Anleitung zu Versuch 23

Ferromagnetische Resonanz
(FMR)

erstellt von
Dr. Thomas Meier

Fortgeschrittenenpraktikum (FOPRA)

Physik-Department
Technische Universität München

Anleitung zuletzt aktualisiert am: December 18, 2020

Contents
1 Introduction 3

2 Theoretical Basics 4
2.1 Ferromagnetism - a brief introduction . . . . . . . . . . . . . . . . 4
2.2 Magnetic Energy and Anisotropy . . . . . . . . . . . . . . . . . . 5
2.2.1 Demagnetizing Energy . . . . . . . . . . . . . . . . . . . . 5
2.2.2 Magnetocristalline Anisotropy . . . . . . . . . . . . . . . . 7
2.2.3 Zeeman Energy . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.4 Total energy density and effective field . . . . . . . . . . . 8
2.2.5 Finding the equilibrium position of the magnetization . . . 8
2.3 Landau-Lifshitz-Gilbert equation . . . . . . . . . . . . . . . . . . 9
2.4 Ferromagnetic resonance - resonance condition . . . . . . . . . . . 11
2.4.1 Principles of FMR . . . . . . . . . . . . . . . . . . . . . . 11
2.4.2 Definition of a suitable coordinate system . . . . . . . . . 11
2.4.3 Calculation of the effective field . . . . . . . . . . . . . . . 12
2.4.4 Solution of the linearized LLG . . . . . . . . . . . . . . . . 14
2.4.5 Resonance condition . . . . . . . . . . . . . . . . . . . . . 14
2.4.5.1 In-plane configuration . . . . . . . . . . . . . . . 15
2.4.5.2 Out-of-plane configuration . . . . . . . . . . . . . 15
2.4.6 Dynamic susceptibilities and line shape of the FMR . . . . 15
2.5 Ferromagnetic resonance - damping . . . . . . . . . . . . . . . . . 18

3 Experimental basics and evaluation 19

3.1 General structure of an FMR-spectrometer . . . . . . . . . . . . . 19
3.2 Microwave technology . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 Evaluation of the resonance spectra . . . . . . . . . . . . . . . . . 22

4 Available samples 22

5 Experimental procedure 23
5.1 Influence of the measurement parameters . . . . . . . . . . . . . . 23
5.2 Frequency dependence in the in-plane configuration on permalloy 24
5.3 Examination of an ytrium-iron-garnet sample . . . . . . . . . . . 24
5.3.1 In-plane configuration . . . . . . . . . . . . . . . . . . . . 24
5.3.2 Perpendicular configuration . . . . . . . . . . . . . . . . . 25

6 Protocol requirements 25

7 Further literature 27
1 Introduction
Ferromagnetic resonance (FMR) is a widely used method for characterization of
ferromagnetic samples. The magnetic moments in the sample are excited by a
small alternating magnetic field with frequency in the microwave range to precess
around their equilibrium position.
Using ferromagnetic resonance, the equilibrium position of the magnetization,
magnetic anisotropies or dynamic properties, such as the gyromagnetic ratio or
the damping of the precession of the magnetization, can be investigated. The
method of ferromagnetic resonance is applicable to a wide range of magnetic
samples from macroscopic samples to ultrathin layers with thicknesses in the
range of single atomic layers or even to nanostructured magnetic devices.
Because of these properties, FMR has also gained great importance in research
for modern computer technology. Computer technology in its current form is lim-
ited in its development towards ever more densely packed and smaller transistors
by the ohmic heat loss in the transistors, which leads to an ever increasing heating
of the components due to the decreasing size of the transistors. This problem is to
be solved in the future by replacing charge currents by spin currents in so-called
spintronic devices. In these devices, information will no longer be transported by
charge but by spin orientation, so that ohmic losses no longer occur. FMR can
be used to characterize the respective samples or to generate spin currents to test
spin effects, such as the spin Hall effect.
Ferromagnetic resonance was first experimentally discovered independently by
J. Griffiths and E. Zavoisky in 1919. Zavoisky in 1946 [1, 2], ,after it was observed
by chance already in 1911 by V. K. Arkad’yev. The first theoretical description
was given by C. Kittel in 1948 [3]. The underlying problem of the relaxation
dynamics of the magnetization vector had already been theoretically treated by
Landau and Lifshitz in 1935, which finally led to the development of the basic
equation of magnetization dynamics, the so-called Landau-Lifshitz-Gilbert equa-
tion (LLG). [4, 5].

Exercise 1: Read the following introduction to the theory of FMR and to

the experimental setup thoroughly and try to understand all essential aspects
before performing the experiment. Also work on the other tasks contained in the
following text.

3
2 Theoretical Basics
In this section the basic theoretical aspects of ferromagnetic resonance shall be
made understandable. It starts with a very short introduction to the ferromag-
netism of 3d transition metals and the introduction of the main energy contribu-
tions in the ferromagnet. In the following the Landau-Lifshitz-Gilbert equation
is discussed and the derivation of the resonance condition is explained. Finally,
a phenomenological description of the damping effects in ferromagnets observed
in experiments is briefly presented. The whole chapter follows in large parts [6].

