Vous êtes sur la page 1sur 52

2

Magnetic Anisotropy of Heterostructures


J
urgen Lindner and Michael Farle
Fachbereich Physik and Center for Nanointegration, Universit
at Duisburg-Essen,
Lotharstr. 1, 47048 Duisburg, Germany
juergen.lindner@uni-due.de

Abstract. The chapter provides a detailed introduction to magnetic anisotropy of


ferromagnetic ultrathin films and its analysis by ferromagnetic resonance on a tutorial level. While the microscopic origins of the magnetic anisotropy as well as recent
developments in its theoretical description are shortly discussed, emphasis is put on
a phenomenological description using the free energy of the system together with
its symmetries. The formalism is used to describe ferromagnetic resonance experiments which present an extremely sensitive method to experimentally investigate
magnetic anisotropy in thin film heterostructures. Expressions for the free energy
and the resonance equations are derived for the most widely used crystal symmetries
such as cubic, tetragonal and hexagonal. The general equations are illustrated by
giving selected examples of current research on thin metallic films on different kinds
of substrates (MgO, GaAs and Cu).

2.1 Introduction
Magnetic thin films have provided a highly successful test ground for understanding the microscopic mechanisms which determine macroscopically
observable quantitities like the magnetization vector, different types of magnetic order (ferro-, ferri- and antiferromagnetism), magnetic anisotropy and
ordering temperatures (Curie, Neel temperature). The success has been
based on the simultaneous development of the following techniques: a) the
preparation of single-crystalline mono- and multilayers on different types of
substrates in ultrahigh vacuum systems, b) the development of vacuum compatible, monolayer-sensitive magnetic analysis techniques, c) the advance in
computing power to provide first-principles calculations of magnetic ground
state properties [1]. Aside from these basic research orientated investigations
the technological exploitation of thin film magnetism has lead to huge increases in the hard disks magnetic data storage capacities [2] and new types
of magneto-resistive angle and position sensitive sensors in the automotive
industry, for example.

J. Lindner and M. Farle: Magnetic Anisotropy of Heterostructures, STMP 227, 4596 (2007)
c Springer-Verlag Berlin Heidelberg 2007
DOI 10.1007/978-3-540-73462-8 2
!

46

J. Lindner and M. Farle

The purpose of this article is twofold: a) an introductory level overview on


magnetic anisotropy, b) characteristic examples of current research using magnetic resonance techniques to quantitatively determine magnetic anisotropy
and explore its microscopic sources.
Various aspects of ultrathin film magnetism [3] have been discussed in extensive reviews and book chapters over the last few years (see for example
[4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15]). There is no way to summarize all
these issues in such limited space and the reader is referred to the reviews
mentioned before. For an overview on the technological relevance or the many
experimental techniques and methods that have been developed to investigate magnetic heterostructures the reader is referred to the series of books
by Heinrich and Bland [6, 7, 8]. This review also excludes laterally structured
samples (for excellent reviews see e.g. [9, 10, 11]) and epitaxially grown films
comprised of two or more elements (double- tri- and multilayers, see the article
of B. Heinrich in this book or [6, 7, 8, 14]) in which coupling effects lead to
phenomena like the tunneling magneto-resistance (TMR), the giant magnetoresistance (GMR) or spin current related effects (see e.g. [16, 17, 18] for a
detailed discussion) such as current induced switching [19], current induced
domain wall movement or spin torque induced magnetic damping [20, 21, 22].
The examples which will be discussed here are strictly restricted to the thickness and temperature dependent magnetic anisotropy of single element ferromagnetic metallic monolayers on different kind of single crystalline substrates
(metals, semiconductors and insulators). It will be shown that epitaxial films
consisting of few atomic layers provide an interesting playground for artificially controlling magnetic properties and hence improving the understanding
of the underlying physical mechanisms.
This chapter is divided into two basic sections. While the first will shortly discuss the sources of magnetic anisotropy energy (MAE), explain Ferromagnetic
Resonance (FMR) and introduce a phenomenological description of the MAE
and its influence on the FMR resonance equations in terms of the magnetic
part of the free energy of the system, the second section will give examples of
FMR investigations of heterostructures. In the framework of a tutorial description, the prototype systems Fe/MgO(001), Fe/GaAs(001) and Ni/Cu(001)
were chosen.

2.2 Origin of Magnetic Anisotropy


Magnetic anisotropy describes the fact that the energy of the ground state of
a magnetic system depends on the direction of the magnetization. The effect
occurs either by rotations of the magnetization vector with respect to the
external shape of the specimen (shape anisotropy) or by rotations relative to
the crystallographic axes (intrinsic or magneto-crystalline anisotropy). The
direction(s) with minimum energy, i.e. into which the magnetization points
in the absence of external fields are called easy directions. The direction(s)

2 Magnetic Anisotropy of Heterostructures

47

with maximum energy are called hard direction. The MAE between two
crystallographic directions is given by the work WMAE needed to rotate
the magnetization from an easy direction into the other direction. The
MAE is a small contribution on the order of a few eV/atom to the total energy (several eV/atom) of a bulk crystal. To estimate the magnitude
of the MAE one can use as a rule of thumb, that the lower the symmetry of the crystal or of the local electrostatic potential (crystal or ligand field) around a magnetic moment, the larger the MAE is. This becomes evident, if one remembers that in a crystal field of cubic symmetry
the orbital magnetic moment is completely quenched in first approximation
[23]. Only by calculating in higher order (2nd) or by allowing a slight distortion of the cubic crystal a small orbital magnetic moment, i.e. a nonvanishing expectation value of the orbital momentums z component is recovered. Without the presence of the orbital momentum which couples the
spin degrees of freedom to the spatial degrees of freedom the MAE would
be zero, since the exchange interaction is isotropic. One should also note,
that the easy axis can deviate from crystallographic directions as for example in the case of Gd whose easy axis is temperature dependent and
lies between the c-axis and the basal plane at T = 0 K [24]. Table 2.1
gives an overview about easy and hard axes and on the magnitude of the
MAE for some elementary ferromagnetic materials with different crystal
symmetry.
There are fundamentally two sources of magnetic anisotropy: (i) spin-orbit(LS)
interaction and (ii) the magnetic dipole-dipole interaction. From the point
of view of quantummechanics both interactions are relativistic corrections
to the Hamilton-operator of the system that lift the rotational invariance
of the quantization axis. Despite the fact that the dipole-dipole interaction
as well as the LS coupling are much weaker than the exchange interaction
( 1 100 eV/atom compared to 0.1 eV/atom), they link the magnetic
Table 2.1. Anisotropic orbital moments, direction of easy axis of the magnetization,
and magnetic anisotropy energy at T = 0 K for the four elemental ferromagnets as
taken from standard references like Landolt-B
ornstein [32]. L is the difference of
the orbital magnetic moment measured along the easy and hardest magnetization
direction. tot is the sum of orbital and effective spin moment. The latter includes
the so-called !Tz " contribution entering the sum rule analysis of x-ray magnetic
circular dichroism measurements
Material

Easy axis

|MAE| (eV/atom)

L /tot

bcc Fe
hcp Co
fcc Co
fcc Ni
hcp Gd

!100"
!0001"
!111"
!111"
tilted hcp

1.4
65
1.8
2.7
50

1.7 104
4.5 104
1.8 104
103

48

J. Lindner and M. Farle

moments to position space. In the following we will use the term magnetic
anisotropy energy or MAE for the magnetic anisotropy energy density, resulting from spin-orbit interaction which includes the so-called magnetocrystalline and also the magneto-elastic contributions [15]. Contributions originating from the magnetic dipole-dipole interaction will be called shape or
magneto-static anisotropy.
We note that the important aspect of magnetic domain formation is not discussed here, since we restrict our discussion to the intrinsic contributions to
the magnetic anisotropy and the main technique which will be discussed in
detail in this chapter is ferromagnetic resonance (FMR). This technique is
in most cases performed in large enough fields that drive the magnetic film
into a single domain state. The reader should keep in mind, however, that the
interplay of magnetostatic energy, exchange energy and magnetic anisotropy
in general leads to an energetically favored multi-domain state. Especially,
the analysis of hysteresis loops is complicated due to domain formation at
small magnetic fields. Special imaging techniques have been developed [25] to
obtain quantitative understanding of the many domain configurations. Due to
the single-domain description used in this chapter, aspects of configurational
magnetic anisotropy [26] which appear in submicron sized nanomagnets due
to small deviations from the uniform state will not be discussed. Similarly,
so-called exchange anisotropies which may arise from different exchange coupling constants along different crystallographic directions in a crystal will not
be specifically addressed, since the experimental observations can be well described by the phenomenological approach presented in Sect. 2.3.2. Also a
discussion of unidirectional anisotropy or exchange bias is beyond the scope
of this chapter. Excellent overviews can be found in [27, 28, 29, 30]. Finally, we
note that the aspect of a non- homogeneous magnetization across the thickness of a several nanometer thick film does not enter the following discussion.
An excellent overview on the magnetization profile across a thin film and its
dependence on the films morphology has been recently given by Jensen and
Bennemann [31].
2.2.1 Spin-orbit Interaction
In a classical picture the orbital motion of the electrons in a perfect crystal is
defined by the potential that is predetermined by the crystal lattice. In case
that there is an interaction between the orbital motion and the spin of the
electrons (i.e. when spin-orbit interaction is present), the spins and thus the
magnetization become coupled to the lattice. By using perturbation theory in
which the LS coupling is described as perturbation of the exchange splitting
Bruno [33] showed that the energy correction and the orbital moment of the
L . Here is the radial part of
minority spins are related as ELS 41 S
is the unit vector along the spin direction, deterthe spin-orbit interaction, S
mining the magnetization direction and L the orbital angular momentum of

2 Magnetic Anisotropy of Heterostructures

49

L is the projection of L on the magnetithe minority spin band, so that S


zation direction. The magnetic anisotropy energy for uniaxial symmetry, for
which the energy differs between the directions parallel ($) and perpendicular
() to one anisotropy axis, is given by the anisotropy of the orbital angular
momentum L = L L $ (or, respectively, the anisotropy of the orbital
$
moment L = L L ):

ELS L =
L .
MAE = ELS
4
4B

(2.1)

According to Bruno the easy axis is the one, where the orbital moment is
largest. This fact which is experimentally often overlooked yields information
on the intrinsic origin of the macroscopically measured MAE by straightforward SQUID magnetometry measurements along different crystallographic
axes. The saturation magnetizations along the easy and hard axes are different! The effect is very small, in bulk crystalson the order of 104 , but
measurable, and well documented (see Table 2.1 and [34]). Here, one should
note that shape anisotropy is not involved. That is to say, that when taking
the shape anisotropy (see Sect. 2.3.2) into account, the equilibrium (easy)
direction of magnetization in zero field may be a hard magnetocrystalline
anisotropy direction with the smaller orbital moment. While Bruno assumed
a fully occupied majority spin band in his model (exchange splitting much
larger than the bandwidth), this restriction was dropped in the later work of
van der Laan [35], who extended Brunos relation by including the majority
spin band orbital moment L :
"
!
L L .
(2.2)
MAE
4B

Although approaches that employ perturbation theory have the advantage


of being less complex, they often yield wrong results on a quantitative basis (in most cases too large values). Ab initio theories that consider the LS
coupling within a fully relativistic ansatz lead to a clear improvement. They
are, however, much more elaborate as the precision of the calculation of the
total energy of the system has to be very high. The reason is that the overall total energy is of the order of 1 eV/atom, while the MAE is very much
smaller and of the order of several eV/atom. Nevertheless, a considerable
progress on ab-initio calculations of the MAE was achieved within the last
1015 years (see e.g. [31, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49]).
The correlation between the anisotropy of the orbital angular momentum and
the MAE becomes strikingly evident in experimental results on ultrathin films
[50] and few atom nanostructures of Co [51], for which orbital anisotropies up
to L /L 20% have been experimentally confirmed. In general, however,
one should note that a direct proportionality between L and the MAE is
only correct for L 0 [41].

50

J. Lindner and M. Farle

2.2.2 Dipole-Dipole Interaction


The magnetic field produced by a dipole i at the position r i is given by:
Hi (r i ) =

3 (r i i ) ri
i
3 .
ri5
ri

(2.3)

Due to this field a second dipole j in the distance rij with respect to the
first has the energy EDip = j H i :
EDip =

i j
3 (rij i ) (r ij j )

.
3
5
rij
rij

(2.4)

Since the dipoles are placed on periodic positions within a crystal lattice, the
axis rij connecting the two dipoles is linked to the crystallographic directions
and, in fact, the interaction energy is connected to the relative orientation of
the crystallographic axes and the direction of the magnetic moments. This in
turn leads to magnetic anisotropy.

2.3 Models of Magnetic Anisotropy


Phenomenologically, the crystallographic easy axis of the magnetization is
determined by the minimum of the free energy F 1 . Before the explicit expressions of F for various crystal symmetries are discussed, the microscopic
origins that explain magnetic anisotropy will be shortly described.
2.3.1 Single Ion Anisotropy
The single ion anisotropy is determined by the interaction between the orbital
state of a magnetic ion and the surrounding crystalline field, when the crystal
field is very strong. The anisotropy is the result of the quenching of the orbital
moment by the crystalline field. As this field has the symmetry of the crystal
lattice, the orbital moments can be strongly coupled to the lattice. This interaction is transferred to the spin moments via the spin-orbit coupling, giving a
weaker electron coupling of the spins to the crystal lattice. When an external
field is applied the orbital moments may remain coupled to the lattice whilst
the spins are more free to turn. The magnetic energy depends on the orientation of the magnetization relative to the crystal axes.
1

In the chapter by B. Heinrich in this book, the Gibbs free energy U is used instead
of F . As long as no external energy contributions are incorporated into F , the
free energy is the appropriate thermodynamic potential. Upon inclusion of the
Zeeman energy of the external field, however, U is the relevant potential. However,
to avoid confusion, we will use the term F throughout the whole chapter.

2 Magnetic Anisotropy of Heterostructures

51

In a magnetic layer, the single-ion anisotropy is present throughout its volume,


and contributes in general to the volume part of the MAE. In transition metals this contribution is usually much smaller than the shape anisotropy but
can be comparable in magnitude in rare earth metals. However, in some cases
also for 3d metals or alloys with small magnetization (and consequently shape
anisotropy) the single ion contribution might overcome the shape anisotropy,
leading to perpendicular magnetic anisotropy, for which the easy axis of magnetization is aligned perpendicular to the film surface.
The single-ion anisotropy can also contribute to the surface anisotropy via
Neel interface anisotropy [52], where the reduced symmetry at the interface
strongly modifies the anisotropy at the interface compared to the rest of the
layer [53]. One may question the relevance of this purely phenomenological approach, which has also been extended in terms of two-ion anisotropy contributions. The importance lies in the fact that a simple model for the temperature
dependence of the MAE and its correlation to the temperature dependence
of the magnetization can be established. More recently, Mryasov et al. [54]
have put this model on solid ground by showing that a model of magnetic
interactions on the basis of first-principles calculations of non-collinear magnetic configurations in FePt effectively contains the observed single-ion and
two-ion contributions and explains the observed unusual scaling exponent
between the magnetization and the MAE (K(T ) M (T ) , see also Sect. 2.4).
Later on, in Sect. 2.5, we discuss that for ultrathin Fe films a single-ion model
( = 3) yields an excellent explanation for the measured correlation of M (T )
and K(T ).
2.3.2 Free Energy Density
As stated above the magnetic anisotropy energy is the work WMAE needed
to rotate the magnetization between two different directions. If this rotation
is performed at constant temperature T , the MAE is given by the difference
of the free energy F of the system with the magnetization pointing along
the two directions. This is easy to see when one considers that for a closed
system (no exchange of particles) dF = dW SdT , S being the entropy, at
constant T #reduces to dF = dW . Setting dW dWMAE this in turn yields
2
F2 F1 = 1 dWMAE = MAE, where 1 and 2 denote the initial (e.g. easy)
and the final direction of the magnetization. Provided that an expression for
the magnetic part of F is given for the system under consideration, it can be
used to interpret the FMR data on thin magnetic films. The phenomenological
expression for F is usually found by symmetry considerations. In the following
we will summarize the expressions for F of the most often used symmetries.
We discuss cubic, tetragonal as well as hexagonal symmetry as these are widely
found in thin film systems. As special case of hexagonal symmetry the uniaxial
one is introduced, which in form of shape anisotropy always occurs in thin films
and which due to stray field minimization favors an easy in-plane alignment of
the magnetization. We will use anisotropy constants having suffixes according

52

J. Lindner and M. Farle

to the symmetry they describe, e.g. K2 for uniaxial (second order) and K4
for cubic symmetry. This seems for us to be more transparent than to just
number the constants in a row. Note that the latter numbering is used by other
authors, so that care has to be taken when comparing results of anisotropy
constants. A first order cubic anisotropy constant (denoted as K4 within our
nomenclature) is often denoted as K1 by other authors.
An expression for the anisotropy field follows from considering the torque
exerted on the magnetization by the effective magnetic field within the sample.
We assume for simplicity that the free energy of the system only depends
on the angle of the magnetization with respect to an anisotropy axis. If
B is the angle of the external field relative to this axis, the free energy
can be written as F = Fa M B cos( B ), where the first term is the
anisotropy energy and the second the Zeeman contribution of the external
field. The equilibrium angle of M can be found from F = 0 = Fa + M

B sin( B ) = Fa + |M B|. This equation means that in equilibrium

the torque M B due to the external field is balanced by the torque due to
the magnetic anisotropy field given by Fa (the opposite sign indicates that

the torques are antiparallel). When B causes a turn of M of , the torque


a
due to the anisotropy field is proportional to and given by F
= c .

