Magnetism, Structure

and
Their Interactions
DISSERTATION
zur Erlangung des akademischen Grades
Doktor rerum naturalium
(Dr. rer. nat.)
vorgelegt
der Fakultat f ur Mathematik und Naturwissenschaften
der Technischen Universitat Dresden
von
Wenxu Zhang
geb. am 26. 06. 1977 in Sichuan, P. R. China
TECHNISCHE UNIVERSIT

AT DRESDEN
2007
1. Gutachter: Prof. Dr. Helmut Eschrig
2. Gutachter: Prof. Dr. Peter Fulde
2. Gutachter: Prof. Dr. J urgen Hafner
Eingereicht am 14. Nov. 2007
Contents
1 Introduction 1
2 Theoretical Background 3
2.1 Electronic Hamiltonian in a solid . . . . . . . . . . . . . . . . . . 3
2.2 Density functional theory and the Kohn-Sham scheme . . . . . . 4
2.3 The Kohn-Sham equations . . . . . . . . . . . . . . . . . . . . . . 7
2.4 Basics of FPLO code . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.5 Model considerations . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5.1 Free electron gas in Hartree Fock approximation . . . . . 12
2.5.2 The Stoner model . . . . . . . . . . . . . . . . . . . . . . 13
2.5.3 An application in FCC iron . . . . . . . . . . . . . . . . . 15
2.6 The band Jahn-Teller eect . . . . . . . . . . . . . . . . . . . . . 16
2.6.1 A one dimensional casethe Peierls distortion . . . . . . 17
2.6.2 A two dimensional casethe square lattice model . . . . 18
3 Band Jahn-Teller eects in Rh
2
MnGe 21
3.1 Introduction to Heusler alloys and related experiments . . . . . . 21
3.2 Calculation details . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.3 Main results of the calculations . . . . . . . . . . . . . . . . . . . 23
3.3.1 The lattice constant and the magnetic moment of the cu-
bic phase . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3.2 Crystal structures at the ground state . . . . . . . . . . . 24
3.3.3 Experimental evidences for the tetragonal phase at low
temperature . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3.4 An interpretation as the band Jahn-Teller eect . . . . . 29
3.3.5 Relativistic eects . . . . . . . . . . . . . . . . . . . . . . 34
4 Magnetic transitions of AFe
2
under pressure 37
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2 Calculational parameters . . . . . . . . . . . . . . . . . . . . . . . 40
4.3 Fixed spin moment schemes . . . . . . . . . . . . . . . . . . . . . 41
4.4 Ground state properties of AFe
2
. . . . . . . . . . . . . . . . . . 41
4.4.1 Common features of the electronic structure . . . . . . . . 41
4.4.2 Specic electronic structures and magnetic moment be-
havior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.4.3 The order of the magnetic transition under high pressure 49
4.5 Relationship between Invar behavior and magnetic transitions . . 53
4.6 Doping eects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
ii CONTENTS
5 Magnetic transitions in CoO under high pressure 61
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.2 A brief introduction to LSDA+U . . . . . . . . . . . . . . . . . . 64
5.3 A Category of insulators . . . . . . . . . . . . . . . . . . . . . . . 65
5.4 Calculational parameters . . . . . . . . . . . . . . . . . . . . . . . 66
5.5 LSDA pictures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.6 LSDA+U pictures . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.6.1 Electronic structures of the ground state . . . . . . . . . . 70
5.6.2 Magnetic transitions in LSDA+U . . . . . . . . . . . . . . 73
5.7 The reason for the magnetic transition . . . . . . . . . . . . . . . 77
5.8 Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6 Summary and Outlook 81
Appendix 91
Acknowledgment 93
List of Figures
2.1 Variant concepts in electronic structure calculations for solids . . 10
2.2 Dierent energy contributions in the free electron model . . . . . 13
2.3 Density of states for FCC iron at a
0
=3.60

A. . . . . . . . . . . . 15
2.4 Plot of Y (m) =
2E
b
mm
versus the magnetic moment (m). . . . . . 16
2.5 Atoms form a linear chain. . . . . . . . . . . . . . . . . . . . . . . 17
2.6 Band structure of the atoms in linear chain before and after the
Peierls distortion . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.7 Atoms form a square lattice with a lattice constant a. . . . . . . 19
2.8 A schematic illustration of dispersion of a p
z
orbital. . . . . . . . 19
3.1 The conventional unit cell of Rh
2
MnGe in a cubic and a tetrag-
onal lattice. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 The contour plot of the energy with respect to the relative volume
and c/a. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3 Energy versus c/a ratios for the tetragonal structure. . . . . . . . 26
3.4 The magnetic moment versus relative volume of dierent phases. 27
3.5 Partial DOS of the Mn-3d state. . . . . . . . . . . . . . . . . . . 27
3.6 The total magnetic moment and the magnetic moment of Mn
versus the c/a ratio. . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.7 Diraction pattern collected at dierent temperatures from T =
20 K to 340 K around 2 = 42
o
. . . . . . . . . . . . . . . . . . . . 29
3.8 Temperature dependence of the lattice parameters in the cubic
and tetragonal phases. . . . . . . . . . . . . . . . . . . . . . . . . 30
3.9 The high symmetry points in the BCT Brillouin Zone. . . . . . . 30
3.10 Angular and magnetic quantum number resolved bands (majority
spin channel) of the cubic phase. . . . . . . . . . . . . . . . . . . 31
3.11 Band structure of the elongated and compressed tetragonal struc-
ture. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.12 Total DOS and DOS of Rh near the Fermi level for dierent c/a
ratios. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.13 Scalar-relativistic (sr) and fully relativistic (fr) band structure of
the cubic phase at V
0
=105.85

A
3
. . . . . . . . . . . . . . . . . . 34
4.1 The unit cell of the cubic Laves phase (C15). . . . . . . . . . . . 40
4.2 DOS of AFe
2
(A=Y, Zr, Hf, and Lu) at the respective equilibrium
volumes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.3 The band structure and DOS of the minority spin channel of ZrFe
2
. 43
iv LIST OF FIGURES
4.4 The corner shared tetrahedrons of Fe and its Kagome net in 111
plane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.5 The fat band and PDOS of the minority state of Fe in ZrFe
2
. . 45
4.6 PDOS of Fe resolved into the irreducible representation (A
1g
and
E
1,2
g
) of the point group D
3d
. . . . . . . . . . . . . . . . . . . . . 46
4.7 Schematic representation of the covalent bonding . . . . . . . . . 47
4.8 Magnetic moment behavior of AFe
2
. . . . . . . . . . . . . . . . 48
4.9 DOS evolution of ZrFe
2
under pressure . . . . . . . . . . . . . . . 49
4.10 Qualitative illustration of rst and the second order quantum
phase transitions. . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.11 The FSM energies of ZrFe
2
at the lattice constant around 6.30

A. 51
4.12 The DOS of nonmagnetic state and ferromagnetic state at a=6.30

A. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.13 The FSM energy of YFe
2
at lattice constant around 6.00

A. . . 53
4.14 The DOS of the nonmagnetic state and ferromagnetic state of
YFe
2
at a=5.98

A. . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.15 A schematic draw of the 2-model. . . . . . . . . . . . . . . . . . 54
4.16 The FSM energy curves of ZrFe
2
near the HS-LS transition regions. 55
4.17 The magnetic moments of Fe
I, II
and Zr versus the atomic con-
centration of Fe. . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.18 The total DOS and the partial DOS of the doped compounds
with composition Zr
0.9
Fe
2.1
at the lattice constant a = 6.825

A. . 58
4.19 The measured and calculated magnetic moments per Fe vs. the
atomic concentration of Zr . . . . . . . . . . . . . . . . . . . . . . 59
5.1 Phase diagram of CoO under pressure summarized from experi-
ments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.2 A schematic illustration of energy levels of two dierent insulators 66
5.3 The unit cell of CoO in the calculation . . . . . . . . . . . . . . . 67
5.4 The LSDA density of 3d states of the Co and 2p states of O. . . 68
5.5 LSDA model DOS of the 3d states of Co in CoO . . . . . . . . . 69
5.6 Magnetic moment of Co and total energy versus the relative vol-
ume. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.7 The results from the LSDA+U with dierent value of U. . . . . . 71
5.8 The Co-3d and O-2p DOS at the equilibrium lattice constant
with U=5 eV. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.9 The LSDA+U model DOS (a) and its variation under pressure (b). 74
5.10 The LSDA+U PDOS at a volume of 90 a.u./f.u., about 70% of
the theoretical equilibrium volume. . . . . . . . . . . . . . . . . . 74
5.11 The Co-3d PDOS at the equilibrium lattice constant resolved into
dierent angular quantum number with respect to the rhombo-
hedral lattice. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.12 The Co-3d PDOS for the two possible low spin solutions. . . . . 76
5.13 The PDOS after the electron with e
g
symmetry ips its spin
decomposed according to the cubic symmetry. . . . . . . . . . . . 76
5.14 The LSDA+U model DOS after the magnetic transition (LS-II). 77
5.15 The total energy and the magnetic moment (inset) versus the
relative volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
LIST OF FIGURES v
5.16 The evolution of the center of gravity of the unoccupied E
g
state
(diamonds) and T
2g
state (circles) under relative volumes with
respect to V
0
=129.4 a.u./f.u.. . . . . . . . . . . . . . . . . . . . . 79
5.17 The enthalpy of CoO under pressure with dierent values of U. . 80
vi LIST OF FIGURES
List of Tables
3.1 Comparison of the calculated lattice constant and magnetic mo-
ments of the cubic structure with experimental results and pre-
vious calculations of the cubic Rh
2
MnGe compound. The cho-
sen method and parameterization of the xc-potential is given in
parenthesis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.1 The experimental values of the lattice constant (a), magnetic
moment (m
s
), and Curie temperature (T
c
) of AFe
2
(A=Y,Zr,Hf,
and Lu) compounds. . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.2 The calculated lattice constants a
FM
in FM state and a
NM
in
NM state, total spin magnetic moments at both theoretical (m
0
)
and experimental (m
exp
) lattice constants of the four compounds. 43
4.3 The spontaneous volume magnetostriction from our LSDA cal-
culations (
s
) and experiments (
exp
s
). . . . . . . . . . . . . . . . 56
4.4 The experimental (a
exp
0
) and theoretical lattice constants (a
LDA
0
)
of Zr
x
Fe
100x
. The experimental values are taken at room tem-
perature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.1 Ground state properties of CoO obtained by LSDA+U calcula-
tions with dierent values of U. Experimental data are given in
the last row. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
viii LIST OF TABLES
Abbreviation
a.u. . . . . . . . . . . . . . . atomic units
ASA . . . . . . . . . . . . . . atomic sphere approximation
ASW . . . . . . . . . . . . . augmented spherical waves
BZ . . . . . . . . . . . . . . . Brillouin zone
CPA . . . . . . . . . . . . . . coherent potential approximation
DFT . . . . . . . . . . . . . density f unctional theory
DOS . . . . . . . . . . . . . . density of states
EOS . . . . . . . . . . . . . . equation of states
FPLO . . . . . . . . . . . . f ull potential nonorthogonal local orbital minimum basis
band structure scheme
FSM . . . . . . . . . . . . . . f ixed spin moment
GGA . . . . . . . . . . . . . generalized gradient approximation
HS . . . . . . . . . . . . . . . high spin
IBZ . . . . . . . . . . . . . . . irreducible Brillouin zone
IR . . . . . . . . . . . . . . . . irreducible representation
LCAO . . . . . . . . . . . . linear combination of atomic orbitals
LDA . . . . . . . . . . . . . . local density approximation
LMTO . . . . . . . . . . . linearized mun-tin orbitals method
LS . . . . . . . . . . . . . . . . low spin
LSDA . . . . . . . . . . . . local spin density approximation
PDOS . . . . . . . . . . . . partial density of states
SIC . . . . . . . . . . . . . . . self interaction correction
Chapter 1
Introduction
Magnetism and structures are two main topics in solid state physics. These two
properties are developed individually but they also interact strongly. Structural
transition leads to changes of magnetism by providing dierent atomic environ-
ment (e.g. dierent coordination of atoms and dierent bond lengths). On the
other hand, changes of magnetic moment unavoidably introduce volume vari-
ations of samples (known as volume magnetostriction). This interaction is of
course technically important. One example is Invar eects, which are believed
to result from the interplay between magnetism and structures, where the in-
crease of volume due to thermal expansions is (partly) compensated by the
demagnetization which causes decrease of the volume. Thus the elastic proper-
ties and/or the thermal expansion coecient in a certain temperature range are
extremely small (invariant volumes), about 10
6
/K in Fe
65
Ni
35
. Behavior of
ferromagnetic shape memory alloys (FSMA) yields examples for the interplay
between magnetism and the structural phase transition (the martensitic phase
transition), where the shape changes due to the movement of twin boundaries
is driven by the rotation of the magnetic moment.
In order to understand the structural trend and magnetism of a solid, a
number of models have be proposed, such as the Stoner model to explain the
itinerant magnetism, the band Jahn-Teller eect to explain some structural
phase transitions, spin and orbital ordering, etc. These models provide us the
physics underlying dierent phenomena. Tight binding approximations in elec-
tronic structure calculation and rigid band model provide some qualitative and
semi-quantitative information about structures, magnetism, and their interac-
tions. At the same time, understanding these phenomena in an ab initio way is
mostly desirable.
Density functional theory (DFT), which was originally invented and devel-
oped by Kohn, Hohenberg, and Sham in the middle of the sixties, provides
a modern tool to study the ground state properties of atoms, molecules, and
solids. It is based on exact theorems, in particular, the Hohenberg-Kohn theo-
rems. Kohn and Sham, later on, put this general theorem into a practical way
where the problem can be solved by a single particle-like Hamiltonian with an
approximated eective potential. The electronic structure calculations provide
a quantitative way to discuss the phase stability at temperature T = 0, and are
even extendable to T ,= 0 with certain model assumptions. They also provide
the microscopic explanation of phase transitions. Bonding characters, energy
2
dispersions or topology of Fermi surfaces etc. all can play a role in the dierent
phase transitions. Magnetic properties are natural outputs of the calculations.
In the non-relativistic case, the magnetic moment is the dierence between the
populations of the spin up and spin down states. Electronic structure calcu-
lations also provide quantitative justication of the model considerations. For
example, in the Stoner model, the density of states and the Stoner parame-
ter are available by DFT calculations. Thus the itinerant magnetism can be
discussed in a more quantitative way. It has been shown that the local spin
density approximation (LSDA) and its extension the general gradient approx-
imation (GGA) are quite successful in understanding itinerant magnetism and
structure trends in metals and intermetallic compounds. The strong electron-
electron interaction seems to be problematic when it is treated in a mean eld
way. For example, some transition metal oxides with partially lled d orbitals
are predicted to be metallic under this local approximation, but are Mott insula-
tors in reality. This strongly correlated state is better treated by a combination
of LSDA and local Coulombic repulsion, the so-called LSDA+U method. This
approach provides us a powerful tool to treat the strongly correlated systems.
On the other hand, developments of high pressure techniques and magnetic
analyzing methods lead to discoveries of many new phenomena. For example, it
was found that the critical temperature of superconductivity may be increased
under pressure, which might be due to the enhancement of electron phonon cou-
pling under pressure. Pressure will surely inuence magnetism. As early as in
1936, Neel gave an estimation of the direct exchange interaction energy between
localized moments located on two neighboring atoms with overlapping orbitals.
This was the rst model to show the large dependence of the molecular eld
on the inter-atomic distance. Taking the homogenous electron gas, which is the
basis of LSDA, as another example, the spin polarized state only exists in a nar-
row electron density range, the electron density parameter r
s
ranging between
75 and 100 Bohr radii. Experiments under pressure provide an unique tool
to characterize materials. Modern experiments can reach hydrostatic pressure
beyond 200 GPa. The pressure changes at least the inter-atomic interaction.
The itinerancy of the electrons will be changed accordingly. This is generally
included in the Stoners model. The improving experimental facilities push us
to extend our theoretical work to high pressures, to understand, and predict
new phenomena.
In this thesis, magnetic and structural transitions of three categories of com-
pounds are investigated by DFT calculations under the LSDA. The thesis is or-
ganized as follows: In Chapter 2, I present a brief introduction to the DFT and
its LSDA. Some model considerations of magnetism and structural transitions
are presented, including the Stoner model, the Peierls distortion, and the band
Jahn-Teller eect in two dimensions. In Chapter 3, two dierent (tetragonal
and cubic) structures of Rh
2
MnGe are investigated and the band Jahn-Teller
eect in this compound is discussed. Four cubic Laves phase compounds (YFe
2
,
ZrFe
2
, HfFe
2
, and LuFe
2
) are investigated in Chapter 4. The interplay between
magnetism and pressures is emphasized. The last material investigated in this
thesis is CoO in Chapter 5, where the pressure induced magnetic transition is
explained by the competition between the ligand eld splitting and the exchange
energy. In the last chapter I give a summary and outlook of the present study.
Chapter 2
Theoretical Background
2.1 Electronic Hamiltonian in a solid [1]
We suppose that properties of a solid can be revealed by nding a wave function
= (R, r) satisfying the Schrodinger equation

H = E with the Hamilto-
nian under the non-relativistic approximation dened as:

H =

H
e
+

H
ion
+

H
elion
+

H
ex
, (2.1)
where

H
e
=

k
p
2
k
2m
+
1
8
0

kk

e
2
[r
k
r
k
[
,

H
ion
=

P
2
i
2M
i
+
1
2

ii

V
ion
(R
i
R
i
)
=

P
2
i
2M
i
+
1
8
0

i=i

Z
i
Z
i

[R
i
R
i
[

H
elion
=

k,i

V
elion
(r
k
, R
i
) =
1
4
0

k,i
Z
i
[e[
[r
k
R
i
[
,
(2.2)
and

H
ex
is due to external elds. The

H
e
,

H
ion
, and

H
elion
are the electron,
ion, and the electron-ion interaction Hamiltonian respectively. The variables
are dened as:

0
, the dielectric constant in vacuum,
M
i
, the static mass of ion i, and m, the static mass of an electron,

P
i
and p
k
, the momentum operator of ion i and electron k, respectively,
e, the electron charge, and Z
i
, the nuclear charge, respectively,
R
i
, the position of ion i, and r
k
, the position of electron k,
and

V
ion
and

V
elion
, the ion-ion interaction and the electron-ion interac-
tion, respectively.
4 2.2 Density functional theory and the Kohn-Sham scheme
Of course, as Dirac had commented, The general theory of quantum me-
chanics is now almost complete. The underlying physical laws necessary for the
mathematical theory of a large part of physics and all of chemistry are thus
completely known, but this equation is too complicate to be solved. Con-
sider a typical solid, there are about 10
23
atoms and the related number of
electrons. The complexity of this problem increase exponentially with the num-
ber of the particles, and soon beyond the capability of storage and computing
of any imaginable computers except for a very small number of particles (less
than several tens). Simplications should be made in order that the problem
is tractable. The rst approximation is the Born-Oppenheimer approximation
(or adiabatic approximation). Considering the large dierence of the rest mass
between electrons and ions (m/M 1/1836), we can assume that the electrons
respond adiabatically to a change in positions of the ions which respond only
slowly to a change in the electron congurations. In this way, the motions of
the electrons and ions are decoupled. As far as we are only concerned about
the motion of the electrons, it is only the instantaneous conguration (R
i
)
of the ions being of interest. The motions of the electrons are governed by the
following Hamiltonian