2.1 Ferromagnetism - a brief introduction

Few subjects in science are more difficult to understand than magnetism.
Encyclopedia Brittanica, 1989
What is already true for magnetism in general is especially true for ferromag-
netism, since it is one of the most important everyday phenomena that can no
longer be described by classical physics. The exchange interaction underlying
ferromagnetism, which leads to a spontaneous parallel alignment of the magnetic
moments in the ferromagnet, is a purely quantum mechanical phenomenon, which
results from the interaction of the Coulomb repulsion of the electrons with the
Pauli principle.
The ferromagnetism of the so-called 3d transition metals iron, cobalt and nickel
is particularly complex, since the magnetic moments are represented by electrons
in 3d states, which are partially delocalized, i.e. can move almost freely through
the crystal lattice. Since the exact explanation by means of band structure models
would require extensive knowledge of solid state physics, a simpler illustrative
explanation shall be presented here.
Based on the Pauli principle of quantum mechanics, electrons with the same
spin must not be located at the same place, since they would then coincide in all
quantum numbers. The closer two electrons are to each other, the greater their
Coulomb energy becomes because of the repulsion of their equal charges. If we
now assume that the electrons have the same spin state, the Pauli principle forbids
that the electrons are in the same place. Therefore, for electrons with parallel
spin, the average distance between the electrons increases and the Coulomb energy
is minimized. According to this consideration, it would be advantageous for all
metals to spontaneously form parallel spins, i.e. ferromagnetism. So there must
still be an opposite effect, which energetically disadvantages the parallel alignment
of the spins. This can also be explained by the Pauli principle. If the electrons
have parallel spin, then according to the Pauli principle, the impulses of the
electrons must not coincide, because otherwise the electrons would be equal in all
quantum numbers. Since the states with low momentum are already all occupied,
states with higher momentum than before must be occupied during the transition
from opposite to parallel spins, so that the average kinetic energy of the electrons
increases. Paragraph based on [7–9].

4
 For most metals, the increase in kinetic energy exceeds the reduction in Coulomb
energy, so that no spontaneous parallel arrangement of spins, i.e., no ferromag-
netism, results. Only for very few materials is the parallel arrangement of spins
more favorable, such as iron, cobalt and nickel. These materials become ferro-
magnetic below a characteristic temperature, the Curie temperature TC , and
show a spontaneous parallel alignment of neighboring spins [8, 9].

2.2 Magnetic Energy and Anisotropy

In this section the most important energy contributions in ferromagnetic samples
are explained. For the samples considered in this practical experiment, the so-
called macrospin approximation shall be applied. This assumes that all magnetic
moments in the sample are aligned in parallel due to the exchange interaction,
and that the magnetic state of the sample can therefore be described by a single
magnetization vector [10, 11]. In general, this approximation is only suitable for
thin samples that are uniformly magnetized in the sample plane.
The first energy contribution in the ferromagnet is the exchange energy, which
is larger the more the magnetic moments deviate from the parallel alignment.
Since the macrospin approximation is used here, where all moments are always
parallel, the exchange energy is constant and need not be considered further.

2.2.1 Demagnetizing Energy

The first energy contribution relevant for the experiment is the demagnetizing
energy. This can be considered as a direct consequence of Maxwell’s equations in
a spontaneously magnetized body.As should already be known from the lectures
on electrodynamics, two magnetic field quantities are needed to describe a ferro-
magnetic material. One is the magnetic induction B, ~ the other is the magnetic
field H.
~ These two field quantities are related to the magnetization by a simple
equation [8]:  
~ = µ0 H
B ~ +M ~ (1)
In this tutorial, SI units will be used instead of the cgs units still commonly used
in FMR theory. The unit for the magnetic induction is therefore [B] = 1 T
(Tesla) and for the magnetic field is [H] = 1 m A
.
Furthermore, for the B-field the following Maxwell equations are valid:
~ =0
~ ·B
∇ (2)

The magnetic field H~ is unlike the magnetic induction B~ non-source-free. Sub-

stituting equation (1) into equation (2) gives thedivergence of H,
~ which in general
does not vanish.
~ ·H
∇ ~ = −∇ ~ ·M~ (3)
The H-field
~ can be decomposed into a sum of two fields [8].
~ =H
H ~ ex + H
~d (4)

5
 Z
H
θH
M
θM
Y
φH
φM

Figure 1: The figure shows the coordinate system XYZ of the sample, which is
used in the calculations of the demagnetizing field of a thin film and in
the calculation of energy densities. The angles describing the magneti-
zation M~ and the magnetic field H ~ are also shown in the graph.

The first part is the external field H~ ex caused by currents or other magnets, the
second part H ~ d is called demagnetizing field inside the sample and stray field
outside the sample [8].
The decomposition is chosen in such a way that the demagnetizing field is a
conservative field and can therefore be represented as a gradient of a potential
generated by so-called magnetic charges. Analogous to the first Maxwell’s equa-
tion for the electric field ∇ ~ = ρ with the charge density ρ the density of the
~ ·E
0
magnetic charges may be identified by comparison with equation 3 [8].

ρ m = −∇
~ ·M
~ (5)

In a homogeneously magnetized sample, the magnetic charges are therefore lo-

cated on the interface, since the magnetization changes there (∇ ~ ·M ~ 6= 0). The
demagnetizing field acts in the opposite direction to the magnetization, which
explains the name [10].
For ellipsoidal homogeneously magnetized samples, the demagnetizing field can
be calculated exactly. It turns out that the demagnetizing field is linearly related
to the magnetization and therefore also homogeneous [8, 11].

~ d = −N M
H ~, (6)

with the demagnetization tensor N , where the trace of the demagnetization tensor
Tr(N ) = 1 [8].

6
 For a thin, widely extended film, the length and width are much larger than
the thickness. Therefore, the components of the demagnetizing field parallel to
the surface (X and Y directions) can be neglected since the magnetic charges are
very far apart [10, 12]. This corresponds to NXX = 0 and NY Y = 0. Since the
trace of the demagnetization tensor is one, NZZ = 1 must apply.
The demagnetizing energy is now the energy of a ferromagnetic sample in
its own demagnetizing field. For a change of the magnetization by dM ~ in any
magnetic field H,
~ the change of the energy density is:

d = −µ0 H
~ · dM
~ (7)
To calculate the demagnetization energy density, the demagnetizing field from
equation(6) is substituted into the last equation in place of H.
~ The magnetization
vector is then split into magnitude and direction, so that M ~ = Mm
~ applies with
the magnitude of the magnetization M and the unit direction vector m. ~ For a
sample magnetized to saturation,the above equation must be integrated from 0
to the saturation magnetization MS in M and the expression follows
1
dem = µ0 MS2 m(N
~ m)
~ (8)
2

Exercise 2: Follow exactly the derivation given in the previous section and
calculate the demagnetization energy density of a thin film as a function of the
angle θM (see Fig. 1).Which equilibrium orientation of the magnetization is pre-
ferred by the demagnetizing energy in a thin film?

