Quantum fluctuations in

antiferromagnetic spin configurations

Bachelor Thesis

Solange Schrijnder van Velzen

Physics and Astronomy
January 2016
Supervised by R.A.Duine
Institute for Theoretical Physics
Utrecht University

Abstract
We consider fluctuations in magnetic systems around the classically-
ordered ground state. We use a Holstein-Primakoff transformation to
calculate the zero-point energy fluctuations in antiferromagnetic lat-
tice configurations using the Heisenberg exchange Hamiltonian. We
show that these quantum fluctuations lower the classically expected
ground state. Furthermore we provide a numerical method to calcu-
late dispersion relations for magnetic systems and specifically show
that our analytical results can be produced. Also, we investigate
the spin-flop transition in a one dimensional antiferromagnetic sys-
tem with anisotropy. Our calculations show that the transition occurs
at B/K ≈ −14, where B is the external field and K the anisotropy
constant. Furthermore, we show that quantum fluctuations decrease
the energy of both the antiferromagnetic and the spin-flop state by
calculating the dispersion relation. Our results imply a shift of the
transition point between the two states. Our results are a first step
towards incorporating the effects of quantum fluctuations in inhomo-
geneous antiferromagnets.
Contents

Contents 1

1 Introduction 2

2 Ferromagnets 4
2.1 Semi-classical approach . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Holstein-Primakoff Transformation . . . . . . . . . . . . . . . 6

3 Antiferromagnets 11
3.1 Semi-classical approach . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Holstein-Primakoff Transformation . . . . . . . . . . . . . . . 13

4 Numerical methods 18
4.1 (Anti-)Ferromagnetic Heisenberg Model . . . . . . . . . . . . . 19

5 Classical ground state of an antiferromagnetic Hamiltonian

with Dzyaloshinskii-Moriya interactions 21

6 Phase transition to the spin-flop state in antiferromagnets 23

6.1 Magnon energies for the spin-flop state . . . . . . . . . . . . . 25
6.2 Magnon energies for the antiferromagnetic state . . . . . . . . 27
6.3 Quantum fluctuations . . . . . . . . . . . . . . . . . . . . . . . 28

7 Conclusion 30

8 Appendix 31

References 34

1
1 Introduction
The magnetic behaviour of a material depends on its structure, particularly
its electronic configuration. Ferromagnetic materials for example exhibit a
long-range ordering which causes the unpaired electron spins to line up paral-
lel with each other which induces a net magnetization in the material. Above
a critical temperature known as the Curie temperature, ferromagnetic order-
ing disappears as a result of thermal agitation. Antiferromagnetic materials
on the other hand have no net magnetization, since their spins align in such a
way that each spin on the lattice is orientated anti-parallel to the neighbour-
ing spin. Above the critical temperature, known as the Néel temperature,
antiferromagnetic order disappears.

Excitations of the ground state of magnetic materials can be described

by spin waves. Spin waves are propagating disturbances in the magnetic
ordering of materials (see Figure 1). In the quantum mechanical picture
the excitations can be described by a quasiparticle, a magnon, which carries
a fixed amount of energy and lattice momentum. The quantum theory of
spin waves enables to study physical properties of magnetic configurations
at nonzero temperatures. This theory was originally introduced by F. Bloch
[1] and the main assumption is that the system has an ordered ground state
and the excited states are described as spin waves. It is shown that the ferro-
magnetic state is an eigenstate of the Heisenberg exchange Hamiltonian for
positive exchange constant J, which is commonly used to describe ferromag-
netic systems. On the contrary the antiferromagnetic Néel state is not an
eigenstate of the Heisenberg exchange Hamiltonian for a negative exchange
constant. This causes quantum fluctuations on top of the classical descrip-
tion of the ground state. Quantum fluctuations describe the deviation from
the classical ground state due to quantum mechanical effects.

In this thesis we discuss the effect of quantum fluctuations in antiferro-

magnetic configurations. We incorporate Heisenberg exchange interactions,
anisotropy, an external field and in some cases also the Dzyaloshinskii-Moriya
coupling. The effect of quantum fluctuations in magnetic models has at-
tracted much interest in recent years [2] [3] [4]. Experimental and theoretical
studies of those systems have revealed a plethora of quantum fluctuation phe-
nomena. In previous research quantum fluctuations in magnetic skyrmions
were investigated using numerical methods [5] [6]. Skyrmions are vortex-like
configurations that occur in magnetic materials. A clear analytical descrip-
tion of skyrmions in antiferromagnetic systems however remains an open
problem. In chapter 3 we start with a description of ferromagnets, both in

2
a semi-classical way and a quantum mechanical way. We will see that there
are no quantum fluctuations that effect the ground state of a homogeneous
ferromagnet. We will also calculate the magnetization of a ferromagnetic
system as a function of the temperature. In the next chapter, chapter 4,
we will discuss the antiferromagnet. As for the ferromagnet, we will give
a description in a semi-classical way and a quantum mechanical way and
calculate the magnetization. We will show that quantum fluctuations lower
the classical ground state for antiferromagnets. In chapter 5 we will discuss
numerical methods to diagonalize the Hamiltonian and find the dispersion
relation for magnetic configurations in general. We will provide a result for
the antiferromagnet and the ferromagnet and show that this confirms our an-
alytic calculations. In chapter 6 we investigate the phase transition from the
antiferromagnetic to the spin-flop state. The pure phase transition occurs
in an antiferromagnetic system without Dzyaloshinskii-Moriya interactions
by increasing the strength of the external field. We explicitly calculate the
dispersion relation for both states and calculate the quantum correction to
the ground states.

Figure 1: In the ground state of a ferromagnet, all spins point in the same
direction. The lowest energy excitation is described by a single spin-wave,
in which the spins precess in a conical motion. A long wavelength spin
wave carries very little energy, because the difference in direction between
neighbouring dipoles (spins) is small. Copyright: Addison-Wesley 2000.

3
2 Ferromagnets
In this chapter we discuss the basics of the ferromagnet. We first perform a
classical calculation using the Heisenberg-exchange Hamiltonian. Secondly,
we perform a Holstein- Primakoff transformation on the same Hamiltonian
to figure out whether quantum fluctuations affect the classical ground state
of the ferromagnet.
The Heisenberg exchange Hamiltonian describes a system of interacting
spins in a magnetic material and can be used to examine the behaviour of
spin waves (magnons). The Heisenberg exchange Hamiltonian is seen in the
equation below, where the sum goes over all neighbouring pairs i and j on
the lattice sites, Si denotes the spin operator at site i and J is the exchange
coupling constant. This coupling constant will be positive in the case of
ferromagnets J = |J|, such that the energy is minimized when all spins are
pointing in the same direction.
X
H = −J Si · Sj . (1)
hi,ji

2.1 Semi-classical approach

We will start with a semi-classical approach to the ferromagnet. Semi-
classical physics refers to a theory in which one part is described classically
whereas the other is treated quantum-mechanically. In this specific case we
use a quantum mechanical description of the spins and the classical equations
of motion described by Ehrenfest’s theorem, as stated in the equation below.

d i
hSk i = − h[Sk , H]i. (2)
dt }
Assume a one dimensional chain of spins pointing along the z-axis, sep-
arated by a distance a. Using the Heisenberg exchange Hamiltonian from
Equation 1, where we assume only nearest neighbour interactions, Ehren-
fest’s theorem leads to
d i X X y y
hSk i = J h[Sk , Six Si+1
x
+ Six Si−1
x
] + [Sk , Si Si+1 + Siy Si−1
y
]
dt } i i
X
+ [Sk , Siz Si+1
z
+ Siz Si−1
z
]i. (3)
i