H
BOA
e
:

H
BOA
e
=

k
p
2
k
2m
+
1
8
0

kk

e
2
[r
k
r
k
[
+

V
elion
(r
k
, R
i
)+const.(R
i
). (2.3)
where const.(R
i
) =

i,j
Z
i
Z
j
/8
0
[R
i
R
j
[ is needed in case of the thermo-
dynamic limit. Thus, we can consider the electronic structure with respect to
the congurations of the ions. Once we have obtained the energy E
e
(R
i
) of
the electronic system by solving the eigen equation:

H
BOA
e
(r) = E
e
(R
i
)(r), (2.4)
we have the Hamiltonian for the ions:

H
BOA
ion
=

P
2
i
2M
i
+
1
2

i,i

V
ion
(R
i
R
i
) +E
e
(R
j
), (2.5)
which is decoupled from the electron coordinates.
Now we have the decoupled electronic and ionic system which can be solved
separately in this adiabatic approximation, but the dimension of the system is
still too large. We still have the huge number of the electrons and the ions.
2.2 The density functional theory and the Kohn-
Sham scheme [2]
We have the electron wave function (x
1
, x
2
, ..., x
N
) where x
i
= (r
i
, s
i
) standing
for the collection of the position r
i
and the spin s
i
of electron i, which depends
on 3N 2 = 6N coordinates. This is untractable for large electron numbers
N. A plausible step to solve the problem for the ground state was realized
by Hohenberg and Kohn [3], who converted the problem of searching for the
electron wave function into the search for a scalar function, the electron density
n(r) in a three dimensional real space. Here we treat the extension to spin
densities n
ss
(r)
2 Theoretical Background 5
Consider the electronic Hamiltonian in the Schrodinger representation
1
,

H
e
[v, M] =
1
2
M

i=1

2
i
+
M

i=1
v
s
i
s

i
(r
i
) +
1
2
M

i=j
w([r
i
r
j
[) (2.6)
for any external spin dependent potential v
s
i
s

i
and integer particle number M.
The function w([r
i
r
j
[) =
1
|r
i
r
j
|
is the electron-electron Coulombic interaction.
As we are dealing with solids, we introduce periodical boundary conditions
which replace the innite position space
3
of electron coordinates by a torus
T
3
with a nite measure.
Dening two sets:
1
N
= v[v L
p
for some ps,

H[v] has a ground state (2.7)
and
/
N
= n(x)[n comes from an N-particle ground state, (2.8)
where v L
p
means that [[v[[
p
= [
_
dx[v(x)[
p
]
1/p
, 1 p < is nite. The
basic Hohenberg and Kohn theorem reads:
Theorem 1 (H-K Theorem) v(x)mod(const.) 1
N
is a unique function of
the ground state density n(x).
In the original paper by Hohenberg and Kohn [3], the 1
N
is conned to
a non-degenerated ground state, and n(x) also belongs to a non-degenerated
ground state, but degeneracy of the ground state is more common in electronic
systems. The theorem was generalized by the following argument by Lieb [4]
where the non-degeneracy is not required.
We dene the density operator which admits ensemble states as:
=

K
[
K
)g
K

K
[, (2.9)
where g
K
0 and

K
g
K
= 1 and [
K
) is for some pure states, which may be
expanded into a xed orthnormal set of eigenstates of particle number operator

N:

N[
M
K
) = [
M
K
)M,
[
K
) =

M
[
M
K
)C
K
M
,

M
[C
K
M
[
2
= 1. (2.10)
The expectation value of particle density
n
ss
(r) = tr( n ) =

K,M
p
K
M
n
M
K,ss
(r), (2.11)
where p
K
M
def
= g
K
[C
K
M
[
2
and n
M
K,ss
(r) = r, s[
M
K
)
M
K
[r, s

). The total particle

number reads
N = tr(

N ) =

s
_
n
ss
(r)dr
=

K,M
p
K
M

s
_
n
M
K,ss
(r)dr
=

K,M
p
K
M
M.
(2.12)
1
from now on, atomic units (a.u.) will be used, which means we put e = m = = 1 and

0
=
1
4
. The superscript BOA will be omitted.
6 2.2 Density functional theory and the Kohn-Sham scheme
The total energy is
tr(

H ) =

K,M
p
K
M

M
K
[

H[v, M][
M
K
), (2.13)
where p
K
M
0,

K,M
p
K
M
= 1. Now, the ground state energy (E) as a functional
of the external potential v, and a function of real particle number N can be
dened as:
E[v, N]
def
= inf

tr(

H )[tr(

N ) = N
= inf
p
K
M

K,M
p
K
M

M
K
[

H[v, M][
M
K
)[

K,M
p
K
M
M = N,
(2.14)
where

N is the particle number operator,

N[
M
K
) = [
M
K
)M. It can be shown
that E[v, N] has the following properties
2
:
1. E[v +const, N] = E[v] +N const (gauge invariance),
2. for xed v, E[v, N] is convex in N, and
3. for xed N, E[v, N] is concave in v.
Starting with the convexity of E[v, N] in N, a Legendre transform

G[v, ] is
dened with a pair of transformations:

G[v, ] = sup
N
N E[v, N],
E[v, N] = sup

N

G[v, ].
(2.15)
Because of the above gauge invariance (Property 1) of E[v, N],

G[v, ] =

G[v , 0]
def
= G[v

]. (2.16)
Then the duality relations of Equ. (2.15) are simplied to
G[v] = inf
N
E[v, N],
E[v, N] = sup

N G[v ].
(2.17)
Since G[v] is convex in v, it can be back and forth Legendre transformed. If we
introduce n as a dual variable to v, then
H[n] = sup
v
(n[v) G[v]
G[v] = sup
n
(v[ n) H[n] = inf
n
H[n] + (v[n)
(2.18)
Insert G[v] of Equ. (2.17) into H[n], we have
H[n] = sup
v
(n[v) + inf
N
E[v, N]
inf
N
sup
v
E[v, N] (n[v)
= inf
N
F[n, N],
(2.19)
2
In this section, we just present conclusions without mathematical proofs in order to outline
the logical base of this theory.
2 Theoretical Background 7
where we dene (rst introduced by Lieb [4])
F[n, N] = sup
v
E[v, N] (n[v) (2.20)
as a density functional. Then the inverse Legendre transformation to Equ.
(2.20) leads to
E[v, N] = sup

N + inf
n
H[n] + (v [n)
= inf
n
H[n] + (v[n)[(1[n) = N.
(2.21)
In Liebs denition Equ. (2.20) of the density functional, the domain of n
is A = L
3
(T
3
), and the dual variable v is limited to A

= L
3/2
(T
3
), where A
and A

are reexive (A

= A).
In the original theorem by Hohnberg and Kohn, the functional F[n] is dened
by
F
HK
[n] = E[v[n]]
_
dxv[n]n, n /
N
. (2.22)
This raises the problem of v-representability (VR).
2.3 The Kohn-Sham equations
Since H[n] as a Legendre transformation is lower semicontinuous on A, it has
a non-empty subdierential for n A. Moreover, a convex function on a nor-
malized space has a derivative, if and only if the subdierenetial consists of a
unique element of the dual space. Hence, if v A

is uniquely dened in the

theory for some given n A for which H[n] is nite, then H[n] has a functional
derivative equal to v, that is

H
n
= v , (2.23)
for non-integer N = (n[1). For integer N, the derivative may jump by a nite
value, constant in r-space.
We put
H[n] = K[n] +L[n], (2.24)
where
K[n] = min

i
,n
i
k[
i
, n
i
][

i
n
i

i
= n, 0 n
i
1
i
[
j
) =
ij
(2.25)
is the orbital dependent functional, and
L[n] =
_
d
3
rn(r)l[n
ss
(r), n
ss
(r)], (2.26)
is the explicit density functional, where
i
are the Kohn-Sham orbitals with an
occupation number n
i
.
With a suitable chosen orbital functional k and function l, one arrives at the
Kohn-Sham equation
(

k +v +v
L
)
i
=
i

i
(2.27)
8 2.3 The Kohn-Sham equations
with

k :
k

i
=

k
i
n
i
,

k Hermitian, (2.28)
and the local Kohn-Sham potential
v
L
=
L
n
, (2.29)
where
i
is the energy of the Kohn-Sham orbital
i
. Especially if
K[n] =
_
T
DM
[n] for n X, n(x) 0
+ elsewhere
L[n] =

ss

_
(v
H
s

s
+v
xc
s

s
+v
s

s
)n
ss
dr,
(2.30)
where T
DM
[n] is the kinetic energy functional of noninteracting electrons:
T
DM
[n]
def
= inf

tr(

T)[ n
= min

i
,n
i

i
n
i
_
dx

i
(
_
2
2
)
i
[

i
n
i
[
i
[
2
= n, 0 n
i
1,
i
[
j
) =
ij
,
(2.31)
v
H
ss
is the Hartree potential: v
H
ss
(r) =
_
d
3
r

n
ss
(r

)w([r r

[), v
xc
ss
is the ex-
change correlation potential and v
ss
is the external potential. Then one arrives
at the Kohn-Sham equation
3
:
(

2
2
+v
eff
s
)
i
=
i

i
,
v
eff
s
= v +v
H
+v
xc
s
.
(2.32)
The v
xc
s
is dened as:
v
xc
s
(r)
def
=

n
s
(r)
E
xc
[n(r)]. (2.33)
The simplest successful approximation to this functional is the local spin density
approximation (LSDA) [5], where the functional dependence is taken to be the
same as in homogenous electron liquid:
E
LDA
xc
[n
+,
(r)] =
_

hom
xc
(n
+,
(r))n(r)dr. (2.34)
This spin dependent xc-functional can be obtained by tting the results
from quantum Monte Carlo simulations of the homogenous electron gas [6].
The presently most precise t is obtained by Perdew and Wang [7]. In this
tting the correlation energy has the form:

c
(r
s
, ) = (r
s
, 0) +
c
(r
s
)
f()
f

(0)
(1
4
)
+ [
c
(r
s
, 1)
c
(r
s
, 0)]f()
4
,
f() =
[(1 +)
4/3
+ (1 )
4/3
2]
2
4/3
2
,
(2.35)
3
Here, we deal with the collinear spin polarization which means that we can nd a global
unitary transformation so that the 22 density matrix (n
ss
) and potential matrix (v
ss
) can
be simultaneously diagonized into n
s
and v
s
, respectively.
2 Theoretical Background 9
and the exchange contribution is

x
(r
s
, ) =
3
4r
s
[
9
4
]
1/3
[(1 +)
4/3
+ (1 )
4/3
]/2, (2.36)
where = (n
+
n

)/((n
+
+ n

)) is the relative spin polarization, r

s
=
[3/4(n
+
+n

)]
1/3
is the density parameter, and
c
(r
s
) =
2

c
(r
s
, = 0)/
2
is the spin stiness. This version of exchange correlation will be used in all of
our LSDA calculations in this thesis.
2.4 Basics of the Full Potential Local Orbital
band structure code (FPLO)
If we choose the periodic boundary conditions for the wave function and consider
an innite crystal with periodically arranged atoms, the electrons feel a periodic
potential V (r) = V (r +R) where R is a Bravais lattice vector. This means that
the Hamiltonian of the electrons has a translational symmetry. The electron
wave function in the crystal has the following property:
Theorem 2 (Bloch Theorem) The eigenstates
nk
(r) of the one-electron
Hamiltonian

H =
2
/2+V (r), where V (r +R) = V (r) for all R in a Bravais
lattice, can be chosen to have the form of a plane wave times a function with
the periodicity of the Bravais lattice:

nk
(r) = e
ikr
u
nk
(r), (2.37)
where
u
nk
(r +R) = u
nk
(r). (2.38)
Because of the boundary condition (BC), let R =

i
a
i
L
i
,

nk
(r +

i
a
i
L
i
) =
nk
(r), (2.39)
where a
i
is the i-th basis vector of the Bravais lattice and L
i
is the number of
the cells along a
i
. The allowed k-points are determined by:
k =

i
(
2m
i
L
i
mod 2)k
0
i
(2.40)
where m
i
Z, the set of integers, 0 m
i
< L
i
and k
0
i
is the basis vector in the
reciprocal space, satisfying k
0
i
a
j
= 2
ij
.
By this periodic BC, we actually convert our problem domain of
3
with
innite measure to a torus T
3
with nite measure. This BC simplies our
treatment of the Coulombic system. For suciently large L
i
s, the physical
properties are not altered by this BC. We can use the mathematically closed
DFT introduced before to treat the electronic structure in the crystal.
In order to solve the Kohn-Sham equations, further approximations in the
DFT calculations must be made. The dierent parts of this set of self-consistent
equations can be approximated in dierent ways. These dierent ways are
summarized in a diagram as in Figure 2.1. Most of the electronic structure
calculations in this thesis are performed with the Full Potential Local Orbital
minimum basis band structure code FPLO-5 [8]. This code has the following
features:
10 2.4 Basics of FPLO code
F
i
g
u
r
e
2
.
1
:
V
a
r
i
a
n
t
c
o
n
c
e
p
t
s
i
n
e
l
e
c
t
r
o
n
i
c
s
t
r
u
c
t
u
r
e
c
a
l
c
u
l
a
t
i
o
n
s
f
o
r
s
o
l
i
d
s
.
T
h
e
f
e
a
t
u
r
e
s
i
n
i
t
a
l
i
c
s
a
r
e
c
u
r
r
e
n
t
l
y
i
m
p
l
e
m
e
n
t
e
d
i
n
F
P
L
O
-
5
.
2 Theoretical Background 11
It is a full potential code, which means, no shape approximation of the
potential is applied. This removes the previous inaccuracy in the results
introduced by ASA, or MT approximations to the potential.
It is an optimized local basis code
4
. Numerical atomic basis states are used
as the basis to expand the Kohn-Sham orbitals. The relatively small basis
set accelerates the calculation. The completeness of the relatively small
number of basis (frequently the minimum basis) is improved by optimizing
the basis.
It is an all electron code, which does not use a pseudopotential.
The basic idea is as follows: The Bloch wave function indexed by (n, k) is ex-
pressed by a superposition of nonorthogonal local orbitals r[R, sL) =
sL
(r Rs)
centered at position R+s with quantum number L ( L = (, l, m), : princi-
ple quantum numbers, l: angular quantum numbers, m: magnetic quantum
numbers):

kn
(r) =

RsL

sL
(r Rs)C
Ls,kn
e
ik(R+s)
. (2.41)
Putting this trial function into the Kohn-Sham equation, multiplying 0s

[r)
from the left on both sides and integrating, we obtain the secular equation:
HC = SC
with
H
s

,sL
=

R
0s

H[RsL)e
ik(R+ss

)
,
S
s

,sL
=

R
0s

[RsL)e
ik(R+ss

)
.
(2.42)
The density (n
s
(r)) obtained as a double lattice sum is reexpressed as a single
lattice sum:
n(r) =
occ.

k,n

kn
(r)
kn
(r) =

Rs
n
s
(r Rs) (2.43)
and the potential (v(r)) is calculated under the LSDA and expressed into local
contributions (v
s
(r)):
v(r) =

Rs
v
s
(r Rs) (2.44)
The exchange correlation potential is again reexpressed as a lattice sum in a
similar way by a partition of unity on the lattice. trick is introducing a parti-
tion function The potential re-enters the Kohn-Sham equations until the self-
consistency is realized.
In this thesis, the notation of the basis in the calculations are like, for ex-
ample:
Fe : 3sp / 3d(4sp) + 4d
They are organized as:
Element : semi core / valence + polarization states
4
The present state of the art is that this optimization is done before calculations. It is
implemented in the new generation of FPLO.
12 2.5 Model considerations
The notation indicates the characters of the used basis orbitals above the core.
For details see http://www.fplo.de.
2.5 Model considerations
It has long been recognized that magnetism and structures (the volume as well
as the crystal symmetry) are related. This can be revealed by the following
model considerations.
2.5.1 Free electron gas in Hartree Fock approximation
Considering free electron gas in a nite volume V , the states are described by
plane waves with wave vector k:

s
k
(r) = (
1

V
e
ikr
)
s
, (2.45)
where
s
is the spin eigenfunction for spin s. Because of the isotropic nature
of the system, we use k instead of k from now on. Using these states to form
the Slater determinant and requiring double occupation of states with k k
F
,
where k
F
is the Fermi wave vector, we arrive at the Hartree-Fock equation for
free electrons, where the energy dispersion relation [9] is:
(k) =
k
2
2

2

k
F
F(
k
k
F
) (2.46)
and
F(x) =
1
2
+
1 x
2
4x
ln [
1 +x
1 x
[. (2.47)
Then the total energy is
E =

k<k
F
(k)
= N[
3
5
k
2
F
2

3
4
k
F
].
(2.48)
The rst term is the kinetic energy E
k
of non-interacting electrons and the
second is the exchange energy E
x
. If we put in the electron density n =
k
3
F
3
2
,
then we obtain dierent function dependence of them on the electron density n,
where E
k
n
2
3
, and E
x
n
1
3
. If we allow dierent Fermi wave vector k
()
F
for the spin up and spin down electrons, then it can be easily shown that for a
completely polarized system, where all the electrons are spin up, k

F
= 2
1/3
k
F
,
the energy amounts to
E
P
= N[
3
5
2
2/3
(k

F
)
2

3
2
2
1/3
k

F
]. (2.49)
The kinetic energy is increased by 2
2/3
and the magnitude of the negative ex-
change energy is lowered by 2
1/3
. A plot of kinetic energy, magnitude of the ex-
change energy, and total energy of the fully polarized and nonpolarized systems
2 Theoretical Background 13
P
0.352
k
(
a
u
)
Figure 2.2: The total energy E
N,P
, the kinetic energy E
N,P
k
and the magnitude
of the exchange energy E
N,P
x
of the fully polarized (P) and nonpolarized (N)
electron systems as a function of the Fermi vector k
F
.
is shown in Figure 2.2. It can be seen that E
P
can be smaller than the nonpo-
larized case E
N
. This happens for small k
F
, i.e. k
F
k
0
F
=
5
2
1
2
1/3
+1
0.352.
Thus, for a low electron density it is expected that the system is spin polar-
ized. This simple model shows that if the volume is shrunk the electron density
will be higher, and the system loses its magnetization. The correlation energy
reduces the tendency to itinerant ferromagnetism. A more exact treatment of
the homogenous electron gas was done by Ceperley and Alder [6] by quantum
Monte Carlo simulations. It was found that the polarized (ferromagnetic) Fermi
liquid is stable between r
s
= 75 and r
s
= 100 where r
s

= 1.92/k
F
is the electron
density parameter. Below that it is normal paramagnetic Fermi uid, and above
that, the electrons crystalize into a Wigner crystal.
2.5.2 The Stoner model
In the original proposal by E. C. Stoner [10], this model accounts for the itinerant
magnetism both at T = 0, and T ,= 0. Here we concentrated on the model at
T = 0, because it provides a criterion for the existence of ferromagnetism in
the ground state. If we have the density of states D(E) by a non-polarized
calculation, we can produce a ferromagnetic state by a rigid shift of the spin-up
and spin-down states [11] as
D