2.2.2 Magnetocristalline Anisotropy

Magnetocrystalline anisotropy means that in a ferromagnetic body certain direc-
tions of magnetization are energetically preferred [12]. The preferred directions
are determined by the symmetry and structure of the crystal lattice [8, 10, 12].
The physical cause of the anisotropy lies in the spin-orbit interaction. The spin-
orbit interaction couples the orbital motion of electrons with their spin [8, 12].
While the permalloy samples studied in this experiment should not exhibit
anisotropies due to the specially tuned composition of this Ni/Fe alloy, the oc-
currence of such magnetocrystalline anisotropies is typical mainly for epitaxially
grown crystalline films. The simplest form of anisotropy is a uniaxial anisotropy
pointing out of the sample plane, described by the following formula
ani = −KU⊥ m2Z . (9)
Where KU⊥ is the associated anisotropy constant (Unit: mJ3 ) and mZ is the Z-
component of the unit magnetization vector (see Fig. 1 for the coordinate sys-
tem). This type of anisotropy often occurs due to interfacial effects or strain in
ultrathin films and can become so strong that the equilibrium position of the
magnetization is perpendicular to the sample plane.

7
2.2.3 Zeeman Energy
If a magnetized sample is in an external magnetic field Hexso the magnetic mo-
ments of the sample interact with the external field and theparallel alignment
of the magnetization and the external field is more energetically favorable. The
corresponding term in the energy density is called Zeeman term and is:
zee = −µ0 MS H
~ ex · m
~ (10)

2.2.4 Total energy density and effective field

The total energy density is calculated as the sum of demagnetization, anisotropy
and Zeeman energy density
1
 = µ0 MS2 m2Z − KU⊥ m2Z − µ0 MS H~ ex · m
~ (11)
2
For further theoretical considerations the effective magnetic field H~ EFF is of
great importance. It is calculated by forming the gradient of the total energy
density with respect to the coordinates of the magnetization M ~ [10, 12, 13].

~ EFF = − 1 ∇
µ0 H ~m
~  (12)
MS

Exercise 3: Calculate the effective field.

2.2.5 Finding the equilibrium position of the magnetization

The equilibrium position of the magnetization is given by a minimum of thetotal
energy density in equation (11) under the constraint that theamount of the
unit magnetization vector is 1. For this purpose, it is convenient to param-
eterize m~ and the external magnetic field H
~ ex in spherical coordinates according
to 1. Then, derive according to the angles θM and ϕM and set both derivatives
to zero.

Exercise 4: Determine the equilibrium position of the magnetization accord-

ing to the above instructions.

At the end, you should get the following result:

Meff cos(θM ) sin(θM ) + H0 sin(θH − θM ) = 0 (13)
ϕH = ϕM , (14)
where H0 is the magnitude of the external magnetic field and the effective mag-
2K ⊥
netization Meff = MS − µ0 MUS is introduced.
It is important to understand that in a ferromagnetic sample, due to anisotropies,
the directions of magnetization and external magnetic field are generally not par-
allel.

8
2.3 Landau-Lifshitz-Gilbert equation
After always assuming a constant magnetization in time in the previous consider-
ations, the time evolution of the magnetization vector shall now be investigated.
If, for example, the magnetization of a thin film is deflected from its equilib-
rium position by an external field, then a precession of the magnetization around
the effective magnetic field can be observed. In this process, the magnetization
returns on a spiral path to its equilibrium position, where it is parallel to the ef-
fective magnetic field. The time scale for this process is in the nanosecond range
[10, 12]. This behavior of magnetization was already described theoretically by
Landau and Lifschitz in 1935 by the Landau-Lifschitz equation [4].
The first part of the Landau-Lifschitz equation describes the precession of the
magnetization around the effective magnetic field[4, 10].

dM~
= −γµ0 M
~ ×H
~ eff (15)
dt

Here γ = 2m eg
e
designates the gyromagnetic relation of the electron with the el-
ementary charge, the electron mass and the g-factoror; g, which is used for free
electrons is approximately 2, 0023. This equation describes in general an elliptic
precession of the magnetization around the effective magnetic field.
However, the above equation cannot yet describe that the magnetization re-
turns to the equilibrium position, a damping mechanism is missing. Already
Landau and Lifschitz inserted a damping term into their equation [4]. However,
since this term gives a physically incorrect behavior for large dampings, Gilbert
introduced an improved damping term in 1955 which gives the correct physical
behavior [5, 10]. By adding this damping term to equation (15), the Landau-
Lifschitz-Gilbert equation (LLG) is obtained. [10, 12]

dm
~ dm
~
= −γµ0 m ~ eff + α m
~ ×H ~ × (16)
dt dt

Here, the unit magnetization vector m~ =M ~ /MS was used and the Gilbert damp-
ing parameter α was introduced, which indicates the strength of the damping.The
second term of the sum is called the Gilbert damping term and leads to the Re-
laxation of the magnetization to the equilibrium position[10]. Figure 2 shows a
numerical solution of the LLG. Under the influence of the LLG, the magnetization
basically moves on a spherical surface. [12].

Exercise 5: Consider the direction in which the individual contributions to

the LLG point. Justify mathematically and graphically why the magnetization
must always move on a spherical surface.

9
 Mz

Mx

My

Figure 2: Numerical solution of the Landau-Lifschitz-Gilbert equation for γ =

200 GHz
T
, α = 0, 05 und µ0 MS = 0, 852 T. Note tha a very high attenu-
ation was chosen for better visualization has been chosen. The effective
field points in the z-direction with a constant strength of 1 T. You can
see the spiral precession of the magnetization towards the equilibrium
position in z-direction (red line). In addition, a spherical surface is
indicated by the black dotted lines.