$
2$
2
Thus, for 0 c = Fa / =0 and the equilibrium condition becomes
$
c + M B sin 2 Fa /2 $=0 + M B = 0. From this equation
the anisotropy field is found to be
$
1 2 Fa $$
Ba =

.
(2.5)
M 2 $=0
Note that the derivative has to be taken at the equilibrium angle, for which
= 0.
It is a quite common though not the only possibility to expand the free energy
of a magnetically saturated single crystal (i.e. no domain walls in the crystal)
as a series of the direction cosines i of the magnetization vector relative to
a rectangular Cartesian system of coordinate axes. The direction cosines are
the projections of M onto the three unit vectors defining the crystal lattice
and given by i = M /M ei (i = 1, 2, 3) where the ei are the unit vectors.
According to Birss one can write [55]:
F = bi i + bij i j + bijk i j k + bijkl i j k l + . . . .

(2.6)

This series is a direct consequence of Neumanns principle stating that any


type of symmetry which is exhibited by the point group of the crystal, i.e.
by the group of symmetry operations that describe the symmetry of the
unit cell of the crystal, is possesed by every physical property tensor. Thus,
the limitations of crystal symmetry must be reflected by the tensors bijk... .
Note, that the higher order terms make F to oscillate rapidly with the angular orientation of the direction of magnetization. Since this is contray to

2 Magnetic Anisotropy of Heterostructures

53

experimental observations the higher order terms must be very small. The
components of the tensors bijk... transform under a rotation of coordinate
axes according to the relations bijk...n = lip ljq lkr . . . lnu bpqr...u (in many cases
this transformation of a tensor is used as definition of the tensor as a physical object). Note that in the equation the Einstein notation is used, i.e.
when a letter occurs as a suffix twice in the same term on one side of the
equation, summation
%3 respect
%3 %3 with
%3 to that suffix is to be understood, i.e.
.
.
.
bijk...n =
p=1
q=1
r=1
u=1 lip ljq lkr . . . lnu bpqr...u . The matrix [lip ]
describes the symmetry operation. As special
%3 case the equation contains the
transformation of a vector given by bi = p=1 lip bp = lip bp . As example a
right-handed rotation of 180 about the z-axis is described by the matrix

0
0
cos 180
1 0 0
0
cos 180 0 = 0 1 0 .
[lip ] =
(2.7)
0
0
cos 0
0 0 1

With this it follows that the requirement that bijk... is a property tensor
and invariant under all permissible symmetry operation appropriate to the
particular crystal class is equivalent to the requirement that the components
bijk... n satisfy the set of equations:
bijk...n = ip jq kr . . . nu bpqr...u .

(2.8)

All the matrices [] correspond to permissible symmetry operations. It can be


shown that there are only 9 so-called generating matrices that are needed to
describe all crystal classes, i.e. all symmetry operations of the point groups can
be described by these matrices and multiplications of them. The matrices are:

100
1 0 0
+
*
+
*
(inv) = 0 1 0
(unit) = 0 1 0
001
0 0 1

1 0 0
1 0 0
*
+
*
+
(2z) = 0 1 0
(2$z) = 1 0 0
0 0 1
001

1 00
10 0
*
*
+
+

(2$z) = 0 1 0
(2z) = 0 1 0
(2.9)
0 01
0 0 1

1 1

010
*
*
+
+
2 2 3 0

(3$z) = 21 3 12 0 (4$z) = 1 0 0
001
0
01

0 1 0
010
+
+
*
*

(4$z) = 1 0 0
(3dia) = 0 0 1 .
0 0 1
100

The first matrix is the unity matrix, the second describes a point inversion
through the unit cell of the crystal. The next two matrices describe a twofold

54

J. Lindner and M. Farle

rotation parallel to the z-axis and to an axis perpendicular to the z-axis,


respectively. The other matrices describe in analogy fourfold and threefold
rotational axes (a bar on top denoting a rotation followed by a point inversion).
Finally, the last matrix describes a threefold rotation parallel to a cube-body
diagonal. Other symmetry operations can be written as multiplications of the
above
E.g.
rotation parallel to the z-axis can be described
-,
- , a six-fold
, matrices.
by (inv) (2$z) (3$z) .
In the following expressions for the free energy in uniaxial, hexagonal,
cubic and tetragonal crystals are derived. The cartesian coordinate system
chosen to describe the crystal is shown in Fig. 2.1. The system is chosen so
that the z-axis coincides with the film normal. Consequently, the x- and y-axes
are located within the film plane. To obtain expressions that are a function
of the external magnetic field B 0 and the magnetization M , the polar angles
B and as well as the azimuthal angles B and are introduced. In case
of an additional distinguished direction (like the direction of step edges), the
angle is defined with respect to the x-axis.
Cubic Symmetry
,
- ,
For crystals of cubic symmetry the generating matrices are: (inv) , (4$[001])
, (4$[111]) and
. While the first matrix describes the fact that the cubic unit
cell is centro-symmetric, the second and third matrix describe fourfold rotational
axes
- parallel to a cube-body edge and diagonal, respectively. Us,
ing (inv) within (2.8) yields bijk...n = bijk...n for all tensors of odd
rank (n is an odd number), thus making them vanish. We note that this

qB
q

B0
M

j
jB

d
x
Fig. 2.1. Cartesian coordinate system used to derive the expressions for the free
energy for the different crystal symmetries

2 Magnetic Anisotropy of Heterostructures

55

also follows directly from time inversion symmetry, which is valid ,as magne-tocrystalline anisotropy is a static property. Using the matrices (4$[001])
,
and (4$[111]) leads to relations showing that for the tensor bij all components, for which i *= j vanish, while b11 = b22 = b33 which expresses the
requirement that for cubic symmetry the energy must not change upon exchanging two i (change of equivalent cubic axes). The first non-vanishing
term is thus given by bij i j = b11 2x + b11 2y + b11 2z . Similarly one obtains the allowed terms for bijkl . Using again the symmetry matrices one gets
b1111 = b2222 = b3333 and b1122 (6) = b2233 (6) = b1133 (6) ((6) means the 6 permutation
that can be
the term). Then,/ one has bijkl i j k l =
.
/ made from
.
b1111 4x + 4y + 4z + 6b1122 2x 2y + 2y 2z + 2z 2x . The factor 6 arises from
the multiplicity implicit in the second relations of the bijkl . The first nonvanishing contributions for cubic crystals therefore are:
.
/
.
/
Fcub = b11 2x + 2y + 2z + b1111 4x + 4y + 4z
.
/
+ 6b1122 2x 2y + 2y 2z + 2z 2x + . . . .
(2.10)
/
.
Using the relations 2x + 2y + 2z = 1 and 1 2 2x 2y + 2x 2z + 2y 2z =
4x + 4y + 4z yields:
.
/
Fcub = K0 + K4 2x 2y + 2x 2z + 2y 2z + K6 2x 2y 2z . . . .

(2.11)

Here the anisotropy constants K0 = b11 + b1111 and K4 = 6b1122 2b1111 were
defined. Note that also the next higher order term was introduced, for which
K6 = 3b111111 45b111122 + 90b112233 (see [55] for details).
Cubic Crystals with (001)-orientation
Considering the coordinate system of Fig. 2.1 the direction cosines can be
written as x = sin cos , y = sin sin , z = cos . In the following we
identify the z-axis with the [001]-direction, the x(y)-axis with the [100]([010])direction of the cubic crystal. Inserting the expressions for the i into (2.11)
one obtains:
.
/
F001 = K4 sin2 cos2 sin2 sin2 + sin2 sin2 cos2 + sin2 cos2 cos2
.
/
(2.12)
= K4 sin2 cos2 + sin4 cos2 sin2 .
This expression can be transformed into another equivalent form:
.
/
F001 = K4 sin2 cos2 + sin4 cos2 sin2
,
.
/
.
/= K4 sin2 1 sin2 + sin4 cos2 1 cos2
,
.
/= K4 sin2 sin4 cos4 cos2 + 1

(2.13)

56

J. Lindner and M. Farle


[001]

[001]

[001]

[110]

[100]
[110]

[100]

[110]

Fig. 2.2. Free energy for a cubic crystal with (001)-orientation with the !100"(left panel), the !111"- (middle panel) and the !110"-axes (right panel) being the
easy ones

= K4 sin sin
2

23
1
1
3 1 1
cos 4 + cos 2 + cos 2
8
2
8 2 2

1
= K4 sin2 K4 (cos 4 + 7) sin4 .
8

The equation shows that for K4 > 0 the +100,-directions are the easy ones,
whereas for K4 < 0 the +111,-directions are the easy ones. This is visualized
in Fig. 2.2, where the free energy has been plotted as polar plot using the
coordinates defined by Fig. 2.1 (the z-axis coincides with the [001]-direction).
For the case that K6 can not be neglegted, the +110,-directions can be the easy
ones (see right panel in Fig. 2.2). Table 2.2 shows the combinations of K4 and
K6 and the corresponding easy, intermediate and hard directions. A graphical
representation in the form of stability or flow diagramms of K4 versus other
anistropy constants have been published by several authors [56, 57, 58, 59].
Cubic Crystals with (011)-orientation
To derive an expression for F in the case of a cubic crystal with (011)orientation, one needs to rotate the (x, y, z)-coordinate system, for which
the axes are parallel to the +100,-directions, yielding a new system (x( , y( , z( )
Table 2.2. Conditions for easy, intermediate and hard axes in cubic symmetry
K4

K6

to
4
9K
4

4
9K
to
4
9K4

Easy
Interm.
Hard

!100"
!110"
!111"

!100"
!111"
!110"

9K4 to

to

9|K4 |
to
4
9 |K4 |

9 |K4 | to
+

!111"
!100"
!110"

!111"
!110"
!100"

!110"
!111"
!100"

!110"
!100"
!111"

9K4
4

2 Magnetic Anisotropy of Heterostructures


z [001]
z
[11
1]

[0

11
]

z [001]

57

x
[11
2]

45

y
[010]

45

y
[010]

45

0]
[11 45
y

[0
]
11

x, x
[100]
(001)

x
[100]

(011)

(001)

(111)

Fig. 2.3. Rotation of the (001)-coordinate system


!

with its z -axis parallel to any of the {110}-planes. As example in Fig. 2.3
!
this rotation is shown for the case that the z -axis becomes parallel to the
[011]-direction. Still the direction cosines in the new systems are given by
x! = sin cos , y! = sin sin , z! = cos . is now defined with
respect to the z( -axis ([011]-direction), is measured against the x( -axis
([100]-direction). For (2.11), however, the direction cosines with respect to
the (x, y, z)-coordinate system are needed, i.e., the direction cosines within
the (x, y, z)-system have to be expressed by means of the direction cosines
within the (x( , y( , z( )-system. From Fig. 2.3 the direction cosines with respect
to the axes of the two systems can be deduced. The relation aresummarized
in Table 2.3.
This yields x = x! = sin cos
, y = 1/ 2 y! +
1/2 z! = 1/ 2 (sin sin + cos ) , z = 1/ 2 y! + 1/ 2 z! =
1/ 2 ( sin sin + cos ). Using these direction cosines within (2.11) one
derives for the free energy of an (011) oriented cubic crystal:
0
1
32
,
K4
sin2
F011 =
cos4 + sin4 sin4 + sin2 (2) + sin2 (2) cos2
.
4
2
(2.14)
We note that this equation describes the same polar plot that (2.12) does with
the difference that it has been rotated such that the [011]-direction forms the
Table 2.3. Direction cosines between (001) and (011)-coordinate system

x
!
y
!
z

cos 0 = 1
cos 90 = 0
cos 90 = 0

cos 90 = 0

cos 45 = 1/ 2

cos 45 = 1/ 2

cos 90 = 0

cos 135 = 1/ 2

cos 45 = 1/ 2

58

J. Lindner and M. Farle


[001]

[100]

[110]
[110]
[100]
[111]
[110]

[112]
[011]

[011]

[100]

Fig. 2.4. Free energy for a cubic crystal with (001)-orientation (uppermost panel),
(111)-orientation (middle panel) and (111)-orientation (lowest panel) for K4 > 0, i.e.
for the !100"-directions being the easy ones. The right column shows a cut of the
polar plots on the left side, for which the [001] , [111] and [011] direction point
out of the paper plane. In this projection the in-plane symmetry can be better seen

film normal (see lowest panel of Fig. 2.4). The azimuthal plane according to
the coordinates introduced in Fig. 2.1 shows a dominating twofold symmetry
as shown in the lowest panel (right plot) of Fig. 2.4.
Cubic Crystals with (111)-orientation
For (111) oriented cubic crystals one changes into a coordinate system
((x, y, z) (x( , y( , z( )), for which the z( -direction is oriented parallel to the
[111]-direction (see Fig. 2.3). This time denotes the polar angle with

2 Magnetic Anisotropy of Heterostructures

59

Table 2.4. Direction cosines between the (001) and (111)-coordinate system

x
!
y
!
z

cos 114.1 = 1/ 6

cos 45 = 1/ 2

cos 54.7 = 1/ 3

cos 114.1 = 1/ 6

cos 135 = 1/ 2

cos 54.7 = 1/ 3


cos 35.3 = 2/ 3
cos 90 = 0

cos 54.7 = 1/ 3

respect to the z( -axis ([111]-direction), the one with respect to the x( -axis
([11
2]-direction). The direction cosines between the two systems are given in
Table 2.4. In analogy to the (011)-plane one gets expressions for the direction
cosines within the (x, y, z)-system as function of the s within the rotated
system:
1
1
1
sin cos sin sin cos

+
+
x = x! + y! + z! =
6
2
3
6
2
3
!
! ""+

1 *
= cos 2 sin sin +
6
3
1
1
1
sin cos sin sin cos

+
y = x! y! + z! =
6
2
3
6
2
3
!
! ""+

1 *
= cos 2 sin cos +
3
3

2
2
1
1
z = x! + z! = sin cos + cos
3
3
3
3
(2.15)
+

1 *
= cos + 2 sin cos .
3

Here the trigonometric expressions for sin (x y) = sin x cos y cos x sin y,
cos (x y) = cos x cos y sin x sin y and sin (/6) = cos (/3) = 1/2 and
cos (/6) = sin (/3) = 3/2 have been used. Equation (2.11) yields for the
free energy of a (111)-oriented cubic crystal:
4
5

2
1
1
cos4 + sin4
sin3 cos cos 3 .
F111 = K4
(2.16)
3
4
3
F111 is visualized in Fig. 2.4 (middle panel). Now the [111]-direction coincides with the z-axis from Fig. 2.1. In the azimuthal plane the free energy is
isotropic (see right plot of the middle panel of Fig. 2.4). An angular dependence within the azimuthal plane does only occur, when the next higher order
term (K6 -term in (2.11)) is considered (see also Sect. 2.3.4 for the azimuthal
dependence of F ).