(E) = D(E + E

),
D

(E) = D(E + E

).
(2.50)
where E

> 0 and E

< 0. Here, D

(E) and D

(E) are the densities of

states for spin-up and spin-down electron subbands, respectively. The energy
shifts E

and E

of D

(E) and D

(E) with respect to D(E) are constrained

by the charge conservation
_
0
E

dED(E) =
_
E

0
dED(E), (2.51)
where the Fermi level E
F
is put at E = 0. The spin magnetic moment is given
by =
1
2
m
B
, where m is the number of unpaired electrons and
B
is the Bohr
14 2.5 Model considerations
magneton. The m can be obtained by counting the electrons in the spin-up and
spin-down subbands, taking care of the charge neutrality given by Equ. (2.51),
as
m =
_
0

[D

(E) D

(E)]dE
= 2
_
E

0
D(E)dE
(2.52)
The energy dierence between a nonmagnetic and a ferromagnetic state consists
of two parts. The rst part is the increase of band energy E
b
due to the
reoccupation of states near the Fermi level. The second part is the decrease of
the exchange energy contribution E
ex
which depends on the Stoner parameter
I. This energy should only have even order terms of m because of symmetry. In
the lowest order approximation, it is proportional to m
2
. Hence, we can write
the total energy dierence as
E
mag
= E
b
+ E
ex
=
_
E

ED(E)dE
1
4
Im
2
.
(2.53)
The instability of the system with respect to onset of ferromagnetism is

2
E
mag
m
2
< 0. (2.54)
From equation (2.51),
D(E

)
E

= D(E

). (2.55)
The rst derivative of the band energy with respect to m reads,
E
b
m
=
E
b
E

m
+
E
b
E

m
=
E

2
,
(2.56)
and the second derivative is

2
E
b
m
2
=
(E

)
2m
=
1
4
(
1
D(E

)
+
1
D(E

)
).
(2.57)
If we dene an average density of state by
1

D(m)
=
1
2
(
1
D(E

)
+
1
D(E

)
), (2.58)
the instability of the nonmagnetic state according to Equ. (2.54) is given by
I

D(0) > 0, (2.59)
2 Theoretical Background 15
0
0
1
1
2
2
D
O
S
.
F
e
4 8 4
Figure 2.3: Density of states for FCC iron at a
0
=3.60

A. The Fermi level is
indicated by the vertical dashed line.
where

D(0) is the averaged nonmagnetic DOS at the Fermi level. This is the
famous Stoner criterion for ferromagnetism at T = 0.
The requirement of stationarity of E
mag
,
E
mag
/m = 0, (2.60)
gives, apart from the trivial solution m = 0, a possible magnetic solution. It
is stable if
2
E
mag
/m
2
> 0. Substituting Equation (2.53) into (2.60), Equ.
(2.60) can be rewritten as
Y (m)
def
=
2E
b
mm
= I =

D
1
(m), (2.61)
and the stationarity requires Y

(m) > 0.
From Equ. (2.58), we can see that if D(E

) and D(E

) are both high for

several ms, we can have several magnetic solutions satisfying Equ. (2.61) [11].
In fact, for each magnetic solution, the Fermi level situating in the valley of
the DOS is mostly favorable for multiple magnetic solutions as will be discussed
in details in Chapter 4.
2.5.3 An application in FCC iron
Here we show an application of the above consideration to face centered cu-
bic (FCC) iron. It was already shown from xed spin moment calculations by
Moruzzi [12] that there are multiple solutions with dierent magnetic moments.
Here we use the avor of equations in the above section to discuss the solutions.
In Figure 2.3, the density of states for the nonmagnetic solution at a lattice con-
stant of a
0
= 3.60

A is shown. In order to nd the magnetic solution, we plot
Y (m) =
2E
b
mm
in Figure 2.4, where E
b
and its derivatives are calculated numeri-
cally by equations from Equ. (2.50) to the rst term in the r.h.s. of Equ. (2.53).
Then the magnetic solutions are obtained by the cross points of a horizontal line
with coordinate equal to the Stoner parameter I and the curve. The stability of
the solutions are determined by the sign of
2
E
mag
/m
2
. As mentioned above,
the stable solutions are the crossing points for which Y

(m) > 0. For FCC iron,

16 2.6 The band Jahn-Teller eect
Figure 2.4: Plot of Y (m) =
2E
b
mm
versus the magnetic moment (m). The
short vertical lines indicate the magnetic solutions at these lattice constants by
self-consistent spin polarized calculations.
the curve of Y (m) corresponding to dierent lattice constant yielding magnetic
solutions: a
0
=3.60, and 3.75

A, respectively are shown in Figure 2.4. We have
also performed self-consistent calculations to nd the magnetic solutions. The
calculated magnetic moments for the dierent lattice constants are shown by the
short red vertical lines on the corresponding curves. It can be seen that results
from our simple model agree with the self-consistent spin polarized calculation.
The horizonal line indicates the I = 0.93 eV [13] from the most recent results.
Earlier result of shows I = 0.92 eV [14]. This scatter of I does not inuence our
discussions here. It is obvious that the nonmagnetic, low spin ( 1.2
B
) and
high spin ( 2.6
B
) solutions can be obtained from the curves with the lattice
constant a
0
=3.60

A. For the other curve where the lattice constant is expanded
by 4%, there is only one magnetic solution with a larger magnetic moment
( 2.7
B
). This agrees semi-quantitatively with the results of Morruzi.
This rigid band consideration provides a birds eye view of the multiple
magnetic solutions with minor calculation eorts. This crude estimation can
give us a rst answer to the question whether there are several magnetic states
existing in the system or not and possible magnetic moments can be estimated.
2.6 The band Jahn-Teller eect
The Jahn-Teller eect is the intrinsic instability of an electronically degenerate
complex against distortions that remove the degeneracy. It was rst predicted
as a very general phenomenon in 1937 by Jahn and Teller [15]. In solids, there
exists a phenomenon equivalent to the Jahn-Teller eect (namely, the band
Jahn-Teller eect): When there are degenerated bands around the Fermi level,
if we allow a slight distortion of the lattice which lowers the symmetry of the
crystal, this distortion will lead to a split of the bands, but does not change their
center of gravity. Thus the distortion is favored if the gain in band energy by
occupation of the lower one of the split bands overcompensates the loss of elastic
energy due to the distortion. We illustrate two model systems to understand
2 Theoretical Background 17
Figure 2.5: Atoms form a linear chain with a regular distance of a. If every
second atom is displaced by , the new periodicity is 2a.
E
0
E
k


-

F
0
k -k
+
F F

|UQ|
|UQ|
Figure 2.6: A schematic band structure of a linear chain of equidistant atoms
(the red broken line curve), and that after the Peierls distortion (the solid black
curves). The Fermi level is shown by a dashed horizontal line.
this eect.
2.6.1 a one dimensional casethe Peierls distortion [16]
Consider a linear chain of atoms with a large length of L, with regular spacing
a as shown in Figure 2.5. For an odd number of electrons, the valence band
will be half lled. The band is assumed to have a shape like the red broken line
curves as in Figure 2.6. The Fermi wave number is k
F
= /2a if there is only
one electron per atom.
If every second atom is displaced by a small distance , this doubles the
periodicity of the chain, and the potential acquires a Fourier component of
wave number Q = /a, which in this case is equal to 2k
F
. Then the states with
k and k Q will be coupled by this potential. The new state can be assumed to
be a linear combination of them. According to the nearly free electron model,
the secular equation is written as:

0
k
U
Q
U

Q

0
kQ

= 0, (2.62)
where
0
k
is the undisturbed dispersion, and U
Q
is the Fourier component of the
crystal potential with wave vector Q =

a
. The two roots of the secular equation
are

(k) =
1
2
(
0
k
+
0
kQ
) [(

0
k

0
kQ
2
)
2
+[U
Q
[
2
]
1/2
, (2.63)
as displayed by the solid curves in Figure 2.6.
18 2.6 The band Jahn-Teller eect
Taking the free electron dispersions:

0
k
=
1
2
k
2
,
0
kQ
=
1
2
(k Q)
2
, (2.64)
At k =

2a
,
0
k
=
0
kQ
, then

=
0
k
[U
Q
[. Namely, at this point, the eigen
energy is split by an amount of 2[U
Q
[ as shown in the gure.
The band energy per unit volume (E
b
) of the distorted system can be eval-
uated by integration from /2a to +/2a, multiplied by L/2, the number of
electron states per unit volume in k-space:
E
b
=
1
L
L
2
2
_
k
F
0
dk

(k)
=
1

_
k
F
0
dk

(k)
=
1

_
k
F
0
dk
1
4
[k
2
+ (k Q)
2
]
k
F
2

(k
Q
2
)
2
+ (
2U
Q
Q
)
2

=
1
6
k
3
F

1

z
k
2
F
4
+
U
2
Q
k
F
ln
z +k
F
/2
z k
F
/2

(2.65)
where z =
_
k
2
F
/4 + 4U
2
Q
/k
2
F
. When the displacement of the atoms () is small,
we can expect that U
Q
is small and proportional to . Then we have a term in
the reduction of the band energy like
E
b
U
2
Q
ln U
Q

2
ln . (2.66)
The tiny displacement will increase the elastic energy by E
el
which is only
proportional to
2
. Thus we can see that if the distortion is small, the logarith-
mic term in the reduction of the band energy E
b
dominates. The distortion
is then favorable for the system. This means that the one dimensional atomic
chain with half occupied bands exhibits a spontaneous distortion which intro-
duces a new periodicity, frequently called a charge density wave (CDW).
2.6.2 A two dimensional casethe square lattice model
The second model system we consider is a square atomic lattice as shown in
Figure 2.7. For simplicity, only the p
z
orbitals with a nearest neighbor inter-
action is included. The dispersion of the related Bloch states is schematically
illustrated in Figure 2.8(a).
In the undistorted square net, a 90
o
rotation around z axis is a symmetry
operation. The wave vectors k =

a
[1, 0] and k =

a
[0, 1] transform into each
other by this operation. The p
z
based crystal orbitals at these two k-points
belong to the same irreducible representation, and are degenerate as shown in
Figure 2.8(a). But for a rectangular net, a 90
o
rotation is not a symmetry
operation any longer. Thus these two orbitals, at k =

a
[1, 0] and k =

a
[0, 1],
are no longer degenerate as shown in Figure 2.8(b). The state at [0,1] has lower
energy in our example. The dierence of the eigenvalues is . If we assume
that each orbital at [1,0] and [0,1] is half lled, in the square lattice the electron
equally occupies these states, as in Figure 2.8(a). In the rectangle lattice the
2 Theoretical Background 19
x
y
o
a
a
Figure 2.7: Atoms form a square lattice with a lattice constant a.
[0,0] [1,0] [0,1] [1,1]
k
E(k)
p
z
+
-
+
-
+
-
+
-
+
-
-
+
-
+
+
-
-
+
+
-
-
+
-
+
+
-
-
+
+
-
+
-
(a) A cartoon band structure of the p
z
band in a square lattice. The states at [1,0]
and [0,1] are degenerate.
(b) A cartoon band structure of the p
z
band in a in a rectangle lattice. The de-
generacy of the orbital at [1,0] and [0,1] is
lifted by an amount of after the distor-
tion.
Figure 2.8: A schematic illustration of the dispersion of the p
z
band along the
high symmetric lines in the reciprocal space of the square and rectangle lattices.
The + and - indicate the phase of the orbitals. Electrons are denoted by
the up arrows.
20 2.6 The band Jahn-Teller eect
degeneracy at [1,0] and [0,1] is removed. Thus, the state at [0,1] will be double
occupied and the orbital at [1,0] will be empty, as denoted in Figure 2.8(b). This
reduces the energy by ln per electron. Thus the half lling of the p
z
orbital
will produce the rectangle lattice distortion stabilized by the band Jahn-Teller
eect. The stability of body centered cubic and face centered cubic structural
phases in some transitional metals and their alloys were recently analyzed from
this perspective [17].
From the previous two pedagogical examples, we see that the band Jahn-
Teller eect happens in high symmetric phases. A lot of interesting phenomena
are proposed to be connected with the band Jahn-Teller eect, such as the
charge density wave, orbital ordering or some structural phase transitions. In
this thesis, stabilization of tetragonal phases of Rh
2
MnGe against cubic phases
is proposed to result from the band Jahn-Teller eect as we will discuss in
Chapter 3.
Chapter 3
Band Jahn-Teller eects in
Rh
2
MnGe
3.1 Introduction to Heusler alloys and related
experiments
Magnetic compounds with a Heusler structure receive a lot of research interests
because of their unique characteristics. For example, Ni
2
MnGa is a typical
magnetic shape memory (MSM) alloy which exhibits large changes in shape and
size in an applied magnetic eld. This deformation can be as large as 10%. It
can nd its applications in actuators and sensors. A martensitic transformation
from a high symmetry cubic (austenitic) phase at higher temperatures to a
lower symmetry phase, for example, a tetragonal (martensitic) phase at lower
temperatures is thought to be a precondition for this large shape or volume
change. Both experimental and theoretical works were conducted to study this
eect [18]. The reason for the phase transition was proposed to be phonon
softening, which in turn originates from Fermi surface nesting [19]. Another
important characteristic of some ferromagnetic Heusler alloys is half-metallicity,
which means that the density of states (DOS) of one spin channel is zero while
the DOS of the other channel is nite. Co
2
MnX (X=Ge, Si, Sn, etc.) are such
compounds. They have the potential to be used as spin injection materials in
spintronics, although at present, the experimental results are not so promising:
the spin injection eciency is much lower than the theoretical prediction [20].
The reason is that the local density of states at the Fermi level close to the
interface is quite susceptible to defects such as atomic disorders, interfaces and
surface segregation, etc.
As a member in the family of Heusler alloy, Rh
2
MnGe with an L2
1
structure
was rst reported, as far as I know, by Hames et al. [21] some thirty years ago.
Later on, extended and systematic experimental works on Rh
2
MnX (X is Al,
Ga, In, Tl, Sn, Pb, and Sb) were reported by Suits [22]. It was found that for
most X in group IV B they are ordered in the L2
1
phase, but for X in group III
B they are crystallized in the disordered B2 phase. The former compounds are
ferromagnetic with a higher Curie temperature above room temperature and
exhibit larger magnetic moments at low temperatures compared with the latter
22 3.2 Calculation details
ones. The Sn hyperne eld in Rh
2
MnGe
0.98
119
Sn
0.02
and Rh
2
MnSn measured
by Dunlap et al. [23] suggested that it is more closely related to the Co based
alloy than those with Ni, Pd or Cu at the Rh site in the sense that Rh atoms
do carry magnetic moments comparable with the Co moments in the Co
2
YZ
Heusler alloys.
Density functional theory (DFT) calculations of this material were reported
by Pugacheva [24], concentrating on the eects of atom substitutions and atomic
disorders. It was shown that the magnetic moment deteriorates by disorder
which may be tuned by the heat treatment. Recently, electronic structures of
a number of Heusler alloys were calculated by Galanakis et al. [25], where the
Slater-Pauling behavior of the magnetic moment in most of these materials was
shown. The Rh
2
MnGe compound does not fall on the Slater-Pauling curve,
because of the nonvanishing DOS of the down spin channel, namely, it is not
a half-metal. All of the electronic structure calculations published, hitherto,
were devoted to the cubic phase of Rh
2
MnGe. Recent experiments by Adachi
et al. [26] showed some indications of a phase transition in Rh
2
MnGe under
hydrostatic pressure of about 0.6 GPa, but the structure of the high pressure
phase was not identied. The density of states of the cubic structure obtained
by Pugacheva [24] and Galanakis [25] shows that the Fermi level is situated at
a peak (van Hove singularity) of the DOS, which implies that the cubic phase
may not be stable at zero temperature. Although the martensitic transition
is widely reported in the Heusler compounds by DFT calculations, not all of
the compounds have the tetragonal structure in the ground state [18]. As re-
ported by Ayuela et al. [18], among Co
2
MnGa, Ni
2
MnAl, Ni
2
MnGa, Ni
2
MnSn,
Ni
2
CoGa, and Fe
2
CoGa, the cubic structure is stable in all Mn alloys expect
Ni
2
MnGa. This is related to the MSM eect in this compound. The phonon
dispersions of Ni
2
MnGa(Ge, Al) and Co
2
MnGa(Ge) were calculated by Zayak
et al. [27]. The softening of the TA
2
mode is proposed to be the reason for
the structure instability. An interesting question is whether the temperature in-
duced structural phase transition in Ni
2
MnGa also takes place in Rh
2
MnGe? A
rst trial to answer this question is to conduct DFT calculations and probe some
possible phases at T=0. Intrigued by the experimental results of Ni
2
MnGa, the
rst non-cubic structure that comes to our mind is the tetragonal phase, which
is obtained by extension or compression of the cubic lattice in one direction
while the other two respond accordingly. We thus carried out density functional
calculations of Rh
2
MnGe in order to study the ground state properties, includ-
ing possible distorted phases, their magnetic moments, and their relations to
electronic structures.
3.2 Calculation details
The Rh
2
MnGe compound crystallizes at room-temperature in the L2
1
structure
which consists of four face centered cubic (FCC) sublattices. The space group
is Fm

3m (No. 225) with the following Wycko positions: Rh: 8c(1/4, 1/4,
1/4), Mn: 4b(1/2, 1/2, 1/2), and Ge: 4a(0, 0, 0). The tetragonal distortion
that results in a change of the space group to I/4mmm (No. 139) with Wycko
positions: Rh: 4d(0, 1/2, 1/4), Mn: 2b(0, 0, 1/2), and Ge: 2a(0, 0, 0). The
results for the cubic structure were also obtained by using the symmetry of the
tetragonal structure, and the consistency of the calculations has been checked.
3 Band Jahn-Teller eects in Rh
2
MnGe 23
Rh Mn
Ge
a
0
(a) the cubic lattice
Rh Mn
Ge
a
c
(b) the tetragonal lattice
Figure 3.1: The conventional unit cell of Rh
2
MnGe in a cubic (a) and a tetrago-
nal (b) lattice. The tetragonal lattice is rotated by 45

around the vertical axis

as compared to the cubic lattice.
Table 3.1: Comparison of the calculated lattice constant and magnetic moments
of the cubic structure with experimental results and previous calculations of the
cubic Rh
2
MnGe compound. The chosen method and parameterization of the
xc-potential is given in parenthesis.
Present work Calculation [24] Experiments
FPLO (PW92) LMTO (vBH)
a
0
(

A) 5.97 6.102 5.99 [21], 6.030 [22]

m
tot
(
B
) 4.51 4.47 4.30 [21], 4.62(4) [22]
m
Rh
(
B
) 0.36 0.39
m
Mn
(
B
) 3.80 3.69
m
Ge
(
B
) -0.016 0.004
In this case, the tetragonal lattice constants (a, c) equal (a
0
/