There is still a sum over i left to sum all pairs. In order to evaluate the
equations of motion for Sxk , Syk and Szk we use the commutation relations

4
given by
[Six , Sjy ] = i}Skz δij . (4)
We now give a detailed evaluation of Equation 3 for the Sxi component of Si .
With the use of Equation 4 we can simplify the expression to Equation 5a,
where the sum over i vanished since the spins always commute when k 6=
i. Furthermore, we made use of the commutation relations to simplify the
expression to Equation 5b.
d x i
hSi i = J h[Six , Siy Si−1
y
] + [Six , Siy Si+1
y
] + [Six , Siz Si−1
z
] + [Six , Siz Si+1
z
]i,
dt }
(5a)
d x y y
hSi i = Jh−Siz (Si−1 + Si+1 ) + Siy (Si−1z z
+ Si+1 )i. (5b)
dt
The same can be done for the Syi and Szi components of Si , but they can
also be found using a cyclic permutation on Equation 5b. The equations of
motion for all three components of Si are
d x y y
S = −JSiz (Si−1 + Si+1 ) + JSiy (Si−1
z z
+ Si+1 ),
dt i
d y
S = −JSix (Si−1
z z
+ Si+1 ) + JSiz (Si−1
x x
+ Si+1 ),
dt i
d z
S = −JSiy (Si−1
x x
+ Si+1 y
) + JSix (Si−1 y
+ Si+1 ). (6)
dt i
These relations can be generalized to dtd Sxi = JSi ×(Si−1 +Si+1 ). Considering
small deviations for S with reverence to the z-axis, such that S z = ~S and
S x , S y  S z , the equations of motion in Equation 6 reduce to:
d x y y
Si ≈ JS(~ − Si−1 − Si+1 + 2Siy ),
dt
d y
S ≈ JS(~ − 2Six + Si−1
x x
+ Si+1 ),
dt i
d z
S ≈ 0. (7)
dt i
An educated guess can now be taken by Sjx = Aeik·Rj −iωt and Sjy = Beik·Rj −iωt ,
where a is the distance between the lattice points and Rj = j·ax̂ indicates
the position in our one dimensional model. Plugging in these equations as
an ansatz yields the following result:
iωA + 2SJ~B (1 − cos(kx a)) = 0,
−iωB + 2SJ~A (1 − cos(kx a)) = 0. (8)

5
Finding the dispersion relation from this set of equations can be done by
using the matrix notation given by
" #" #
iω 2SJ~(1 − cos(kx a)) A
= 0. (9)
2SJ~(1 − cos(kx a)) −iω B

Calculating the zero points of the determinant will give the dispersion relation
of a one dimensional ferromagnet. This results in ωk = 2SJ~(1 − cos(kx a)).
Extracting the eigenvectors of this matrix will provide us the states a one
dimensional ferromagnet can occupy. The two resulting eigenvectors are
given by
   
1 i 0
v1 = √ , v2 = . (10)
2 1 0

Eigenvector v2 describes the static state where all the sites have their spin
pointing along the positive z-axes. For eigenvector v1 we find an eigenstate
with magnons. Using v1 for A and B in the used ansatz, the real part of this
eigenstate for a = 0 gives

Sjx ∝ sin ωt,

Sjy ∝ cos ωt. (11)

This describes a precessing spin around the z-axis. If you now include all the
sites you see that each site picks up an extra phase shift of kx a. In fact you
now have a spin wave, or magnon (see Figure 1). For k  1, the dispersion
reduces to ωk ≈ SJ~(kx a)2 . As a magnon with wavevector k costs an energy
~ωk > 0 we conclude that the ground state of a one dimensional ferromagnet
has no magnons.

2.2 Holstein-Primakoff Transformation

We will now include quantum mechanical effects in our calculations. To
study the behaviour of magnons in our ferromagnetic system using spin-
theory we would like to make them explicitly appear in the Hamiltonian.
This can be done by using the Holstein-Primakoff transformation [7], which
is a mapping from angular momentum operators to bosonic creation and
annihilation operators [8]. Explicitly, the rising and lowering operators of S
can be written in terms of creation and annihilation operators:

6
 q √
Si+ =~ 2S − a†i ai ai ' 2S ~ai ,
q √
Si− =~a†i 2S − a†i ai ' 2S ~a†i ,
Siz =~S − ~a†i ai , (12)

where a†i and ai obey the commutation relation [ai , a†j ] = δij . The operator
n̂i = a†i ai counts the number of bosons placed at site i and is therefore
often called the number operator for site i. The boson number must satisfy
the constraint hn̂i i ≤ 2S, since otherwise the expression in the square root
would become negative and hence imaginary. Also, we performed a Taylor
expansion on Si+ and Si− in powers of 1/S to simplify the expression. Keep in
mind that by doing so we removed the critical information that a state with
more than 2S magnons per lattice site would be non-physical. Furthermore,
our final expansion is only accurate in the limit where S is large. In order to
introduce these Holstein-Primakoff transformations we write the Hamiltonian
in Equation 1 in terms of the rising and lowering operators of S. Also,
since we only consider nearest neighbour interactions, we write j = i + δ,
where δ is a vector connecting nearest neighbour sites. To avoid double
counting, we choose to run δ over only half the nearest neighbour vectors.
As a model of our system we use hypercubic lattices. This means that in one
dimension δ = +x̂, in two dimensions δ = +x̂, +ŷ and in three dimensions
δ = +x̂, +ŷ, +ẑ. These modifications result in the following Hamiltonian:
XX 1 1
H =−J Siz Si+δ
z
+ Si+ Si+δ
+
+ Si− Si+δ
−
. (13)
i δ
2 2

Consider first the fully ferromagnetic state, on which we act with the
operator S to flip a single spin. The term Siz Si+δ z
leaves the reversed spin
at the same place and only gives an energy cost to reverse it. The terms
Si+ Si+δ
+
+ Si− Si+δ
−
will move the reversed spin to another site, such that it will
create a spin wave with a certain dispersion. Now we perform the Holstein-
Primakoff transformation, as described in Equation 12, to study the Heisen-
berg ferromagnet and its excitations. We only include terms up to quadratic
order, as can be seen in Equation 14. Terms in the Hamiltonian which are
higher order contain four or more boson operators and therefore represent
interactions between bosons. However, since these are suppressed by at least
by a factor 1/S compared to the non-interacting terms, their effects are small.
Certainly at large S or when the boson number is small they can either be
neglected to a first approximation or be treated as weak perturbations on

7
the non-interacting theory. After the Holstein-Primakoff transformation we
find that
XX
H = − J~2 S 2 − Sa†i ai − Sa†i+δ ai+δ + Sa†i+δ ai + Sa†i ai+δ . (14)
i δ