10
2.4 Ferromagnetic resonance - resonance condition

2.4.1 Principles of FMR

As already indicated in the introduction, measurements of ferromagnetic reso-

nance can be used to determine certain properties of magnetic materials, such
as the g-factor or the Gilbert damping parameter α or the magnetic anisotropies
[12, 13].
In an FMR experiment, the precession of the magnetization is excited with
a high-frequency magnetic field in the microwave range. The sample is in an
external time-independent magnetic field H ~ 0 .The precession of the magnetization
around the effective field absorbs energy from the microwave field. Ferromagnetic
resonance occurs at a fixed microwave angular frequency ω in a certain external
magnetic field HFMR and is detected by a maximum of microwave absorption.
The evaluation of FMR measurements focuses on the resonance position and the
line width of the measured resonance curve.

2.4.2 Definition of a suitable coordinate system

In order to simplify the calculation of the resonance condition in the following

subchapters, a suitable coordinate system is to be introduced. As in the previous
chapters, XYZ is the coordinate system whose z-axis is normal to the sample
plane. Similarly, the angle θM of the magnetization with respect to the Z-axis
and the angle ϕM between the projection of the magnetization in the X-Y plane
and the X-axis are again used.For the calculation of the resonance condition of
FMR, the coordinate system XYZ is not convenient and therefore the coordinate
systemxyz is used in the following with its x-axis pointing in the direction of M
~ .
In Fig. 3 , the two coordinate systems and the magnetization vector are shown.
The coordinate transformation from the system XYZ into the system xyz, with
which a vector in the system XYZ can be transferred into the system xyz, is given
by the following equation

   
Mx MX
My  = TXY Z  MY  = (17)
  xyz  

Mz MZ
sin θM cos ϕM sin θM sin ϕM cos θM
  
MX
=  − sin ϕM

cos ϕM 0   MY 


− cos θM cos ϕM − cos θM sin ϕM sin θM MZ

The following calculations also require the transformation from system xyz to

11
 Z

z
y
θM M
x

z'
Y
φM y'

x'
X
Figure 3: In dieser Abbildung sind die beiden für die weiteren Berechnungen
benötigten Koordinatensysteme, ein Hilfskoordinatensystem und der
Magnetisierungsvektor dargestellt.

system XYZ. This is given by

   
MX Mx
 MY  = Txyz My  = (18)
  XY Z  

MZ Mz
sin θM cos ϕM − sin ϕM − cos θM cos ϕM
  
Mx
=  sin θM sin ϕM cos ϕM − cos θM sin ϕM  My  .
  

cos θM 0 sin θM Mz

Exercise 6: Based on your knowledge of linear algebra, work out how these
two transformation matrices come about. You do not have to calculate the trans-
fer matrices exactly.

2.4.3 Calculation of the effective field

The effective field, which you have already calculated in a preparatory task in the
coordinate system XYZ, must now be transferred to the coordinate system xyz
relevant for the FMR. This is done by multiplication with the transfer matrix in

12
equatio 17, so that µ0 H
~ eff = TXY
xyz
Z (µ0 HEFF ). This gives:
~

µ0 Heff,x =µ0 sin(θM ) cos(ϕM )Hex,X + µ0 sin(θM ) sin(ϕM )Hex,Y +

µ0 cos(θM )Hex,Z − µ0 cos(θM )Meff mZ (19)
µ0 Heff,y = − µ0 sin(ϕM )Hex,X + µ0 cos(ϕM )Hex,Y (20)
µ0 Heff,z = − µ0 cos(θM ) cos(ϕM )Hex,X − µ0 cos(θM ) sin(ϕM )Hex,Y +
µ0 sin(θM )Hex,Z − µ0 sin(θM )Meff mZ (21)

Now the magnetization components Z must be replaced by the magnetization

components in the coordinate system xyz. This is done by using the back trans-
formation Txyz
XY Z
from equation(18), so that

mZ = cos(θM )mx + sin(θM )mz (22)

is valid. Furthermore, the static external field H

~ 0 is to be expressed by spherical
coordinates and a small microwave excitation field h has to be added with the
angular frequency ω.

H0 sin(θH ) cos(ϕH )
 

~ ex
H =  H0 sin(θH ) sin(ϕH )  + ~h exp(iωt) (23)
 

H0 cos(θH )

Using the addition theorems for the trigonometric functions, the effective field
can be calculated from equations (22) and (23) simplifying the following expres-
sion

µ0 Heff,x =µ0 H0 (sin(θM ) sin(θH ) cos(ϕH − ϕM ) + cos(θM ) cos(θH )) −

µ0 Meff cos(θM ) (cos(θM )mx + sin(θM )mz ) + µ0 hx exp(iωt) (24)
µ0 Heff,y =µ0 H0 sin(θM ) sin(ϕH − ϕM ) + µ0 hy exp(iωt) (25)
µ0 Heff,z =µ0 H0 (− cos(θM ) sin(θH ) cos(ϕH − ϕM ) + sin(θM ) cos(θH )) −
µ0 Meff sin(θM ) (cos(θM )mx + sin(θM )mz ) + µ0 hz exp(iωt) (26)

In FMR experiments, the microwave excitation is usually so small that the

magnetization is never strongly deflected from its equilibrium position (i.e. mx
≈1 and my,mz  1), so using the equilibrium position of the magnetization found
in section2.2.5, the effective field can be further simplified [10].

µ0 Heff,x =µ0 H0 cos(θH − θM ) − µ0 Meff cos(θM ) cos(θM )mx +

sin(θM )mz + µ0 hx exp(iωt) (27)
µ0 Heff,y =µ0 hy exp(iωt) (28)
µ0 Heff,z = − µ0 Meff sin2 (θM )mz + µ0 hz exp(iωt) (29)

13
2.4.4 Solution of the linearized LLG
The driven motion of the magnetization vector under a high frequency magnetic
field is described by the Landau-Lifshitz-Gilbert equation. Since this is a nonlin-
ear differential equation, an analytical solution is generally not possible. However,
for small excitation fields ~h a linearization according to the following approach
for the unit magnetization vector is possible.

1
 

m my exp(iωt)
~ = (30)


mz exp(iωt)

Exercise 7: Substitute the approximation from equation (??) into the LLG
(16) and simplify as much as possible (without using the effective field explicitly).
Linearize the obtained equations for all components as much as possible (i.e.,
neglect quadratic orders in my and mz ).