2, a
0
), where a
0
is the cubic lattice constant. The cubic and tetragonal structures are shown in
Fig. 3.1 (a) and (b), respectively. The number of k-points in the irreducible
part of the Brillouin zone (IBZ) is 67685 in the calculations. The basis is chosen
as follows: Rh: 4sp/4d5sp+5d, Mn: 3sp/3d4sp+4d, Ge: 3spd/4spd. The local
spin density approximation is applied and the parameterization of the exchange
and correlation potential by Perdew and Wang [7] was chosen. The convergence
of the self-consistent iterations is checked with respect to both the density (10
6
in code specic units) and the total energy (10
8
Hartree).
3.3 Main results of the calculations
3.3.1 The lattice constant and the magnetic moment of
the cubic phase
The calculated lattice constant (a
0
), total (m
tot
) and local (m
Rh
, m
Mn
, and
m
Ge
) magnetic moments of the cubic structure at zero temperature are listed
in Table 3.1 and compared with previous experimental and theoretical values.
24 3.3 Main results of the calculations
The lattice constant is in good agreement with the experiments, especially
with the measurement of Hames [21]. The dierence is within 1% for both
experiments. It is within the error bound of LSDA. The somewhat larger theo-
retical value reported in Ref. [24] is probably due to the less accurate numerical
method (LMTO-ASA) that obviously yields an (over-shooting) error cancelation
of the overbinding of LSDA.
The magnetic moments of the cubic structure show a relatively larger dis-
crepancy. Theoretical result (m
tot
= 4.47
B
) by Pugacheva [24] agrees better
with ours. Nevertheless, our result is between the experiments. The dierences
between the atomic spin moments, which are not well dened both experimen-
tally and theoretically, is probably due to dierent projections. There is a
ferrimagnetic coupling between Ge and the other atoms. Similar results were
also reported in other Heusler alloys by K ubler [28] and Fujii [29]. It is from the
hybridization (covalent bonding) of the magnetic 3d states and the nonmagnetic
4sp states. This hybridization, because of the dierent energy of the 3d up and
down spins, results in a spin split of the 4sp states, which couple ferrimagneti-
cally with the 3d spin moment. This covalent bonding will be discussed in the
next chapter.
3.3.2 Crystal structures at the ground state
In order to obtain the relative stability of the cubic and the tetragonal structures,
we relaxed c/a ratios under dierent volumes under our consideration. We have
the energy contour plotted as a function of the c/a and relative volume (V/V
0
,
where V
0
is the volume per formula unit of the energy minimum of the cubic
phase) shown in Figure 3.2. Two local minima that are denoted with +
signs in the gure at c/a=1.39, volume = 105.54

A
3
and c/a=1.60, volume =
105.30

A
3
are obtained. Here and after, we refer to these two phases as tetI and
tetII, respectively. When the relative volume is between 94% and 106% of the
equilibrium volume of the cubic phase, the two energy minima are present, and
the cubic structure gives a local maximum of E(V). Typical energy variations
with respect to the c/a ratio under a xed volume are plotted in Figure 3.3.
In Figure 3.3(a) the volume is 105.32

A
3
, we have two energy minima with
c/a=1.61 and 1.38. The cubic phase with c/a=

2 is a local maximum. At
volumes smaller than 94% of the equilibrium volume of the cubic phase, the
c/a=

2 is an inexion point. Then only a global minimum with extended c-

axis exists. As an example, the energy versus c/a curve at the volume of 100.10

A
3
are shown in Figure 3.3(b).
In the experiments, the cubic L2
1
phase was detected at elevated tempera-
tures, while our calculation shows a local energy maximum for this cubic phase
at zero temperature. Although the calculated energy dierences are tiny and
within the accuracy of the LSDA, there is no doubt a at region and a tendency
towards tetragonal distortions. Regarding to this, we propose a possible phase
transition caused by temperature. The electronic original of the predicted phase
transition is the band Jahn-Teller eect which will be justied in detail in the
following section. More detailed electronic structure discussions come right after
the following pressure eect.
The volume dependence of total and local magnetic moments is shown in
Figure 3.4. The total magnetic moment is contributed mainly from the 3d
electrons of Mn. The decrease of the Mn magnetic moments with the decrease
3 Band Jahn-Teller eects in Rh
2
MnGe 25
c
/
a
v/v
0
(%)
c/a=
+
+
Figure 3.2: The contour plot of the energy with respect to relative volume and
c/a. The dashed horizonal line with c/a=

2 indicates the cubic phase. The two

crosses indicate the two local energy minima in the plot corresponding to the
two tetragonal structures. V
0
=105.85

A
3
is the volume of the energy minimum
of the cubic phase.
of volume is common for itinerant magnetism, as band broadening deteriorates
the magnetism.
Comparing the spin moment of the cubic phase and the tetII phase at the
same volume, the magnetic moment is reduced by about 10% as shown in Figure
3.4. The electronic reason of this reduction can be understood by checking the
partial DOS of Mn at dierent c/a ratios. It is shown in Figure 3.5. Considering
the conventional unit cell of the tetragonal phase in Figure 3.1 (b), it is found
that the nearest neighbors of Mn are eight Rh atoms forming the vertices of
tetragonal lattice. The next nearest neighbors are the apical and in-plane Ge
atoms. Among the majority spin states there are strong bonding and antibond-
ing states between d
x
2
y
2,d
xz
, and d
yz
orbitals of Mn and the d orbitals of the
Rh. The d
xy
and d
3z
2
r
2 orbitals form nonbonding states. In the cubic phase
states formed by the rst three orbitals are degenerate, and the latter two are
also degenerate as shown in Figure 3.5(c). Under tetragonal distortion, the rst
three degenerate states split into E
g
(xz, yz) and B
1g
(x
2
y
2
) according to the
irreducible representation of the point group (4/mmm) of Mn. Thus the anti-
bonding E
g
states of the majority spin move above the Fermi level under the
distortion and the magnetic moment is reduced. With the increase of the c/a
ratio, the crystal eld splitting is enlarged as can be seen by comparing Figure
3.5(a) to (b), and the magnetic moment is decreased as shown in Figure 3.6,
because of the more de-occupation of the majority spin states.
26 3.3 Main results of the calculations
1.4 1.5 1.6 1.7
c/a
-14.0
-13.5
-13.0
-12.5
E
+
1
2
8
1
4
.
2

(
m
H
a
r
t
r
e
e
/
f
.
u
.
)
cubic
(a) Energy vs. c/a for the tetragonal structure with volume of
105.32

A
3
.
1.4 1.5 1.6 1.7 1.8
c/a
-10
-9
-8
-7
E
+
1
2
8
1
4
.
2

(
m
H
a
r
t
r
e
e
/
f
.
u
.
)
cubic
(b) Energy vs. c/a for the tetragonal structure with volume of
100.10

A
3
. The local minimum at smaller value of c/a is sup-
pressed.
Figure 3.3: Energy versus c/a ratios for the tetragonal structure.
3 Band Jahn-Teller eects in Rh
2
MnGe 27
0.88 0.94 1.00 1.06
Relative volume (V/V
0
)
3.6
3.8
4
4.2
4.4
4.6
S
p
i
n

M
o
m
e
n
t

(

B

)
Mn mom: cubic
Total mom: cubic
Total mom: tetrI
Total mom: tetrII
Figure 3.4: The magnetic moment versus relative volume of dierent phases.
The reference volume (V
0
) is taken to be the volume (105.85

A
3
) of the energy
minimum of the cubic phase.
1
0
1
2
xy
3z
2
-r
2
xz,yz
x
2
-y
2
-4 -2 0 2
E (eV)
2
1
0
1
2
xy,3z
2
-r
2
xz,yz
x
2
-y
2
1
0
1
2
(a) c/a=1.68
(b) c/a=1.52
(c) c/a=1.414
P
D
O
S

(
1
/
s
p
i
n
.
e
V
)
Figure 3.5: Partial DOS of the Mn-3d state resolved into real spherical harmon-
ics contributions at dierent c/a ratios at the same volume. Figure (a) and (b)
are from the tetragonal phase and share the same legend. (c) is from the cubic
phase.
28 3.3 Main results of the calculations
1.4 1.5 1.6 1.7 1.8 1.9 2
c/a
3
3.5
4
4.5
S
p
i
n

M
o
m
e
n
t

(

B
)
Mn moment
Total moment
Figure 3.6: The total magnetic moment and the magnetic moment of Mn de-
crease with the increase of the c/a ratio at the constant volume.
3.3.3 Experimental evidences for the tetragonal phase at
low temperature
In order to check the proposed low temperature tetragonal phases, a polycrys-
talline sample of Rh
2
MnGe was synthesized by arc-melting
1
. Subsequently, the
sample was homogenized at 950
o
C for ve days in argon. The measurements
[30] were performed using an X-ray diractometer D5000 (Siemens AG) with
monochronised Cu K
1,2
radiation (

1,2
= 1.54178

A) in Bragg-Brentano. The
diractometer was equipped with a low temperature chamber (APD Cryogen-
ics). The Rh
2
MnGe powder was mixed with Standard Silicon Powder 640c
(NIST) as an intrinsic line position reference material.
Diraction patterns in the angle range of 2 = 25

to 80

with a step width

of (2) = 0.02

and a recording time of t = 10 s per step were recorded in the

temperature range from 20 to 340 K. In Figure 3.7, the split of the peak around
2 = 42

corresponding to a cubic to tetragonal phase transition is observed

when the temperature is lowered.
From the Rietveld renement the temperature dependent lattice parameters
were obtained and are shown in Figure 3.8. Above a temperature of 180 K the
diraction patterns can be described by the cubic Heusler structure L2
1
(space
group 225, Fm

3m). Below, a phase transition occurs, becoming visible by a

splitting of various peaks in the diraction patterns (see Figure 3.7). The phase
transition is characterized by a symmetry reduction from a cubic to a tetragonal
lattice with space group symmetry I4/mmm (No. 139). The phase transition at
around 180 K (circle) can clearly be seen. Below the transition temperature the
lattice parameter a
0
of the cubic phase splits into the tetragonal parameters a
and c. The temperature dependence of c/a points to a second order transition.
The largest c/a ratio of the tetragonal phase could be determined to be 1.423
at a temperature of 20 K. This ratio is considerably smaller than the predicted
splitting (1.60). However, the calculated energy dierence E between the
two phases tetI and tetII is below k
B
20K, and imperfections like chemical
disorder might be responsible for this dierence, as they tend to broad the
1
Experiments in this section were done by my cooperators S. Nils (sample preparations)
and Z. Manuel (XRD measurements).
3 Band Jahn-Teller eects in Rh
2
MnGe 29
Figure 3.7: Diraction pattern collected at dierent temperatures from T = 20
K to 340 K around 2 = 42
o
[30].
DOS.
3.3.4 An interpretation as the band Jahn-Teller eect
The tetragonal distortion in Rh
2
MnGe has been predicted by our calculation
and veried by our preliminary experiments. The discovered structural phase
transition can be understood by the band Jahn-Teller eect. The general Jahn-
Teller theorem [9] asserts that if an ion is at a crystal site of such high symmetry
that its ground-state degeneracy is not the Kramers minimum, then it will be
energetically favorable for the crystal to distort such that the degeneracy is
lifted. The band Jahn-Teller eect is a special case of the theorem as explained
in Chapter 2, Section 2.6. If there are degenerate bands very close to the Fermi
level, and one allows for a slight distortion of the lattice lowering the symmetry
of the crystal, this distortion will lead to a splitting of the degenerate bands,
but does not change their center of gravity. Occupation of the lower branches of
the split levels must therefore lead to a lower band energy. If the increase of the
elastic energy does not exceed the decrease of the band energy, the distortion is
favored. The competition between FCC and BCC phases in transition metals
and alloys was extensively studied in the spirit of this band Jahn-Teller eect
[17].
Full Heusler alloys form in an FCC Bravais lattice. When this lattice is
distorted by elongating or shrinking the c axis, it transforms into body centered
tetragonal (BCT) lattice. Then former equivalence of the x, y and z directions in
the cubic structure is now lifted. Thus the Brillioun Zone (BZ) is also changed.
The BZ of the BCT lattice are shown in Figure 3.9. The high symmetric points
X(
2

2a
,
2

2a
, 0) and Z(0, 0,
2
c
) in the BZ of BCT structure are equivalent if
the BZ is of FCC lattice (c =

2a). The direction Z
3
and X P are
also equivalent directions in the latter case. But if the tetragonal distortion is
exerted, the points Z and X are no longer equivalent.
30 3.3 Main results of the calculations
.
.
.
.
.
0
L
a
t
t
i
c
e

p
a
r
a
m
e
t
e
r
Temperature
Figure 3.8: Temperature dependence of the lattice parameters in the cubic and
tetragonal phases. The phase transition at around 180 K (circle) can clearly
be seen. The lattice parameter a
0
of the cubic phase splits into the tetragonal
parameters (a,c).
z
M
Figure 3.9: The high symmetry points in the BCT Brillouin Zone.
3 Band Jahn-Teller eects in Rh
2
MnGe 31
-2
Figure 3.10: Angular and magnetic quantum number resolved bands (majority
spin channel) of the cubic phase. The symmetry lines correspond to tetragonal
notation as shown in Figure 3.9.
The reduction of symmetry will inuence the electronic structures. To see
this, rstly the fat bands
2
of the cubic phase in the BCT BZ are shown in
Figure 3.10. In this gure only the bands of Rh with m
l
= 2 and m
l
= 0 of
the 4d orbitals are shown, because it is the main contribution to the majority
DOS peak around the Fermi level which will be relevant to our discussion. We
see that the eigen energies at X and Z are degenerate in the cubic phase. Also,
the dispersions along XP and
3
Z are the same because of the equivalence
of these directions.
When we compress the c-axis, the eigenvalue at X is shifted downwards,
while the eigenvalue at Z is shifted upwards because of the antibonding nature
of the d
z
2 electrons. The degeneracy of the bands is lifted. An elongation of
the c-axis produces a shift of the eigenvalues in the opposite direction. Both
situations are shown in Figure 3.11. The lift of the band degeneracy can also be
observed in the density of states, where a splitting of the van Hove singularity
close to the Fermi level occurs as shown in Figure 3.12(a). Because the main
contribution of the DOS near the Fermi level is from Rh atoms, only the DOS
of Rh majority spin channel near the Fermi level for dierent c/a values is
shown in the Figure 3.12 (b) which is a zoomed-in plot of (a). It is clear that
there is a high peak in the vicinity of the Fermi level in the cubic phase. Under
tetragonal distortions, the peak is split into two with dierent heights due to the
dierence of X and Z points. Thus, compression leads to an asymmetric DOS
with respect to elongation. In particular, the elongation shifts the higher peak
above the Fermi level which decrease the energy more than the compression.
The split increases under the larger compression or extension in the c-direction.
The van Hove singularities show a square-root divergency which is the char-
2
Energy dispersions weighted by occupations of states.
32 3.3 Main results of the calculations
c/a=1.407
c/a=1.420
RhMnGe
Figure 3.11: Band structure of the elongated (black, c/a=1.420) and compressed
(red, c/a=1.407) tetragonal structure. Only bands from the majority spin chan-
nel are shown.
acteristic of one dimensional electronic state. This means the bands in the other
two dimensions are non-dispersive. This can be clearly observed in the band in
the plane of X-M-P. We can see that in both Figures 3.10 and 3.11 along X-P,
and X-M, there are large portions of the bands which are at.
Here we see that the band Jahn-Teller eect produces a stabilization of the
tetragonal ground state. One remark should be added here: whether or not
the band Jahn-Teller eect lowers the energy depends on the position of the
degenerate bands, namely, whether the Jahn-Teller active bands are near the
Fermi level or not. If they are far away from the Fermi level, the distortion still
lifts the degeneracy, but does not lower the energy. The position of the bands
are largely dependent on the chemical composition of the compound or the band
lling. On the other hand, the degeneracy of the Jahn-Teller active orbitals is
of course dependent on the crystal eld. It is also very sensitive to magnetic
polarization. It is the magnetic exchange that put the van Hove singularity of
the majority spin near the Fermi level. Atomic disorder, atomic substitutions,
or spin disorder will destroy the local symmetry, which will suppress this eect.
This might be the reason that the distortion in our experiments is much smaller
than the theoretical prediction. We strongly propose that more experiments
should be done in samples with better quality to investigate the properties of
this compound.
There are other Heusler compounds showing a tetragonal ground state as
mentioned in the rst section of this chapter. The mechanism for the distortion
from the cubic structure may be dierent. For example, the lattice instabilities
in Ni
2
MnGa [19] are attributed to the Fermi surface nesting, which leads to a
complete softening of the TA
2
transverse acoustic phonon branch along 110).
3 Band Jahn-Teller eects in Rh
2
MnGe 33
8
4
0
4
D
O
S

(
1
/
e
V
.
f
.
u
.
)
c/a=1.407
8
4
0
4
D
O
S

(
1
/
e
V
.
f
.
u
.
)
c/a=1.414
-8 -6 -4 -2 0 2 4
E (eV)
8
4
0
4
D
O
S

(
1
/
e
V
.
f
.
u
.
)
c/a=1.420
(a) The total DOS of Rh
2
MnGe for dierent c/a ratios: 1.407,
1.414, and 1.420.
-0.4 -0.2 0 0.2 0.4
E (eV)
0
1
2
3
4
D
O
S

(
1
/
e
V
.
s
p
i
n
) c/a=1.414
c/a=1.407
c/a=1.420
(b) DOS from the majority spin state of Rh near the Fermi level.
Figure 3.12: Total DOS and DOS of Rh near the Fermi level for dierent c/a
ratios: 1.407, 1.414, and 1.420 at a volume of 105.85

A
3
.
34 3.3 Main results of the calculations
fr
sr
sr
Figure 3.13: Scalar-relativistic (sr) and fully relativistic (fr) band structure of
the cubic phase at V
0
=105.85

A
3
.
On the other hand, Fujii [29] argued that the band Jahn-Teller eect cause the
lattice transformation in Ni
2
MnGa and Co
2
NbSn by comparing the DOS of the
cubic and tetragonal structure. In our case, the tetragonal distortion is clearly
driven by the band Jahn-Teller eect.
3.3.5 Relativistic eects
We have seen that the band Jahn-Teller eect can explain the instability of the
cubic phase versus the tetragonal one. Can the spin-orbital coupling also split
the degeneracy? We have checked this point by performing a full relativistic
calculation of the cubic phase. The bands from scalar relativistic and fully rel-
ativistic calculations at the equilibrium volume, V
0
=105.85

A
3
are shown in
Figure 3.13. The quantization axis of spin in the full relativistic calculation
points along the z-direction. It is clear that the degeneracy at X and Z is still
there. The spin-orbit coupling, which is small in our compound, lifts some de-
generacy of the d-orbitals at point, which is highlighted by the green ellipsoids
in the gure. Thus we can conclude that a split of the DOS peak around the
Fermi level can solely arise from a structural distortion.
The magnetic crystalline anisotropy, dened as E
[100]
E
[001]
, amounts about
5 Hartree/f.u. when c/a=1.61 at the equilibrium volume.
The fully relativistic calculation also yields an orbital contribution to the
magnetic moment. The calculated total orbital moment and spin moment are
0.037
B
and 4.503
B
, respectively, giving a total moment of 4.545
B
/f.u. in
the cubic phase, when the spin quantization axis is in [001] direction. The orbital
moment comes mainly from Mn atoms (0.030
B
). The tetragonal distortion
3 Band Jahn-Teller eects in Rh
2
MnGe 35
increases the orbital moment because of the reduction of the symmetry. At
c/a=1.61, the total orbital moment is 0.046
B
, of which 0.043
B
is contributed
by the Mn atoms.
36 3.3 Main results of the calculations
Chapter 4
Magnetic properties of
AFe
2
(A=Y, Zr, Lu, Hf )
under pressure
4.1 Introduction
In 1898, Charles Edouard Guillaume discovered that the thermal expansion
coecient of Fe-Ni alloys exhibits a sharp minimum with atomic concentration
of Ni at around 35 at.%. This particular alloy is called Invar alloy because
its length is invariant with temperature within a certain temperature region.
Guillaume was awarded the Nobel Prize for physics in 1920, in recognition
of the service he has rendered to precision measurements in Physics by his
discovery of anomalies in nickel steel alloys. The Invar eect was proposed to be
connected with magnetism because the plateau of the volume expansion starts
almost at the Curie temperature below which ordered magnetic moments are
established. The reason for the almost zero volume expansion is that the Invar
alloy has a spontaneous volume magnetostriction large enough to compensate
normal thermal expansions due to lattice vibrations.
It has been found that quite a number of alloys and compounds show large
spontaneous volume magnetostriction. The most famous ones are Fe
65
Ni
35
and
Fe
72
Pt
28
. These are both Fe-rich ferromagnetic FCC iron alloys. Some of the
Laves phase intermetallic compounds of transition metals such as AFe
2
and
ACo
2
also show large spontaneous volume magnetostriction. Shiga [31] reported
for the rst time experimental evidences of the Invar eect in some of the Laves
phase intermetallic compounds. The spontaneous volume magnetostriction
s
in HfFe
2
was reported to be 810
3
, in ZrFe
2
1010
3
, and in YFe
2
to be small
and negative. Here the
s
is dened in terms of the ratio of the equilibrium
volumes in the ferromagnetic (FM) and the paramagnetic (PM) state
1

s
=
V (FM) V (PM)
V (PM)
. (4.1)
1
In experiment the volume of PM states is obtained by extrapolating volume versus tem-
perature curves from high temperatures down to low temperatures (T=0).
38 4.1 Introduction
Table 4.1: The experimental values of the lattice constant (a), magnetic moment
(m
s
), and Curie temperature (T
c
) of AFe
2
(A=Y,Zr,Hf, and Lu) compounds.
AFe
2
a m
s
(at 4.2K) T
c
(