To diagonalize this Hamiltonian we introduce Fourier-transformed boson op-

erators by
1 X −ik·ri
ai = √ e ak ,
N k
1 X ik·ri †
a†i = √ e ak . (15)
N k

An important property of this transformation is that it is canonical. This

means that it preserves the commutation relations. The operators ak will
obey the same kind of commutation relations as the original boson operators:
a†k0 ]P
[ak , P = δk,k0 . Before we insert the transformed operators we note that
−J i δ ~2 S 2 is the same as −JN z2 ~2 S 2 , where z is the number of nearest
neighbours. In our formulation z = 2d, where d denotes the dimension of
the system. To calculate the rest of the sum we start by looking at the term
P P †
i δ ai ai+δ .
XX 1 X X X −ik·ri † ik’·(ri +δ)
a†i ai+δ = e ak e ak’ ,
i δ
N i δ kk’
1 X X X −i(k−k’)·ri ik’·δ †
= e e ak ak’ ,
N kk’ δ i
1 XX
= N δkk’ eik’·δ a†k ak’ ,
N kk’ δ
XX
= eik·δ a†k ak . (16)
k δ
P −i(k−k’)·r
In the third line of this calculation we used that ie
i = Nδ
kk’ .
In a similar way we calculate the other terms. In the appendix the results
of the operator multiplications are shown. By insirting the results of those
multiplications the Hamiltonian in Equation 14 transforms to
z X † 
H = − J~2 S 2 N + J~2 Sz

ak ak 1 − γk , (17)
2 k

where γk = z2 δ cos(k · δ). In one dimension this means that γk = cos(kx a),
P
where a is again the spacing between lattice points. Finally, we can write

8
the Hamiltonian for ferromagnets in the compact diagonalized form:
X
H = E0 + ~ωk a†k ak , (18)
k

where ωk = SJ~z(1 − γk ) and E0 = −J~2 S 2 N z2 . In one dimension we find

that ωk = 2SJ(1 − cos(kx a)). This is the same dispersion relation as the
one resulted from our semi-classical approach. Equation 18 tells us that we
can describe our system as a sum of independent harmonic oscillators, each
with its own wavevector k. The quanta of the harmonic oscillator are called
magnons. They are just quantized spin wave excitations with energy ~ωk . As
a magnon with wavevector k costs an energy ~ωk , the minimized energy (the
ground state) has no magnons. The ground state energy is therefore simply
the situation where all spins point in the same direction with projection ~S
along the z−axis.
Increasing the temperature of the system will lead to magnons, originating
as thermally excited states of the ground state. When we ignore the ground
state energy we can derive a distribution function for magnons by using the
methods of statistical physics. We will first turn to the partition function for
magnons, which is given by
Y Y 1
Z= Zk = .
k k
1 − e−β~ωk

We can calculate the total energy E of the system by

∂ X ~ωk X
E =− ln Z = β~ωk − 1
≡ nk ~ωk ,
∂β k
e k

where nk is the well known Bose Einstein distribution, which describes the
statistical behaviour of bosons. Note that the chemical potential of magnons
is zero so that the Bose-Einstein distribution reduces to a Planck distribution
function, like for photons. If there are a few magnons in the system, such
that they won’t interact, we can describe them as bosons. WePwill now look
at the magnetization. The magnetization is given by M = i Si and is a
natural parameter to describe the magnetic ordering of a system. The larger
M , the stronger the ferromagnetic order. The total number of magnons at
temperature T is given by [8]
Z   32 Z ∞ √
1 N T xdx
n= d3 knk = 2 .
(2π)3 4π 2SJ 0 ex −1

9
Each magnon will reduce the magnetization. The magnetization will thus be
dependent on temperature:
  23
T
M (T ) = M (0)[1 − const. ].
2SJ

We are now finished with our investigation of the ferromagnet.

10
3 Antiferromagnets
In this chapter we will discuss the antiferromagnet. The classical ground
state of an antiferromagnet exist of neighbouring spins pointing in opposite
directions: | ↑↓↑↓↑↓i. This can be interpreted as two sublattices, such that
sites on one sublattice have as their nearest neighbours only sites from the
other. We choose our labelling such that sites in sublattice A have spins
pointing up and sites in the other sublattice, B, have spin down. The anti-
ferromagnetic case corresponds to a J < 0 in Equation 1. So we will use the
Hamiltonian described by
X
H=J Si · Sj , (19)
i,j

where J is taken to be positive such that the energy of the system is mini-
mized when the sites in sublattice A have spin up and the sites in sublattice
B have spin down.
In the following paragraphs we shall first discuss the classical features
of the antiferromagnet and afterwards discuss the effects of the quantum
fluctuations by performing a Holstein-Primakoff transformation to include
quantum effects in our calculations.

3.1 Semi-classical approach

We start with a semi-classical approach to the antiferromagnet. The change
of the spin operator in time can be summarized in the following equation:
∂Sj ∂H
= −Sj × , (20)
∂t ∂Sj
If we rewrite the Hamiltonian in Equation 19 in terms of sublattice A and B
and rewrite the sum over i and j we find
JX A JX B
H= Sj · [SB B
j+1 + Sj−1 ] + Sj · [SA A
j+1 + Sj−1 ].
2 j∈A 2 j∈B

We now use Equation 20 to describe the evolution of the system in time,

which results in
∂SAj
= −JSA B B
j × [Sj−1 + Sj+1 ],
∂t
∂SBj
= −JSB A A
j × [Sj−1 + Sj+1 ]. (21)
∂t

11
Using small deviations for S with respect to the z-axis, the spins in sublattice
A and B can be described by
 A,x 
−δSiB,x
 
δSi
SA SB
 A,y  B,y 
i = δSi  , i = −δSi . (22)


~S −~S

We now can construct equations of motion using Equation 21 considering

small deviations described by Equation 22.

δ ṠjA,x = −JS~ − 2δSjA,y + δSj−1

B,y B,y


+ δSj+1 ,
δ ṠjA,y = −JS~ 2δSjA,x − δSj−1
B,x B,x


− δSj+1 ,
−δ ṠjB,x = −JS~ − 2δSjB,y + δSj−1A,y A,y


+ δSj+1 ,
B,y B,x A,x A,x


−δ Ṡj = −JS~ 2δSj − δSj−1 − δSj+1 . (23)

Using the ansatz described by SjA,x = Ax eik·Rj −iωt , SjA,y = Ay eik·Rj −iωt ,
SjB,x = Bx eik·Rj −iωt and SjB,y = By eik·Rj −iωt , we write the equations of mo-
tion in matrixform. In this formulation a is the distance between the lattice
points and Rj = j·ax̂ indicates the position in a one dimensional model.
  
iω 2J~S 0 2J~S cos(kx a) Ax
−2J~S iω −2J~S cos(kx a) 0   Ay 
  

    = 0.

 0 2J~S cos(kx a) −iω 2J~S  B 
  x
−2J~S cos(kx a) 0 −2J~S −iω By
(24)

Diagonalizing this matrix we find that ωk = 2J~S sin(kx a), which for small
values of kx gives a linear dependence. There are four eigenvectors associated
with this Hamiltonian, which for k = 0 are given by

       
−i i −i i
1 −1 1 1
 1 1
 1 −1
v1 =  , v2 =  , v3 =  , v4 =  . (25)
2 i  2 i 2 −i 2  −i 
1 1 1 1

With the use of the ansatz we we find the eigenstates of the system. We
look at the real part with a = 0. The result for v1 is given in Equation 26.
This can be interpreted as a precession in sublattice A and a precession in
sublattice B. Those precessions both go clockwise and have the same phase.