The effective field from equations (27) to (29) is now substituted into the result
you calculated and again all quadratic orders in my and mz are neglected. Fur-
thermore, due to the small microwave excitation field, mixing terms between the
components of ~h and my and mz can also be neglected. With these simplifications,
the x-component of the LLG becomes trivial and we are left with the following
two equations [10, 12, 13]:

!
ω ω
0 =i my + Beff + αi mz − µ0 hz (31)
γ γ
!
ω ω
0 = − i mz + µ0 Heff + αi my − µ0 hy (32)
γ γ

Here, the effective magnetic induction Beff and the effective magnetic field Heff
were introduced, so that [10, 12, 13]

Beff = µ0 H0 cos(θH − θM ) − µ0 Meff cos(2θM ) (33)

µ0 Heff = µ0 H0 cos(θH − θM ) − µ0 Meff cos2 (θM ) . (34)

Exercise 8: Solve the system of linear equations for my and mz .

2.4.5 Resonance condition

The resonance condition of the FMR can be read from the results for my and mz .
Neglecting the usually very small damping α, the maximum of the precession
amplitude of the magnetization is given by the zero of the denominator of my
or mz . This generally leads to the following resonance condition, which specifies

14
the microwave angular frequency for which resonance occurs at a given magnetic
field HFMR [10, 12, 13]:
!2
ω

= (35)

µ0 Beff Heff 
γ H0 =HFMR

The two most important special cases are discussed below:

2.4.5.1 In-plane configuration In the in-plane configuration, both the external

field and the magnetization lie in the sample plane (θH = 90◦ and θM = 90◦ ) and
are parallel (ϕH = ϕM ). There must be no in-plane anisotropies in the sample.
By substituting these angles into equation (33) and (34), the following resonance
condition is found, also known as the Kittel formula due to its discovery by C.
Kittel in 1948 [3]:

ω q
= µ0 HFMR (µ0 HFMR + µ0 Meff ) (36)
γ

2.4.5.2 Out-of-plane configuration Here, the applied magnetic field is per-

pendicular to the sample plane θH = 0◦ . It is assumed that magnetization and
external field are parallel in this process. For in-plane magnetized samples this is
ensured by applying a very high external field in which the sample is saturated,
for out-of-plane magnetized samples the assumption is always fulfilled. By setting
the angles in equation (33) and (34) to zero the following resonance condition is
obtained:
ω
= µ0 HFMR − µ0 Meff (37)
γ

Exercise 9: Verify the resonance conditions presented here using your results
for my and mz .

2.4.6 Dynamic susceptibilities and line shape of the FMR

While in the last subchapter the position of the ferromagnetic resonance was
derived, now the shape of the absorption curves measured in the experiment
shall be derived. First of all, the dynamic susceptibilities are defined by

~ = χ~h
m (38)

with the susceptibility tensor χ.

Using the results for my and mz from the previous subsection, the components

15
of the susceptibility tensor can be determined.
 
−µ0 Beff + iαω
γ
χyy =  2    (39)
ω
γ
− Beff + iαω
γ
µ0 Heff + iαω
γ
−µ0 iω
γ
χyz = −χzy =  2    (40)
ω
γ
− Beff + iαω
γ
µeff Heff + iαω
γ
 
−µ0 Heff + iαω
γ
χzz =  2    (41)
ω
γ
− Beff + iαω
γ
µ0 Heff + iαω
γ

The microwave absorption in the sample and the measured line shape is given
by the imaginary part of the susceptibilities. In the following, the line shape of
χyy is derived as an example. The calculations can be carried out for the other
susceptibilities in an analogous way.  2
Equation (39) for χyy is assumed. For ωγ the resonance condition from
equation (35) is used. Then the numerator and denominator are developed around
the resonant field HFMR in H0 to linear order and terms proportional to α(H0 −
HFMR ) or α2 are neglected. This yields:
  
µ0 BFMR + ∂Beff 
∂H0 H0 =HFMR
 δH + iµ0 ∆H
χyy =    
BFMR ∂H eff 
∂H0 H0 =HFMR
 + HFMR ∂Beff 
∂H0 H0 =HFMR
 µ0 δH + (BFMR + µ0 HFMR )iµ0 ∆H)
(42)
Here, the effective
 field and the effective magnetization in the resonance case,
HFMR = Heff  and BFMR = Beff  .δH is the difference between
H0 =HFMR H0 =HFMR
external field and resonant field H0 − HFMR . In addition, the linewidth (half
width at half maximum) was defined as µ0 δH = αω γ
.
By calculating the derivatives using equations (34) and (33) and multiplying
by the complex conjugate of the denominator, the line shape of the susceptibility
χyy is obtained.
∆H(H0 − HFMR ) − i∆H 2 µ0 ∆H
!
BFMR
χyy = + (43)
(BFMR + µ0 HFMR )∆H (H0 − HFMR )2 + ∆H 2 BFMR
Thus, it can be seen that the real part of the susceptibility χyy has the shape
of an antisymmetric Lorentz curve, while the imaginary part has the shape of
a symmetric Lorentz curve. The line shape of all susceptibilities is exemplarily
shown in figure 4.
It can be shown that the microwave power absorbed by the sample is deter-
mined by the imaginary parts of χyy and χzz . Therefore, FMR resonance curves
have the form of symmetric Lorentz curves. For the mean absorbed power, [12, 14]
holds:
1  
P̄ = µ0 MS ω =(χyy )h2y + =(χzz )h2z (44)
2

16
Figure 4: Line shape of real and imaginary part of susceptibilities. For the map-
ping, ω = 2π 10 GHz, γ = 200 GHz T
, µ0 MS = 1 T, and α = 0.007 are
used. The real parts of χyy and χzz have the form of an antisymmetric
Lorentz curve and the imaginary parts are symmetric Lorentz curves.
For χyz the opposite is true
.