A) (
B
/f.u.) (K)
YFe
2
7.363 2.90 545
ZrFe
2
7.060 3.14-3.46 625
HfFe
2
7.020 3.36 612
LuFe
2
7.217 2.97 610
A large spontaneous volume magnetostriction means that the magnetic mo-
ment and the lattice constants are strongly coupled. Experimentally this cou-
pling can be clearly seen by measurements, such as hyperne eld measurements
[32], X-ray magnetic circular dichroism (XMCD) [33], and nuclear forward scat-
tering (NFS) [34] under pressure.
As proposed to be in the family of Invar, AFe
2
was extensively investigated
both experimentally and theoretically. The reported lattice constants, satura-
tion magnetization and Curie temperatures for the four compounds in the C15
Laves phase are listed in Table 4.1. No other stable AFe
2
compounds in the C15
Laves phase at T = 0 were reported. Wortmann [35] showed that the hyperne
eld decreases to zero at about 40 GPa (d
FeFe
2.37

A) in LuFe
2
, and 50 GPa
(d
FeFe
2.40

A) in YFe
2
at room temperature by NFS. At low temperature,
the loss of magnetic moments in YFe
2
took place at about 90 GPa (d
FeFe

2.30

A). The pressure dependence of the hyperne eld (H
0
) of YFe
2
and ZrFe
2
was measured up to 0.8 GPa by Dumelow [36] by Mossbauer spectroscopy. The
values of d ln H
0
/dP were 4.20.210
4
kbar
1
and 7.30.110
4
kbar
1
at 4.2 K, respectively, while at room temperature, it is 8.5 10
4
kbar
1
in
YFe
2
, as reported by Riedi [37]. It agrees with the calculated values [36] of
8.96 and 4.88 10
4
kbar
1
, respectively for these compounds where only
the Fermi contact term has been taken into account in the calculation.
Kai [38] measured the magnetic moment of several groups of Fe compounds in
the Laves phase, including C14 and C15 phases: group I: TFe
2
, T=Sc, Ti, Zr, Hf;
group II: RFe
2
, R=Lu, Tm, Er, Ho, Dy, Tb, Gd; and group III: M

, M

Fe
2
=Ta,
Nb, U, Pu, Ce and Sm. A linear relationship between the Fe-Fe distance and the
magnetic moment was found in each group. The magnetic moment decreases
with the reduction of the Fe-Fe distance. The magnetic moment behavior under
pressures of these compounds was reported by Armitage [39]. The pressure
dependent magnetization
ln
P
was obtained by the thermodynamic relationship
from the measured forced volume magnetostriction
ln V
B
:
ln
P
=
40

0
M
(
ln V
B
),
where is the magnetization per unit mass, P is the pressure, V is the volume
of the sample, B is the magnetic eld, and
0
M is the magnetization per unit
volume. The values at T = 4.2 K are 8.2 0.4 and 6.3 0.3 10
4
kbar
1
for YFe
2
and ZrFe
2
, respectively. This is quite agreeable with the hyperne
eld measurements and calculation by Dumelow, which suggests that in these
4 Magnetic transitions of AFe
2
under pressure 39
compounds the hyperne eld at the Fe site is proportional to the magneti-
zation of the compounds. The pressure dependence of the Curie temperature
(T
c
) of ZrFe
2
was measured by Brouha [40]. The negative dT
c
/dP up to hy-
drostatic pressure of 35 kbar indicated the characteristics of itinerant ferromag-
netism. All of these experiments suggest strong magneto-volume couplings in
these compounds.
Theoretically, Asano [41] investigated the phase stability by comparing the
total energies of dierent magnetic states (nonmagnetic, ferromagnetic, or anti-
ferromagnetic states) of C14 or C15 Laves phases by LMTO, where the magnetic
state refers to the spin arrangement of the iron atoms, and not to the coupling
between iron and the other element. They concluded that the ground state
of Y, Zr, and Hf compounds are the ferromagnetic C15 Laves phase, which
is in agreement with experiments. Yamada [42] has calculated the high eld
susceptibility
hf
by tight binding approximations. It was found that in ZrFe
2

hf
= 5.810
4
emu/mol, and in YFe
2

hf
= 5.5710
4
emu/mol which agree
with the experimental values of 6.1 10
4
emu/mol, and 1.55 10
4
emu/mol
in the order of magnitudes. Klein et al. [43] discussed the electronic struc-
ture, superconductivity, and magnetism in the C15 compounds ZrX
2
(X=V,
Fe, and Co). Their results showed that the simplied Stoner theory, which is
basically a rigid band model, is quantitatively inaccurate in describing the mag-
netic properties in stoichiometric and non-stoichiometric compounds because
of a signicant covalent bonding. This binding mechanism in ZrFe
2
was rst
proposed by Mohn [44]. The consequence of this binding is that the weights of
DOS of the majority and minority electrons changes, rather than only a rigid
shift of the two spin subbands as assumed in the Stoner model. Thus reliable
conclusions can only be available by full self-consistent calculations. The sim-
ilar total energy of paramagnetic and ferrimagnetic state, where the magnetic
moments of Fe and Zr are antiparallel but do not compensate, at small lattice
parameter in the calculation by Mohn [44] indicated that the magnetic moment
would collapse in ZrFe
2
under pressure, but no detailed information about the
magnetic transition was given there.
In the previous twenty years, a lot of work has been spent on clarifying the
magnetic structure of these compounds. One of the main questions is whether
the A atom carries a magnetic moment or not. All theoretical calculations gave
the same answer that the A atom has a small induced antiparallel magnetic
moment because of covalence with iron [44, 45]. But the magnetic form factor of
the A atoms might be too small to be detected by scattering methods. Later on,
neutron scattering experiments of YFe
2
by Ritter [46] conrmed the theoretical
prediction that the Yttrium carries a negative magnetic moment, but no reports
about the other compounds.
However, theoretical calculations based on the density functional theory
(DFT) to study the pressure eect in these compounds in detail are still not
reported according to our best knowledge. The experimental evidence that un-
der high pressures the magnetic moment of YFe
2
, HfFe
2
, and LuFe
2
collapses,
has been reported by Wortmann, as we have mentioned. For ZrFe
2
, if we com-
pare the three isostructural compounds: YCo
2
, ZrFe
2
, and YFe
2
, which have
93, 92, and 91 electrons, respectively, YFe
2
is ferromagnetic with magnetic mo-
ments about 2.90
B
/f.u., while YCo
2
is metamagnetic. If we assume that their
electronic structures are similar, it can be expected that the middle compound
shows a magnetic deterioration under moderate pressures. Collecting the direct
40 4.2 Calculational parameters
and indirect magnetic moment information under pressure we may expect to
obtain a magnetic transition under pressure in all these four compounds.
The deterioration of the magnetic moment in itinerant systems is universal
and can be qualitatively understood with the help of the Stoner model as ex-
plained before: In a simplied version of this model, a magnetic state is stable
if IN(E
F
)>1, where I is the Stoner parameter, which is only weakly dependent
on the interatomic distance, while N(E
F
), the density of states at the Fermi
level, decreases as the band width increases under pressure. Taking Heines [47]
general result that the width of d bands (W) is proportional to R
5
, where R is
the nearest neighbor atomic distance, the N(E
F
) should accordingly decrease
with R if we take the rectangular DOS model. At a certain pressure, the Stoner
criterion is no longer satised, then the system becomes nonmagnetic.
In this chapter, the magnetic moment behavior of the four ordered stoichio-
metric compounds (YFe
2
, ZrFe
2
, LuFe
2
, and HfFe
2
) in the cubic Laves phase
are investigated by density functional calculations. The main problems we are
going to address are the order of the quantum magnetic phase transition at
high pressures and the related Invar eect. Two of them (ZrFe
2
and HfFe
2
)
were reported to show Invar anomaly, while the other two were not. A natural
question would be: What is the dierence between them? Partial answers to
these questions are provided by our density functional calculations. Addition-
ally, non-stoichiometric compounds of Zr
x
Fe
1x
is studied in order to under-
stand the volume and magnetic moment trend with respect to doping in the
homogenous region.
4.2 Calculational parameters
The space group of the cubic Laves phase is Fd

3m (C15 phase). It consists

of two formula units in the conventional unit cell (u.c.). The A atom occupies
the Wycko position 8a: (
1
8
,
1
8
,
1
8
), and Fe is at 16d: (
1
2
,
1
2
,
1
2
), as shown in
Figure 4.1. The LSDA exchange correlation functional chosen is parameterized
Figure 4.1: The unit cell of the cubic Laves phase (C15). The Fe atoms form a
pyrochlore sublattice as shown by the bonds between Fes.
by Perdew and Wang [7]. The number of k-points in the irreducible part of the
4 Magnetic transitions of AFe
2
under pressure 41
Brillouin zone (IBZ) is 897 if not specied in the context. The valence basis
states are Fe: 3sp/3d4sp +4d, Zr and Y: 3d4sp/4d5sp, Lu: 4f4d5sp/5d6sp and
Hf: 4d5sp/5d6sp, respectively.
4.3 Fixed spin moment schemes
In order to probe possible local minima of the total energy with respect to the
spin moment, a xed spin moment (FSM) scheme was proposed by Schwarz and
Mohn [48]. In this scheme, the total spin moment M is xed. In general, the
number of valence electrons Z
val
and the spin moment M per unit cell are given
by:
Z
val
= N

+N

=
_
E

(E)dE +
_
E

(E)dE (4.2)
M = N

=
_
E

(E)dE
_
E

(E)dE (4.3)
where the D

(E) and D

(E) are the densities of states for majority and minority

spins, respectively. The variable N
()
is the total number of spin up (down)
electrons. We can take the M as a parameter. Then the Fermi levels of the up
and down spins E
()
F
are determined by the equations above. Their dierence
corresponds to the Zeeman energy with respect to a ctitious applied magnetic
eld. By this constraint search, we are able to evaluate the total energy (E) of
the ground state as a function of M together with some other thermodynamic
variables such as V (volume): E = E(M, V, ...). In reality, for a xed volume,
we might obtain several energy minima as a function of M. As a function of
volumes, the local minimum might change its positions, disappear or appear.
This gives complicate magnetic behaviors of the system under pressure [12]. In
summary, the FSM scheme provides us an additional freedom to explore the
phase space. Interesting physics such as metamagnetism can be discussed on
the basis of this scheme [49].
4.4 Ground state properties of AFe
2
Although the magnetic moments of A and Fe are ferrimagnetically coupled, we
still use the terminology FM to indicate the state where the Fe sublattice has a
nite parallel spin moment. The calculated lattice constants in FM (a
FM
) and
NM (a
NM
) states, total spin moment (m
0
) at the theoretical equilibrium lattice
constant (a
FM
) and that (m
exp
) at the experimental lattice constant are listed in
Table 4.2. Comparing with the experimental parameters shown in Table 4.1, we
can state that the results are in good agreement with the experiments within
the error bar of the LSDA. The use of LSDA results in an underestimation
of the lattice constants by about 3% for 3d transition metals. The relatively
large underestimation of the magnetic spin moment if we use the theoretical
equilibrium lattice parameter results from the sensitivity of the moment with
respect to the lattice constant as we will see in the following discussions.
4.4.1 Common features of the electronic structure
The DOS of the four compounds, evaluated at the respective theoretical equi-
librium lattice constants (a
FM
in Table 4.2), are shown in Figure 4.2. Common
42 4.4 Ground state properties of AFe
2
-
6
-
4
-
2
0
2
E

(
e
V
)
2
0
1
0 0
1
0
2
0
DOS (1/eV.u.c.)

F
e

Y
T
o
t
a
l
(
a
)
D
O
S
o
f
Y
F
e
2
-
6
-
4
-
2
0
2
E

(
e
V
)
2
0
1
0 0
1
0
2
0
DOS (1/eV.u.c.)

F
e

Z
r
T
o
t
a
l
(
b
)
D
O
S
o
f
Z
r
F
e
2
-
6
-
4
-
2
0
2
E

(
e
V
)
2
0
1
0 0
1
0
2
0
DOS (1/eV.u.c.)

F
e

L
u
T
o
t
a
l
(
c
)
D
O
S
o
f
L
u
F
e
2
-
6
-
4
-
2
0
2
E

(
e
V
)
2
0
1
0 0
1
0
2
0
DOS (1/eV.u.c.)

F
e

H
f
T
o
t
a
l
(
d
)
D
O
S
o
f
H
f
F
e
2
F
i
g
u
r
e
4
.
2
:
D
O
S
o
f
A
F
e
2
(
A
=
Y
,
Z
r
,
H
f
,
a
n
d
L
u
)
a
t
t
h
e
r
e
s
p
e
c
t
i
v
e
e
q
u
i
l
i
b
r
i
u
m
v
o
l
u
m
e
s
.
T
h
e
v
e
r
t
i
c
a
l
d
a
s
h
e
d
l
i
n
e
s
i
n
d
i
c
a
t
e
t
h
e
F
e
r
m
i
l
e
v
e
l
.
4 Magnetic transitions of AFe
2
under pressure 43
Table 4.2: The calculated lattice constants a
FM
in FM state and a
NM
in NM
state, total spin magnetic moments (m
0
) at a
FM
and that (m
exp
) at experi-
mental lattice constants of the four compounds. The numerical accuracy of the
lattice constant is within 0.002

A.
AFe
2
a
FM
a
NM
m
0
m
exp
(

A) (

A) (
B
/f.u.) (
B
/f.u.)
YFe
2
7.040 6.927 2.57 3.10
ZrFe
2
6.838 6.771 2.38 3.14
HfFe
2
6.824 6.746 2.86 3.17
LuFe
2
6.931 6.874 2.53 2.95
features of these DOS are quite obvious: they are highly peaked although the dis-
tance between the nearest neighbors of irons is small, for example d
FeFe
4.56
a.u. in the ZrFe
2
compared with d
FeFe
4.69 a.u. in BCC Fe. Just above
the Fermi energy, there is a relatively high DOS in the minority spin channel.
This feature is dominated by anti-bonding 3d states of Fe. The bonding and
antibonding states of the minority spin channel are separated by a deep and
wide valley with a width of about 1 eV. The zoomed-in band structure and
the DOS of the minority states in ZrFe
2
are shown in Figure 4.3. The peak in
the minority states just above the Fermi level comes from almost non-dispersive
states. This energy range is highlighted by the rectangular block with height of
0.06 eV. The high DOS around the peak stems from the narrow bands in the
directions along W K and W U. Because of the similarity of the electronic
Figure 4.3: The band structure and the DOS of the minority spin channel of
ZrFe
2
at the theoretical equilibrium lattice constant. The rectangular block with
height of 0.06 eV indicates the energy window around the pronounced DOS.
structure in these four compounds only the bands and the DOS of ZrFe
2
are
shown here.
The van Hove singularity just above the Fermi level in the spin down channel
is closely related to the magnetic properties which we are going to discuss, so
we explore the origin of it here. It was long realized that the Fe atoms in
the C15 Laves phase form a so-called pyrochlore structure with corner shared
44 4.4 Ground state properties of AFe
2
(a) The atom net of Fe in the plane of
{111}.
X
Z
o
Y
x
z
o
y
1
2
4
3
(b) The corner shared tetrahedrons of Fe.
Figure 4.4: The Kagome net of Fe (a) and the corner shared tetrahedrons (b) of
the Pyrochlore net. The O-XYZ is the global coordinate system, and o

is the local coordinate system used to project the partial DOS in a trigonal
lattice.
tetrahedrons as shown in Figure 4.1. If we look at the net in the 111 layer, it
is a Kagome net with alternatingly connected triangles and hexagons as shown
in Figure 4.4(a). As already shown by Johnston and Homann [50], the high
peaks in the DOS in a Kagome net of iron atoms come from narrow bands
with d character. One band tight binding model calculation by Isoda [51]
discovers two non-dispersive degenerated states along the XW line. Further,
there are two additional non-dispersive degenerated antibonding states along
all high symmetry directions. These results indicate that the spiking DOS is
closely related to the topology of the nets.
If we plot the m
l
-resolved fat band and the partial DOS (PDOS) of Fe
as shown in Figure 4.5, it is clear that at -point the states with m
l
=0 (d
z
2)
and 2 (d
x
2
y
2) are degenerate and m
l
=-2 (d
xy
), -1 (d
yz
), and 1 (d
xz
) are also
degenerate. These two groups are denoted by E
g
, and T
2g
, respectively.
But they are not the irreducible representation of the cubic symmetry, because
the site symmetry of Fe is D
3d
. It can be shown that the PDOS is divided
into two catalogues: One is the three d orbitals (m
l
=-2, -1, and 1) and the
other is two d orbitals (m
l
=0, 2) if we choose the coordinates as in Figure
4.4, but the dispersions are dierent. It is shown that the strongest van Hove
singularity in the PDOS just above the Fermi level of the minority spin state is
solely from the T
2g
states. From the fat band, it can be observed that the
states at the point are mainly contributed from the T
2g
state. It is quite
understandable because the T
2g
states form quite strong binding states
along each atomic chain with nearest neighbor interactions. For example, the
d
xy
orbitals from Fe
1
2
and Fe
4
, or Fe
2
and Fe
3
form -orbitals. The antibonding
state of these orbitals is the source of the quite spiking feature of the DOS in
the cubic Laves phase compounds as analyzed by Johnston and Homan [50].
From another point of view, if we rotate the coordinates and put the z-direction
to the diagonal direction of the cube as shown in Figure 4.4 (b), then the lattice
2
Atoms are numbered as in Figure 4.4(b).
4 Magnetic transitions of AFe
2
under pressure 45

2
g
E[eV]
(
d
)
z
,

x
-
y
2
2
2
(
d
)
x
z
,

y
z
,

x
y
F
i
g
u
r
e
4
.
5
:
T
h
e

f
a
t

b
a
n
d
a
n
d
P
D
O
S
o
f
t
h
e
m
i
n
o
r
i
t
y
d
-
s
t
a
t
e
o
f
F
e
i
n
Z
r
F
e
2
.
T
h
e

