be expressed by the spin-orbital interaction, which will
try to rotate the atomic orbitals in order to align them with the spins. This rotation of the
atomic orbitals, obviously, will repulse the oxygen ion between the 2 sites via Coulomb force,
and result in ferroelectricity. Mathematically, the vector D~ij is found to be propotional to
R~ij × x~ij , where R~ij is the position difference of the 2 sites and x~ij is the displacement of
the oxygen ion.

6
 4.Lone pair multiferroics: in some perovskite materials, the 6p orbital of the A site and the

′
FIG. 5: Spin frustration structures appear under certain range of J /J in both Heisenberg
model (spiral spin) and Ising model ( ↑↑↓↓). This picture is taken from ref.5.

2p orbital of the O atoms will have a hybridization due to strain from surfaces, which leads
to the distortion of the A sites so that ferroelectricity appears7 . Ferromagnetic structure is
due to the B sites.

FIG. 6: Induced polarization induced by strain on the (111) plane of BiM nO3 as indicated by
ref.7.

7
 B. Reversal symmetry

Ferromagnetic and ferroelectric have the same reversal symmetries as magnetic and
electric field. From the Maxwell equations

~ = 4πρ
∇·E
~ (1)
~ ∂B
∇×E =−
∂t
it is clear that electric fields are invariant under time reversion but will get a minus sign
under spatial reversion, while magnetic fields are invariant under spatial reversion but will
get a minus sign under time reversion. And this totally opposite property in symmetry is
one of the reasons for people to think ferromagnetism and ferroelectricity exclude each other.

FIG. 7: Reversal symmetries table for ferroics from ref.5.

C. Difficulty of Traditional Landau Theory

~ H
Traditional materials under external fields E, ~ have free energies expressed by the Lan-
dau phase transition theory
1 1
−F (E, H) = ǫij Ei Ej + µij Hi Hj + αij Ei Hj
2 2 (2)
+ βijk Ei Hj Hk + γijk Hi Ej Ek + ...

and one easily obtains

Pi = αij Hj + βijk Hj Hk
(3)
Mi = αji Ej + βijk Ej Ek

However, these expressions are not prefered in the research of multiferroics. The reason is
that those coupling parameters will display field hysteresis, which make them dependent

8
on the history of the material. People are trying to use the resultant field, instead of the
external one, to describe the state of multiferroics.
Another challenge arise from the higher order terms. These non-linear terms will dominate
in large fields while the magnetoelectric coupling is already complicated. What’s more, the
effects of strain, which is not included in the above equation, can also be significant.
All of the above strongly desire a microscopic and quantum machenical, instead of macro-
scopic and thermodynamical, model in order to study ferroics theoretically. However, it
seems to be very difficult to establish a simple but accurate enough microscopic model for
all the ferroics.

D. Microscopic Doulbe-Exchange Model

A complete microscopic model should contain:

1.Electron hopping between sites.
2.Coulomb interaction between electrons.
3.Super-exchange coupling between spins.
4.Electron-phonon coupling.
Bad news: these 4 kinds of interactions do not mean that there are only 4 terms in the
Hamiltonian. The complexity increases rapidly with the number of orbitals, and 2 body
interaction should be expressed by 4 Fermion operators in a many body basis. For example,
although there is only one term for the Coulomb interaction in the single orbital problems
(Hubbard model), we need 4 terms in a two orbital problem:
X X X X
S~iα · S~iβ + J c†α↑ c†α↓ cβ↓ cβ↑
′ ′
U niα↑ niα↓ + U niα niβ + J (4)
i,α i,α6=β i,α6=β i,α6=β

The similar thing happens to the super-exchange coupling. As stated above, the multifer-
roics are mainly transition metal oxides, and multiorbital problems are really common.
Good news: there are particular systems in which the multiorbital problems can be easier.
Maganites, in which the 5 d orbitals are split into 3 degenerate t2g orbitals with lower en-
ergy and 2 degenerate eg orbitals with higher energy, is a system that can be studied by
significantly simplified model. People found that in such systems, first of all, the Coulomb
interaction between electrons are so strong that any orbital is hardly double occupied. Sec-
ondly, the Hund coupling (in fact one part of the Coulomb interaction) is also much stronger