17
2.5 Ferromagnetic resonance - damping
The dependence of linewidth on microwave frequency ω in ferromagnetic reso-
nance measurements is described by the following empirical equation [10, 12]:
αω
µ0 ∆H = µ0 ∆H(0) + (45)
γ
The αωγ
part of the damping is thereby generated by intrinsic mechanisms and
can already be accounted for in the LLG by the Gilbert damping parameter α. It
is mainly caused by Eddy currents (Foucault’s currents), direct magnon-phonon
scattering, or relaxation via itinerant electrons [10, 12]. The offset ∆H(0) comes
from extrinsic mechanisms and is caused by defects or inhomogeneities in the
sample [10].

18
3 Experimental basics and evaluation
3.1 General structure of an FMR-spectrometer

Koaxialkabel Schottky-
Gauss- Diode
Mikrowellen-
Hall-
Meter
generator
sonde in out
Polschuh Lock-In-
Koplanarer Verstärker
Wellenleiter Spule Spule Ref out

Modulations- Probe
spulen Leistungs-
verstärker

Magnet- PC
Steuerung und Messwerterfassung
netzteil

Figure 5: Construction of the FMR spectrometer used in experiment. A fre-

quency generator produces microwaves, which are guided via coaxial
cables through the waveguide and finally to the Schottky diode. On
the waveguide, the sample is in the field of an electromagnet and a
small modulation coil. The intensity of the microwaves is detected by
a Schottky diode and measured by a lock-in amplifier. Fig. after [6]

In this experiment, a simple FMR spectrometer is available to understand the

basic principles of ferromagnetic resonance. The following description of an FMR
spectrometer follows [6] in essential parts.
In FMR measurements, the magnetic moments in the sample are excited to pre-
cess around the effective magnetic field by an irradiated high-frequency magnetic
field ~h in the microwave range [10, 12]. The sample is thereby in the constant
magnetic field H ~ 0 of an electromagnet. At a certain frequency and magnetic field
strength H ~ FMR resonance occurs and the absorption of microwaves becomes max-
imum [10, 12, 14]. To measure the resonance of the sample, the experiment varies
the magnetic field at constant microwave frequency and simultaneously measures
the absorption of microwaves in the sample using a Schottky diode.
The sample is located between the pole pieces of an electromagnet on a waveg-
uide so that the sample is approximately centered and level on the waveguide.
The block diagram shown in Figure ?? can be used to illustrate the setup. The
power supply of the magnet generates a coil current of up to 20 A, which causes
a magnetic field of approximately 490 mT between the pole pieces of the magnet.

19
To measure the field strength, a Hall probe is placed as centrally as possible be-
tween the pole pieces. The Hall probe was calibrated in advance of the experiment
using a DPHH (2,2-diphenyl-1-picrylhydrazyl) sample. In this complex organic
molecule, exactly one unpaired electron occurs, so its g-factor is approximately
that of a free electron. Using the known g-factor, the resonant field strength can
be calculated at a fixed microwave frequency, and the experimentally measured
field strength can be matched with the calculated one. A Gaussmeter measures
the values for the field strength and transmits them to the measurement program
on the PC. During the measurement, the magnetic field is increased linearly with
time by the measurement program.

Exercise 10: Find out how magnetic field measurement with a Hall sensor
works. Why is calibration necessary to determine the exact magnetic field at the
sample position?

A microwave field is required to excite the precession of the magnetization. In

the FMR setup used, the microwaves are generated by a microwave generator
that allows both the power and the frequency to be varied over a wide range.
Frequencies from 2 to 26 GHz and microwave powers up to 13 dBm can be
generated. The microwaves are transported via coaxial cables from the microwave
generator to the coplanar waveguide. The sample is placed on the waveguide and
is therefore in its magnetic field. The sample absorbs part of the irradiated power.
The intensity of the remaining microwave radiation is measured using a Schottky
diode.
Since the transmitted power changes only slightly due to microwave absorption
in the sample, the change in diode voltage is measured with a lock-in amplifier.
For measurements with a lock-in amplifier, the measurement signal must be mod-
ulated with a reference signal. For this purpose, a sinusoidal reference signal with
the frequency 620 Hz is generated with a frequency generator, which is already
integrated in the lock-in amplifier in the setup considered here. Care must be
taken to select a frequency that is not a multiple of the mains frequency of 50 Hz.
The reference signal is amplified by a power amplifier and applied to the modu-
lation coils. This results in a small modulation of the external magnetic field H~ 0.
The lock-in amplifier now filters out from the measurement signal the component
with the same frequency and a defined phase relationship with respect to the
reference signal, which significantly improves the signal-to-noise ratio.

Exercise 11: Find out about the operation and internal structure of a lock-in
amplifier. Make a block diagram of a lock-in amplifier. Also find out about the
different types of noise and their frequency spectrum. For example, you can use
the following application note from Stanford Research Systems for this purpose
http://www.thinksrs.com/downloads/PDFs/ApplicationNotes/AboutLIAs.pdf

20
3.2 Microwave technology

ferromag. Probe
Schicht

Irf/2 Irf Irf/2

Figure 6: Principle of a coplanar waveguide. The coplanar waveguide consists of

three parallel conductive paths through which a high-frequency current
flows. This results in the microwave excitation field ~h around the traces,
which is outlined by the red field lines. The sample is placed on the
waveguide with the ferromagnetic layer facing downwards, so that its
magnetic moments can be excited to precession by ~h.

Here, the microwave components used in the experiment will be briefly ex-
plained. In the FMR experiments performed here, the sample is placed on a
coplanar waveguide. A coplanar waveguide is a device consisting of three parallel
metallic conductors [10]. In figure 6 such a waveguide is shown schematically.
The middle of the traces is the signal line on which the microwave power is
transported [10]. The two outer lines are ground lines [10]. The magnetic field
direction in the vicinity of a coplanar waveguide can be easily determined by the
right-hand rule. This results in the field directions shown in figure 6. Since the
sample covers the complete waveguide, the microwave magnetic field will be under
different directions with respect to the sample. Here, the excitation field points
perpendicular between the waveguides and parallel to the sample plane on the
waveguides. Thus, for a microwave field parallel to the equilibrium position of the
magnetization, no precession of the magnetization is excited in the x-direction in
the coordinate system xyz (the dynamic magnetization components my and mz
are independent of hx ), the direction of the signal line of the waveguide should
be parallel to the external field so that its alternating field points perpendicular
to the field direction and equilibrium magnetization.
The intensity of the microwaves is measured with a Schottky diode. This is
a special form of semiconductor diode that works with a metal-semiconductor
contact.