12
Note that the spins in sublattice A point in the positive z-direction and the
spins in sublattice B in the negative z-direction.
1 1
δSxA = − sin ωt, −δSxB = sin ωt,
2 2
1 1
δSyA = − cos ωt. −δSyB = cos ωt. (26)
2 2
Interpreting the results for the other eigenvectors we find that v2 describes a
state where the precessions in sublattice A and B have a phase difference of
π, but they both spin clockwise. The real part of v3 describes a state where
both precessions go anticlockwise but have a phaseshift of π. The real part
of v4 describes a state where the precessions both go anticlockwise and have
no phase shift.

3.2 Holstein-Primakoff Transformation

We will now include quantum mechanical effects in our calculations. As the
ground state for classical spins is described by the Néel state, one might
naively guess that the ground state for the quantum case is analogous. It
turns out that the Néel-state is actually not the ground state, and is in fact
not even an eigenstate for the antiferromagnetic Heisenberg model. Quantum
fluctuations play an important role in the antiferromagnetic case, as they
change the ground state (and therefore its energy) away from the classical
result.
To illustrate this effect we will again use a Holstein-Primakoff transforma-
tion to make the magnons appear specifically in our Hamiltonian, described
by Equation 19. On the A sublattice, where the sites all have spin pointing
upward, we use the standard expressions in Equation 27. In the limit of large
S we may approximate them.
!
√ † √
a a i
Si+A = ~ 2S 1 − i ai ' 2S~ai ,
2S
!
√ † √
† a ai
Si−A = ~ 2Sai 1 − i ' 2S~a†i ,
2S
SizA = ~S − ~a†i ai . (27)

The sites in sublattice B have spin pointing down, so we must modify the
Holstein-Primakoff expressions accordingly to reflect this. The modifications

13
correspond to replacing S z → −S z and S + ↔ S − :

b†j bj
!
√ †
√
+B
Sj = ~ 2Sbj 1 − ' 2S~b†j ,
2S
b†j bj
!
−B
√ √
Sj = ~ 2S 1 − bj ' 2S~bj ,
2S
SjzB = − ~S + ~b†j bj . (28)

Since we can split our system in sublattice A and B, we write the Hamil-
tonian given by Equation 19 in terms of A and B. Also we will replace
the sum over nearest neighbours i and j by a sum over i and δ, as we did in
the Holstein-Primakoff transformation for ferromagnets. The Hamiltonian in
terms of the lowering and rising operators of S that reflects our modifications
is given by Equation 29
XX 1 1
−B
H =J Si+A Si+δ + Si−A Si+δ
+B
+ SizA Si+δ
zB

i∈A δ
2 2
XX 1 1
−A
+J Sj+B Sj+δ + Sj−B Sj+δ
+A
+ SjzB Sj+δ
zA
. (29)
j∈B δ
2 2

In this definition the sum over i and j runs over the lattice sites belonging
to sublattices A and B respectively. Inserting the Holstein-Primakoff trans-
formations described by Equation 27 and Equation 28 and considering only
terms up to quadratic order in operator products, our Hamiltonian trans-
forms to
XX
H = J~2 S(ai bi+δ + a†i b†i+δ ) − S 2 + S(a†i ai + b†i+δ bi+δ )
i∈A δ
XX
+ J~ 2
S(bj aj+δ + b†j a†j+δ ) − S 2 + S(b†j bj + a†j+δ aj+δ ). (30)
j∈B δ

We will now Fourier transform the operators in order to diagonalize the

Hamiltonian. We will choose the vectors k to lie in the Brillouin zone as-
sociated with each sublattice. The Brillouin zone is identical for sublattice
A and B. Therefore −π2
< k < π2 and the Fourier transformed operators are
given by
1 X ik·rj 1 X ik·ri
aj = √ e ak , bi = √ e bk . (31)
Na k Nb k

14
 P
Including our transformed operators in the term δ ai bi+δ we find
XX 1 X X X i(k+k’)·ri ik’δ
ai bi+δ = e e ak bk’ ,
i δ
N aN b i δ k,k’
XX
= δk-k’ eik’δ ak bk’ ,
δ k,k’
XX
= e−ikδ ak b-k . (32)
δ k

The same can be done for the other components in the Hamiltonian. The
results of these operator multiplications are given in the appendix. The
Hamiltonian in terms of these transformed operators, as seen below, is not
diagonal.
z X † X 
H = −J~2 N S 2 + J~2 Sz ak ak + b†k bk + J~2 Sz γk ak b−k + a†k b†−k ,
 
2 k k
(33)

with γk = z2 δ cos(k · δ). In contrast to the ferromagnetic case, performing

P
a fourier transform does not diagonalize the Hamiltonian. To diagonalize
the Hamiltonian we will now introduce Bogoliubov transformations [9]. The
basic idea of the Bogoliubov transformations is introducing a new set of
operators α and β such that the Hamiltonian has only terms proportional to
α† α and β † β [10]. The operators α and β are defined in Equation 34, where
uk and vk are assumed to be real functions of vector k.

αk = uk ak − vk b†−k ,
βk = uk bk − vk a†−k . (34)

The transformations are canonical, in the sense that the new operators
will obey the Bose commutation rules. From this we conclude that, since
[αk , αk† 0 ] = [βk , βk† 0 ] = δk,k0 , it follows that u2k − vk2 = 1. This will be auto-
matically satisfied when uk = cosh θk and vk = sinh θk . The requirement
[αk , βk† 0 ] = 0 gives another condition, namely uk v−k = u−k vk . This is solved
by setting uk = u−k and v−k = vk . These conditions are assumed to hold in
the inverse transformations, given by Equation 35.
†
ak = uk αk + vk β−k ,
†
bk = uk βk + vk α−k . (35)

If we now express the Hamiltonian from Equation 33 in terms of the α and

β bosons and use the corresponding commutation relations we can rewrite

15
the Hamiltonian to
X
H = H0 + J~2 Sz [u2k + vk2 + 2γk uk vk ](αk† αk + βk† βk ) + 2(vk2 + γk vk uk )
k
†
+ [γk (u2k + vk2 ) + 2uk vk ](β−k αk + β−k αk† ) ,

(36)

where H0 = −J~2 N z2 S 2 . Now we choose uk and vk such that the terms

†
involving β−k αk and β−k αk† cancel. Inserting uk = cosh θk and vk = sinh θk
the constraint for θk is described by

γk = − tanh 2θk . (37)

This constraint simplifies the Hamiltonian to

z X
H = − N J~2 S 2 + J~2 SZ u2k + vk2 + 2γk uk vk αk† αk + βk† βk + 1


2 k
X
2 2 2


+ J~ Sz vk − uk . (38)
k

After some manipulations we find the diagonal Hamiltonian given by

X
~ωk αk† αk + βk† βk ,


H = E0 + (39)
k
p
where ωk = J~Sz 1 − γk2 and

1 X
E0 = − N J~2 (S 2 + S)z + ~ωk . (40)
2 k

In this diagonalized form of the Hamiltonian it becomes clear that the op-
erators α and β create magnon excitations with wavevector k and energy
~ωk . We first note that for each k there are two types of magnons (α and β),
which are degenerate in energy. Furthermore, note that while k approaches
zero ωk → 0, as in the ferromagnetic case. But unlike the ferromagnetic
case, for which a quadratic dispersion was found in this limit, in the antifer-
romagnetic case we find a linear dispersion for small values of k. So in a one
dimensional antiferromagnet where kx  1 the dispersion relation is given
by ωk ≈ J~Sz|kx |a. The ground state energy E0 , given in Equation 40, gives
the energy in the limit T → 0. The term − 21 N J~2 S 2 z is just the ground state
energy of a classical antiferromagnet with nearest-neighbour interactions be-
tween the spins. The other terms represent the quantum corrections to the
classical ground state energy. The contribution of the quantum corrections is