9
than the hopping, and then one can make the approximation that all electrons on the same
site must have the ”same” spin. The most drammatic thing is that numerical calculations
showed that for spin-3/2, quantum spin and classical spin give very close results, so that
the t2g spin can be treated classically (the 3 electrons form a spin-3/2 state by the infinite
Hund coupling) in the super-exchange coupling.
Finally, we obtain the so called double-exchange model for maganites:
X X
H= ti,j,α,β Ωi,j c†i,α,σ cj,β,σ + JAF S~i · S~j
i,j,α,β,σ i,j
X 1
+ [λ(−Q1,i I2×2 + Q2,i τix + Q3,i τiz ) + (γQ21,i + Q22,i + Q23,i )] (5)
i
2
X
+ D~ij · S~i × S~j
i,j

Where Ωi,j is called barry phase, comes from the fact that the hopping amplitude will be
dependent on the spin directions of the 2 sites due to infinite Hund coupling. The Q′ s
are normal modes of the so called Jahn-Teller phonon (there are other 3 modes with no
significance), and the gamma, which is usually taken as 2, appears due to the Q1 mode has
a spring constant different from Q2 and Q3 .

E. The Phase Diagrams

Now let us pay attention to some phase diagrams of the multiferroics.

F. The Future

1.The first challenge is to find a efficient machenism via which electric polarization can
induce magnetic structures. Although an external magnetic field can strongly affect the
electric properties of a multiferroic material, external electric fields have rare effect on the
magnetic properties of the material. Because electric polarization does not break the time
reveral symmetry, it can not produce magnetical order in the same way as how magnetic
frustrations induce ferroelectricity.
2.For practical applications, it is important to find more and more room temprature mul-
tiferroics. However, the magnetic coupling is ususally not strong enough to hold at room
temprature (but it is not impossible, for example, hexaferrite Ba0.5 Sr1.5 Zn2 F e12 O22 ).

10
 FIG. 8

FIG. 9: New phase found in large JAF by Shuai Dong appears to be ferroelectric.

3.For theoretical physics, more detailed and comprehensive understanding to the microscopic
mechanism and dynamics of the magnetoelectric coupling is desired. This requires both a
more detailed study via double-exchange model on the lucky systems and a developement
of brilliant ideas to explore what we can do for the more complicated systems. An alterna-
tive way is to develope more advanced numerical methods in order to calculate multiorbital
many body problems more accurate and efficient, which is always important, not only for

11
FIG. 10: Polarization of T bM nO3 under external magnetic field along the b axis. A critical field
5T turns the polarization from c axis to a axis.(ref.5)

FIG. 11: Dielectric constant of DyM n2 O5 along the b axis under applied magnetic field along
the a axis. Under 25T , the dielectric constant is strongly field-dependent, but then it becomes
almost field-independent and a peak appears at 38T .(ref.5)

multiferroics but also for all the other problems in computational physics.

1 W. Eerenstein, et al. Nature 442, 759 (2006).

2 T. Fukushima, et al. Arxiv:1002.2556v2.
3 T. Kimura, et al. Nature 426, 55 (2003).

12
4 H. Schmid. Ferroelectrics 162, 317 (1994).
5 S. Cheong, et al. Nature materials 6, 17 (2007).
6 J. Wang, et al. Science 299, 1719 (2003).
7 A.J. Hatt, et al. Eur. Phys. J. B 71, 435 (2009).
8 N. Hur, et al. Nature 429, 392 (2004).
9 N. Ikeda, et al. Nature 436, 1136 (2005).
10 D. Efremov, et al. Nature 3, 853 (2004).
11 C. Fennie, et al. Phys. Rev. B 72, 100103 (2005).
12 M. Mostovoy, et al. Phys. Rev. Lett 96, 067601 (2006).

13