T
2
g

s
t
a
t
e
s
a
r
e
f
r
o
m
s
t
a
t
e
s
w
i
t
h
m
l
=
-
2
,
-
1
a
n
d
1
,
w
h
i
l
e
t
h
e

E
g

s
t
a
t
e
s
a
r
e
f
r
o
m
m
l
=
0
a
n
d
2
.
46 4.4 Ground state properties of AFe
2
-4 -2 0 2
E (eV)
1
0.5
0
0.5
1
D
O
S

(
1
/
e
V
.
F
e
)
E
g
2
(d
xy, x
2
-y
2)
E
g
1
(d
xz, yz
)
A
1g
(d
z
2)
Figure 4.6: PDOS of Fe resolved into the irreducible representation (A
1g
and
E
1,2
g
) of the point group D
3d
.
can be described by a trigonal lattice. The point group of the atom on the new
z-axis (Fe
1
) is D
3d
. There is a one dimensional irreducible representation A
1g
with basis of d
z
2. It turns out that the d
z
2 orbital comes solely from the T
2g

orbitals. As discussed by Isoda [51] by single orbital tight binding calculations,

which is a one dimensional representation, the antibonding orbitals are non-
dispersive in all high symmetric directions of the BZ. This implies that the one
dimensional representation with d
z
2 orbitals as its basis should give also quite
spiking feature in the DOS. This is evident if we plot the PDOS resolved into
the irreducible representation of D
3d
as shown in Figure 4.6.
Hybridization between the states of the Fe and A atoms in this cubic Laves
phase was extensively studied by several authors, e.g. K ubler [28], and Brooks
[45]. The covalent bonding scenario was generally accepted in order to explain
the ferrimagnetic coupling between the iron atom and the A atoms. The model
density of states shown in Figure 4.7 depicts this process. The atomic iron d
states are spin polarized and the A atoms are not. The split 3d states are much
lower in energy than the 4d or 5d of the A atoms (The atomic eigenenergy of the
d states are
Fe
d
16.54 eV,
Y
d
6.80 eV,
Zr
d
8.46 eV,
Lu
d
6.63 eV
and
Hf
d
8.14 eV) [52]. When they form a bond, because the energy of the
majority 3d states is about 2 eV lower than that of the minority 3d states by the
exchange splitting, the hybridization between the 3d and 4(5)d majority states
is weaker than between the minority states. By covalency charges transfer more
from the majority spin state of A atom to Fe than from the minority spin state.
Thus the minority 4(5)d states become more occupied than the majority states.
In this way the magnetic moments of the A atoms are antiparallel to the Fe.
This covalent interaction between A and Fe explains the antiparallel coupling
between the spin moments of the two elements in the compounds.
4 Magnetic transitions of AFe
2
under pressure 47
E
F
Fe
E
F
A A
2
Fe
covalency
E
F
Figure 4.7: Model DOS of the d states of A and Fe before and after hybridization.
The green blocks show the occupied states of the A atoms while the yellow is of
the Fe. The red arrows (up and down) indicate the majority and minority spin
of the electrons.
4.4.2 Specic electronic structures and magnetic moment
behavior
Because of the dierences of the A atoms, we can naturally expect some dier-
ences among these compounds. Firstly the lattice constants of these materials
are more or less determined by the atomic volume of A. Taking the atomic
volume, dened by (atomic weight/mass density), of the elements: Y=19.89,
Zr=14.06, Lu=17.78, and Hf=13.41 (cm
3
/mol), respectively [53], we can see
that the lattice constants in Table 4.2 follows the same tendency.
Secondly their magnetic moments have dierent behaviors under pressure.
The dependence of the magnetic moment on lattice parameters are shown in
Figure 4.8. The corresponding hydrostatic pressures are shown on the upper
abscissas. Obviously all of them show a decrease of the magnetic moment with
the decrease of the lattice constant as expected from the itinerant electron mag-
netism, but the Hf and Zr compounds show a more rapid decrease of the moment
at a lattice constant around 6.8

A (in the vicinity of the equilibrium lattice con-
stant), while the other two show a gradual decrease at this low pressure. At
high pressure, all four compounds show at least one rst order transition to a
lower or zero spin state. The dierences are quite understandable by examining
the dierences of the electron numbers of the compounds under the assumption
that the electronic structure is not so much inuenced by the dierence of the
A atoms. YFe
2
, ZrFe
2
, HfFe
2
, and LuFe
2
show basically similar DOS as dis-
cussed before. The dierence of the electron number shifts the Fermi level in
these systems. Zr(4d
2
) and Hf(5d
2
) have one more d-electron than Y(4d
1
) and
Lu(5d
1
), so the Fermi levels of the former are shifted towards higher energy,
situating closer to the pronounced peak of the minority spin DOS as shown in
Figure 4.2. This accounts for the low pressure instability of the moment.
The four compounds show multi-step magnetic transitions. This process can
be understood by the particular DOS of these compounds. Taking ZrFe
2
as an
example, the DOS at dierent lattice constants are shown in Figure 4.9 (ad).
The lattice constants of each gure are indicated by the arrows in Figure 4.8(b)
with the corresponding labels of (a), (b), (c), and (d). From the Figure 4.9(a),
it is obvious that at the experimental lattice constant the DOS of the up spin,
contributed mainly from the iron, has a gradual increase below the Fermi level,
48 4.4 Ground state properties of AFe
2
6 6.5 7 7.5
Lattice constant ()
0
0.5
1
1.5
2
M
a
g
n
e
t
i
c

m
o
m
e
n
t

(

/
F
e
)
2.74 13.47 42.27
Pressure (GPa)
(a) Magnetic moment vs. volume of
YFe
2
6.2 6.4 6.6 6.8 7 7.2
Lattice constant ()
0.0
0.5
1.0
1.5
2.0
M
a
g
n
e
t
i
c

m
o
m
e
n
t

(

/
F
e
)
(a)
(b)
(c)
(d)
0.56 4.64 11.78 -3.07 -4.91
Pressure (GPa)
(b) Magnetic moment vs. volume of
ZrFe
2
6 6.5 7 7.5
Lattice constant ()
0
0.5
1
1.5
2
M
a
g
n
e
t
i
c

m
o
m
e
n
t

(

/
F
e
)
Pressure (GPa)
-3.18 -0.25 9.52 40
(c) Magnetic moment vs. volume of
LuFe
2
6 6.2 6.4 6.6 6.8 7
Lattice constant ()
0
0.5
1
1.5
2
M
a
n
g
e
t
i
c

m
o
m
e
n
t

(

/
F
e
)
-3.13
Pressure (GPa)
0.03 5.15 13.3 26.34
(d) Magnetic moment vs. volume of
HfFe
2
Figure 4.8: Magnetic moment behaviors of AFe
2
(A=Y, Zr, Lu, Hf) under
hydrostatic pressures. The experimental and the theoretical equilibrium lattice
constants are shown by the solid and dotted vertical lines respectively. The
pressure at the corresponding lattice parameters is shown on the upper abscissas
of each gures. These theoretical pressures are obtained from the high spin state.
The labels (a), (b), (c), and (d) in gure (b) marks the positions where we show
the DOS in Figure 4.9.
while the DOS of the down spin has a wide ( 0.8 eV) dip below and a sharp
increase just above E
F
. Applying pressure will broaden the band and reduce
the width of the dip and decrease the magnetic moment. Thus the exchange
splitting is reduced. The DOS of the up spin and down spin moves towards each
other. This gradual decrease of the magnetic moment is shown in Figure 4.8(b)
between the arrows (a) and (b). The gradual decrease of the magnetic moment
continues until the Fermi level passes through the high DOS peak of the minority
spins, see Figure 4.9(b) and (c). Then the magnetic moment is rapidly reduced,
as shown in Figure 4.8(b) when the lattice constant is between 6.85 (arrow (b))
and 6.80 (arrow (c))

A. If we take the rigid band model [54, 55] and assume
that the band width W and the interatomic distance R follow Heines model
W R
5
= constant [47], the magnetic moment variation under the change of
volumes is directly related to the averaged DOS at the Fermi level by
V
M
[
M
V
]
H=0
=
5
3
I
N
1
eff
I
, (4.4)
where
N
1
eff
=
1
2
(1/N
+
+ 1/N

), (4.5)
4 Magnetic transitions of AFe
2
under pressure 49
15
10
5
0
5
10
15
D
O
S

(
1
/
e
V
.
u
.
c
.
)
Fe
Zr
Total
-5 -4 -3 -2 -1 0 1
E (eV)
15
10
5
0
5
10
15
-4 -3 -2 -1 0 1 2
E (eV)
(b)
(c)
(a)
(d)
Figure 4.9: The total DOS and partial DOS of ZrFe
2
at dierent lattice con-
stants. From (a) to (d), the lattice constants are 7.08, 6.85, 6.80, and 6.70

A,
respectively. In (a), the partial DOS of Fe and Zr are also shown. The Fermi
level is indicated by the dashed vertical line at E=0 eV.
and I is the Stoner parameter, N
+()
is the DOS at Fermi level of the spin
up (down), and M, V are the magnetic moment and the volume of the sample
respectively. From this, we can see, the larger the average DOS (the Fermi level
around the peaks), the more rapidly the magnetic moment varies.
The occurrence of an intermediate magnetic solution in ZrFe
2
at lattice con-
stant between 6.32 and 6.58

A is due to the details of the DOS. It is impossible to
give an argument without calculations, but one thing is essential: narrow peaks
form around the Fermi level, so that multiple magnetic solutions can exist [56],
and this has been discussed in our previous introduction to the Stoner model.
In this range of lattice parameters, the material becomes weakly ferromagnetic,
with rather small magnetic moments: 0.25
B
/Fe. Further compression will
suppress the magnetism. A question arises whether the transition is rst order
or second order. This transition can also be induced by doping. For example,
in a recent experiment on Hf(Fe
1x
Co
x
)
2
by Amako [57] a collapse of the fer-
romagnetic moment has been observed between x=0.6 and 0.65. Further, the
doping can tune the pressure dependence of the Curie temperature (T
c
), and
induce a transition from the ferromagnetic state to the metamagnetic state. We
postpone the detailed discussions to the following section. From the experimen-
tal point of view, the interesting point here is that the transition pressure is in
the laboratory accessible range (tens of GPa), see Figure 4.8.
4.4.3 The order of the magnetic transition under high
pressure
We can see from the previous gures (Figure 4.8) that under moderate pressure
(tens of GPa), there is a nite small magnetic moment in these compounds.
With further increase of pressure, the magnetic moment eventually becomes
50 4.4 Ground state properties of AFe
2
zero. The transition of the magnetic moment can be either a rst order (a
discontinuous transition) or a second order (a continuous transition). The free
energy (E) landscape at T = 0 of rst and second order phase transition is
schematically illustrated in Figure 4.10. This dierent behavior when approach-
m
p
p
c
E
m
m
E
m
p<
p
c
p=
p
c
p>
p
c
m
p
p
c
(1)
(2)
E
m
m
E
m
p<
p
c
p=
p
c
p>
p
c
E
E
Figure 4.10: Qualitative illustration of rst order (1) and second order (2) tran-
sitions of the magnetic moment (m) under pressures, and their corresponding
free energy (E), respectively, adapted from Peiderer [58].
ing the quantum phase transition can be of interest both experimentally and
theoretically. For the second order phase transition, the eect of uctuation
was shown to lead to novel electronic ground states in magnetic metals such
as magnetically mediated superconductivity, partial or quadrupolar order and
non-Fermi liquid phases. For the rst order phase transition, as summarized
by Peiderer [58], it is interesting for a number of reasons: It can drive novel
electronic states, novel types of low lying excitations, or signal the existence of
subtle quantum correlation eects. In general, peaks of DOS (van Hove singu-
larities) near the Fermi level in all real materials result in a ragged free energy
landscape. The topology of the DOS thus has a connection with the order of
the quantum phase transition. It is necessary to give some hints about the
order of the transition based on our calculations of electronic structures. Take
the simplied Stoner model, the magnetic free energy of the system in the rigid
band model is expressed by [59]
E(m) =
_
m
0
(m

)dm

1
4
Im
2
, (4.6)
where (m) is exchange splitting as a function of magnetic moments m, and
I is the Stoner parameter. The (m) can be expanded as a power series of m:
(m) = a
1
m+a
3
m
3
+a
5
m
5
(4.7)
where
a
1
=
1
2

N
1
1
(4.8)
a
3
=
1
3
(3

N
2
2

N
5
1


N
3

N
4
1
) (4.9)
a
5
=
2
5!
(105

N
4
2

N
9
1
105

N
3

N
2
2

N
8
1
+10

N
2
3
N
7
1
+ 15

N
4

N
2

N
7
1


N
5

N
6
1
). (4.10)
4 Magnetic transitions of AFe
2
under pressure 51

N
i
is dened as the (i 1)-th order derivative of the density of states at the
Fermi level with respect to the energy
3
.
Then the free energy is
E(m) =
1
2
(a
1

I
2
)m
2
+
1
4
a
3
m
4
+
1
6
a
5
m
6
. (4.11)
The stability of the phase can be discussed in line with Landaus theory of
second order phase transitions. Magnetic instability is necessarily given by the
condition that a

1
= a
1

I
2
0, which is equivalent to the Stoner criterion
IN(E
F
) 1 by considering Equ. (4.8).
The necessary condition to have a rst order transition is a
1
I > 0, a
3
< 0,
and a
5
> 0
4
if higher order terms than m
5
are neglected in Equation (4.11).
This means the DOS at the Fermi level should be suciently small (the Stoner
criterion is not fully satised) and the curvature of the DOS at E
F
is positive
and large, so that

N
3
is positive and large enough to give negative a
3
, otherwise,
if

N
3
< 0, a
3
is denitely positive. These rst two conditions require that the
Fermi level is at a narrow valley of the DOS.
Let us replace this qualitative analysis by direct FSM calculation results and
the corresponding DOS to analyze the transition. The rst example is ZrFe
2
,
which shows a rst order transition to the non-magnetic state. The FSM energy
curves are shown in Figure 4.11 at lattice constants around the transition point.
The E(m) curve at a=6.30

A is enlarged in the inset. It clearly shows two energy
0 0.05 0.1 0.15 0.2 0.25
Magnetic moment (

/Fe)
-0.10
-0.05
0
0.05
0.10
0.15
0.20
E
n
e
r
g
y

(
m
H
a
r
t
r
e
e
/
f
.
u
.
)
0 0.05 0.1
Magnetic moment (

/Fe)
0
1
2
E
n
e
r
g
y

(

H
a
r
t
r
e
e
/
f
.
u
.
)
6.42 ()
6.34 ()
6.30()
6.28 ()
Figure 4.11: The FSM curves of ZrFe
2
at the lattice constants around 6.30

A.
The inset shows the enlarged curve at the lattice constant a=6.30

A. It clearly
shows that magnetic and nonmagnetic solutions coexist at this lattice constant.
The data in this gure are obtained with 3107 k-points in the IBZ.
minima at m=0 and m=0.085
B
/Fe. The DOS of the related nonmagnetic and
magnetic solutions is shown in Figure 4.12. It is clear that the Fermi level (the
3
This is just the Taylors expansion of Equation (2.53) in Chapter 2. This analytic de-
scription breaks down if the van Hove singularity crosses the Fermi level.
4
a
5
or some higher a
n
should always be positive in order to have a nite moment solution,
although it is dicult to determine the sign of it by the topology of the DOS.
52 4.4 Ground state properties of AFe
2
dashed vertical line in the gure) is at a dip (between two peaks marked by two
ellipses) of the nonmagnetic DOS. In the magnetic solution, the two subbands
are shifted against each other as shown by the dashed horizontal arrows.
-0.2 -0.1 0 0.1 0.2
E (eV)
4
2
0
2
4
6
D
O
S

(
1
/
e
V
.
u
.
c
.
)
ferromagntic
nonmagnetic
Figure 4.12: The DOS of nonmagnetic state (dashed lines) and ferromagnetic
state (red lines) of ZrFe
2
at a=6.30

A. The horizontal dashed arrows show
the relative shift of the DOS of the up and down spin subbands. The two
ellipses indicate the two peaks around the Fermi level which cause the rst
order magnetic transition.
The other example is YFe
2
where the magnetic transition is of second order.
The FSM curves are shown in Figure 4.13. The energy minimum moves to
zero when compressing the lattice as shown in the gure. The energy curve
at a=5.99

A is zoomed in and shown in the inset. The FSM energy dierence
for small magnetic moments reaches the accuracy limit guaranteed by the code.
This is the reason that we should resort to the DOS in order to discuss the
possible magnetic solutions. The DOS of nonmagnetic and ferromagnetic states
are shown in Figure 4.14. It is clear that the valley character around the
Fermi level is missing compared with Figure 4.12. Rather, E
F
is situated at a
plateau which can not have more than one magnetic solutions. The other two
compounds, LuFe
2
and HfFe
2
show a similar second order transition.
Thus in the Y, Hf and Lu compounds, we obtain a second order quantum
phase transition (QPT), but in the Zr compound, we obtain a rst order QPT.
It will be quite interesting for the experimentalist to perform high pressure (tens
of GPa) measurements, comparing the magnetic and transport properties in this
series of compounds. It can help to reveal the analogies and dierences in the
QPT.
4 Magnetic transitions of AFe
2
under pressure 53
0 0.05 0.1 0.15
Magnetic moment (

/Fe)
-10
0
10
20
30
40
50
E
n
e
r
g
y

(

H
a
r
t
r
e
e
/
f
.
u
.
)
0 0.02 0.04
Magnetic moment (
B
/Fe)
0.1 (Hartree)
5.90
5.99
6.00
6.02
Figure 4.13: The FSM energy of YFe
2
at lattice constants around a=6.00

A.
The inset shows the zoomed-in curve at the lattice constant of 5.99

A with
an error bar of 0.01 Hartree. The data in this gure are obtained with 8797
k-points in the IBZ.
-0.4 -0.2 0 0.2 0.4
E (eV)
10
5
0
5
10
15
D
O
S

(
1
/
e
V
.
u
.
c
.
)
ferromagnetic
nonmagnetic
Figure 4.14: The DOS of the nonmagnetic state (dashed lines) and the ferro-
magnetic state (solid lines) of YFe
2
at a=5.99