60

J. Lindner and M. Farle

Cubic Films with In-plane Magnetization


For thin layers of a cubic material and for the case that the magnetization
is confined to the film plane, the direction cosines for the (001)-, (011)- and
(111)-plane take the forms listed in Table 2.5. For the free energy also listed in
the table in addition to the K4 -term the next higher order term (K6 ) according
to (2.11) was considered. Table 2.5 shows that for the (001)-orientation a
fourfold symmetry in the plane exists, while the (011)-orientation shows also
twofold terms. For (111)-oriented films the in-plane anisotropy in first order
vanishes and only to the next higher order shows a sixfold symmetry, which,
however, in most cases is very small.
Tetragonal Symmetry
In case of tetragonal symmetry only the (001)-orientation will be discussed
as this is the one most widely dicussed one in literature. The transformation to other orientations can be performed in the same way as discussed
for
cubic
matrices for tetragonal systems are:
- , symmetry.
,The symmetry
, (inv)
, (2z) and (4$z) . The rotational axis perpendicular to the [001]
direction now presents only a twofold symmetry and thus the terms in the
expansion of the .free energy
/ reflect this lowering of symmetry. The first allowed term is b11 2x + 2y + b33 2z and thus, the twofold symmetry does not
vanish as for cubic symmetry. The first terms are given by:
.
/
.
/
Ftet = b11 2x + 2y + b33 2z + b1111 4x + 4y + b3333 4z +
.
/
(2.17)
+ 6b1122 2x 2y + 6b1133 2x 2z + 2y 2z + . . . .
Table 2.5. Direction cosines and free energy for the case that the magnetization is
confined in the film plane
Plane

Direction cosines

Free energy

(001)

x = cos
y = sin
z = 0

(011)

x = cos
y = 12 sin
1
sin
z =
2

F001 = K4 cos2 sin2


= K44 sin2 2
= K84 (1
. cos 4)
/
F011 = K44 sin4 + 4 sin2 cos2
K6
4
2
+ 4. sin cos /
= K44 sin4 + sin2 2
6
+K
sin2 sin2 2
16
K4
= 32 (7 4 cos 2 3 cos 4)
1
K6 (2 cos 2 2 cos 4 + cos 6)
+ 128
.
/
6
F111 = K44 + K
9 cos2 24 cos4 + 16 cos6
54
K4
K6
= 4 + 108 (1 + cos 6)

(111)

x =
y =
z =

cos

+ sin
6
2

cos
sin

2
6
3
cos
2

2 Magnetic Anisotropy of Heterostructures

61

One can see that for b11 = b33 , b1111 = b3333 and b1122 = b1133 , i.e. for the
case where x-,y- and z-axes are equivalent, the
for
. expression
/
. cubic symmetry
/
is retained. Using the relations 2z = 1 2x + 2y and 2x 2z + 2y 2z =
. 4
/
1
1
4
4
2 2
2 2 x + y + z x y the equation transforms to:
. 4
/
(
(
(( 2 2
(
Ftet = K0( K2
x + 4y + K4$
2z + K4$
x y + K4
4z + . . . , (2.18)
(
(
((
= b11 b33 , K4$
= b1111 3b1133 , K4$
=
with K0( = b11 + 3b1133 , K2
(
6 (b1122 b1133 ) and K4 = b3333 3b1133 . A further simplification is made
.
/2
.
/2
through the relation 4x + 4y = 2x + 2y 22x 2y or 2x 2y = 12 2x + 2y
.
/
1
4
4
2 x + y , leading to:

.
/ 1
1
Ftet = K0 K2 2z K4$ 4x + 4y K4 4z + . . . ,
2
2

(2.19)

((
(
((
= b11 + 3b1122 , K2 = K2
K4$
= b11 b33 +
with K0 = K0( + 12 K4$
(
((
6 (b1122 b1133 ), K4$ = 2K4$ + K4$ = 2b1111 + 6b1133 + 6 (b1122 b1133 )
(
((
and K4 = 2K4
K4$
= 2b3333 + 6b1133 6 (b1122 b1133 ). Using the
polar coordinates according to Fig. 2.1 finally yields:

1
1
Ftet = K2 cos2 K4 cos4 K4$ (3 + cos 4) sin4 .
2
8

(2.20)

Uniaxial and Hexagonal Symmetry


- ,
- ,
,
The symmetry matrices for hexagonal systems are (inv) , (2z) , (2$z)
, (3$z) and
. The matrix (3$z) describes threefold rotational symmetry about
the z-axis (note that the sixfold symmetry of the hexagonal unit cell can be
described by combinations of the matrices).
- As for the other cases, we have a
,
centrosymmetrical unit cell (due to (inv) ) and thus all odd rank tensors van,
ish. The use of (2z) within (2.8) further shows that the s are separately
non-zero only if i = p or j = q or . . . In addition, the product ip jq kr . . . ip
is 1 when the subscript 2 appears an odd number of times (note that the
number of s within the product must be even). Since this means that
dijkl... = dijkl... all coefficients, in which ,the subcript
2 appears an odd
number of times, must vanish. Similarly, if (2$z) is used, it can be shown
that all coefficients, in which the subscript 3 appears must vanish. These two
restrictions then imply that the coefficients, in which any subscript appears an
odd number off times vanish. Thus, the first .non-vanishing
term is the same
/
as for tetragonal symmetry, i.e. bij i j = b11 2x + 2y + b33 2z . The last matrix yields several relations between the remaining dijkl... s, leading finally to
.
/2
.
/
bijkl i j k l = b1122 2x + 2y + 6b1133 2x + 2y 2z + b3333 4z (see [55] for
details of the calculation). Taking even the next higher order term one obtains:
.
/
.
/2
.
/3
Fhex = K0 + K2 2x + 2y + K4 2x + 2y + K6 2x + 2y +
.
/.
/
+ K6$ 2x 2y 4x 142x 2y + 4y .
(2.21)

62

J. Lindner and M. Farle

One sees that for K4 << Ki<4 there is no in-plane anisotropy and the hexagonal symmetry converts to an uniaxial one. Using polar coordinates yields:
Fhex = K2 sin2 + K4 sin4 + K6 sin6 + K6$ sin6 sin 6 .

(2.22)

Figure 2.5 shows a polar plot according to the coordinate system of Fig. 2.1
for (left panel) K2 < 0, for which the [0001]-direction is the hard one and
for (right panel) K2 > 0, for which the [0001]-direction is the easy one. K6$
leads to the sixfold anisotropy within the azimuthal plane.
Shape Anisotropy
At each lattice point within a cubic lattice the dipole fields of all neighbors
cancel out. This is, however, only true for an infinite system. As soon as
surfaces are present, magnetic poles develop and thus, the dipole-dipole interaction leads to anisotropy. Since the shape of the sample determines this
anisotropy, one usually calls it shape anisotropy. In systems with reduced dimension such as thin films the shape anisotropy might be even the dominating
contribution to the overall MAE. For a thin disc (realized by the thin film),
the shape anisotropy has, phenomenologically, the form of uniaxial anisotropy
(being a special case of hexagonal anisotropy):
shape
=
Funi

/
0 .
N N$ M 2 cos2 ,
2

(2.23)

where N and N$ are the demagnetizing factors parallel and perpendicular


to the film plane. In the limit of a homogeneous disc with infinite diameter
N = 1 and N$ = 0 holds and thus shape anisotropy always favors an easy
in-plane orientation of the magnetization. The only exception occurs for rough
[0001]

[0001]

Fig. 2.5. Polar plot of the free energy surface for a hexagonal system for (left
panel) easy axis parallel to the hexagonal axis and (right panel) perpendicular to
the hexagonal axis. For uniaxial systems (K6i = 0) the polar plot is the same with
the only difference that no sixfold symmetry in the basal plane perpendicular to the
uniaxial axis is present

2 Magnetic Anisotropy of Heterostructures

63

films. This situation was theoretically studied by Bruno [60], who modeled the
roughness by two parameters and , the former being the mean vertical deviation from a reference plane, the latter describing the average lateral size of
flat areas. The calculation shows that the stray field energy for perpendicular
magnetization of a rough film is given by:
1 23
0

shape
= 0.50 MS2 d 1 f
Frough
.
(2.24)
2d

Here d is the average thickness, i.e. the thickness of the reference plane. MS
is the saturation magnetization. The function f has the value 1 for a flat film,
= 0, and it approaches 0 for increasing roughness, / 1. This model
shows that film roughness gives rise to a small dipolar surface anisotropy contribution of magnitude / that favors an out-of-plane easy axis of magnetization. Such a contribution was indeed found in [61] for rough Ni films on
Cu(001).
Another point raised by Heinrich and Cochran [4] and already earlier by
Benson and Mills [62] is that for very thin films of only a few atomic layers the
continuum approximation fails to describe the dipolar shape anisotropy. The
discreteness of the atomic moments results in a variation of the dipolar field
across the sample which depends on the number of atomic layers involved. The
dipolar field of a given layer decreases exponentially away from its surface with
a decay length corresponding to the in-plane lattice spacing. Thus, the dipolar field inside the film decreases when approaching the sample surface from
inside the film and the value of the average dipolar field decreases strongly
when the thickness of the film is reduced towards the monolayer regime. The
larger the lattice spacing, the stronger this effect will be. The reduced dipolar
field will appear as a reduced shape anisotropy and the reduction can be written as a reduced demagnetizing factor. For bcc(001) films the demagnetizing
factor is given as N = 1 0.425
N , while for the more densely packed fcc(001)
surface N = 1 0.234
,
N
being
the number of atomic planes of the film. The
N
case of hcp structure is discussed in [24]. Except for very thin films with a
thickness of few monolayers this correction is less than 1% and will thus not
be considered in the following.
Uniaxial Symmetry Surface Anisotropy
In thin films the presence of the symmetry breaking surface and interface to
the substrate also introduces an uniaxial anisotropy term that can be written
in the form:

= K2 sin2 = K0 K2 cos2 = K0 K2 2z ,
Funi

(2.25)

with K0 = 1 . When an uniaxial distortion of the crystal lattice is present


in the volume of the material, such a term may also contribute to volume
2

Note that as a potential energy contribution F is defined as energy difference, so


that adding constant (angular independent) terms has no influence on the value

64

J. Lindner and M. Farle

anisotropy. In general, magnetic anisotropies of thin films can be decomposed


in volume contributions, that are independent of thickness and surface state
and can be explained as a superposition of shape, magnetocrystalline and
residual strain anisotropies, and surface contributions, which scale with 1/d
and depend sensitively on the state of the surface. In some cases Neels phenomenological anisotropy model has provided a useful connection between
different components of surface anisotropies. For example, the role of epitaxial strain for the anisotropies of ultrathin films was demonstrated for the
case of Co(0001) films on W(110) by measurements of anisotropies using torsion oscillation magnetometry combined with measurement of the strain by
high-angular-resolution low-energy electron diffraction. Up to a thickness of
d = 2 nm, the films were found to grow in a state of constant strain, governed
by pseudomorphism with a growth relation [1100] Co$ [110] W, which results
in a true volume-type strain anisotropy. Above 2 nm, a relaxation of strain
is observed which scales roughly with 1/d and, therefore, results in an apparent surface-type contribution to strain anisotropy, superimposed on a reduced
volume contribution [63].
Besides uniaxial anisotropy parallel to the film normal uniaxial contributions
frequently appear along a direction in the film plane. This can be caused e.g.
by preferential interactions due to oriented hybridization at the film-substrate
interface or an uniaxial in-plane distortion in the volume. Such a term can be
described by:
$
(2.26)
Funi = K2$ sin2 cos2 ( ) ,
where is measured with respect to the [100]-direction (x-axis). To include
any possible in-plane easy axis the angle was defined as shown in Fig. 2.1.
shape

, Funi
(and in
One should note the following: The anisotropy due to Funi
$
many cases also Funi ) are direct results of the fact that the specimen has the
shape of a thin film with interfaces that break the translational symmetry of
the system. Thus, all these contributions are inherently connected with the
film surface itself and do not depend on the crystallographic direction of the
film normal. Consequently, for other crystallographic orientations of the film,
$
shape

, Funi
and Funi has to be made, in contrast to
no transformation of Funi
the case of crystalline anisotropy resulting from the volume symmetry of the
film material as discussed for cubic symmetry in Sect. 2.3.2.
In order to separate the different anisotropy contributions into volume and
surface contributions one needs also to perform thickness dependent measurements. The thickness dependence of each anisotropy constant can be fitted
by a constant term representing a volume contribution (Kiv ) and an effective
surface/interface contribution (Kis,eff ) being proportional to 1/d where d is
the thickness of the film.

of F . Therefore, the constant term K0 does not contribute to magnetic anisotropy


and can be made to vanish upon normalizing F (i.e. subtracting the constant K0 ).

2 Magnetic Anisotropy of Heterostructures

Ki = Kiv +

Kis,eff
d

i = 2, 4.

65

(2.27)

Magnetoelastic Contributions
pt Once one realizes that the MAE is a quantity describing the interaction
between the electron spin and the lattice, it is intuitively clear that changes of
the lattice constant will affect the magnetic properties. In a magnetized body
one has energy terms that depend both on the strain and the magnetization
direction: the magneto-elastic energy. Although resulting from the same origin, namely the spin-orbit interaction, magneto-elastic anisotropies only exist
when stress is exerted on the magnetic system. For instance in iron, the effect
of tension on a single crystal is to create a preferred direction of magnetization
parallel to the direction of stress. The experimentally obtained magneto-elastic
constants are significantly larger than the crystalline anisotropy constants [64].
As a consequence, even small strains may give rise to an important anisotropy
contribution. Moreover, this phenomenon may be of importance in epitaxial
structures, where considerable strains may result from the epitaxial growth of
the film on a substrate or adjacent layers having a different lattice parameter. With respect to the film material, the strain in epitaxial films is given by
= (asub af ilm ) /af ilm , i.e. the misfit is determined by the lattice constants
ai , which describe the atomic distances on the relevant surface orientation. If
the lattice mismatch is not too large, below a critical thickness dc (coherent
regime), the misfit is accommodated by introducing a tensile strain in one
layer and a compressive strain in the other such that both adopt the same
in-plane lattice magnetic anisotropy parameter. A reasonable epitaxial match
of the film lattice with respect to the substrate one can also be achieved by
the rotation of the two lattices against each other. This happens e.g. for bccFe(001) growing on fcc-Ag(001), for which the lattices are rotated by 45 with
respect to each other. For relatively thin films the strain and the magnetoelastic coupling are independent of thickness. Above the critical thickness dc ,
it becomes energetically more favorable to introduce misfit dislocations, which
partially accommodate the lattice misfit, allowing the uniform strain to be reduced (incoherent regime). In the incoherent regime, the contribution to the
magneto-elastic energy contains a reciprocal thickness dependence [65]. The
film strain can be isotropic in the plane of the film (i.e. %11 = %22 = , where
the %ii are the in-plane component of the strain tensor along two axes that are
perpendicular to each other) or also anisotropic (%11 *= %22 ) The latter might
occur when preferentially oriented misfit dislocations have formed or a strong
interface hybridization between the film and substrate along specific directions
is present. According to continuum elasticity the in-plane strain leads to an
out-of-plane variation of the lattice. From the requirement of a minimum of the
elastic energy one can calculate the strain component %33 perpendicular to the
film plane. Table 2.6 lists the results for several symmetries of the film lattice.
The corresponding change in volume is given by V /V = (%11 + %22 + %33 ).

66

J. Lindner and M. Farle

Table 2.6. Calculated out-of-plane strain "33 as function of in-plane strain components "11 , "22 (from [15]). The elastic stiffness constants cij are tabulated in [66]
Cubic (001)
1:[100], 2:[010]
c12 ("11 +"22 )
c11

Cubic
-(011)
,
1: 011 , 2:[100]

Cubic (111)
1,2: arbitrary axes

Hex. (0001)
1,2: basal plane

(c11 +c12 2c44 )("11 +2c12 "22 ) (c11 +2c12 2c44 )("11 +"22 ) c13 ("11 +"22 )
(c11 +c12 +2c44 )
(c11 +2c12 +4c44 )
c33

The magneto-elastic part to the magnetic anisotropy for a cubic system can
be written as [15]:
.
/
cub
(2.28)
= B1 %11 2x + %22 2y + %33 2z
FMEL
+ 2B2 (x y %12 + y z %23 + x z %31 ) + . . . .