16
 P p 
thus described by Eqc = − 12 N J~2 Sz + k ~ωk = JS~2 z k 1 − γk2 − 1 .
P

Since this is always smaller than zero we conclude that quantum fluctuations
lower the energy of an antiferromagnetic system.
Looking at the magnetization of the system it is convenient to look at
the sublattice magnetization. The total magnetization can not be used since
it turns out to be zero in the presence of antiferromagnetic order. The cor-
rection for P
the sublattice magnetization to the classical result S is given by
∆MA = N k ha†k ak i. Using the Bogoliubov transformations in Equation 35
1

on the operators a we find

   
1 2 X 1 1 1 2 X 1 JSz
∆MA = − + nk + p =− + nk + .
2 N k 2 1 − γk2 2 N k 2 ωk
(41)

In the limit that T → 0 we find that ∆MA 6= 0. Even at zero temperature the
magnetization of an antiferromagnet is reduced due to quantum fluctuations.
This concludes our analytical description of the ferromagnet and anti-
ferromagnet. We now proceed numerically, such that we can consider more
complicated magnetic configurations.

17
4 Numerical methods
In this chapter we construct a method to numerically calculate the ener-
gymodes of any magnetic configuration. Specifically we will calculate the
dispersion for antiferromagnets and ferromagnets and compare this with our
analytic description in chapter 3 and 4. We would like to modify the Hamil-
tonian such that we obtain a matrixform. Let us assume a two dimensional
lattice n x m, where we label the site at the bottom left corner with 1. Now
we count the cells in the positive x-direction on each row, such that the
site in the upper right corner will be labeled nm. The matrixform of the
Hamiltonian can now be written as [11]
 
†

a 1
H = a a Hm † − T r[Hm ], (42)
a 2
 
†

† † †

A B
where a a = a1 , a2 , ..., aN , a1 , a2 , ..., aN and Hm = , with A
B ∗ A∗
and B matrices. In order to obtain only physical modes/magnons, Hm is
constrained to be positive-definite. The bosonic operators a can be trans-
formed using a para-unitary 2N × 2N matrix T , such that the matrix Hm
will be diagonalized according to the following scheme [9]:
     
†

a †

† † −1 −1 a †

γ
a a Hm † = a a T (T ) Hm T T = γ γ D † , (43)
a a† γ

where D = (T † )−1 Hm T −1 = 21 ~ diag(ω1 , ω2 , ...., ωN , ω1 , ω2 , ..., ωN ) with all ωi

positive. This transformation is similar to the Bogoliubov transformations
discussed previously, but is now performed in real space, rather than in mo-
mentum space after a Fourier transform. This more general form is needed
in case there is no translation invariance such that a Fourier description will
not useful. By doing so we defined a new basis as shown in Equation 44.
The para-unitary property of T ensures that the operators γ are bosonic
creation/annihilation and therefore satisfy [γ, γ † ] = δij .
   
γ a
γ † γ = a† a T † .



† =T † , (44)
γ a

We are now interested in finding T −1 . Since T is para-unitary it must satisfy

T † φ = φT , where φ is the para-unit matrix given by φ =diag(1, 1, ..., −1, −1).
Also, we know D from Equation 43. Therefore we can conclude that

D = diag(λ1 , λ2 , ..., λN , −λN 1 , −λN +2 , ..., −λ2N ) = φ Λ, (45)

18
where λi = 12 ~ωi for i from 1 to N and λi = − 12 ~ωi for i from N + 1 to 2N .
Also, Λ is introduced as a list of λi to simplify the notation. Multiplying
both sides of Equation 45 with φT −1 gives φDT −1 = ΛT −1 . From which can
easily be seen that each column of T −1 satisfies

(φD − λp )ωp = 0, (46)

where ωp denotes the column p of T −1 . The para-unitary matrix T −1 can now

be obtained by finding the eigenvectors ωp of the matrix φD and use them to
create T −1 = (ω1 , ω2 , ..., ω2N ). In summary we find that our transformation
matrix is given by

U V∗
 
−1
T = , (47)
V U∗



where we define U = u1 u2 u3 ... uN and V = v1 v2 v3 ... vN ,
with up and vp column vectors. Those can be found by calculating ωp =

0
up vp , where the prime indicates that it is a column vector.

4.1 (Anti-)Ferromagnetic Heisenberg Model

To show that this method actually works, we calculate the spin wave spec-
trum for a one dimensional ferromagnet and antiferromagnet consisting of
64 spins. We use the Heisenberg model in Equation 1 for ferromagnets and
in Equation 19 for antiferromagnets. We will diagonalize those Hamiltonians
using the matrix form explained in this chapter and also in ref [5].

We first perform the calculations on a ferromagnet and find that in the

classical limit the energy spectrum gives a sinusoidal shape, as shown in Fig-
ure 2. This is in accordance with our analytical dispersion relation calculated
in chapter 3. Every mode is associated with a wave vector k, but every spin
wave going in the opposite direction has the same energy due to symmetries
of the model, such that the values of ωk P appear twice. The total energy of
this spectrum is 128. The contribution 21 k ωk is therefore the same as the
classical ground state energy, JS~N z2 with J = 1, S = 1, ~ = 1, N = 64 and
z = 2. So we can again conclude that the ground state remains unchanged
due to quantum fluctuations in the ferromagnet.
In the antiferromagnetic case we find the results shown in Figure 2. The
quadruple effect appears as result of degenerate eigenmodes, described in
Equation 25. The dispersion relation analytically calculated in chapter 4
produces the same energy spectrum as the numerical results. The classical
ground state energy is the same as for the ferromagnet. The total energy of

19
the antiferromagnetic dispersion relation gives 81.4. We therefore conclude
that the quantum fluctuations lower the ground stateP energy of the one di-
z 1
mensional antiferromagnet by Eqc = −JS~ 2 + 2 k ωk , which for J = 1,
S = 1 and ~ = 1 gives Eqc = −23.3.

3
E

0
0 10 20 30 40 50 60
mode

2.0

1.5
E

1.0

0.5

0.0
0 10 20 30 40 50 60
mode

Figure 2: Energy spectrum for a one dimensional ferromagnet (top) and

antiferromagnet (bottom) consisting of 64 spins. We used S = 1, J = 1,
~ = 1 and a = 1. The energy eigenvalues are plotted by ordering the energies
from small to large.