Exercise 12: Inform yourself about the phenomenological functioning of a

diode. What is the voltage drop across a diode when it is connected in series with
a resistor to an AC voltage source? Find out about the special characteristics of a

21
Schottky diode compared to other semiconductor diodes. The physical operation
of a diode (band diagrams) don‘t need be understood .

3.3 Evaluation of the resonance spectra

To record an FMR spectrum, at constant microwave frequency ω the external DC
magnetic field H~ 0 is slowly increased and in parallel the absorption of microwaves
is measured via the Schottky diode voltage. The absorption of microwaves during
the ferromagnetic resonance of the sample is described by the imaginary part of
the susceptibility χyy , which has the form of a symmetric Lorentz curve [10, 12,
14]. However, the total absorption is proportional to an asymmetric Lorentz
curve because, in addition to the symmetric line shape of the absorption, there
is a dispersive component which, like the real part of χyy , has an antisymmetric
line shape.
The measurement of the FMR spectra is performed with the help of a lock-in
amplifier. This does not measure the absorption itself, but the change of the
absorption with the external DC magnetic field H0 . Therefore, the measured
curves correspond to the derivative of the absorption with respect to the field H0 .
Introducing the FMR amplitude A, the fit function f (H0 ) for the FMR spectra
can now be defined.
d
!
f (H0 ) ∝ =(χyy ) cos  + <(χyy ) sin 
dH0
 
2∆H 3 (H0 − HFMR ) cos  ∆H 2 (∆H 2 − (H0 − HFMR )2 ) sin  
=A−  2 +  2
(H0 − HFMR )2 + ∆H 2 (H0 − HFMR )2 + ∆H 2
(46)
In addition to the last equation, an offset and a slope must be considered
when fitting the curves [14]. In the following figure 7 a typical FMR spectrum
fitted with equation (46) is shown. The two most important quantities that can
be determined from the fit curve are the resonant field strength HFMR and the
linewidth δH [10, 12]. To perform the experiment, a fit program is provided to
determine the resonant field HFMR and the linewidth ∆H.
For nearly antisymmetric FMR spectra, the resonance field strength and linewidth
can also be read directly from the measured spectra (see Fig. 7). The field at
which the spectrum intercepts th field axis indicated the value of the resonance
field HFMR . The line width
√ is proportional to the distance ∆H pp between the two
extrems, so that ∆H = 3/2 ∆H [10, 12, 14].
pp

4 Available samples
• puttered permalloy film (an alloy with composition Ni81 Fe19 ) with a film
thickness of 50 nm protected against oxidation with a gold layer.

22
 differenzierte Absorption (a.u.)
0,5

0,0

-0,5

0,2 0,25 0,3 0,35 0,4

Magnetfeld µ0H0 (T)

Figure 7: The figure shows a typical FMR spectrum. The points drawn in red
represent the recorded measured values, the black line shows the fit
according to equation (46). Using the fit, the resonant field strength
µ0 HFMR = 303.2 mT and the linewidth µ0 δH = 13.8 mT can be deter-
mined. The figure shows a typical FMR spectrum. The points drawn
in red represent the recorded measured values, the black line shows the
fit according to equation (46). Using the fit, the resonant field strength
µ0 HFMR = 303.2 mT and the linewidth µ0 δH = 13.8 mT can be deter-
mined. Fig. from[6].

• An yttrium iron garnet (YIG) layer of about 20 nm thickness grown by

sputter deposition on gadolinium gallium garnet.

5 Experimental procedure
5.1 Influence of the measurement parameters
• Insert the specimen holder for the in-plane configuration, on which the
permalloy specimen is already glued, into the setup.
• Set a frequency of 10 GHz and a power of 0 dBm on the microwave genera-
tor. Record a spectrum with a modulation current of 5 mA, a time constant
of 0.1 s, and a sweep speed of 1 mT
s
. Use this correctly recorded spectrum
as a reference below.
• Vary the modulation current strength up and down starting from the value
given in the previous point and observe the effect on the spectrum. De-
termine the line width for each spectrum. Record at least five curves with
current strengths ranging from 0.5 mA to 200 mA. Change the sensitivity
of the lock-in amplifier in case the signal is too small or cut off.

23
 • Set the values mentioned at the beginning again. Then vary the time con-
stant at the lock-in amplifier up and down starting from the value mentioned
at the beginning and observe the influence on the spectrum. Determine the
line width for each spectrum. Record at least four curves with time con-
stants ranging from 1 ms to 10 s.

• Set the values given at the beginning. In the following, vary the sweep
speed up and down starting from the value mentioned in the previous point
and observe the effect on the spectrum. Determine the line width for each
spectrum. Record at least five curves with velocities from 0.1 mT
s
to 50 mT
s
.

5.2 Frequency dependence in the in-plane configuration on

permalloy
• Again, set reasonable parameters for modulation field, sweep velocity and
time constant. Now change the microwave frequency in the range from 2
to 20 GHz and follow the resonance line of Permalloy. You can do this by
confining the magnetic field to the range where the resonance is expected
to occur to save time.

• Evaluate the measured resonance lines in terms of resonance field strength

and line width using the fit program.

• Interpret the resonance fields using the Kittel formula in equation (36). Use
it to determine µ0 Meff . Critically analyze your fit results and try to find
reasons for deviations from the expected values. Note that the g-factor of
electrons in permalloy is about 2.14 (corresponding to γ = 188 GHz T
) [12].

• Interpret the line widths using equation (45). Determine the intercept of
the line width and the Gilbert "damping" parameter α.

5.3 Examination of an ytrium-iron-garnet sample

5.3.1 In-plane configuration
• Now place the YIG specimen on the in-plane specimen holder instead of
the permalloy specimen.