For hexagonal systems the magneto-elastic contribution is [15]:


.
/
hex
= B1 %11 2x + %22 2y + %12 x y
FMEL
.
/
.
/
+ B2 1 2z %33 + B3 1 2z (%11 + %22 )

(2.29)

+ B4 (y z %23 + x z %13 ) + . . . ,

where the %ij (i, j = 1, 2, 3) are the strain components, the Bi the magnetoelastic coupling constants and the i the direction cosines (see coordinate
system in Fig. 2.1) given by 1 = sin cos , 2 = sin sin and 3 = cos .
One should note that while in general the Bi of ultrathin films are different
than in the respective bulk material one finds in the case of Ni films that the
strain induced anisotropy contributions can be explained even as a function
of temperature by the respective bulk values [5]. In Sect. 2.5 we will, however,
show that the Bi -values for Fe films on GaAs differ from the ones of Fe bulk.
In order to calculate the magneto-elastic part one needs to measure the Bi
(volume values for 3d ferromagnets are found in e.g. in [15]). In (2.29) if i = j,
the strain is along the cubic +100, axes, for i *= j the strain is along the +110,
axes. While the former type of strain leads to a change of the volume of the
unit lattice cell, the latter is equivalent to a shearing of the lattice keeping
the volume constant.

2.3.3 Landau-Lifshitz Equation of Motion and General Resonance


Equation
Magnetic excitations from the ground state that occur in the microwave
regime and are detected within an FMR experiment (see Sect. 2.5 for details on the FMR technique) are usually described within the Landau-Lifshitz
formalism. Due to the high number of spins that take part in the absorption
process and the large quantum numbers associated with it classical and quantummechanical description lead to identical results [67] and thus a classical
formulation of the process is usually considered.

2 Magnetic Anisotropy of Heterostructures

67

Upon considering the total angular momentum J n as a classical vector with


continuous possibilities of adjustments, the time dependence of J n is accordn
ing to Newtons law given by the torque D n = dJ
dt . In our case the torque
is given by a magnetic field acting on the magnetic moment n that results
from J n . Considering a magnetic part of the rf-field due to the microwave
int
excitation brf
n , an external field B 0 and also an internal field B n within the
sample (e.g. magnetic anisotropy fields) !
as contributions to
" the overall magint
rf
netic field, the torque is given by n B n + B 0 + bn . Taking the two
expression for the torque together and using n = J n with =
the gyromagnetic ratio (gJ : Lande g-factor) one obtains:
"
!
1 dn
rf
.
= n bint
n + b0 + bn
dt

gJ B
!

being

(2.30)

Summing
internal field
% up all magnetic moments yields the macroscopic %
B int = n B int
,
that
acts
on
the
total
magnetic
moment

=
n . Taking
t
n
into account that the magnetization M is defined as magnetic moment per
unit volume (t /V ) and assuming a homogeneous microwave field brf over the
sample, one has:
.
/
dM
= M B eff .
(2.31)
dt
Here the abbreviation B eff = B int + B 0 + brf was used. The latter equation
is known as the Landau-Lifshitz(LL)-equation. We note that it can be also
derived in the framework of quantummechanics [68]. The LL equation can
be extended to include magnetic damping, leading to a finite linewidth of
the FMR signal. However, throughout this paper, which focusses on magnetic
anisotropy and thus on the field needed for resonance (resonance field) only,
damping will be neglected.
A straightforward but complex way to describe FMR is to solve the LL
equation for given anisotropy fields. There is, however, an alternative route,
which uses the LL equation to formulate a general equation on the basis of
the magnetic part of the free energy of the system. As this approach is rather
general, it is described in some detail in the following. The method to calculate
the resonance frequency or, equivalently, the resonance field for the uniform
FMR mode (collective precession of all magnetic moments) was introduced
by Smit and Beljers and independently by Suhl ([69, 70]). In this formalism
the equation of motion is described by the free energy F . The same result
was obtained by Gilbert by solving a Lagrange-Equation for the motion of M
[71]. The magnetization is considered as classical gyroscope with moment of
inertia I. Figure 2.6 shows the transformation from the laboratory (x, y, z)!
!
!
!
coordinate system to another cartesian one (x , y , z ), in which the z -axis
rotates with the magnetization. The transformation is uniquely given by the
three Euler Angles , , .
!
"2
The kinetic energy of the system is given by Ekin = I + cos .
2

From this the Lagrangian function of the system follows to be


L = Ekin Epot (, ), Epot being the potential energy. The Langrangian

68

J. Lindner and M. Farle


z
z
M
y

"

Fig. 2.6. Euler Angles to describe the rotation of the coordinate system

d L
L
equation of motion then is dt
qi qi = 0, where the qi are the generalized
coordinates, which in our case are given by , and . This approach yields
the following three equations:
"
+ E
d * !
pot
I + cos cos +
=0
dt

!
"
Epot
I + cos sin +
=0
(2.32)

*
!
"+
d
I + cos = 0.
dt

As e.g. shown in [72] the term I( + cos ) describes the angular momentum
!
!
!
of the magnetization within the (x , y , z )-system and thus, the last equation
shows the time invariance of the angular momentum. Considering the LLG
equation of motion the angular momentum of the magnetization is given by
MS /. Therefore, the two remaining Lagrangian equations yield:
MS
sin =

MS

sin =

F
.

(2.33)

Here it was taken into account that Epot is given by the free energy F of
the system. We now assume that the precession angle of the magnetization is
small, so that
. only /small variations and with respect to the equilibrium
orientation 0 , 0 occur. This approach must be modified for high power
microwave excitations which cause large precession angles and non-linear responses. In the small angle regime we have = 0 + , = 0 + and one
can expand the first derivatives of F around the equilibrium position into a
series of and , in which only the linear terms have to be considered:
F = F + F , F = F + F .

(2.34)

2 Magnetic Anisotropy of Heterostructures

69

$
The derivatives have to be taken at the equilibrium positions, i.e. F $0 and
so on. Periodic solutions , exp(it) have to be found, as the deviations
around the equilibrium position are driven by the periodic excitation due to

the microwave field with a frequency . This yields


/ and = i.
. = i
Taking into account that for small deviations sin 0 + = sin 0 cos +
cos 0 sin sin 0 + cos 0 sin 0 , one has from (2.33):
1
2
iMS
sin 0 F F = 0
(2.35)

2
1
iMS
sin 0 F F = 0 .
(2.36)

In matrix formulation this can witten as:


5 1 2
4
0
F
F i

MS sin

=0.
0

F
F + i

M
sin

.
S

(2.37)

The condition for a solution is F2 F F + 2 2 MS2 sin2 0 = 0, yielding


the following equation for the resonance frequency:
"
!
6!
1 22
"
F F F2

2
F
F
.
=

=
0

MS2 sin2 0
MS sin 0
(2.38)
If one has an expression for the free energy, from which also the equilibrium
angles 0 and 0 can be determined, (2.38) yields the resonance condition
(B) or an expression for the resonance field Bres as function of the angles
of the external field for a fixed frequency. Equation (2.38) shows that FMR is
sensitive to the curvature of the free energy surface. As this surface strongly
depends on the anisotropy fields, FMR is a very useful tool to quantitatively
determine magnetic anisotropy. We finally note that (2.38) presents a singularity at the angle 0 = 0. This problem is removed from the resonance
equation by adding the first derivatives as discussed in [73], the improved
form being:
7
22 8
1 22
1
2 1
F

1
cos 0
cos 0 F
F
= 2 F
+
F

MS
sin2 0
sin 0
sin 0
sin 0 sin 0
(2.39)
For 0 = /2, i.e. for the in-plane configuration (see Fig. 2.1) the latter resonance equation has the same form as the original one, since the prefactor
cos 0 / sin 0 vanishes. Also for other angles, except 0 = 0, the original form
is still numerically correct, since at equilibrium the first derivatives F and
F are zero. The original equation is, however, not convenient anymore as the

70

J. Lindner and M. Farle

different terms of the free energy are mixed, covering the symmetries. The
form by Baselgia et al. [73] is therefore favorable compared to the older version of the resonance equation, in particular when out-of-plane dependencies
are considered.
2.3.4 Resonance Equations for Angular and Frequency
Dependent FMR
In the following the resonance equations for tetragonal, cubic and hexagonal
films will be explicitly given. As for thin films shape anisotropy and uniaxial inplane terms play an important role, they are included (see (2.26) and (2.23)).
Moreover, the Zeeman energy FZee = M B due to the presence of the external field is included. Only the equations for the so-called saturated modes
are given, being solutions, for which the precessional motion of the magnetization is mainly determined by the external field. So called unsaturated (or
not aligned) modes are solutions, for which the motion of the magnetization
vector is strongly influenced by internal fields (e.g. anisotropy fields). Such
modes usually occur for small external magnetic fields and only a numerical
description is possible. One should keep in mind that in the following resonance equations not aligned modes are excluded. This implies that we set
B = 0 for the out-of-plane geometries (meaning that the magnetization is
confined to the same plane, in which the external magnetic field is varied) and
B = 0 for the in-plane geometry.

Cubic and Tetragonal Symmetry: Out-of-plane Geometry


For the out-of-plane geometry the external magnetic field is varied in a plane
that comprises the film normal and one principal in-plane crystallographic
axis (see also Fig. 2.1 for the coordinate system in use).
(001)-Orientation
The free energy used to derive the resonance equations for tetragonal films
(of which cubic ones are a special case) is (see Sect. 2.3.2):
F = M B0 (sin sin B cos ( B ) + cos cos B )
"
! .
/
0
+ K2$ sin2 cos2 ( )
N N$ M 2 K2 sin2
2
1
1
4
K4 cos K4$ (3 + cos 4) sin4 .
(2.40)
2
8
The equation includes cubic systems, as can be seen by setting K4 = K4$ =
K4 . Except for a constant term, which does not lead to anisotropy this yields
the expression for cubic symmetry as given in Sect. 2.3.2. Then, (2.39) for

2 Magnetic Anisotropy of Heterostructures

71

the out-of-plane geometry, for which the external field is varied from the film
normal [001] to the [100]-direction (0 = B = 0) yields:
0
1
1 22
2
K4$

K4
res

= B cos + Meff +
cos 2 0

M
M
1
3
2
K4$
K4
0
+
+
(2.41)
cos 4 + u1
M
M
1
0
2
4K4$
res
cos + Meff
B
cos2 0
M
1
3
2
2K4$
2K4$
2K4
4 0
+
+ u2 u3 ,
+
cos +
M
M
M
.
/
with = 0 B and Meff = 2KM2 0 N N$ M denoting the effective
out-of-plane anisotropy field. For Meff < 0 (> 0) the easy axis of the system
lies in (normal to) the film plane. Note that according to Fig. 2.1 the angles
of the magnetization and B of the external field are measured with respect
to the film normal, while the in-plane angles and B were defined with
respect to the [100]-direction. The terms ui resulting from an uniaxial in-plane
anisotropy are listed in Table 2.7. The set of the ui being appropriate to a
given out-of-plane geometry is determined by the in-plane angle of the external
magnetic field B (being equal to the equilibrium angle of the magnetization
0 for the reasons mentioned at the beginning of Sect. 2.3.4). For cubic systems
one has to set K4 = K4$ = K4 within the equation.
For the out-of-plane geometry, for which the external field is varied from the
film normal [001] to the [1
10]-direction (0 = B = /4) the following
equation results:
0
1
2
1 22
K4$

K4
res

= B cos + Meff +
cos 2 0

M
2M
1
3
2
K4$
K4
0
+
+
(2.42)
cos 4 + u1
M
2M
0
1
2
K4$
res
B
cos + Meff +
cos2 0
M
1
3
2
2K4$
2K4 K4$
4 0
+
+ u2 u3 .
+
cos
M
M
M
Again the replacement K4 = K4$ = K4 leads to the special case of cubic
symmetry.
(011) and (111)-Orientation
In this case the same free energy expression is used as for the (001)-orientation
(2.40) with the only difference that the cubic anisotropy contribution being
proportional to K4 is now given by (2.14) for (011)-oriented films and by (2.16)

72

J. Lindner and M. Farle

in case of (111)-orientation. This yields for the case that the external field is
varied from the film normal ([011]- or [111]-direction) to the in-plane direction
([01
1] or B = 90 in case of (011)-orientation and [110] or B = 90 for the
(111)-orientation):
1 22
, res

= B
cos + Meff cos 20 + a + u1

, res
B
cos + Meff cos2 + b + u2 u3 .

(2.43)

While the ui are the same as given by Table 2.7, the terms a and b are listed
in Table 2.8. For the case that the external field is varied from the film normal
([011]-direction) to the [100]-direction (B = 90 ) different values for a and b
result which are also given in Table 2.8.
Cubic and Tetragonal Symmetry: In-plane Geometry
Using the free energy expressions according to Table 2.5 within the general
resonance equation ((2.39)), one obtains for the case that the magnetization
is restricted to the film plane:
1 22 *!
"+ *
+

B$res cos + b u2 c2 ,
= B$res cos Meff + a u1

(2.44)
with = 0 B . The relations for a, b and c are summarized in Table 2.9. If
/
.
2K
an uniaxial in-plane anisotropy is present, the terms u1 = M2# cos2 0
.
/
2K
and u2 = M2# cos 2 0 have to be added.
One should note that the angles and are measured with respect to
different crystallographic axes for the different orientations, i.e. is measured
either against the [111]-, the [011]- or the [001]-direction, against the [100]direction in case of the (011)- and (001)-orientation and with respect to the
[11
2]-direction in case of the (111)-orientation.
Table 2.7. Uniaxial in-plane terms contributing to the resonance equations for a
tetragonal (cubic) thin film with (001)-orientation. The equilibrium angle 0 from
minimizing the free energy given by (2.40)
B = 0 u1
0
$$
$ $
4
$$
$ $
2

2K2#
M
2K2#
M
2K2#
M

u2
cos2 cos 20
/
.
cos2 4 + cos 20
cos2 cos 20

2K2#
M
2K2#
M
2K2#
M

u3
.

/
cos2 cos2 0 cos 2
/
. 2. /
cos 4 + cos2 0 +sin 2
. 2
/
sin cos2 0 + cos 2

2
K2#

M2
2
K2#
M2
2
K2#
M2

cos2 0 sin2 2
cos2 0 cos2 2
cos2 0 sin2 2

2 Magnetic Anisotropy of Heterostructures

73

Table 2.8. Cubic terms within the resonance equations for (011)- and (111)-oriented
cubic systems. The equilibrium angle 0 from minimizing the free energy given by
(2.40)
Plane

K4
0
2 0
cos

3M
160 2 sin cos
3 0
10 2 sin +/ 28 cos
4
27 cos 0 K
M
.
/
K4
12 cos4 0 13 cos2 0 + 2
M
.
0

[111]
to[1
10]
[011]
to[100]

/
.
4
8 cos4 8 cos2 + 1
2K
M

[011]

to[011]

.
K4

cos 0 4 2 sin 0 cos2 0 /


3M
10 2 sin 0 + 7 cos3 0 3 cos 0

K4
M

/
3 cos4 0 7 cos2 0 + 2

/
.
4
K
3 cos4 3 cos2 1
M

Hexagonal Symmetry
For hexagonal symmetry and (0001)-oriented films the resonance equation for
the in-plane variation of the external field (in the plane perpendicular to the
c-axis) is given by the same equation as for tetragonal symmetry ((2.44)) with
a and b listed in Table 2.9. For the out-of-plane geometry one can in most cases
neglect the very small sixfold anisotropy in the azimuthal plane given by K6$
and only consider the out-of-plane constant of highest order (K2$ ). Then, the
resonance equation has the form of (2.42) and (2.43) when one sets K4i = 0.