A. The horizontal dashed arrows
show the relative shift of the DOS of the up and down spin subbands. The
Fermi level is shown by the vertical dashed line.
4.5 Relationship between Invar behavior and mag-
netic transitions
Invar alloys have their importance in modern industry, especially in precise
instruments. The inventor Ch. E. Guillaume was award the Nobel prize in
54 4.5 Relationship between Invar behavior and magnetic transitions
LS
HS
V
V
LS
HS
(b)
HS
LS
CS
(f)
V
0
(e)
HS
LS
CS
(h)
HS
LS
CS
(g)
LS
HS
V
V
LS
HS
c) ( E
n
e
r
g
y
V
o
l
u
m
e
T(arb. unit)
V(arb. unit)
LS
HS
V V
LS HS
(d)
V
0
(a)
Figure 4.15: A schematic illustration of the 2-model. (a) is for simple metal,
(b) an Invar system, (c) an anti-Invar system, and (d) a spin glass state. The
corresponding volume vs. temperature behaviors are shown in (e), (f), (g), and
(h). V
0
, V
LS,(HS)
denote the equilibrium volume for normal metal, LS (HS)
states. CS refers to the composite LS and HS behaviors. This gure is adapted
from Moruzzi [61].
physics in 1920. Understanding the Invar eect, however, has been a problem for
a half of the century. More than twenty dierent models have been published in
the past 50 years for the explanation of the Invar eect. A general review about
the Invar eect can be found, for example, in handbooks edited by Buschow
and Wohlfarth [56], and references therein. One model called 2-model [60] is
based on the hypothesis of Weiss that there exist two separated energy minima
with dierent volumes and magnetic states: HS-high-volume and LS-low-volume
states. Here we introduce this model qualitatively and extend this model a little
bit to include all possible relative energy minima of the two states. The binding
curves of the two states are schematically shown in the upper panel in Figure
4.15 together with the corresponding thermal expansion curves in the lower
panel. In Figure 4.15(a) nonmagnetic metal binding curves in a rigid lattice is
shown. When the temperature dependent lattice vibration is included, we have
the volume versus temperature curve in (e). At high temperatures, the thermal
volume expansion coecient is a constant. When extrapolating to T = 0, it
derives from linearity. When there are two magnetic states (HS and LS) in the
binding curves, with slightly dierent energy minima, we can expect Invar or
anti-Invar eect depending on which brunch is lower in energy. If the energy
minimum of the LS state is slightly higher than that of the HS state as in (b),
we have the volume versus temperature curve like (f). At low temperature, the
HS state with larger volume is the ground state. Increase of the temperature
will excite the state into LS state. Because this state has a lower volume, the
thermal expansion of the volume is compensated. Thus a plateau in the volume
versus temperature curve is developed. This is the Invar eect in the Weiss
2- model. The energy dierence was found to be 0.0355 eV for Fe [60]. If
the HS state and LS state have the binding curves with reversed energy minima
4 Magnetic transitions of AFe
2
under pressure 55
order like (c), the thermal expansion is expected to be enhanced as shown in
(g). This is called anti-Invar. If the minima of the energy in the LS and HS are
degenerate, the system then consists of a matrix of droplets with very dierent
magnetic behaviors and with large internal stresses at the droplet boundaries.
Such a system might show a spin-glass behavior.
First principle calculations for Fe
3
Ni by Entel [62] and other authors sup-
ported the 2-model. Entel argued that the special position of the Fermi level in
the minority band, being at the crossover between nonbonding and antibond-
ing states, is responsible for the tendency of most Invar systems to undergo
a martensitic phase transition. Two minima binding curves should lead to
some discontinuity (a rst order transition) in the pressure dependence of some
physical properties, such as volume, magnetic moment etc., but this kind of
discontinuity has never been observed in Invar alloys. This gives an obstacle in
applying the 2-model to explain the Invar eect.
The HS-LS transition can also be continuous and it is in the Invar alloy
like ZrFe
2
and HfFe
2
as in Figure 4.8, according to our LSDA calculation. This
point can be illustrated by our FSM calculations. In the FSM energy curves, the
energy minimum shifts to the lower magnetic moments as the lattice constant
is decreased as in Figure 4.16. Here the FSM energy curves of ZrFe
2
is taken
as an example. The quite at FSM energy curves, which means a large spin
susceptibility, near the transition region are because that the average DOS at
the Fermi level, dened in Equ. (4.5), is large. The reciprocal susceptibility,

1
M
= E

(M), is given by a simple formula [49]

1
M
=
2
B
(2N
1
eff
I), (4.12)
where I is the Stoner parameter. Thermal excitations cause loss of the magnetic
0 1 2 3 4 5 6 7
Magnetic moment(
B
/unit cell)
-35
-30
-25
-20
-15
E
(
-
1
2
2
7
3

m
H
a
r
t
r
e
e
/
u
.
c
.
)
a=6.96 ()
a=6.92 ()
a=6.88 ()
a=6.84 ()
a=6.80 ()
a=6.70 ()
Figure 4.16: The FSM energy curves of ZrFe
2
near the HS-LS transition regions.
moment leading to a magnetic transition from the HS state to the LS state.
Therefore, increasing the temperature leads to a gradual loss of the spontaneous
volume expansion associated with the ferromagnetic state. This gradual process,
contrary to the two states (HS and LS) in some Invar alloy (e.g. Fe
3
Ni), will
not cause any discontinuity in the pressure dependence of physical properties.
The spontaneous volume magnetostriction is calculated by Equation (4.1).
The volumes of dierent magnetic states are provided in Table 4.2. The results
56 4.6 Doping eects
are listed in Table 4.3, together with the experimental data available [31]. The
theoretical values agree with the experimental ones in the sense that they are at
the same order. The overshooting of the spontaneous volume magnetostriction
Table 4.3: The spontaneous volume magnetostriction from our LSDA calcula-
tions (
s
) and experiments (
exp
s
).
AFe
2

s

exp
s
(10
3
) (10
3
)
YFe
2
27 -
ZrFe
2
31 10
HfFe
2
35 8
LuFe
2
25 -
(
s
) can partly be from the non-vanishing local magnetic moment above the
transition temperature in the experiments, while in our model it is in a Pauli
paramagnetic state where the spin moment is zero. The cure for this problem
requires a more realistic treatment of the paramagnetic phase. It has been
shown that a noncollinear [63] or a disordered local moment (DLM) [64, 65]
model gives a better agreement with the experiments. Nevertheless, the results
presented here show the major characteristics of Invar alloy: Compared with the
compounds where no Invar anomaly is observed, the
s
is larger. In typical Invar
alloy, such as Ni
35
Fe
65
and Fe
72
Pt
28
,
s
(10
3
) = 18 and 14.4 [31], respectively.
But this is not the full story. We see that the values of
s
of YFe
2
and LuFe
2
are also large. Why do they not show Invar anomalies? In order to show the
Invar anomaly, the rapid decrease of the magnetic moment should be near the
equilibrium lattice constant at ambient conditions. This requirement excludes
the Y, Lu compounds to be Invar alloy. In our compounds ZrFe
2
and HfFe
2
the
gradual decrease of the magnetic moment is the essential dierence, compared
with the discontinuity present in a typical Invar system as Fe
3
Ni. How to develop
an unied Invar theory to include the dierences of the Invar alloys is still an
open question.
4.6 Doping eects
Doping in ferromagnetic compounds can introduce interesting phenomena, such
as metamagnetism or suppressions of ferromagnetism. Concerning doping into
ZrFe
2
there was a long-standing problem, namely, the presence of a homogeneity
range of the ZrFe
2
Laves phase which was already pointed out some 40 years
ago. Then, a certain scatter in the properties was reported, caused by uncertain-
ties in compositions. Reports on the binary Fe-Zr phase diagram also exhibit
discrepancies with respect to the phases formed as well as to the extension of
homogeneity ranges. In order to clarify this problem, experiments [66] were
carried out recently. The main experimental results show that the homogeneity
regime extends to 74 at.% Fe content without formation of a secondary phase
or a structure change. The substitution takes place at the Zr site: by doping
with Fe, the Zr is partially substituted. The magnetic moment per Fe and also
4 Magnetic transitions of AFe
2
under pressure 57
the ferromagnetic Curie temperature increase with the increase of Fe content,
on the other hand, the lattice constant decreases.
In order to understand these behaviors within the homogeneity region, we
performed LSDA calculations where the atomic substitution has been modeled
by Coherent Potential Approximations (CPA). The general idea of the CPA
approach is to formulate an eective (or coherent) potential which, when placed
on every site of the alloy lattice, will mimic the electronic properties of the
actual alloy. Detailed implementation of CPA in FPLO can be found in the
paper by Koepernik et al. [67]. The valence basis sets comprised 3sp/3d4sp
states for iron and 3d4sp/4d5sp states for zirconium. The local spin density
approximation (LSDA) in the parameterization of Perdew and Wang 92 [7] was
used in all calculations. The Fe is at 16d site, and Zr and the doped Fe are both
at 8a site with the corresponding concentrations according to the CPA setups.
The number of k-points in the irreducible wedge of the Brillouin zone was set
to 200. Energy convergence at the level of 10
7
Hartree was achieved during
the self-consistent iterations.
The calculated lattice constants and the experimental ones are listed in the
Table 4.4. It is quite obvious from Table 4.4 that doping with irons results in a
Table 4.4: The experimental (a
exp
0
) and theoretical lattice constants (a
LDA
0
) of
Zr
x
Fe
100x
. The experimental values are taken at room temperature.
Zr
x
Fe
100x
a
exp
0
(

A) a
LDA
0
(

A)
Zr
33
Fe
67
7.0757 6.85
Zr
30
Fe
70
7.0570 6.82
Zr
28
Fe
72
7.0342 6.77
decrease of the lattice constant. This can be explained by the fact that Fe has
a smaller atomic radius than Zr. As we discussed before, the lattice constant
of this compound is determined by the volume of the A atom. Doping atoms
with a smaller volume decrease the average atomic volume at the A site, so the
lattice constant is decreased.
In the stoichiometric compound ZrFe
2
, the calculated moments amount to
1.65
B
for Fe
I
atoms at 16d sites and -0.50
B
for the Zr
II
atoms at 8a sites,
respectively. In the non-stoichiometric compounds, the excess Fe
II
atoms at
the 8a sites exhibit an enhanced magnetic moment which increases further with
the Fe content as shown in Figure 4.17. The magnetic moment of Fe
II
is close
to that of the BCC Fe (2.2
B
). The related site-resolved density of states
shows the characteristics of strong ferromagnetism in the Fe
II
sublattice (fully
occupied majority subband, see Figure 4.18). It has no humps at the Fermi
level and only a small tail above it. The nearest neighbor of Zr in the C15
phase is Fe
I
and the second nearest neighbor is the doped Fe
II
. This leads to a
stronger hybridization between the nonmagnetic Zr and Fe
I
than that between
Zr and Fe
II
, which is quite visible in the DOS shown in Figure 4.18. The DOS
of Zr resonates mostly with that of Fe
I
instead of Fe
II
. Moreover, the nearest
neighbors of the doped Fe
II
atoms are Fe
I
atoms with a distance larger than
the distance between Fe
I
and its neighboring Fe
I
. Thus the atomic volume of
Fe
II
is larger than that of Fe
I
and the related bands are narrower. Both facts
58 4.6 Doping eects
6768697071727374
Fe-contentxinFeZr
x1-x
Zrat8a
Feat16d
I
Feat8a
II
M
a
g
n
e
t
i
c

m
o
m
e
n
t

(
)
m
B
2.5
2.0
1.5
-0.50
-0.55
Figure 4.17: The magnetic moments of Fe
I, II
and Zr versus the atomic concen-
tration of Fe. The results are obtained at the experimental lattice constants.
-8 -6 -4 -2 0 2 4
E (eV)
10
5
0
5
D
O
S

(
1
/
e
V
.
s
p
i
n
.
f
.
u
.
)
Fe
I
Zr
10*Fe
II
Total
Figure 4.18: The total DOS and the partial DOS of the doped compounds with
composition Zr
0.9
Fe
2.1
at the lattice constant a = 6.825

A.
provide a reason for the larger spin moment on the Fe
II
sites in comparison
with the Fe
I
sites, as in Figure 4.17. The averaged total magnetic moments per
iron atom calculated at the respective experimental lattice constants are shown
in Figure 4.19 in comparison with the data measured at 5 K and 300 K. The
4 Magnetic transitions of AFe
2
under pressure 59
Figure 4.19: The measured and calculated magnetic moments per Fe versus the
atomic concentration of Zr. The lines are for guiding the eyes. The results are
obtained at the experimental lattice constant.
deviations in the absolute values are within 5% and the LSDA underestimates
the magnetic moment. About half of this dierence can be attributed to the
neglect of the orbital moments in the present scalar relativistic calculations. The
linear composition dependence of the magnetic moment is in a good agreement
with the experimental data showing almost the same slopes.
60 4.6 Doping eects
Chapter 5
Magnetic transitions in
CoO under high pressure
5.1 Introduction
Behavior of transitional metal monoxides has attracted a lot of experimental
and theoretical interest. As a property of ground states, band gaps and posi-
tions of the Fermi level should be reproduced by DFT. The band gap opening
due to strong electron correlations which is not properly reproduced by LDA
was partly remedied by combination of LDA and model approaches, namely the
LDA+U. Some other functionals were also invented to treat correlation eects
in solids, such as self-interaction corrections (SIC), hybrid functionals, etc. This
development has proven to be quite successful in exploring and explaining in-
teresting physics in strongly correlated systems. For a review on these aspects,
see Ref. [68] by Anisimov et al.
Pressure plays a unique role in tuning correlation eects in solids. The
importance of the electron correlation is measured, to some extent, by the ratio
U/W, where U is the correlation energy of the localized orbitals and W is
the related band width. Generally, W is increased when the pressure becomes
enhanced because of the increased hopping probability of the electrons due to
larger orbital overlap. So the ratio U/W becomes smaller under pressure. On
the other hand, pressure can lead to structural phase transitions, because it
alters the bonding character. Thus the local environment, especially the crystal
eld or ligand eld is changed. All of these changes will impact on magnetism.
How the magnetic moment changes under pressure is an important feature of
magnetic systems. We have already seen that the magnetic moment can behave
dierently in itinerant electron systems. The transition to a nonmagnetic state
can be of a rst or a second order. There are at least two reasons for such a
transition. On one hand, it can arise because the bands broaden, so that the
Stoner criterion is no longer satised. On the other hand, it can arise because
of structural transitions, so that the local environment is changed.
The magnetic moment behavior under pressure of a Mott insulator, the
transition metal monoxides, can be yet more interesting and complicate. Con-
sidering the parameters which inuence the properties, the electron correlation
energy U, the electron band width W, the spin pairing energy J, the ligand eld
62 5.1 Introduction
strength
l
, and the p d charge transfer energy
pd
, all of these parameters
are of the same order and enter the model Hamiltonian to describe electronic
properties. Especially, the last four parameters are very susceptible to pressure.
Because of this complex situation, quite a lot of work has already been done.
MnO and FeO under high pressure were studied by Fang et al. [69] using GGA
and supplemented by LDA+U. Three spin congurations, nonmagnetic (NM),
ferromagnetic (FM), and antiferromagnetic (AFM) were investigated and the
crystal structures were optimized. It was found that, at high pressure, MnO
should take the NiAs (nB8) structure being either FM or AFM, and in the in-
termediate pressure range, the NM rhombohedrally distorted B1 (rB1) phase
with a stretched distortion can be realized. For FeO under high pressures it
was predicted to have an inverse NiAs (iB8) structure. It transfers from a Mott
insulator to a band insulator under this high pressure. Recently, pressure-driven
magnetic moment collapse in MnO was extensively studied by Kasinathan et
al. It was found that the spin state transforms from S=
5
2
to S=
1
2
under high
pressure [70]. The low spin state was obtained by a spin ip within the e
g
states,
producing a quite anisotropic spin density but a relatively isotropic charge den-
sity. The specic ground state is not only sensitively dependent on U which is
the usual case, but also on J, which inuences the gain in exchange energy from
the anisotropic part of the spin density. The results from calculations using
dierent functionals (LDA+U, SIC, and hybrid functionals) [71] indicate that
the high pressure state (metallic or insulating) and the phase transition pres-
sures are quite sensitive to the functionals. Nevertheless all related calculations
predicted a rst order magnetic transition from HS (S=
5
2
) to LS (S=
1
2
) state.
In CoO the interactions between pressure, structure and magnetism are com-
plicate, leading to dierent structural and magnetic phases under dierent pres-
sures. The transition to a non-magnetic state shows a rst order behavior.
The basic phase diagram summarized from several recent experimental results
is shown in Figure 5.1. At ambient conditions, CoO is a paramagnetic insulator
Pressure (G )
0
80
133
R
e
s
i
s
t
a
n
c
e
(
)

0.6
1.0
9
10
10
Cub.I Rhom.I Rhom. II Cub. II
Metal
Insulator
Insulator
to metal
Pa
0.1
160
97 140 90 43

I
Figure 5.1: Electrical resistance, structure and magnetic transition under pres-
sure as summarized from recent experiments [72, 73, 74, 75]. I is the satellite
intensity of the X-ray emission spectra (XES).
with a rock salt structure. Below its Neel temperature (T
N
=290 K), it or-
ders antiferromagnetically along the 111) direction and preserves an insulating
state. Accompanying the magnetic order, it distorts into a tetragonal structure
with a small rhombohedral distortion. The magnetic structure [76] investigated
5 Magnetic transitions in CoO under high pressure 63
by synchrotron magnetic X-ray scattering reveals signicant reections by the
magnetic moment propagating along the tetragonal as well as along the rhom-
bohedral axis. Thus, the magnetic symmetry is monoclinic, in accordance with
the distorted lattice symmetry which is also monoclinic. On applying hydro-
static pressure, the ambient temperature rock salt crystal transforms into a
rhombohedral structure at about 43 GPa [72]. When the pressure is increased
to 90 GPa, a sudden decrease of the volume by 2.7% is observed by highly pre-
cise synchrotron X-ray diraction [73], which results in a transition to a denser
rhombohedral phase (Rhom. II). Above a pressure as high as 133 GPa, the com-
pound returns to its cubic crystal structure (Cub. II). The Neel temperature
T
N
is also increased under pressure. But the pressure induced structural phase
transition cannot be so simply explained by the onset of magnetism with the
increase of the Neel temperature, because of the lack of tetragonal distortion
under pressure. Under pressure as high as about 140 GPa, a magnetic transition
from high spin to low spin state takes place [74]. The magnetic moment persists
with a lower value but not equal to zero which is evident from the non-vanishing
intensity of K

emission line (3p1s) from the transitional metal atom. The

change of the intensity of the X-ray emission spectroscopy (XES) under pressure
is shown in Figure 5.1, where the intensity of the satellite line drops to 60% of
the intensity at ambient pressure. The existence of the satellite lines indicates
that the magnetic moment does not totally collapse. On releasing the pressure
the magnetic HS state recovers at a pressure of about 97 GPa. The magnetic
moment transition shows hysteresis.
The collapse of the magnetic moment was proposed to be related with a
re-occupation from an E
g
majority spin state to a T
2g
minority spin state, i.e.,
from a HS state with spin S=
3
2
(T
5
2g
E
2
g
) to a LS state with S=
1
2
(T
6
2g
E
1
g
). Recent
electrical resistance measurements at room temperature [75] show that under
pressure above 133 GPa, the material shows metallic behaviors (increase of the
resistance under increase of the temperature). Below that, the material shows
a transition state between insulators and metals. There is a drastic decrease
of the resistance at a pressure of about 43 GPa, as shown in Figure 5.1, which
corresponds to the structural phase transition from the Cub. I phase to the
Rhom. I phase, but the electronic reason is unrevealed.
Summarizing the results, the experiments show that the crystal structure
and the magnetic structure under pressure are complicate. There are several
successive structural, magnetic and electronic phase transitions, which are re-
lated to each other. With the hope to elucidate the behavior of CoO under
pressure we resort to the theoretical description.
Early theoretical calculations for CoO were mainly concentrated on its elec-
tronic structure at ambient condition. LSDA+U was rst used by Anisimov [77]
to study the band gap opening problem in this material due to the on-site corre-
lation. Lattice dynamics of CoO was studied by LSDA+U [78] where a sensitive
U dependence of the lattice dynamics was found. The acoustic branches of the
phonon dispersion agree well with the experiments. Recently, hybrid exchange-
correlation energy functionals were used and proved to be quite successful in
this compound as an alternative to LSDA+U to study strongly correlated elec-
trons [79]. LSDA (GGA) calculations by Cohen et al. [80] show that under a
pressure of about 18 (88) GPa, the magnetic moment collapses to zero, which
does not agree with recent experiments. At some intermediate pressures, they
found that there exists a low spin solution with m 0.3
B
/Co obtained from a
64 5.2 A brief introduction to LSDA+U
generalized Stoner theory, but the details of the electronic structure of this phase
and how it is developed were not shown. Of course the insulating behavior at
ambient pressure and moderate pressures was not reproduced by LSDA (GGA).
Thus, the LSDA might also fail to describe the magnetic behavior, especially
the processes of this transition.
In this chapter, the LSDA+U functional is used to investigate the magnetic
moment behavior in this strongly correlated electron compound CoO. The re-
sults reveal that ligand eld splits of the 3d electrons play an important role in
the magnetic moment transition at variance with the discussions based on the
LSDA results of Cohen et al. [80].
5.2 A brief introduction to LSDA+U
In order to improve the treatment of strongly correlated systems, the combi-
nation of the LSDA and many body theory (mainly the Hubbard model) was
proposed by Anisimov [77, 81]. The energy functional is generally expressed by
E
LSDA+U
= E
LSDA
+E
int
E
dc
, (5.1)
where E
LSDA
is the LSDA functional, E
int
is the Coulombic interaction between
the correlated local orbitals, and E
dc
is the double counting term which excludes
the energy already included in the LSDA. Depending on the treatment of E
dc
,
there are two main versions of the functionals, namely the around the mean
eld (AMF) version invented by Anisimov himself and the atomic limit (AL)
introduced by Czy zyk and Sawatzky [82]. The latter version roughly shifts the
energy of the unoccupied orbitals upwards by (UJ)/2 and that of the occupied
orbitals downwards by the same amount, while the former version gives nearly
nothing for a half-lled fully spin polarized shell.
The U-functional is incorporated into the Kohn-Sham scheme through the
orbital related K-part of the functional as explained in Section 2.3.
K = t +e
H
+e
U,AMF(AL)
(5.2)
where
t +e
H
=