• At a microwave frequency of 2 GHz, try to record an initial spectrum. Use

a magnetic field range of 15 to 35 mT. Use a low sweep speed of 0.3 mT s
due to the expected low linewidths, and reduce the modulation current to
3 mA and the time constant to 30 ms.

• Based on this spectrum, consider reasonable measurement parameters for

the following spectra. Consult with the supervisor if you are unsure of your
choice.

24
 • In the following, vary the microwave frequency from 2 to 11 GHz and deter-
mine resonance field and linewidth for each spectrum. Record ten spectra.
For each spectrum, determine the resonance position by reading and ap-
proximately averaging over all peaks present.

• Interpret the results by finding the resonance field using equation (36) and
the linewidth using equation (45). Determine the effective magnetization
µ0 Meff of the YIG sample. Use the literature value g ≈ 2.01 for YIG
(γ ≈ 177 GHz
T
) [15].

5.3.2 Perpendicular configuration

• Here, the same specimen is to be measured again in the vertical configura-
tion. To do this, insert the specimen holder for the vertical configuration,
equipped with an identical specimen, into the setup.

• First, try to record a spectrum at a frequency of 2 GHz. Use the same

parameters as for the in-plane configuration.

• Vary the microwave frequency from 2 GHz to 8 GHz and determine the
resonant field approximately as described in the previous section. Record
ten spectra.

• Interpret the results for the resonant field using equation (37) and determine
the effective magnetization µ0 Meff and the gyromagnetic ratio γ. Compare
the result with the measurements in the in-plane configuration.

6 Protocol requirements
The protocol should include the following aspects:

• Magnetic energies in the ferromagnet

• Theory of ferromagnetic resonance. Start from the LLG and cover the
essential aspects of the theory. Skip complex derivations (e.g., coordinate
transformations in the effective field, derivation of the line shape). Derive
the resonance conditions.

• Describe the FMR experiment and the equipment used in the experiment.

• Description of the performed measurements

– Influence of the measurement parameters. Discuss the influence of the
different measurement parameters on the line width of the resonance.
Try to establish criteria to find the optimal measurement parameters.

25
 – Frequency dependence on a permalloy sample. Evaluate the recorded
spectra in terms of linewidth and resonance field strength and deter-
mine the material parameters of permalloy.
– Measurements on YIG sample. Evaluate the spectra in terms of reso-
nance field strength by manual averaging over all peaks and determine
the effective magnetization. Compare the measurements in the in-
plane and perpendicular configurations.

26
7 Further literature
• M. Härtinger, Untersuchung magnetischer Materialien mit Methoden der
Ferromagnetischen Resonanz, Dissertation, Universität Regensburg, 2016,
verfügbar online unter https://epub.uni-regensburg.de/35434/1/Thesis.pdf

• G. Woltersdorf, Spin-Pumping and Two-Magnon Scattering in Magnetic

Multilayers, Dissertation, Simon Fraser University, 2004, verfügbar online
unter https://epub.uni-regensburg.de/14960/1/Woltersdorf-PhD04.pdf

• S. Blundell, Magnetism in Condensed Matter

• J. M. D. Coey, Magnetism and magnetic materials

• C. P. Slichter, Principles of Magnetic Resonance

References
[1] J.H.E. Griffiths. Anomalous High-frequency Resistance of Ferromagnetic
Metals. Nature, 158:670–671, November 1946.

[2] E. Zavoisky. Spin Magnetic Resonance in The Decimetre-Wave Region. J.

Phys. USSR, 10:197, November 1946.

[3] C. Kittel. On the Theory of Ferromagnetic Resonance Absorption. Phys.

Rev., 73:155–161, Jan 1948.

[4] L.D. Landau und E. Lifshitz. On the theory of the dispersion of magnetic
permeability in ferromagnetic bodies. Phys. Z. Sowjetunion, 8:153–169, 1935.

[5] T.L. Gilbert. A Lagrangian formulation of the gyromagnetic equation of the

magnetization field. Phys. Rev., 100:1243–1244, 1955.

[6] T.N.G. Meier. Temperaturabhängige Messungen des inversen Spin-Hall-

Effekts. Bachelorarbeit, Universität Regensburg, Fakultät Physik, Oktober
2012.

[7] T.N.G. Meier. Thermodynamical properties of fluctuating magnetic stripe

domains in ultrathin Fe/Ni/Cu(001) and Ni/Fe/Cu(001) films. Masterarbeit,
Universität Regensburg, Fakultät Physik, Dezember 2014.

[8] J.M.D. Coey. Magnetism and Magnetic Materials. Cambridge University

Press, 2009.

[9] S. Blundell. Magnetism in Condensed Matter. Oxford University Press, 2001.

[10] M.O. Härtinger. Magnetische Eigenschaften neuer ferromagnetischer Maeri-

alien. Masterarbeit, Universität Regensburg, Fakultät Physik, August 2011.

27
[11] J. A. Osborn. Demagnetizing Factors of the General Ellipsoid. Phys. Rev.,
67:351–357, Jun 1945.

[12] G. Woltersdorf. Spin-Pumping and Two-Magnon-Scattering in Magnetic

Multilayers. Dissertation, Simon Fraser University, Department of Physics,
August 2004.

[13] M.O. Härtinger. Untersuchung magnetischer Materialien mit Methoden

der Ferromagnetischen Resonanz. Dissertation, Universität Regensburg,
Fakultät für Physik, März 2016.

[14] Z. Celinski, K.B. Urquhart, und B. Heinrich. Using ferromagnetic resonance

to measure the magnetic moments of ultrathin films. Journal of Magnetism
and Magnetic Materials, 166(1):6 – 26, 1997.

[15] Seongjae Lee, Scott Grudichak, Joseph Sklenar, C. C. Tsai, Moongyu Jang,
Qinghui Yang, Huaiwu Zhang, und John B. Ketterson. Ferromagnetic reso-
nance of a YIG film in the low frequency regime. Journal of Applied Physics,
120(3):033905, 2016.

28