2.4 Temperature Dependence of Magnetic Anisotropy


The macroscopic anisotropy energy density is temperature dependent. This
statement holds for the anisotropy contributions due to dipole-dipole and spinorbit interaction. The shape anisotropy which is proportional to the square
of the magnetization (see Sect. 2.3.2) vanishes at the Curie temperature TC .
Table 2.9. Resonance equations for a cubic thin film with different crystallographic
orientations as well as for an (001)-oriented tetragonal and a (0001)-oriented hexagonal system. The equilibrium angle 0 from minimizing the free energy given by
(2.40)
Plane

b
.

/
cos 40 + 3
.
/
cos 40 + 3

K4
2K4
(001)
cos 40
2M
M
2K4#
(001) K4#
cos 40
M
tetra. 2M.
/
.
/
K4
4
3 cos4 0 + cos2 0 2 K
12 cos4 0 11 cos2 0 + 1
(011)
M
M
4
(111) K
0
M
4K4 +6K6 +6K6# sin 60
36K
(0001)

M6# sin 60
M
hex.

c
0
0
0

2K4
M

sin30

74

J. Lindner and M. Farle

Also, the intrinsic magneto-crystalline (spin-orbit) MAE is temperature dependent (see e.g. [5, 12, 34, 55]) and vanishes at TC . This has been often overlooked in the comparison of theoretical studies (usual performed at T = 0 K)
and experimental investigations (usually conducted at room temperature). In
regard to the microscopic origin of MAE, i.e. the anisotropy of the orbital moment, it is surprising to experimentally measure a temperature dependence of
the MAE. When one considers that the spin-orbit interaction (approximately
70 meV in 3d ferromagnets) is temperature independent, and smearing out
the exchange split states at the Fermi level (order of eV) [38] does not affect
the easy direction of the magnetization, one has to conclude that the difference of the orbital magnetic moment along the easy and the hard magnetic
axis persists above TC . Unfortunately, there is no direct evidence for this,
since the magnetic moment fluctuates too vividly in space and time above
TC . Most techniques will measure an averaged magnetic moment only. However, susceptibility measurements and paramagnetic resonance measurements
(which actually measure the susceptibility at microwave frequencies) proof
the existence of atomic magnetic moments above TC even in an intinerant
ferromagnet like Ni. As the magnetic moment above TC is the same (except
for polarization of the conduction electrons) as the one measured below TC
(for T = 0 K), it is reasonable to conclude that the orbital magnetic moment
is unchanged in the paramagnetic state. A direct proof of the existence of
the orbital magnetic moment and its anisotropy in the paramagnetic state
is obtained by angular dependent measurements in the paramagnetic phase
of a ferromagnet in magnetic fields of several kOe. To our knowledge such
measurements can be performed by electron spin resonance (ESR, EPR) only
[74]. Here, the deviations of the spectroscopic splitting factor which is proportional to the ratio of orbital to spin magnetic moment was found to be
different for different crystallographic directions and could be well described
in the framework of crystal field theory.
How can one resolve the conceptual problem that the macroscopically measured MAE is temperature dependent while its microscopic origin is not? The
classical theory of the temperature dependence of the intrinsic anisotropy
(see for example [75] and references therein) was worked out based on the
assumption that around each lattice site there exists a region of short-range
magnetic order in which the local anisotropy constants are temperature independent. Due to thermal motion, the local instantaneous magnetizations
of these regions will be distributed randomly, and they produce the average
magnetization of the crystal as a whole which vanishes at TC . This does not
mean that the magnetic moment vector vanishes, but it fluctuates so quickly
and uncorrelated to other moments that the spatially and timely averaged
moment vanishes. Hence, also the macroscopically measurable MAE vanishes,
it averages out above TC . This hand-waving argument has been quantified
by expanding the MAE in a series of spherical harmonics Ylm (, ), which
reflects the role of crystal field and spin-orbit interaction with temperature
dependent coefficients k2l,m (T ):

2 Magnetic Anisotropy of Heterostructures

F =

2l
9
9

k2l,m Y2l,m .

75

(2.45)

l=0 m=2l

From a theoretical point of view the advantage of using spherical harmonics


is the fact that they are orthogonal. As anisotropy is an even function of the
magnetization, polynomials off odd degree vanish in the expansion. Accordingly, the expansion for cubic systems is given by [55]
Fcub = K0 + k0,0 Y0,0 + k4,0 Y4,0 + k4,4 Y4,4 + k6,0 Y6,0 + k6,4 Y6,4
1
1
(21K4 + K6 ) Y0,0
(11K4 + K6 ) Y4,0
= K0 +
(2.46)
105
55
1
2
2
(11K4 + K6 ) Y4,4 +
K6 Y6,0
K6 Y6,4 ,

9240
231
41580
while for hexagonal symmetry the expansion is given by
Fhex = K0 + k0,0 Y0,0 + k2,0 Y2,0 + k4,0 Y4,0 + k6,0 Y6,0 + k6,6 Y6,6
2
(35K2 + 28K4 + 24K6) Y0,0
= K0 +
105
2
8
(11K4 + 18K6 ) Y4,0
(7K2 + 8K4 + 8K6 ) Y2,0 +
21
385
16
1
K6 Y6,0 +
K6$ Y6,6 ,

(2.47)
231
10395
with the Y2l,m listed in Table 2.10. The equation for hexagonal symmetry
includes the special case of uniaxial symmetry. For uniaxial symmetry perpendicular to the film plane, the equation is also valid when one sets k6,6 = 0.
For uniaxial anisotropy in the film plane (see (2.26)) and perpendicular to the
film plane (see (2.25)) the expression to first order is:
(2.48)
Funi = k0,0 Y0,0 + k2,0 Y2,0 + k2,2 Y2,2
/
/
1 .
1
1 .
70K2 + 35K2$ Y0,0
14K2 + 7K2$ Y2,0 + K2$ Y2,2 .
=
105
21
6
Table 2.10. Spherical harmonics used for the expansion of the free energy of cubic,
uniaxial and hexagonal systems
Y0,0 = 1
Y1,0 = z
Y1,1 = x
Y1,1 = y
.
/
4
2
Y4,0 = 18 35
z +3
. 4z 30
/
2 2
Y4,4 = 105 x + 4y
x y /
. 6
2
2
Y4,4 = 105 (4x y ) x y

.
/
Y2,0 = 12 . 32z 1/
2
2
Y2,2 = 3 x y
Y2,2 = 3 (2x y )
.
/
2
1
2316z 3154z + 105
Y6,0 = 16
z 5 /
.
/
.
4x + 4y
112z 1 /
Y6,4 = 945
x y / .
2
. 4
2
2
2
945
Y6,4 = 2 (4
. x y ) /x. y 11z 1 /
Y6,6 = 10395 2x 2y . 4x 14/2x.2y + 4y /
Y6,6 = 10395 (2x y ) 32x 2y 2x 32y

76

J. Lindner and M. Farle

The relationship between the temperature variations of the anisotropy coefficients ki and the magnetization M was theoretically [76] and experimentally
[77] found to have the form:
M (T )
k2l,m (T )

,
k2l,m (0)
M (0)

(2.49)

where = l(l + 1)/2, l being the order of the sperical harmonics. This gives
for example k2,m M (T )3 , k4,m M (T )10 . The Callen-Callen model does
not identify the microscopic origin of the anisotropy coefficients, but includes
the contributions from magneto-elastic as well as magnetostrictive properties entering into the spin hamiltonian through the combination of spin-orbit
coupling and crystal field splitting.
As the relation above holds for the anisotropy coefficients k2l,m , one has to
be careful when comparing to temperature dependencies of the experimentally
measured anisotropy constants Ki . The relations between the k2l,m and the
Ki can be found from (2.46), (2.47) and 2.48). Assuming a typical temperature dependence of the magnetization one can plot the anisotropy coefficients
as shown in Fig. 2.7. One sees that the k2,0 and k4,0 decrease monotonically
with increasing temperature and vanish at TC . If one confuses these temperature dependent ki,0 with the usual magnetic anisotropy parameters Ki ,
one would draw the conclusion that a temperature change of the easy axis of
magnetization is not possible [5]. However, one finds that if one rearranges
the cos and sin terms in the Spherical harmonics in terms of increasing powers that the new parameters Ki (the ones used in the experimental analysis)
can vary in sign so that their temperature dependent change of sign in Co

K
k

Ki (arb. units)

Ki /

arb.units

4
2

4$

0
KK2 2(T) = 1.47 k2 - 3.3 k4(T)

K4$(T) = 3.85 k4(T)

-2

100

200
300
Temperature (K)

400

Fig. 2.7. Temperature dependence of the coefficients k2l,m used when expanding the
free energy into spherical harmonics and the expected experimental Ki anisotropy
parameters that are coefficients of an expansion into direction cosines (reproduced
from [5])

2 Magnetic Anisotropy of Heterostructures

77

arb. units

or Gd can be quantitatively understood. An illustrative example is plotted in


Fig. 2.7 showing that K2 changes its sign. This behavior becomes also clear
from (2.47), where one can see that k2,0 is a linear combination of K2 and
K4 (for K6 = 0).
There are only few experimental results on the power law dependence
of the second order normalized MAE on the magnetization. = 2.6(5)
for 5.6 monolayer (ML) Fe on Cu(100) [78] and = 6.5 for W(110)/Fe
6 nm/W(110) [79] was reported. While the value for Fe on Cu(100) shows
reasonable agreement with the theory, the reason for the large value in the
case of Fe on W(110) is unclear. One may speculate that higher order contributions of the anisotropy were not properly accounted for. An unusual
exponent = 2.1 was also reported for bulk like FePt films [80, 81] and
found to be the result of delocalized induced Pt moments leading to a two-ion
anisotropy. This result has been explained by ab initio electronic structure theory for L10 ordered FePt [47, 54]. The importance of separating higher order
terms from second order terms for this type of analysis was recently shown by
Zakeri et al. [83]. Here, the thickness of Fe layers on GaAs(001) was tuned to
a critical thickness so that higher order anisotropies present in thicker layers
vanished. In this case perfect agreement was found within the error bar of the
experiment with the theoretical prediction = 3 (Fig. 2.8). A linear power law
correlation with = 2.9 is observed in the experiment. The inset shows the
deviations from the linear behaviour, i.e. deviations from = 2.9, for other
film thicknesses in which K4 contributions become important. Similarly, the

Fig. 2.8. Temperature dependence of the uniaxial out-of-plane anisotropy K2


(filled circles) and the magnetization M (open squares) for 5 ML Fe/GaAs(001).
The inset shows the dependence for Fe films from 5 ML to 20 ML (Reprinted
figure with permission from Kh. Zakeri et al., Phys. Rev. B, Vol. 73, 052405 (2006).
Copyright (2006) by the American Physical Society)

78

J. Lindner and M. Farle

temperature dependence of the surface anisotropy and its relation to the surface magnetization following a different temperature dependence than the bulk
one have been analyzed [5].

2.5 Selected Experimental Results


Before discussing the experimental results, a typical FMR-signal ist shortly
explained. Within an FMR experiment, the specimen is placed in a cavity, into
which microwaves are coupled that excite the magnetic system. To generate
resonant absorption from the microwave field inside the cavity, the experiment is performed in an external magnetic dc-field that is varied while the
microwave frequency is kept constant. Detailed descriptions of various setups
may be found elsewhere [4, 5, 7, 14]. FMR absorption spectroscopy measures
the imaginary part of the high frequency susceptibility = mrf /hrf . mrf
is the dynamic contribution of the magnetization that is created due to the
high frequency magnetic field hrf of the microwaves and, thus, determines
the response of the magnetic system to the excitation (see [4, 5, 7, 14] for
details).
A typical FMR signal from a thin film measured at a microwave frequency
of 9 GHz is shown in Fig. 2.9. While the main plot shows the derivative of
the signal obtained from the lock-in detection procedure, the inset shows the
integral, i.e. the absorption signal itself. Three pieces of information can be
directly extracted: (i) The resonance field Bres that includes information on
the internal fields, such as anisotropy fields. (ii) The linewidth B that yields
information on magnetic damping and the distribution of internal magnetic

c (arb. units)

0.6

0.2

-0.2
-0.4
-0.6

250

300

350

Bres

DBpp

300

B0|| (mT)
400 500

600

Bres

Isotropic
resonance field

dc/dB (arb. units)

0.4

DBpp
400

450

500

550

600

B0|| (mT)
Fig. 2.9. Typical FMR spectrum of a thin film. The spectrum is measured as
derivative of the high frequency susceptibility with respect to the external magnetic
field The inset shows the integral of the spectrum (reproduced from [82])

2 Magnetic Anisotropy of Heterostructures

79

fields and (iii) the intensity of the signal that is proportional to the number of
magnetic moments taking part in the resonance absorption. In the following
we focus only on the analysis of the resonance field, as this quantity is the
one used to investigate the MAE in thin film systems. From Fig. 2.9 it can
be seen that the resonance field is moved away from the so-called isotropic
resonance field (dotted vertical line in Fig. 2.9), which is the field in case that
no anisotropy field is present. For a given microwave frequency = 2, the
isotropic field is given by / = B0 . This follows directly from (2.39) when
one uses the expression for the free energy F without anisotropy terms ((2.40)
with Ki = N = N$ = 0).
2.5.1 Films on Semiconducting and Insulating Substrates
In the following sections some examples of different systems are given, for
which FMR was used to determine the MAE. Only metallic thin films were
chosen that were epitaxially grown on different kind of single crystal substrates.
Fe on MgO(001)
Fe films on MgO(001) are known to be a prototype system for epitaxial growth
of a metal on an insulating substrate. The main reasons for this are the
rather simple preparation of monoatomically flat MgO substrates and the
fact that the interface hybridization and intermixing between Fe and MgO is
very small [84, 85]. The lattice mismatch of the Fe and the MgO lattice is
reduced due to a rotation of the Fe lattice by 45 with respect to the MgO
(meaning that the +110,-directions of Fe are parallel to the +100,-directions of
nm) and MgO
the MgO) [86]. With the lattice constants of Fe (aF e = 0.287

2a

Fe
(aMgO = 0.421 nm) this leads to a lattice misfit of M gO
= 3.6%.
aM gO
We do not want to give an overview over the magnetism of epitaxial Fe films
on MgO(001). For details we refer to some of the many publications on this
system [85, 87, 88, 89]. In this chapter we will focus on how FMR is used to
extract the MAE in form of the anisotropy constants that were introduced in
the preceeding sections.
In Fig. 2.10 (a) FMR spectra of 10 nm Fe on MgO(001) are shown that were
taken at room temperature and at constant microwave frequencies of 9.2 and
24 GHz. The external magnetic field B0 was applied along the direction given
in the Figure (note that the directions refer to the Fe lattice, see also the inset
of Fig. 2.10 (b), where a hard sphere model of the Fe/MgO(001) system is
shown). While at 24 GHz only one peak can be measured for a given angle of
the external field, at 9.2 GHz two peaks are detectable. These two peaks only
appear for external field angles close to the Fe +110,-directions, while for other
angles no signal is observed. This can be seen better in Fig. 2.10 (b) and (d),
where the complete in-plane angular dependence of the resonance fields measured at the two frequencies is plotted (the angle is measured with respect

J. Lindner and M. Farle

100

Fe[110]

b) Fe[110]
(001)

Fe

80

MgO

60

[110]

f=9.2GHz
T=RT

40

24 GHz@RT

9.2 GHz@RT
0

30

100

200

saturated mode
unsaturated mode

20
-60

300

-30

30

60

!B ()

B0 (mT)
340

c)

25

320

20

300

Bres(mT)

%(GHz)

B0 || [110]

B0 || [100]

B0 || [110]

Bres(mT)

d&''/dB0 (arb. units)

a)

[1
10

80

15

d)

Fe[110]

Fe[110]

saturated mode

f=24GHz
T=RT

280
260

10
B0|| [110] (hard in-plane axis)
B0|| [100] (easy axis)

Fe[100]

240
220

2.5

100

200
300
B0 (mT)

400

500

3.0

Fe[001]

e)

-90

out of field range

2.0

-60

-30

!B ()

30

60

90

Fe[001]

f)

out of field range

2.5

f=9.2GHz
T=RT

1.5

1.0

saturated mode
unsaturated mode

0.5
0

-90

Fe[110]