20
5 Classical ground state of an antiferromag-
netic Hamiltonian with Dzyaloshinskii-Moriya
interactions
In the previous chapters we calculated the dispersion relations for ferro- and
antiferromagnets by assuming only the coupling of nearest neighbours. To
gain a more realistic result we also have to include the influence of an external
field, anisotropy and Dzyaloshinskii-Moriya interactions in our Hamiltonian.
In this chapter we will discuss the effects of Dzyaloshinskii-Moriya coupling
in an antiferromagnet. This coupling is also known as the antisymmetric
exchange and describes a contribution to the total magnetic exchange inter-
action between neighbouring spins that occurs when inversion symmetry is
broken. We will see that whirling types of magnetization in chiral magnets
1
are induced by the Dzyaloshinskii-Moriya interaction. An example of such
a whirling magnetic configuration is a skyrmion. The skyrmion is a vortex-
like configuration and was originally introduced in particle physics by T.H.R.
Skyrme in 1962 [12]. Although they were predecited they were discovered
experimentally just recently [13] [14]. Skyrmions are potentially good for
data storage, because of their topological stability and easy manipulation
with very low electrical currents [15] [5]. Therefore skyrmions are an active
field of research.
We start with a ferromagnetic Hamiltonian, containing only the Heisen-
berg exchange coupling and Dzyaloshinskii-Moriya interaction given by
JX DX
H= Si · Si+1 + Si · Si−1 − (Si × Si+1 )y − (Si+1 × Si )y , (48)
2 i 2 i
where J is the Heisenberg exchange coupling constant and D is the Dzyaloshinskii-
Moriya interaction constant. For simplicity we use natural units during the
calculations. After expanding the Fourier transform of the Hamiltonian for
small q and using the continuum limit the Hamiltonian is given by
d2 S(x)
Z Z
J D dS(x)
H=C− S(x) · 2
dx + S(x) × dx. (49)
2a dx a dx
By minimizing the variation of the Hamiltonian we find that the correspond-
ing Euler equations are solved by
S sin( D
 
J
x)
S= 0 . (50)
D
S cos( J x)
1
Chirality is a form of asymmetry of the system. If the atomic structure of a magnet
lacks inversion symmetry we call them chiral magnets.

21
To find the ground state of the antiferromagnet we use a staggered notation
of the antiferromagnetic spin, given by Sstag
i = Si (−1)i . In this way we
can use the ferromagnetic Hamiltonian. By minimizing the ferromagnetic
Hamiltonian for all three components of the spin we find three constrains:
y y
Si+1 = Si−1 ,
x x z z
J(Si+1 + Si−1 ) = D(Si−1 − Si+1 ),
z z x x
J(Si+1 + Si−1 ) = D(Si+1 − Si−1 ). (51)

From the continuous result given by  Equation 50 we derive

 an educated guess
i
for this discrete case: Si = S(−1) sin(aqi), 0, cos(aqi) , where the constraint
for aq is given by Equation 51 and will therefore be aq = arctan D J
. We see
that Si makes a precession with period p = 2π/arctan D J
. For 64 spins,
J = 1, S = 1, a = 1 and choosing D such that there is exactly one period,
the configuration for the staggered ground state of an antiferromagnet is
shown in Figure 3b. The ground state of an antiferromagnet can now easily
be obtained and is shown in Figure 3a.

(a)

(b)

Figure 3: (a) Ground state of a one dimensional antiferromagnetic configu-

ration where D = 0.09849, J = 1 and S = 1. The ground state is described
by one magnon. (b) As for (a) but staggered.

22
6 Phase transition to the spin-flop state in
antiferromagnets
We thoroughly treated the quantum effects in the ferro- and antiferromegnet
considering only Heisenberg exchange interactions. Also, we constructed a
numerical method to calculate the dispersion relations in magnetic config-
urations. In this chapter we will use this knowledge to calculate the effect
of quantum fluctuations in the transition between an antifromagnetic Néel
state and a spin-flop state. In Figure 4 a classical picture of both spin-
states is shown. We first examine the classical interpretation of the transi-
tion in two dimensions. The Hamiltonian used contains exchange coupling,
anisotropy, magnetization and Dzyaloshinskii-Moriya interaction. The phase
transition from an antiferromagnetic state to a spin-flop state occurs when
the Dzyaloshinskii-Moriya interaction is zero [16]. In our calculations we
used natural units and the effective energy in Equation 52 for simplicity.
An effective Hamiltonian is a Hamiltonian that acts in a reduced space and
only describes a part of the eigenvalue spectrum. The anisotropy factor K
in the Hamiltonian will make sure that our system has an ‘easy-axis’ along
which it would like to magnetize either up or down, which in our case will be
the z-axis. The strength of the external field B, oriented in the z-direction,
eventually forces the antiferromagnetic state to change to the spin-flop state.
The transition-value for B is found at the crossing of the minimized energy
for both phases at D = 0.

Hef f = JSij · (Sij+x + Sij+y ) + K(Szij )2 − BSzij (52)

In a pure antiferromagnetic state we describe the spin by SAF ij given in

AF M
Equation 53. This gives an effective energy, Hef f = −2J +K. The external
field has no influence on the effective energy as it remains constant by in-
creasing B. In the pure spin-flop state SSF
ij from Equation 53 can be used to
describe the spinstates. After minimizing the effective energy in Equation 52
SF B2
as function of θ, we find Hef f = −2J − 4(4J+K) .

   
0 (−1)i+j cos[φ] sin[θ]
SAF
ij = 0  SSF
ij =  (−1)i+j sin[φ] sin[θ]  (53)
(−1)i+j cos[θ]

The minimized energy in the antiferromagnetic state is independent of

the strength of the external field B. However, the minimized energy in the

23
spin-flop state has a quadratic dependence on B. In Figure 5 the mini-
mized energy for both phases is plotted. Increasing B causes the spin-flop
state to eventually become the favoured state of the system and there-
fore a phase transition will occur at B ≈ 0.796. The same calculations
can be done in one dimension, which results in a transition at B ≈ 0.56.

Figure 4: Classical picture of the spin-flop and antiferromagnetic state.

-2.00

-2.02

-2.04
E

-2.06

-2.08

0.0 0.2 0.4 0.6 0.8 1.0 1.2

B

Figure 5: Effective energy of the antiferromagnetic state (red) and the spin-
flop state (blue) in two dimensions. The crossing happens at B ≈ 0.796.
Here we used J = 1, S = 1 and K = −0.04.

24
6.1 Magnon energies for the spin-flop state
Increasing B will cause the antiferromagnetic state to eventually change to
the spin-flop state. It is therefore interesting to calculate the dispersion rela-
tion of the spin-flop state such that the effect of quantum fluctuations on the
ground state can be obtained. In the next section we will also calculate the
dispersion relation of the antiferromagnetic state such that we can investigate
the effects of quantum fluctuations on the transition between the two states.
As before, a Holstein-Primakoff transformation can be used to calculate the
energy spectrum. The Holstein-Primakoff representation of the spin operator
given by Equation 12 discriminates only in one direction. Therefore, in the
case of non-collinear classical ground states such as the spin-flop state, it is
convenient to define a rotated local coordinate system as follows [11]:

êi1 = cos θi cos φi x̂ + cos θi sin φi ŷ − sin θi ẑ

êi2 = êi3 × êi1
êi3 = sin θi cos φi x̂ + sin θi sin φi ŷ + cos θi ẑ, (54)

or, in more compact form:

êia = (Ra,b
i
)−1 r̂b , (55)

where θi and φi are the angles that characterize the spin direction of the clas-
i
sical ground state at site i, Ra,b is a rotation matrix and r̂b are the Cartesian
axis. The Hamiltonian that includes nearest neighbour interactions, external
field and anisotropy can now be expressed in this frame:
X X
H =J (Si · êiα )(Sj · êjβ )Rα,γ
i j
Rβ,γ −B (Si · êiα )Rα,3
i

<ij> i
X
+K (Si · êiα )(Si · êiβ )Rα,3
i i
R3,β , (56)
i

where we included that B = B ẑ and chose ẑ as the anisotropic easy axis.