Bres(T)

Bres(T)

2.0
1.5

f=24GHz
T=RT

sat. mode [110]


sat. mode [100]

1.0
0.5

-60

-30

qB ()

30

60

90

0
-90
Fe[110]

-60

-30

30

60

90

qB ()

Fig. 2.10. In-plane angular dependence for 10 nm Fe/MgO(001) at room temperature at (b) 9.2 GHz and (d) 24 GHz. The solid lines are fits to the data (see text).
a) shows typical FMR spectra measured along the given directions of the external
field, c) shows the calculated dispersion relation. In (e) and (f) the out-of-plane
dependence is shown for the two frequencies

to the Fe [100]-direction). The two signals at 9.2 GHz appear within an angle
range of only 10 around the Fe +110,-directions. At 24 GHz one clearly observes a fourfold in-plane symmetry of Bres (note that only half of the whole
dependence is shown) that is slightly disturbed by a twofold one. The latter
leads to somewhat smaller resonance fields along the Fe [110]-direction com-

2 Magnetic Anisotropy of Heterostructures

81

The same symmetry is observed at 9.2 GHz. The minimum


pared to Fe [110].
of the dependence at 24 GHz indicates that the Fe +100,-directions are the
easy ones of the film.
When the external field is varied in the plane given by either the Fe [110]direction and the film normal ([001]-direction) or the Fe [100]-direction and the
film normal, the situation is different. Figure 2.10 (e) shows the experimental
result for 9.2 GHz, where (two) resonance signals are visible only within the
[1
10]/[001]-plane, while in Fig. 2.10 (f) shows the data for 24 GHz, where the
angular dependence within the [1
10]/[001]-plane (filled circles) as well as in
the [100]/[001]-plane (open circles) can be measured. All out-of-plane angular
dependencies show a twofold symmetry with a minimum resonance field in the
plane of the film. This immediately implies that the film normal is the hard
axis of the system. The twofold symmetry results from the effective magnetization Mef f that is the sum of the demagnetizing field 0 M and the intrinsic
uniaxial out-of-plane anisotropy field 2K2 /M (see Sect. 2.3.2).
The solid lines within the dependencies shown in Fig. 2.10 (b) (d)(f) are
fits of the measured dependence according to (2.42) and (2.43) in case of the
out-of-plane variation and according to (2.44) in case of the in-plane dependence. The equations were used for cubic(001) systems, including the effective
magnetization Mef f , a fourfold term K4 as well as an uniaxial in-plane term
K2$ as fitting parameters. For the gyromagnetic ratio a g-factor of 2.09
(bulk Fe) was used. The equilibrium positions for a given angle of the external field was found from numerically minimizing the free energy. All four
angular dependencies could be fitted with the same set of parameters that
are Mef f = 2.1 T, 20 K4 /M = 55 mT and 20 K2$ /M = 2 mT. The first
two contributions are very close to the Fe bulk value, showing that the film
has a dominating fourfold volume anisotropy with an in-plane easy direction
due to shape anisotropy. The surprising contribution is the twofold in plane
one. From the fit one can conclude that its easy direction is parallel to Fe
[110] or parallel to MgO [100]. As MgO substrates preferentially exhibit steps
parallel to this direction, the small anisotropy with twofold symmetry can
be attributed to steps on the substrate. The example shows that FMR is
a very sensitive tool to disentangle and quantitatively determine even small
anisotropy contributions.
The nature of the modes detected at the two frequencies can be finally deduced
from plotting the resonance dispersion (B0 ) using the anisotropy constants
that have been determined from the angular dependencies. The resulting
dispersion is shown in Fig. 2.10 (c). The expected resonance fields can be
extracted from the intersections of the horizontal dashed lines with the dispersion branches. The upper frequency branch is the one for the case that
the external field is aligned parallel to the easy +100,-directions, while the
lower branch gives the dispersion for the case that the external field is aligned
parallel to the hard in-plane axes (+110,-directions). One sees that at 24 GHz
one resonance signal along the easy as well as hard direction is expected,
while for the lower frequency of 9.2 GHz no signal along the easy axes can be

82

J. Lindner and M. Farle

measured as the dispersion for this case starts at frequencies above the one
used for the FMR experiment. The behavior of the dispersion along the hard
axes leads to the occurence of the two resonances that have been observed
in experiment. The change of a negative to a positive dispersion results from
the fact that at small fields a not aligned mode occurs, for which the magnetization precesses around the internal anisotropy field. As the external field
strength increases, this mode becomes an aligned mode that precesses around
the external magnetic field.
Fe on GaAs(001)
Fe/GaAs(001) can be considered as a prototype system for metallic films on a
semiconducting substrate and was investigated by many groups, mainly due
to its possible application within spintronic devices. A good review about the
different approaches used to grow high quality thin films and determine their
magnetic properties may be found in [90]. In [91] Fe films grown on the Ga
rich 4 6-reconstructed GaAs(001) surface were investigated using in situ
ultrahigh vacuum (UHV)-FMR. This technique can be used within the UHV
environment and thus allows for the study of films without protective layers
that usually have to be deposited on top of the film to avoid contamination.
UHV-FMR is thus well suited for the analysis of surface anisotropy that might
strongly differ from the volume contribution.
In the insets of Fig. 2.11 typical FMR spectra at 9.3 GHz with the external field parallel to the [1
10]-direction are shown for (a) a 5 ML and (b) a
20 ML thick Fe film grown on 4 6-reconstructed GaAs(001). Note the very
small linewidth B=1.8 mT of the bulk like 20 ML film shown in the insets of Fig. 2.11 (b) indicating excellent magnetic homogeneity of the films.
Figure 2.11 shows the polar angular dependence of the saturated resonance
signal, Bres , for (a) 5 ML Fe and (b) 20 ML Fe measured at room temperature. The solid lines are fits using (2.43) for cubic symmetry (i.e. for
K4$ = K4 ) with the parameters given in Table 2.11. The maximum of the
resonance field along the film normal indicates that the magnetization of the
films favors an in-plane alignment at both thicknesses. For the 20 ML film the
difference between the resonance field in parallel and perpendicular configuration is larger than for the 5 ML film implying that the thin film has a larger
anisotropy.
The fits directly yield the anisotropy fields 2Ki /M . The g-value in (2.43) was
chosen to be g = 2.09, which is the Fe-bulk value. From the anisotropy fields
the anisotropy constants were extracted using the bulk saturation magnetization (M =1.71 106 A/m). The results for the magnetocrystalline anisotropy
constants are listed in Table 2.11 for different film thicknesses (d=520 ML).
The value of K2 in the first column of Table 2.11 does not dominate over
the shape anisotropy given by 12 0 M 2 (i.e. Mef f < 0). Consequently, the
magnetization lies in the film plane for all films. As K4 is positive for films
above 11 ML, the magnetization is aligned along the [100]-direction within

2 Magnetic Anisotropy of Heterostructures

83

1.0

9.24 GHz
4.03 GHz

d&''/dB [a.u.]

1.2

Bres (T)

0.8
0.6

5 ML
9.24 GHz

0
-1
225
B [mT]

250

a) 5 ML

0.4
0.2
0.0

-80 -60-40-20 0 20 40 60 80
polar angle " '

0.9

9.24 GHz
4.03 GHz

d&''/dB [a.u.]

1.2
20 ML
9.24 GHz

1
0

x32

-1

Bres (T)

-2
0

0.6

20

40 100
B [mT]

120

b) 20 ML

0.3
0.0

-80 -60-40-20 0 20 40 60 80
polar angle " '

Fig. 2.11. Out-of-plane angular dependence of the resonance field for (a) 5 ML
Fe and (b) 20 ML Fe/GaAs(001) measured at room temperature at microwave
frequencies of (open squares) 4 GHz and (open circles) 9.2 GHz. The inset shows
typical FMR spectra measured in the film plane (B = 90 ) (from [91])

the thicker films, which is the easy axis of bulk-Fe. For thinner films d7 ML
a strong thickness dependent in-plane uniaxial anisotropy K2$ is observed
(Table 2.11). The interplay between K4 and K2$ leads to a change of the easy
axis from the [100]- towards the [110]-direction around d7 ML. Qualitatively,
the strong uniaxial in-plane anisotropy may be understood by considering the
twofold surface symmetry of the Fe-GaAs interface due to the 46 reconstruction. The rectangular surface cell is supposed to be directly connected to the
Fe-Ga and Fe-As bonds at the interface and thus to the atomic configuration
[92]. The uniaxial anisotropy could therefore be related to an uniaxial stress
within the Fe film or to a change of the Fe band structure at the interface due
to hybridization. We come back to this point later.
To investigate the in-plane anisotropy in more detail the in-plane angular dependence of the resonance field close to the [1
10]-direction was investigated.
The result is presented in Fig. 2.12 (a), where the resonance field as a function

84

J. Lindner and M. Farle

Table 2.11. Magnetic anisotropy constants of Fe on GaAs(001) for different thickness at room temperature. The conversion to eV/atom is given by 1105 J/m3
=7.4 eV/atom
Meff
2

K4
M

K2#
M

Thickness
(ML)

K2
(105 J/m3 )

K4
(10 J/m3 )

K2%
(10 J/m3 )

(mT)

(mT)

(mT)

Bulk
20
15
11
7
6
5

2.68
2.88
4.8
11.1
11.52
11.53

0.47
0.46
0.44
0.3
0
0
0

0.043
0.08
0.29
0.59
0.85
1.02

1072
910
900
785
420
390
390

27.5
27
26
17.6
0
0
0

2.5
3.5
6.0
32
56
60

of the in-plane angle is plotted for the 5 ML thick Fe film. With rotation of the
magnetic field in the film plane the saturated resonance mode (open squares)
moves to lower fields, whereas the unsaturated resonance mode (solid circles)
moves to higher field values and within 2 of rotation the FMR signal disappears. Upon comparison to the case of Fe/MgO(001) (see Sect. 2.5.1) this
directly shows that the [1
10]-direction is the hard (in-plane) axis of the system
and that the lower field mode in an unsaturated one. Using the anisotropy
constants, which have been determined by the out-of-plane angular dependent
measurements the fits for both saturated (closed circles) as well as unsaturated (open circles) modes reproduce the in-plane angular dependence around
the hard direction very well.

10
3

120

-47

-10
-20

-46

[1 1 0]

-44

in-plane angle jB

-43

5 d (ML)

10
5
0
-5
-10
-15
-20

x20

-5
-15

110

8 7

Bres (mT)

130

20 15 11

b)

K4
K2II
K2$
0.03

0.06 0.09 0.12


-1
)*d (nm )

K i ((eV/atom)

KI (10 J/m )

140

15

a)

0.15

Fig. 2.12. (a) In-plane angular dependence of a 5 ML thick Fe film on GaAs(001)


measured at a microwave frequency of 9.3 GHz. (b) Surface and volume anisotropies
for Fe/GaAs(001). The open triangles denote the uniaxial out-of-plane anisotropy,
the open circles the uniaxial in-plane anisotropy and the filled squares the fourfold
anisotropy (from [91])

2 Magnetic Anisotropy of Heterostructures

85

Table 2.12. Surface/interface and volume contributions to the magnetic constants


of Fe on GaAs(001) at room temperature. The conversion to eV/atom is given by
1103 J/m2 =515 eV/atom
v
K2
5
(10 J/m3 )

K4v
(10 J/m3 )

v
K2%
(10 J/m3 )

s,eff
K2
3
(10 J/m2 )

K4s,eff
(105 J/m2 )

s,eff
K2%
5
(10 J/m2 )

1.7

0.66

0.18

1.17

6.1

8.9

In order to separate volume and surface/interface anisotropy contributions we plot in Fig. 2.12 (b) the anisotropy constants as a function of film
thickness according to (2.27). The volume and surface/interface anisotropy
constants resulting from this analysis are summarized in Table 2.12. We start
the discussion with the uniaxial out-of-plane contribution K2 . The volume
v
contribution K2
= 1.70.8105 J/m3 (136 eV/atom) is very small. It
v
is close to the value for Au capped Fe/GaAs reported by McPhail et al. (K2
5
3
v
= 1.2 0.710 J/m = 96 eV/atom) [93]. The negative sign of K2 indicates a preferential alignment of the magnetization in the film plane due to this
volume term, which is enhanced by the larger contribution to Mef f , mainly
resulting from the shape anisotropy. The much larger surface/interface term
s,eff
K2
=1.170.1103 J/m2 (60050 eV/atom) is a superposition of the Fevacuum surface and Fe-GaAs interface anisotropy. As shown by the positive
s,eff
the interface anisotropy favors an easy axis out-of-plane, which
sign of K2
is well known also to be the case for thin Fe films on Cu(001) [94]. As the interface contribution gets more important for thinner films, the reduced value
of Mef f at lower film thicknesses is a direct result of the interface anisotropy
of second order.
The analysis of Fig. 2.12 (b) shows that the fourfold anisotropy vanishes below
7 ML indicating a transition from cubic to predominantly uniaxial symmetry.
This transition has also been observed by Brockmann et al. at a film thicknesses of 81 ML [95]. The thickness dependence of the fourfold anisotropy
constant (Fig. 2.12 (b)) yields a negative surface/interface contribution (K4s,eff
= 6.10.1105J/m2 =311 eV/atom and a positive volume contribution (K4v = 0.660.1105 J/m3 = 4.80.75 eV/atom) close to bulk iron
(K4,bulk = 0.47105 J/m3 =3.5 eV/atom). These values indicate that the
interior part of the thinner films exhibits a rather moderate strain. K4v is responsible for the alignment of the magnetization parallel to the [100]-direction
(d>7 ML), which is the easy axis for bulk-Fe. The decrease of K4 at smaller
film thickness results from the negative interface contribution K4s,eff . The negative sign indicates that the +110, directions are the favored easy axes, which
is different from the bulk easy axes +100, (positive K4 ).

86

J. Lindner and M. Farle

Similar to the uniaxial out-of-plane anisotropy, the thickness dependence


of the uniaxial in-plane anisotropy K2$ shows a very small value for the
v
volume contribution K2$
= 0.180.25105J/m3 (1.31.85 eV/atom), proving that the uniaxial anisotropy is an interface effect. It should be noted that
v
is approximately zero. The surwithin the error bar this volume value of K2$
face/interface contribution is, however, large. From the fit in Fig. 2.12 (b)
s,eff
one gets K2$
= 8.9 0.4105 J/m2 (46.32 eV/atom). This value

s,eff
was also found for Au capped samples studied by McPhail et al. (K2$
5
2
= 10 110 J/m =526 eV/atom) [93] and also by Brockmann et al.
s,eff
(K2$
= 12 2105J/m2 =6211 eV/atom) [95]. The negative sign of

s,eff
K2$
shows that the easy axis given by the uniaxial in-plane anisotropy is
the [110]-direction, whereas the [1
10] is the hard in-plane direction of our Fe
films. One should note that some confusion concerning the identification of the
crystallographic in-plane directions occurred in the literature (see e.g. [96]),
where [1
10]- and [110]-direction were erroneously exchanged. The correct description is given in [97]. The change of the easy axis for thicker films towards
the [100]-direction, which occurs at about 7 ML results therefore from the
increasing influence of the volume part of the fourfold anisotropy K4v .

Using a magneto-elastic model the anisotropy constants can be investigated in more detail. As shown by quantitative studies of the stress evolution
during Fe deposition [98] the Fe films were found to present a compressive
stress of 3.5 GPa at the initial stage of growth (first 23 ML). This stress is
even larger than the one resulting from an ideal coherent growth, for which
the stress would be given by the 1.36% misfit between Fe (a =0.2866 nm) and
GaAs (a/2 =0.2827 nm) yielding a compressive stress of 2.8 GPa [98]. This
enhancement was explained in terms of surface stress changes when the substrate reconstruction changes to the new interface consisting of Fe, Ga and As
atoms. Within the model used in [91] Fe is assumed to be uniaxially strained
at the interface. The compression of the lattice parallel to the [110]-direction
cub
per unit volume
results in a contribution of the magneto-elastic energy FMEL
to the overall free energy density (see Sect. 2.3.2 for the discussion of magnetoelastic contributions to the free energy). For Fe/GaAs(001), the strain is
given by %12 parallel to the (110)-direction. This leads to a contribution of
cub
FMEL
given by 2B2 1 2 %12 . Note that %12 is negative in our case due to
the compressive stress [98]. In [99] the magneto-elastic constants of Fe on
Ga-terminated GaAs(001) were measured. For 25 nm thick Fe films values
of B1 = 3.5 106 J/m3 and B2 = 7.2 106 J/m3 were obtained, while the
Fe-bulk values are given by B1 = 3.44106 J/m3 and B2 = 7.62106 J/m3 ,
respectively. Using the direction cosines 1 = sin cos and 2 =sin sin
one gets FMEL = +B2 %12 along the (110)-direction (1 = 2 = 2/2) and
10)-direction (1 = 2 = 2/2). With B2 > 0
FMEL = B2 %12 along the (1
and %12 < 0 a total energy reduction along the (110)-direction results, whereas

2 Magnetic Anisotropy of Heterostructures

87

the (1
10)-direction becomes a hard one in excellent agreement to our results for the films with d<7 ML. The contribution of the magneto-elastic
anisotropy is directly given by the energy difference for the two directions, i.e.
MEL
= FMEL,110 FMEL,110 = 2B2 %12 . For %12 = 1.7% (corresponding to
K2$
MEL
a stress of 3.5 GPa)K2$
= 2.6 105 J/m3 results. This has to be com-

s,eff
pared to the experimental value for the uniaxial interface anisotropy K2$
/d
5
3
= 6.210 J/m , where d is the thickness of the monolayer (d = 0.1433 nm).
The value is of the correct order of magnitude and thus strain at the interface
plays a crucial role for the uniaxial anisotropy in the film plane that dominates at small film thicknesses below 7 ML. The fact that the experimental
value is larger than the estimated one due to strain indicates, however, that
other sources possibly contribute to the uniaxial interface anisotropy as well.
One reason could be hybridization of the Fe bands with the ones of GaAs,
in particular when interface intermixing is present. However, to verify this
model, theoretical support is needed.
While the interface obviously exhibits a strain of uniaxial character, in the
volume of the films no uniaxial strain persists as can be concluded from the
v
. The experiments in [98] showed that the Fe films stay
vanishing value of K2$
compressed even for film thicknesses up to 20 nm. The magnitude of this stress
is small due to the incorporation of misfit dislocations. An x-ray absorption
fine structure experiment in [100] showed that a 10 ML thick film is strained
by 1.1%. Since the experiments in [91] show that no uniaxial strain is present
in the volume, the strain in the volume is governed by compressive strain of
biaxial character. This is consistent with a nearly cubic bcc environment of
the Fe in the inner part of the films, and unlike at the interface the twofold
symmetry of the reconstructed GaAs substrate has no effect. Hence, the inplane biaxial strain in the volume is given by %11 = %22 %. In [15] it has
been shown that a biaxial strain does not lead to an in-plane anisotropy for
cubic (001) systems. It leads, however, to an uniaxial anisotropy contribution
perpendicular to the film surface given by B1 (% %33 ). The strain component
perpendicular to the surface %33 can be calculated from elastic theory to be
12
2 cc11
% [15]. cij are the elastic constants of the material. For Fe c12 c11 , so
that %33 2% and one obtains a (perpendicular) uniaxial anisotropy term
v
given by K2
3B1 %. Using the experimental value for B1 = 3.5 106 J/m3
v
from [99] and % 1% according to [100] K2
1.0105 J/m3 results.
v
This is within the error bar the same value as the experimental one (K2
5
3
=1.70.810 J/m ) with the same negative sign showing that strain is the
v
origin for K2
. The fact that the film strain becomes smaller for thicker films
is confirmed by the observation of the fourfold volume anisotropy term K4v
that is very close to the bulk Fe value.

88

J. Lindner and M. Farle

2.5.2 Films on Metallic Surfaces


The growth of metallic films on metallic substrates will be discussed using the
example of the system Ni/Cu(001). This system became the focus of many
studies, as it presents an out-of-plane easy axis of magnetization (see [5]). Usually the shape anisotropy overcomes the magneto-crystalline contribution, so
that the magnetization will be aligned in the film plane. Due to the small magnetic moment of Ni, shape anisotropy for Ni films is small. Such spin reorientation transitions (SRTs) of the magnetization in ultrathin ferromagnetic films
have attracted much interest, experimentally [101, 102, 103, 104, 105, 106] as
well as theoretically [57, 107, 108, 109, 110, 111, 112]. One reason for the interest is that temperature driven SRTs may become of technological importance,
leading to new magnetic thin film sensors or switches. From a basic research
point of view, one should note that the existence of a perpendicular magnetization is non trivial since the demagnetizing energy and the entropy of disorder
[9] favors in-plane magnetization in thin films. In the case of Ni/Cu(001) it is
not the surface contribution to the intrinsic anisotropy, but the volume part,
that lead to the easy axis out of the film plane. The complicated interplay
between all contributions can be experimentally manipulated yielding direct
access to the thickness of the reorientation transition from in to out-of-plane.
Ni/Cu(001)
Ni films on Cu(001) with thicknesses below 1011 ML exhibit a film magnetization with easy axis in the film plane. For larger thicknesses Ni films
show a spin reorientation transition (change in sign of Mef f ), which drives
the magnetization out of the film plane [5]. This transition is shifted down
to about 7.5 ML when capping the films with Cu [113]. In addition, the TC
is reduced by a Cu overlayer. The complete out-of-plane angular dependent
measurement for Ni at room temperature is shown in Fig. 2.13, where Bres
as a function of B for 8 ML of Ni/Cu(001) (a) and 9 ML of Ni/Cu(001) (b)
Meff,8 = -4.80(5) kG

Meff,9 = -1.9(1) kG

4
2

Meff,8 = 1.336(8) kG

B res (102 mT)

B res (102 mT)

-45

45

qB

a)
90

Meff,9 = 2.56(4) kG
-90 -45

45

b)
90

qB

Fig. 2.13. Out-of-plane angular dependence for (left panel) 8 ML and (right panel)
9 ML Ni/Cu(001) before (filled circles) and after (open circles) capping the films
with 4 ML of Cu (reproduced from [14])

2 Magnetic Anisotropy of Heterostructures

89

is plotted. In each panel we show the results before (filled circles) and after
capping the film with 4 ML of Cu (open circles). For the bare Ni films the
negative values of Mef f indicate an easy axis in-plane. The smaller value of
the Ni9 film3 shows that this film is closer to the spin reorientation at about
1011 ML, where Mef f is close to zero and thus, the higher order anisotropy
constants become important [5]. After capping the two films with Cu the minimum of Bres moves from B = 90 to B = 0 . This means that the easy
axis has switched from in the film plane towards the film normal. This change
of the easy axis is also reflected in the positive sign of Mef f . Besides the sign
of Mef f its value is larger for the Cu4 Ni9 film compared to the Cu4 Ni8 film,
which means an increase of the out-of-plane anisotropy within the Cu capped
Ni films.
This trend is continued with increasing Ni thickness (Fig. 2.14). Figure 2.14
(c) shows the dispersion curves according to (2.42) with K4i = K2$ = 0 as
determined from experimentally measured Bres values for Ni films with thicknesses ranging between 8 and 12 ML already capped with Cu. The external
field was aligned along the in-plane direction (B = 90 ). The points of intersection of the 9 GHz line with the dispersion yield the resonance fields.
Figure 2.14 (a) shows the original FMR spectra for the smallest (8 ML) and
the largest (12 ML) Ni thickness. For 12 ML two resonances are observed
(unsaturated and saturated mode, see Sect. 2.5.1 for details). This behavior can be explained with an increase in the perpendicular anisotropy given
by K2 and corresponds to the case of a film having an out-of-plane easy
axis of magnetization. The complete analysis presented in Fig. 2.14 (c) indicates that the positive values of Mef f increase with Ni thickness. Therefore,
two resonances are observed for thicknesses larger than 10 ML. The transition between observing one and two signals is observed for the 10 ML film,
which was investigated at different temperatures in the range 300400 K.
The transition between observing one and two signals can be seen directly
in Fig. 2.14 (b), where the measured resonance fields of the film are plotted as a function of the temperature. At room temperature one observes
two signals, which are also shown in the inset for T = 313 K. The resonance field for the signal at low field values moves towards smaller fields
until it vanishes at about T = 375 K. Above 375 K only one resonance is
observed as shown in the inset for T = 380 K, which stays even up to the
highest temperature of 400 K. As Fig. 2.14 (d) shows, this behavior can be
explained quantitatively with the help of the dispersion calulations. Due to
the decrease of Mef f at higher temperatures the left branch of the dispersion
curve moves below the 9 GHz line, so that only one signal is observed at larger
T values.
We have seen that capping the Ni films with Cu induces a shift of the thickness, at which (for constant temperature) the reorientation of the easy axis
of the magnetization occurs. Cu tends to decrease the critical thickness. This
3

In the following the number of monolayers is given as subscript.

90

J. Lindner and M. Farle


6

dc/dB (arb. units)

(102 mT)

a)

res

Cu4Ni12/Cu(001)

dc/dB (arb. units)

Cu4Ni10/Cu(001), f=9GHz, B0 II [110]

Cu4Ni8/Cu(001)

200

a)

20

400

500

B0|| (mT)

600

0
300

Cu4Ni12/Cu(001), Meff=532(5) mT
Cu4Ni10
403(5) mT
Cu5Ni9
256(4) mT
Cu4Ni8
133.6(8) mT

320

340

360

T (K)

380

10

Microwave
frequency

c)
200

400

600

B0|| (mT)

800

1000

400

T=310K, Meff =372(5) mT


360K,
302(5) mT
380K,
220(5) mT
400K,
206(5) mT

20

10

0
0

T=313 K
400
600
B0|| (mT)

b)

700

% (Ghz)

% (Ghz)

300

T=380 K
0

d)
0

200

400

B0|| (mT)

600

800

Fig. 2.14. a) Spectra for 8 ML and 12 ML thick Ni films capped by 4 ML of Cu.


b) Resonance fields as a function of the temperature for the two signals observed in
a 10 ML thick Ni film capped by 4 ML of Cu. One of the signal vanishes at higher
temperatures. The inset shows the spectra at two temperatures. c) Microwave resonance frequency as a function of the external field for Ni films with thicknesses from
812 ML. d) Microwave frequency as a function of the external field for various temperatures calculated for the same film shown in a). The calculation was performed
such that the observed resonance fields are best reproduced by the intersections of
the calculated curve with the 9 GHz line being the experimental frequency. From
the calculation the anisotropy is derived despite the fact that not the whole angular
dependence could be measured (reproduced from [14])

effect was the motivation by several groups to investigate capping layers of


different kind and their influence on the reorientation thickness. Upon capping
the Ni films with Cu and also investigating the same film without a protective
layer within an UHV environment, one can disentangle the Ni/vacuum from
the Ni/Cu interface anisotropy. In [113] the influence of CO on the interface
anisotropy and in [114] the Ni films were grown on O-reconstructed Cu(001).
The latter procedure interestingly leads to a chemically bonded O layer floating on top of the Ni film. The changes of the interface anisotropy are listed
in Table 2.13. One sees that the growth on the O-reconstructed surface is
most effective. Using H2 has the advantage that the procedure is reversible as

2 Magnetic Anisotropy of Heterostructures

91

Table 2.13. Interface anisotropies for Ni films on Cu(001) with various capping
int
is 10% and 15% for O/Ni. The values for CO and H2 are
layers. The error for K2
from [113]. Note that the measurements in [113] are performed at room temperature
Interface

int
K2
(eV/Atom)

dC (ML)

Vakuum/Ni
Cu/Ni
CO/Ni
H2 /Ni
O/Ni

107
59
81
70
17

10,8
7,6
7,3
6,8
4,9

by heating the hydrogen can be driven out of the film. This example shows
that adsorbates and (or) chemically bonded layers can be effectively used to
influence interface anisotropies and thus taylor magnetic properties of thin
films.
The reason for the out-of-plane reorientation of Ni films can be deduced
from thickness dependent measurements. The result, plotted as function of
the reciprocal Ni film thickness, for bare Ni/Cu(001) (filled circles), Cu
capped films (open circles) and films grown on a O-reconstructed Cu(001)
surface (open squares, see [114] for details) are shown in Fig. 2.15. According to (2.27) one extracts a negative surface anisotropy (that is the sum of
the two interfaces) and a positive volume anisotropy that even overcomes the
25
20
15

K2^(eV/atom)

10 (0 M2
2
5
0
Ni/Cu(001)
-5

CuNi/Cu(001)
Ni on O/Cu(001)

-10

t=0.75

-15
-20
0

0.05

0.10
0.15
1/d (1/ML)

0.20

Fig. 2.15. Surface and volume uniaxial out-of-plane anisotropy for (filled
circles) Ni/Cu(001), (open circles) Cu/Ni/Cu(001) and (open squares) Ni on
O-reconstructed Cu(001) (reproduced from [82])

92

J. Lindner and M. Farle

shape anisotropy for thicknesses above the critical thickness of the reorientation. Within the error the volume contribution is not affected by capping the
Ni film. A negative surface anisotropy favors an in-plane easy alignment and,
thus, it is the volume anisotropy that leads to the out-of-plane easy axis. This
is contrary to many other systems, where the surface MAE favors an out-ofplane easy axis (as e.g. for Fe/GaAs(001), see Table 2.12). In such a case,
for a thickness of only a few ML, the surface contribution does not dominate
anymore and the easy axis turns into the film plane. The fact that Ni films
on Cu(001) show the opposite effect make them a very interesting system. In
[115] it was shown that the positive volume anisotropy is a direct consequence
of the tetragonal distortion of the Ni films, showing again that the structure
and symmetry of the system plays a major role for the MAE.

Acknowledgments
We thank all our past and present coworkers who have contributed to the
results presented here. Although we have tried to include all the relevant
literature of the many groups working in this field, the list of references must
be incomplete due to the large number of publications and we apologize for
any missing citation. This work was supported by the DFG, Sfb 290 and 491.