We choose êi3 along the spindirection at every site i such that we can easily
use a Holstein-Primakoff transformation for the small fluctuations around
the spin direction. In the specific case of a spin-flop state this results in
B
φi = πi and θi = arccos [ 2(zJ+K) ], where z is the number of neighbours. The
value of φi is chosen such that each neighbouring spin has opposite sign in
the direction of the chain, as can be seen in Figure 4. The constant value of
θi can be found by minimizing the energy as function of θi , as we did in the
previous section for Equation 52. Implementing those values of φi and θi in
i
Equation 54, a rotation matrix Ra,b can be constructed from Equation 55.

25
 i
The exact form Ra,b for the spin-flop state can be found in Equation 62 in
the appendix. To diagonalize the resulting Hamiltonian we use the Holstein-
Primakoff transformation described by the combined use of Equation 12 and
the following expressions[11]:
Si± = Si · êi1 ± iSi · êi2 ,
S − a†i ai = Si · êi3 . (57)
We will proceed in one dimension, so z = 1. By keeping only terms up to
the second order in bosonic operators α, α† we find the Hamiltonian
√ below.
B2 B
In this formulation E0 = −J − 4(2J+K) , x = 2(2J+K) and y = 1 − x2 .
X
H = E0 + S(Bx + 2J(1 − 2x2 ) − 2Kx2 − y 2 K(−1)i ) a†i ai
i
X X
+ JSy 2 a†i a†i+1 + ai ai+1 − x2 JS a†i ai+1 + ai a†i+1
i i
2
y KS X
− ai ai + a†i a†i . (58)
2 i
Now we are able to write the Hamiltonian in the matrix form explained
in Equation 42. We then use the diagonalization techniques described in
chapter 4 to find the excitation energies for the spin-flop state. In Figure 6
the results for a one dimensional chain of spins are shown.

2.0

1.5
E

1.0

0.5

0.0
0 20 40 60 80 100 120
mode

Figure 6: Energy spectrum for the spin-flop state considering a one dimen-
sional system consisting of 128 spins. In this graph S = 1, J = 1, B = 1 and
K = −0.04.

26
6.2 Magnon energies for the antiferromagnetic state
Since we are interested in the effects of quantum fluctuations on the transition
between the antiferromagnetic and the spin-flop state we will now calculate
the energy spectrum for the antiferromagnet. We use the Hamiltonian in
Equation 56 as we did for the spin-flop state. We again choose êi3 along the
spindirection at every site i such that we can easily use a Holstein-Primakoff
transformation for the small fluctuations around the spin direction. The ro-
tation matrix for the antiferromagnet can be constructed from the combined
use of Equation 54 and Equation 55 with θi = πi and φi = 0. The exact form
i
Ra,b for the antiferromagnetic state is given in Equation 63 in the appendix.
The resulting Hamiltonian can now be transformed using Equation 12 and
Equation 57. Here we show the one dimensional form:

X X
H = E0 + (2JS − 2KS + B(−1)i ) a†i ai − JS (ai ai+1 + a†i a†i+1 ), (59)
i i

where E0 = KS 2 N − JS 2 N . The dispersion relation that results from diag-

onalization is shown in Figure 7.

3.0

2.8
E

2.6

2.4

0 20 40 60 80 100 120
mode

Figure 7: Energy spectrum for the antiferromagnetic state considering a one

dimensional system consisting of 128 spins. In this graph S = 1, J = 1,
B = 1 and K = −0.04.

27
6.3 Quantum fluctuations
After obtaining the dispersion relation for the antiferromagnetic and spin-
flop state we are now able to calculate the influence of quantum fluctuations
in the system. The contribution to the groundstate energy due to quantum
fluctuations can be calculated by Eqc = 2 k ωk − 21 Tr(Hm )[11]. We will
1
P
calculate the contribution due to quantum fluctuations in a one dimensional
chain of 128 spins for S = 1, J = 1 and K = −0.04.
First we calculate the fluctuations in the spin-flop state, where the ma-
trix Hm was constructed from Equation 58. We found that the quantum
fluctuations lower the energy of the classically known ground state. Note
that by changing the value of the external field the quantum fluctuations
also change. We calculated the quantum fluctuations for several values of
B. Unfortunately we faced the problem that the matrix Hm was not always
positive definite and therefore some values for B could not be used for the
calculation. Afterwards we calculate the quantum fluctuations in the anti-
ferromagnetic state by constructing Hm from Equation 59. We found that
quantum fluctuations also lower the ground state of the antiferromagnetic
state.
In figure Figure 8 the effect of quantum fluctuations on the transition
between the two states in one dimension is shown. The quantum fluctuations
for both states decrease for increasing B. Due to the effect of quantum
fluctuations the transition seems to be moved and the crossing now appears
to happen around B = 0. These results imply that the one dimensional
antiferromagnetic state would only be favoured at B = 0, if at all. Due to
uncertainties we cannot exclude the possibility that there is no crossing at
all. There could also be other physical aspects that influence the ground
states, such that the classically known ground states change accordingly.
More research is needed to provide the decisive answer.

28
 -120

-140
Minimal Energy

-160

-180

-200

0.0 0.5 1.0 1.5

B-value

Figure 8: The solid lines represent the classical ground states of the antiferro-
magnetic (red) and spin-flop (blue) state. The effect of quantum fluctuations
on the ground state is indicated by the dots. In this one dimensional sys-
tem the classical crossing happens at B ≈ 0.56. We used J = S = 1 and
K = −0.04.

29
7 Conclusion
In this thesis we discussed the impact of quantum fluctuations on the classical
description of magnetic configurations. We started with a general investiga-
tion of the ferro- and antiferromagnet. We found that quantum effects lower
the energy of an antiferromagnet and that the energy of ferromagnet remains
unchanged. Afterwards we constructed a numerical method to obtain the en-
ergy spectrum of magnetic configurations, which could be used for further
research. In the end we investigated the spin-flop state. The spin-flop state
appears in antiferromagnets after increasing the external field. This phase
transition is classically well understood and therefore we used a numerical
method to investigate the influence of quantum fluctuations on this transi-
tion. With the energy spectrum of the antiferromagnetic and the spin-flop
state we calculated the quantum correction to the ground state. We found
that the quantum fluctuations lower the energy of both states. Our results in-
dicate the antiferromagnetic state is only the ground state, if at all, around
B = 0. This could mean that in one dimension the spin-flop state is the
ground state for all values of the external field except zero. Due to uncer-
tainties our results do not give a definite answer to the question whether
there is a transition point at all.

For further research it would therefore be interesting to optimize our cal-

culations and investigate how quantum fluctuations influence the phase tran-
sition from an antiferromagnetic to a spin-flop state in one dimension. We
also suggest to calculate the dispersion relations for both phases in momen-
tum space using the Hamiltonian given in Equation 64 in the appendix and
diagonalizing it with a Bogoliubov transformation. by doing so an analyti-
cal description of the quantum fluctuations could be obtained. Also, higher
dimensions could be explored. In two and three dimensions our expectation
is that the quantum fluctuations will lower the classical energy and the tran-
sition point will be shifted. The classical transition value in two and three
dimension happens at a higher value for B than in the one dimenional case.
Therefore, we expect the antiferromagnetic state to still be the groundstate
for low values of B in the two and three dimensional cases. Nevertheless, we
expect the transition to spin-flop state to be shifted to a lower value of B
in those cases. Furthermore, other phase transitions in the antiferromagnet
could be investigated using real space numerical diagonalization. For exam-
ple the transition from the antiferromagnetic to spiral state, which happens
at B = 0. Another example is the transition from the spin-flop to the spiral
state, which occurs for positive anisotropy [6].