References

1. J.M. Stocks, M. Eisenbach, B. Ujfalussy,


B. Lazarovits, L. Szunyogh,
P. Weinberger: Progress in Materials Science 52, 371 (2007) 45
2. H.J. Richter: J. Phys. D: Appl. Phys. 40, R149 (2007) 45
3. U. Gradmann: Magnetism in ultrathin transition metal films. In: Handbook of
Magnetic Materials, Vol. 7, ed. by K.H.J. Buschow (Elsevier Science Publishers
B.V. 1993) 46
4. B. Heinrich, J.F. Cochran: Adv. Phys. 42, 523 (1993) 46, 63, 78
5. M. Farle: Rep. Prog. Phys. 61, 755 (1998) 46, 66, 74, 76, 78, 88, 89
6. J.A.C. Bland, B. Heinrich (Eds.): Ultrathin magnetic structures I An Introduction to the Electronic, Magnetic and Structural Properties, 2nd edn (Springer,
Berlin Heidelberg New York 2005) 46
7. B. Heinrich, J.A.C. Bland (Eds.): Ultrathin magnetic structures II Measurement Techniques and Novel Magnetic Properties, 2nd edn (Springer, Berlin
Heidelberg New York 2005) 46, 78
8. J.A.C. Bland, B. Heinrich (Eds.): Ultrathin magnetic structures III Fundamentals of Nanomagnetism, 2nd edn (Springer, Berlin Heidelberg New
York 2005) 46
9. B. Hillebrands, K. Ounadjela (Eds.): Spin Dynamics in confined magnetic
Structures I, Topics in Applied Physics 83 (Springer, Berlin Heidelberg New
York 2003) 46, 88
10. B. Hillebrands, K. Ounadjela (Eds.): Spin Dynamics in confined magnetic
Structures II, Topics in Applied Physics 87 (Springer, Berlin Heidelberg New
York 2003) 46

2 Magnetic Anisotropy of Heterostructures

93

11. B. Hillebrands, K. Ounadjela (Eds.): Spin Dynamics in confined magnetic


Structures III, Topics in Applied Physics 101 (Springer, Berlin Heidelberg New
York 2003) 46
12. K. Baberschke, M. Donath, W. Nolting (Eds.): Band-Ferromagnetism, Lecture
Notes in Physics 580 (Springer, Berlin Heidelberg New York 2001) 46, 74
13. M. Donath, W. Nolting (Eds.): Local-Moment Ferromagnets Unique Properties for Modern Applications, Lecture Notes in Physics 678 (Springer, Berlin
Heidelberg New York 2005) 46
14. J. Lindner, K. Baberschke: J. Phys.: Condens. Matter 15, R193 (2003);
J. Lindner, K. Baberschke: J. Phys.: Condens. Matter 15, S465 (2003) 46, 78, 88, 90
15. D. Sander: J. Phys.: Condens. Matter 16, R603 (2004) 46, 48, 66, 87
16. J.C. Slonczewski: J. Magn. Magn. Mater. 159, L1 (1996) 46
17. Y. Tserkovnyak. A. Brataas, G.E.W. Bauer: Phys. Rev. Lett. 88, 117601 (2002)
46
18. M.D. Stiles, A. Zangwill: Phys. Rev. B 66, 014407 (2002) 46
19. J.A. Katine, F.J. Albert, R.A. Buhrman: Phys. Rev. Lett. 84, 3149 (2000) 46
20. B. Heinrich, Y. Tserkovnyak, G. Woltersdorf, A. Brataas, R. Urban, G. Bauer:
Phys. Rev. Lett. 90, 187601 (2003) 46
21. K. Lenz, T. Toli
nski, E. Kosubek, J. Lindner, K. Baberschke: phys. stat. sol.
(c) 1, 3260 (2004) 46
22. M. Zwierzycki, Y. Tserkovnyak, P.J. Kelly, A. Brataas, G.E.W. Bauer: Phys.
Rev. B 71, 64420 (2005) 46
23. Ch. Kittel: Introduction to solid state physics, 8. edn (Wiley and Sons 2005) 47
24. M. Farle, A. Berghaus, and K. Baberschke: Phys. Rev. B 39, 4838 (1989) 47, 63
25. W. Szmaja: Advances in Imaging and Electron Physics 141, 175 (2006) 48
26. R.P. Cowburn, A.O. Adeyeye, M.E. Welland: Phys. Rev. Lett. 81, 5414 (1998)
48
27. J. Nogues, I.K. Schuller: J. Magn. Magn. Mater 192, 203 (1999) 48
28. A.E. Berkowitz, Kentaro Takano: J. Magn. Magn. Mater 200, 552 (1999) 48
29. M. Kiwi: J. Magn. Magn. Mater 192, 584 (2001) 48
30. J. Nogues, J. Sort, V. Langlais, V. Skumryev, S. Suri
nach, J. S. Mu
noz,
M.D. Bar
o: Physics Reports 422, 65 (2005) 48
31. P.J. Jensen, K.H. Bennemann: Surface Science Reports 61, 129 (2006) 48, 49
32. Landolt-B
ornstein, Vol. III/19a, ed. by H.P.J. Wijn (Springer Berlin 1986) 47
33. P. Bruno: Phys. Rev. B 39, 865 (1989) 48
34. P. Poulopoulos, K. Baberschke: J. Phys.: Condens. Matter 11, 9495 (1999) 49, 74
35. G. van der Laan: Phys. Rev. Lett. 82, 640 (1999) 49
36. P. Strange, J.B. Staunton, B.L. Gyorffy, H. Ebert: Physica B 1720, 51 (1991) 49
37. J.B. Staunton, P. Strange, B.L. Gyorffy, M. Matsumoto, J. Poulter, H. Ebert,
N.P. Archibald: Theory of Magnetocrystalline Anisotropy. In: The effects of relativity in atoms and molecules and the solid-state, ed. by S. Wilson, I. P. Grant
and B. L. Gyorffy (Plenum Press 1991) pp 295 49
38. B. Nonas, I. Cabria, R. Zeller, P.H. Dederichs, T. Huhne, H. Ebert: Phys. Rev.
Lett. 86, 2146 (2001) 49, 74
39. J.G. Gay, Roy Richter: Phys. Rev. Lett. 56, 2728 (1986) 49
40. Ding-Sheng Wang, Ruqian Wu, A.J. Freeman: Phys. Rev. Lett. 70, 869 (1993);
Phys. Rev. Lett. 71, 2166 (1993) 49
41. O. Hjortstam, J. Trygg, J.M. Wills, B. Johansson, O. Ericksson: Phys. Rev. B
53, 9204 (1996) 49

94

J. Lindner and M. Farle

42. B. Ujfalussy,
L. Szunyogh, P. Bruno, P. Weinberger: Phys. Rev. Lett. 77,
1805 (1996) 49
43. J. Dorantes-D
avila, G.M. Pastor: Phys. Rev. Lett. 77, 4450 (1996) 49
44. C. Uiberacker, J. Zabloudil. P. Weinberger, L. Szunyogh, C. Sommers: Phys.
Rev. Lett. 82, 1289 (1999) 49
45. E. Sj
ostedt, L. Nordstr
om, F. Gustavsson, O. Eriksson: Phys. Rev. Lett. 89,
267203 (2002) 49
46. J. Dorantes-D
avila, H. Dreysse, G.M. Pastor: Phys. Rev. Lett. 91,
197206 (2003) 49
47. J.B. Staunton, S. Ostanin, S.S.A. Razee, B.L. Gyorffy, L. Szunyogh,
B. Ginatempo, E. Bruno: Phys. Rev. Lett. 93, 257204 (2004) 49, 77
48. M. Kosuth, V. Popescu, H. Ebert, et al., Europhys. Lett. 72, 816 (2005) 49
49. L.F. Yin, D.H. Wei, N. Lei, L.H. Zhou, C.S. Tian, G.S. Dong, X.F. Jin,
L.P. Guo, Q.J. Jia, R.Q. Wu: Phys. Rev. Lett. 97, 067203 (2006) 49
50. D. Weller, J. St
ohr, R. Nakajima, A. Carl, M.G. Samant, C. Chappert, R. Megy,
P. Beauvillain, P. Veillet, G.A. Held: Phys. Rev. Lett. 75, 3752 (1995); Comment on Phys. Rev. Lett. 75, 3752 (1995) by H. A. D
urr, G. van der Laan,
B. T. Thole: Phys. Rev. Lett. 76, 3464 (1996) 49
51. P. Gambardella, S. Rusponi, T. Cren, N. Weiss, H. Brune: Comptes Rendus
Physique 6, 75 (2005) 49
52. L. Neel: J. Physique Radium 15, 225 (1954) 51
53. H.J. Elmers, T. Furubayashi, M. Albrecht and U. Gradmann: J. Appl. Phys.
70, 5764 (1991) 51
54. O.N. Mryasov, U. Nowak, K.Y. Guslienko, R.W. Chantrell: Europhys. Lett. 69,
805 (2005) 51, 77
55. R.R. Birss: Symmetry and Magnetism. In: Series of Monographs on selected
Topics in Solid State Physics, ed by E. P. Wohlfarth (North-Holland Publishing
Co., Amsterdam, John Wiley and Sons, New York 1966) 52, 55, 61, 74, 75
56. A. Berghaus, M. Farle, Yi Li, K. Baberschke: In: Springer Proceedings in
Physics, Vol. 50 (Springer Berlin 1990) pp 61 56
57. P.J. Jensen, K.H. Bennemann, Phys. Rev. B 52, 16012 (1995) 56, 88
58. M. Farle, B. Mirwald-Schulz, A.N. Anisimov, P. Poulopoulos, K. Baberschke:
Phys. Rev. B 55, 3708 (1997) 56
59. Y.T. Millev, H.P. Oepen, J. Kirschner: Phys. Rev. B 57, 5837 (1998) 56
60. P. Bruno P: J. Appl. Phys. 64, 3153 (1988) 63
61. P. Poulopoulos, J. Lindner, M. Farle, K. Baberschke: Surf. Sci. 437, 277 (1999)
63
62. M. Benson, D.L. Mills: Phys. Rev. 178, 839 (1969) 63
63. H. Fritzsche, J. Kohlhepp, U. Gradmann: Phys. Rev. B 51, 15933 (1995) 64
64. P. Bruno: Physical origins and theoretical models of magnetic anisotropy. In:
ulich
Vorlesungsmanuskripte des 24.IFF-Ferienkurses im Forschungszentrum J
(Forschungszentrum J
ulich 1993) 65
65. C. Chappert, P. Bruno: J. Appl. Phys. 64, 5736 (1988) 65
66. R.F.S. Hearmon: In: Landolt-B
ornstein Numerical Data and Functional Relationships in Science and Technology Group III, vol 2, (Berlin, Springer 1969);
R.F.S. Hearmon: The Elastic Constants of Crystals and Other Anisotropic Materials. In: Landolt-B
ornstein Numerical Data and Functional Relationships in
Science and Technology Group III, vol 18, (Springer, Berlin 1984) 66
67. J.M. Luttinger: Helv. Phys. Acta 21, 480 (1948) 66

2 Magnetic Anisotropy of Heterostructures


68.
69.
70.
71.
72.
73.
74.

75.
76.
77.
78.
79.
80.
81.
82.

83.
84.
85.

86.
87.
88.
89.

90.
91.
92.
93.
94.
95.
96.

95

J.H. van Vleck: Phys. Rev. 78, 266 (1950) 67


J. Smit, H.G. Beljers: Phillips Res. Rep. 10, 113 (1955) 67
H. Suhl: Phys. Rev. 97, 555 (1955) 67
T.L. Gilbert: Phys. Rev. 100, 1243 (1955) 67
H. Goldstein: Klassische Mechanik, 11. edn (AULA-Verlag, Wiesbaden 1991) 68
L. Baselgia, M. Warden, F. Waldner, S.L. Hutton, J.E. Drumheller, Y.Q. He,
P.E. Wigen, M. Mar
ysko: Phys. Rev. B 38, 2237 (1988) 69, 70
A.N. Anisimov, M. Farle, P. Poulopoulos, W. Platow, K. Baberschke, P. Isberg,
R. W
appling, A.M.N. Niklasson, O. Eriksson: Phys. Rev. Lett. 82, 2390 (1999)
74
S.V. Vonsovskii: Magnetism, (Wiley, New York 1974) pp 946ff 74
H.B. Callen, E. Callen: J. Phys. Chem. Solids 27, 1271 (1966) 76
W.J. Car, Jr: Phys. Rev. 109, 1971 (1958) 76
D.P. Pappas: J. Vac. Sci. Technol. 14, 3203 (1996) 77
O. Fruchart, J.-P. Nozieres, D. Givord: J. Magn. Magn. Mater. 165, 508 (1997)
77
S. Okamoto, N. Kikuchi, O. Kitakami, T. Miyazaki, Y. Shimada, K. Fukamichi:
Phys. Rev.B 66, 024413 (2002) 77
J.-U. Thiele, K.R. Coffey, M.F. Toney, J.A. Hedstrom, and A.J. Kellock: J.
Appl. Phys. 91, 6595 (2002) 77
J. Lindner: Ferromagnetische Resonanz an ultrad
unnen magnetischen Einfach
und Mehrfachlagen der 3d-Ubergangsmetalle
Statik und Dynamik. PhD thesis, Freie Universit
at Berlin, Berlin (dissertation.de 2003) 78, 91
Kh. Zakeri, Th. Kebe, J. Lindner, M. Farle: Phys. Rev. B 73, 052405 (2006) 77
C. Li, A.J. Freeman: Phys. Rev. B 43, 780 (2002) 79
Yu.V. Goryunov, N.N. Garifyanov, G.G. Khaliullin, I.A. Garifullin,
L.R. Tagirov, F. Schreiber, Th. M
uhge, H. Zabel: Phys. Rev. B 52, 13450 (1995)
79
Th. M
uhge, A. Stierle, N. Metoki, U. Pietsch, H. Zabel: Appl. Phys. A 59,
659 (1994) 79
G.A. Prinz, G.T. Rado, J.J. Krebs: J. Appl. Phys. 53, 2087 (1982) 79
R. Meckenstock, K. Harms, O. von Geisau, J. Pelzl: J. Magn. Magn. Mater.
148, 139 (1995) 79
M. Klaua, D. Ullmann, J. Barthel, W. Wulfhekel, J. Kirschner, R. Urban,
T.L. Monchesky, A. Enders, J.F. Cochran, B. Heinrich: Phys. Rev. B 64,
134411 (2001) 79
G. Wastlbauer, J.A.C. Bland: Adv. Phys. 54, 137 (2005) 82
Kh. Zakeri, Th. Kebe, J. Lindner, M. Farle: J. Magn. Magn. Mater. 299,
L1 (2006) 82, 83, 84, 86, 87
T. Toli
nski, K. Lenz, J. Lindner, E. Kosubek, K. Baberschke, D. Spoddig,
R. Meckenstock: Sol. Stat. Comm. 128, 385 (2003) 83
S. McPhail, C.M. C
urtler, F. Montaigne. Y.B. Yu, M. Tselepi, J.A.C. Bland:
Phys. Rev. B 67 24409, (2003) 85, 86
W. Platow, A.N. Anisimov, M. Farle, K. Baberschke: phys. stat. sol. (a) 173,
145 (1999) 85
M. Brockmann, M. Z
olfl, S. Miethaner, and G. Bayreuther: J. Magn. Magn.
Mater. 198, 384 (1999) 85, 86
M. Gester, C. Daboo, R.J. Hicken, S.J. Gray, A. Ercole, and J.A.C. Bland: J.
Appl. Phys. 80, 347 (1996) 86

96

J. Lindner and M. Farle

97. R. Moosbuhler, F. Bensch, M. Dumm, G. Bayreuther: J. Appl. Phys. 91,


8757 (2002) 86
98. G. Wedler, B. Wassermann, R. N
otzel, R. Koch: Appl. Phys. Lett. 78,
1270 (2001) 86, 87
99. G. Wedler, B. Wassermann, R. Koch: Phys. Rev. B 66, 064415 (2002) 86, 87
100. R.A. Gordon, E.D. Crozier, D.-T. Jiang, T.L. Monchesky, B. Heinrich: Phys.
Rev. B 62, 2151 (2000) 87
101. U. Gradmann: Annalen der Physik 17, 91 (1966) 88
102. D.P. Pappas, K.-P. K
amper, H. Hopster: Phys. Rev. Lett. 64, 3179 (1990) 88
103. R. Allenspach, A. Bischof: Phys. Rev. Lett. 69, 3385 (1992) 88
104. Dongqi Li, M. Freitag, J. Pearson, Z.Q. Qiu, and S.D. Bader: Phys. Rev. Lett.
72, 3112 (1994) 88
105. A. Berger, H. Hopster: Phys. Rev. Lett. 76, 519 (1996); A. Berger, H. Hopster:
J. Appl. Phys. 79, 5619 (1996) 88
106. G. Garreau, E. Beaurepaire, K. Ounadjela, M. Farle: Phys. Rev. B 53,
1083 (1996) 88
107. D. Pescia, V. L. Pokrovsky: Phys. Rev. Lett. 65, 2599 (1990) 88
108. A. Moschel, K.D. Usadel: Phys. Rev. B 49, 12868 (1994) 88
109. D.K. Morr, P.J. Jensen, K.H. Bennemann: Surf. Sci. 307309, 1109 (1994) 88
110. S.T. Chui: Phys. Rev. B 50, 12559 (1994) 88
111. X. Hu, R. Tao, Y. Kawazoe: Phys. Rev. B 54, 65 (1996) 88
112. Y. Millev, J. Kirschner: Phys. Rev. B 54, 4137 (1996) 88
113. S. van Dijken, R. Vollmer, B. Poelsema, J. Kirschner: J. Magn. Magn. Mater
210, 316 (2000) 88, 90, 91
114. J. Lindner, P. Poulopoulos, R. N
unthel, E. Kosubek, H. Wende, K. Baberschke:
Surf. Sci. Lett. 523, L65 (2003) 90, 91
115. W. Platow, U. Bovensiepen, P. Poulopoulos, M. Farle, K. Baberschke,
L. Hammer, S. Walter, S. M
uller, K. Heinz: Phys. Rev. B 59, 12641 (1999) 92

Vous aimerez peut-être aussi