30
8 Appendix
Operator Multiplications
A list of operator multiplications used to express the Hamiltonian in Equa-
tion 14 in terms of a†i ai :

XX zX †
a†i ai =
ak ak ,
i δ
2 k
XX † zX †
ai+δ ai+δ = a ak ,
i δ
2 k k
XX † XX
ai+δ ai = e−ik·δ a†k ak . (60)
i δ k δ

A list of operator multiplications used to express the Hamiltonian in

Equation 30 in terms of a†i ai and b†i bi :
XX † † XX
ai bi+δ = eikδ a†k b†-k ,
i δ δ k
XX XX
ai+δ bi = eikδ ak b-k ,
i δ δ k
XX XX
a†i+δ b†i = e−ikδ a†k b†-k ,
i δ δ k
XX XX zX †
a†i+δ ai+δ = a†i ai =
a ak ,
i δ i δ
2 k k
XX XX † zX †
b†i+δ bi+δ = bi bi = b bk . (61)
i δ i δ
2 k k

Rotation Matrices
i
Rotation matrix Ra,b for the spin-flop state is obtained when φi = πi and
θi = arccos x:

x(−1)i y(−1)i
 
0
i
Ra,b = 0 (−1)i 0 , (62)
 

−y 0 x
B B
√
where x = 2(4J+K)
in 2D, x = 2(2J+K)
in 1D and y = 1 − x2 .

31
 i
Rotation matrix Ra,b for the antiferromagnetic state is obtained when
θ = πi and φ = 0:
 
(−1)i 0 0
i
Ra,b = 0 −1 0 . (63)
 
i
0 0 (−1)

Matrixform of the Hamiltonian in momentum space

Up to this point we did calculations for the spin-flop state in real space. At
the end of the research program we were working on an analytical descrip-
tion of the wave spectrum for the spin-flop state. Implementing the rotation
i
matrix Ra,b , for the spin-flop state given in Equation 62, in the Hamiltonian
from Equation 56 results in the Hamiltonian in Equation 58. √In this nota-
B
tion x = 2(zJ+K) with z the number of neighbours and y = 1 − x2 . We
performed a Fourier transformationP on the Hamiltonian
P iQaj in the processπ of di-
j
agonalizing. We used the fact that j (−1) = j e , where Q = 2a with
a the spacing between the lattice points. The result of the transformation
B2
is given in the Hamiltonian given below, where E0 = −J − 4(zJ+K) and z
denotes the number of neighbours.
z X
H = E0 + J Sy 2 cos [ka](ak a−k + a†k a†−k )
2 k
z 2X
− J Sx cos [ka](a†k ak + ak a†k )
2 k
x X †
+ (BS − KSx2 − JzS(−1 + 2x2 )) (ak ak + a†−k a−k )
2 k
a2 X †
− KS (ak ak−Q + ak a†k+Q + a†k a†−k+Q + ak a−k−Q ). (64)
2 k

We already tried to make a symetric and hermetian matrixform of this

Hamiltonian by constructing a 8N × 8N notation, by making use of cyclic
boundaries, as can be seen in Equation 65. For further research a total
diagonalization of this Hamiltonian using Bogoliubov transformations would
provide a way to analytically calculate the dispersion relation.

    
 †  A B χ 1 A B
H = E0 + χ χ − Tr ∗ , (65)
B ∗ A∗ χ† 2 B A∗
 
† †
where χ = ak ak+Q a−k a−k−Q and

32
  2

L 2
− a8 KS 0 − a8 KS
 2 
0
 a2 
 − KS 0 0 
A= 8 ,
0
 
L

 0 0 2


2
− a8 KS 0 0 0

2 JSa2 cos (ka) 2

 
0 − a8 KS 2
− a8 KS
 
 a2
 − 8 KS 0 0 0


B=
 JSa2 cos (ka)
,


 2
0 0 0 

2
− a8 KS 0 0 0

2
with L = (BS x2 − KSx2 − JzS(−1 + 2x2 ) − JS x2 cos (ka)) and a the
spacing between the spins.

33
References
[1] F. Bloch, Z. Phys. 61, (1930) 206.

[2] Mattia Crescioli, Paolo Politi, Ruggero Vaia, Re-entrant spin-flop transi-
tion in nanomagnets, Phys. Rev. B 90, 100401(R) (2014)

[3] N. B. Ivanov, J. Ummethum, J. Schnack, Phase diagram of the

alternating-spin Heisenberg chain with extra isotropic three-body exchange
interactions,Eur. Phys. J. B 87 (2014) 226

[4] C. A. Lamas, D. C. Cabra, P. Pujol, G. L. Rossini, Diagnosing order by

disorder in quantum spin systems, Eur. Phys. J. B 88 (2015) 176

[5] P. van Dieten, Quantum fluctuations and degeneracies of two-dimensional

magnetic skyrmions, Bachelor Thesis, Utrecht University (2015).

[6] M. Raaijmakers, Inhomogeneous groundstates in antiferromagnets with

Dzyaloshinskii- Moriya interactions, Bachelor Thesis, Utrecht University
(2015).

[7] T. Holstein and H. Primakoff, Dependence of the Intrinsic Domain Mag-

netization of a Ferromagnet, Physical Review 58, (1940).

[8] D. I. Khomskii, Basic Aspects of the Quantum Theory of Solids, Cam-

bridge University Press, New York (2010) 1973: Random House, N.Y.

[9] J. H. P. Colpa, Diagonalization of the quadratic boson Hamiltonian, Phys-

ica 93A, 327-353 (1978).

[10] H.T.C. Stoof, K.B. Gubbels and D.B.M. Dickerscheid, Ultracold quan-
tum fields, 207-211, Springer (2009).

[11] A. Roldán-Molina, M.J. Santander and Á.S. Núñez, J. Fernández-

Rossier, Quantum theory of spin waves in finite chiral spin chains, Phys.
Rev. B 89, 054403 (2014).

[12] T. H. R. Skyrme, A unified field theory of mesons and baryons, Nucl.

Phys.31, 556-569 (1962).

[13] S. Mühlbauer, B. Binz, F. Jonietz, C. P eiderer, A. Rosch, A. Neubauer,

R. Georgii, P. Böni, Skyrmion Lattice in a Chiral Magnet, Science 323,
915-919 (2009).

34
[14] C. Pappas, E. Lelièvre-Berna, P. Falus, P. M. Bently, E. Moskvin, S.
Grigoriev, P. Fouquet, B. Farago, Chiral Paramagnetic Skyrmion-like
Phase in MnSi, Phys. Rev. Lett. 102, 197202 (2009).

[15] F. Jonietz, S. Mühlbauer, C. Pleiderer, A. Neubauer, W. Münzer, A.

Bauer, T. Adams, R. Georgii, P. Böni, R. A. Duine, K. Evershor, M.
Garst, A. Rosch, Spin Transfer Torques in MnSi at Ultralow Current
Densities, Science 330, 1648-1651 (2010).

[16] S. Mahdavifar, M. R. Soltani, A. A. Masoudi, Quantum corrections

of the Dzyaloshinskii-Moriya interaction on the spin- 12 antiferromagnet-
Heisenberg chain in an uniform magnetic field, Eur. Phys. J. B 62, 215-
220 (2008).

[17] S. J. Venema,Magnetic phase diagrams of spin systems with inversion

asymmetry, Bachelor Thesis, Utrecht University (2014).

35