i
n
i

i
[

t[
i
) +

ij
n
i
n
j
2

i

j
[ w[
i

j
), (5.3)
e
U,AMF
=
1
2

R,

(,

-[ w[,

-)( n

)( n

)+
[(,

[ w[,

) (,

[ w[

, )]( n

)( n

)
=
1
2

(,

-[ w[,

-) n

+
[(,

[ w[,

) (,

[ w[

, )] n

1
2

R
U(N n

) J(N

)N

, (5.4)
e
U,AL
= e
U,AMF
+
1
2

R
(U J)(1 n

)N

, (5.5)
5 Magnetic transitions in CoO under high pressure 65
where w is the screened Coulombic interaction, and [) indicates the correlated
orbitals with spin . R is the position of the correlated orbital. N is the
number of electrons occupying the whole correlated shell, and N

is that for one

spin sort.
i
is the Kohn-Sham orbitals with an occupation number n
i
. U and J
are the correlation and exchange parameters which are input parameters in this
scheme. They are used to evaluate the Coulombic integrals via their relationship
with the Slater integrals F
l
. In the above equations, the occupation matrix n
of the correlated shells is to be determined from the Kohn-Sham orbitals and
their occupation numbers:
n
mm

i
(m[
i
)n
i

i
[m

), (5.6)
where [m) are the localized correlated orbitals used in the program. It can be
diagonalized by a unitary transformation
n
mm
=

U

m
n

m
, (5.7)
where U

m
is an unitary matrix. The averages over correlated shells and spins
read:
n

=
1
2l
c
+ 1

, n =
1
2
( n
+
+ n

), (5.8)
with 2l
c
+ 1 the total number of the correlated orbitals.
With the functionals above, the U-potentials of the two versions are:
V
AMF

=
e
U,AMF
n

(,

-[ w[,

-)( n

)+
[(,

[ w[,

) (,

[ w[

, )]( n

) (5.9)
and
V
AL

=
e
U,AL
n

= V
AMF

(U J)( n

1
2
). (5.10)
Both these versions of the LSDA+U were implemented into the current
FPLO code. Technical details can be found in Reference [2] and [83].
5.3 A Category of insulators
Restrictions to zero temperature and small external elds allow us to identify
two basic categories of insulators [84], namely,
insulators due to the electron-ion interaction, and
insulators due to the electron-electron interaction.
The rst category subsumes all insulators which can be understood in terms of
single-electron theory. The second one, which includes all the rest, is concep-
tually dierent. In this category we will have to deal with the many-electron
problem. The Mott insulator is due to the electron-electron interaction. We have
seen in the previous section that the strong electron correlation will split the
half lled d band. This splitting is modeled by a parameter U in the Hubbard
66 5.4 Calculational parameters
U
E
F
charge
gap
d-band
E
p-band
U
E
D =| - | e e
dp
D
d-band
U
U
charge
gap
(a)
(b)
pd
pd
p-band
Figure 5.2: A schematic illustration of energy levels for (a) a Mott-Hubbard
insulator and (b) a charge-transfer insulator generated by the Coulumbic inter-
actions of localized d-orbitals. This gure is adapted from Ref [85].
model or the related mean-eld approximation, like LSDA+U. The system is
called a Mott-Hubbard insulator as long as U is smaller than energy splitting of
the O-2p and 3d bands
pd
= [
d

p
[. In this case, the energy gap is situated
in the d band. In contrast, if
pd
is smaller than U, the gap is between the O-2p
and the upper Hubbard part of the 3d band. Thus, any hole excitation goes
into the O-2p shell. This type of compound is called a charge-transfer insulator.
The energy levels for these two kinds of Mott insulators are schematically drawn
in Figure 5.2.
5.4 Calculational parameters
CoO crystalizes into the rock salt (B1) structure at room temperature. The
magnetic moments of Co are ferromagnetically coupled in the 111 planes and
antiferromagnetically coupled between the planes. If we ignore the tetragonal
distortion, we obtain a trigonal space group (R

3m). The two antiferromagneti-

cally coupled Co atoms occupy the Wycko positions 1a (0, 0, 0) and 1b (1/2,
1/2, 1/2) respectively, while the O atoms occupy 2c (1/4, 1/4, 1/4). The con-
ventional unit cell is shown in Figure 5.3. The exchange correlation functional
is parameterized according to Perdew-Wang 92 [7]. The number of k-points in
the irreducible wedge of the BZ is set to be 11076. The energy convergency
criterion is set to be 10
8
Hartree. The atomic limit version of the U-functional
is used with the orthogonal projection method. This choice of functional gives
a better agreement with the experiments for the HS state compared with the
AMF functional.
Before presenting the details of our calculation, we illustrate the related co-
ordinates and the symmetry frequently used in the following discussions [71].
Because of the antiferromagnetic coupling between the 111 layers of Co, the
cubic crystalline symmetry is reduced to rhombohedral one if we leave out the
5 Magnetic transitions in CoO under high pressure 67
Co
Co
O
O
Figure 5.3: The conventional unit cell of CoO in the antiferromagnetic state.
The red and green Co atoms are antiferromagnetically coupled. The Z axis is
the diagonal axis in the cubic rock salt crystal.
lattice distortions. From now on we use capital letters (T
2g
, E
g
) to indicate the
representations of the cubic symmetry, and lower cases (e
g,1
, e
g,2
, a
g
) for the
representation of the rhombohedral symmetry. In the cubic symmetry the Co
3d states split into irreducible representations (IRs) denoted by T
2g
(threefold
degenerate) and E
g
(twofold degenerate). The rhombohedral site symmetry re-
sults in the three IRs a
g
, e
g,1
and e
g,2
with the rst one being one dimensional
and the latter two being both twofold degenerate. In our calculation, the coor-
dinates of the cubic lattice and the rhombohedral one (with superscript c and
r, respectively) are related by:
_
_
x
r
y
r
z
r
_
_
=
_
_
_
1

6
1

2
1

2
0
1

3
1

3
1

3
_
_
_
_
_
x
c
y
c
z
c
_
_
. (5.11)
68 5.5 LSDA pictures
-8 -6 -4 -2 0
E (eV)
8
4
0
4
8
D
O
S

(
1
/
e
V
.
a
t
o
m
)
O-2p
Co-3d: T
2g
Co-3d: E
g
Figure 5.4: LSDA-DOS of the 3d states of Co and 2p states of O. The d states
are resolved into the irreducible representations of the cubic symmetry. The
Fermi level is shown by the dashed vertical line at E=0 eV.
Thus the 3d orbitals are transformed like:
d
r
xy
=
1

3
(d
c
xy
d
c
yz
d
c
x
2
y
2), (5.12)
d
r
yz
=
1

6
(d
c
yz
d
c
xz
)
_
2
3
d
c
x
2
y
2, (5.13)
d
r
xz
=

2
3
d
c
xy

1
3

2
(d
c
xz
+d
c
yz
)
_
2
3
d
c
z
2, (5.14)
d
r
x
2
y
2 =
1
3
(d
c
xz
+d
c
yz
2d
c
xy
)
1

3
d
c
z
2, (5.15)
d
r
z
2 =
1

3
(d
c
xy
+d
c
yz
+d
c
xz
). (5.16)
The orbital quantum number is corresponding to the symmetrized d orbitals
as shown in the Appendix: [m
l
[ = 2 d
r
xy
, d
r
x
2
y
2
, [m
l
[ = 1 d
r
xz
, d
r
yz
, and
m
l
= 0 d
r
z
2
. Note that if there are components of the crystal eld that are
not diagonal in the L=(2,m
l
) basis, these states will be mixed. For example,
the IRs e
g,1
and e
g,2
are a mixture of [m
l
[ = 2 and [m
l
[ = 1, while a
g
contains
only m
l
= 0.
5.5 LSDA pictures
By LSDA, we obtain an equilibrium volume of 118.7 a.u. per formula unit
(f.u.), which is much smaller (9%) than the experimental one (130.5 a.u./f.u.).
The density of states of the Co-3d and O-2p bands are presented in Figure 5.4
according to the irreducible representations of the site symmetry (E
g
and T
2g
).
The calculation has been carried out at the LSDA equilibrium volume. It is
quite clear that the ligand eld separates the E
g
and T
2g
bands. This splitting
amounts to about 1 eV. The exchange splitting is about 2.5 eV. A schematic
plot of the 3d states is shown in Figure 5.5. In our LSDA calculations, all ve
5 Magnetic transitions in CoO under high pressure 69
E
g
T
2g
E
g
T
2g
E
F
l


ex
LSDA
~2.5eV
~1eV
~1eV
l


Figure 5.5: LSDA model DOS of the 3d states of Co in CoO
Co 3d states of the up spin channel are occupied, which is favorable by Hunds
rule. The states from the O-2p and Co-3d are not so much hybridized. The
O-2p states are located about 3 eV below Co-3d states.
Under hydrostatic pressure, the magnetic moment decreases, and at last
vanishes under compression of the volume to about 80%. The magnetic moment
under pressure is shown in Figure 5.6. Here we basically reproduce the LSDA
(GGA) results by Cohen [80]. The collapse of the magnetic moment is because
the Stoner criterion is no longer satised. With the increase of the bandwidth
under compression, the exchange is suppressed as discussed in the Stoner model
before. The magnetic moment jumps from a high value to zero.
As discussed by Terakura [86], the partially lled T
2g
state gives rise to
metallic behavior in the LSDA model. The exchange and crystal eld splitting
can produce an insulating state in MnO and NiO, but not in FeO and CoO.
The Fermi level crosses the T
2g
states, because there are two electrons and three
available degenerate orbitals in the minority spin channel. Terakura proposed
that the large orbital moment in these latter two compounds can induce a band
gap, because the orbital and spin interaction will lead to a population imbalance
in the three T
2g
orbitals. A gap will open in the band when two of the states
are occupied and one is empty. They concluded that the band gap problem is
from the LSDA, not from the band picture of the solid. An electronic structure
calculation by Norman [87], where orbital polarization and spin orbit coupling
were included, conrmed this proposal. But the resulting total moment is too
large and the band gap is too small. If the orbital moments were corrected
to its experimental value (about 50% of the calculated value), it would be too
small to open the gap. So Norman concluded that more sophisticated orbital
polarization functionals were desired in order to correctly describe the insulating
state.
70 5.6 LSDA+U pictures
0.6 0.8 1 1.2
V/V
0
0
0.1
0.2
0.3
0.4
E
+
2
9
3
1
.
8
5

(
H
a
r
t
r
e
e
/
u
.
c
.
)
magnetic
Nonmagnetic
0.6 0.8 1 1.2
V/V
0
0
1
2
3
S
p
i
n

M
o
m
e
n
t

(

B
/
C
o
)
Figure 5.6: Magnetic moment of Co and total energy versus the relative volume.
The reference volume V
0
=118.7 a.u..
5.6 LSDA+U pictures
Already the LSDA approach gives the HS to NM transition. Further, using
LSDA in strongly correlated systems under high pressure has its justication:
because of the increase of the hopping probability under pressure, the correlation
plays now a less important role compared with the case of ambient pressure. As
mentioned in the introduction, the importance of correlation is measured by
U/W, where U is the correlation energy of the localized states (3d states of
Co here) and W is the related bandwidth. Nevertheless, the physical picture
of the moment collapse can be dierent if we treat the correlation dierently
(for example, by LSDA+U). In our view, less important is dierent from
non-existing. More importantly, the physical mechanism of the transition
uncovered by the LSDA+U approach may also be dierent from the LSDA
picture.
Dierent view points on the behavior of CoO under pressure were taken in
the literature. As discussed by Cohen [80], the physics underlying the magnetic
collapse could be the Stoner model: the competition between the band energy
and the exchange energy. As proposed by Ohnishi [88], the crystal-eld split-
ting is another candidate for the HS-LS transition in some transitional metal
complexes. As mentioned in the previous section, all parameters (U, W, J,
l
)
are comparable and may play their role in determining the electronic state, so
it is necessary to take as many as possible parameters into account in order
to uncover the physics underlying dierent phenomena. Very recently, a two
orbital model Hamiltonian involving U, J,
l
was solved by Millis et al. [89]
showing that all of these parameters strongly inuence the HS to LS transition.
5.6.1 Electronic structures of the ground state
The energy versus of CoO volume curves obtained by LSDA+U calculations
are shown in Figure 5.7(a), together with the spin moment of Co in the inset.
The curves for dierent value of U (3 eV, 5 eV, and 7.8 eV) and the same J=1
5 Magnetic transitions in CoO under high pressure 71
eV are presented in order to get an impression how the value of U inuences
the results. The smaller the U is, the smaller the magnetic moment is and
the faster the moment under pressure decreases as shown in the inset of Figure
5.7(a). Thus, the reduced screening stabilizes the moment.
70 80 90 100 110 120 130 140 150
Volume (a.u./f.u.)
0
0.1
0.2
0.3
0.4
0.5
E
+
2
9
3
1
.
8

(

H
a
r
t
r
e
e
/
f
.
u
.

)
U=7.8 eV
U=5 eV
U=3 eV
80 100 120 140
Volume ( a.u./f.u. )
1
1.5
2
2.5
3
S
p
i
n

m
o
m
e
n
t

o
f

C
o

(

B
)
(a)
70 80 90 100 110 120 130 140 150
Volume (a.u./f.u.)
0
100
200
300
400
500
P
r
e
s
s
u
r
e

(
G
P
a
)
U=7.8 eV
U=5 eV
U=3 eV
(b)
Figure 5.7: The total energy, spin moment (a) and hydrostatic pressures (b)
under dierent volumes from LSDA+U with dierent value of U (3, 5, and
7.8 eV respectively) and the same J=1 eV. The AL version of LSDA+U is
adopted in the calculations.
The energy versus volume curves are tted to Birchs EOS [90]:
E(V ) = E
0
+
9
8
B
0
V
0
[(
V
0
V
)
2/3
1]
2
+
9
16
B
0
V
0
(B

0
4)V
0
[(
V
0
V
)
2/3
1]
3
, (5.17)
where V
0
and E
0
are the equilibrium volume and energy respectively, and B
0
and
B

0
are the bulk modulus and its pressure derivative respectively. The pressure
dependence on volume is shown in Figure 5.7(b). The equilibrium volume, the
bulk modulus and its derivatives obtained from dierent values of U are listed
in Table 5.1, together with the experimental results in the lowest row. The
LSDA+U improves the equilibrium volume and the bulk modulus compared
72 5.6 LSDA+U pictures
-6 -4 -2 0 2 4
E (eV)
3
2
1
0
1
2
3
D
O
S

(
1
/
e
V
.
a
t
o
m
)
O-2p
Co-3d: T
2g
Co-3d: E
g
Figure 5.8: The Co-3d and O-2p DOS at the equilibrium lattice constant with
U=5 eV. The Co-3d DOS is decomposed into T
2g
and E
g
representations. The
O-2p DOS is shown by the area lled with slashes.
with the LSDA (U = 0). In our case U=5 eV reproduces the experiments best
and this value will be used in further calculations unless specied explicitly.
Because of the strong correlation between the 3d electrons, the occupied and
unoccupied T
2g
levels will be split by an amount of U J as expected in the
AL version of LSDA+U. An insulating state is obtained with a band gap
about 2.6 eV at U=5 eV, which agrees relatively well with the optical band gap
of about 2.4 eV [68].
Table 5.1: Ground state properties of CoO obtained by LSDA+U calculations
with dierent values of U. Experimental data are given in the last row.
U V
0
B
0
B

0
Band gap
(eV) (a.u./f.u.) (GPa) (eV)
0 118.7 235 4.5 0
3 126.3 202 4.2 1.2
5 129.4 198 4.2 2.6
7.8 128.9 208 4.1 3.4
exp 130.5 180 3.8 2.4
The Co-3d DOS in the high spin state is shown in Figure 5.8. The ligand
eld split of the unoccupied T
2g
orbitals and the E
g
orbitals is estimated about
1.0 eV by evaluating the center of gravity of the unoccupied states as shown in
Figure 5.16. Compare with the DOS obtained in the LSDA, Figure 5.4: In the
LSDA, the O-2p states are situated well below (3 eV) the 3d states, while in
the LSDA+U the O-2p states are spread from -7 eV up to 0 eV. The correlation
energy U pushes the occupied d states downwards, thus hybridizations with the
O-2p orbitals are much stronger. We observe that the states between -1 and
0 eV in the majority spin channel originate mainly from the oxygen p states.
5 Magnetic transitions in CoO under high pressure 73
This shows the charge transfer character of the compound. At the same time,
the DOS of the minority d states is also present in the same energy region. So
both of them can contribute to the optical excitations. Experimentally, detailed
resonant photoemission spectroscopy (RPES) measurements [91] aimed at 2p
and 3p core-levels conrms the charge transfer character of CoO. Recent soft
X-ray scattering investigations show that the nal state is 3d
8
L, where L means
a hole in the O-2p shell. The change transfer energy (
pd
) is determined to
be 4.0 eV where the single impurity Anderson model was used to simulate the
experimental spectroscopy [92], which is smaller than the value U=5 eV used
in our calculations. This agrees with the condition of our schematic illustration
of the charge transfer insulator (
pd
< U).
5.6.2 Magnetic transitions in LSDA+U
From the electronic structure calculations, the energy diagram of the 3d states
of Co can be drawn as in Figure 5.9(a). In this and the following schematic
drawings, we neglect the eect of hybridizations between the Co-3d and O-2p
states. This is because, depending on U, the hybridization between O-2p and
Co-3d states is dierent and this will cause noticeable charge transfer between
Co and O. The charge transfer eect is orbital dependent. This dierence of
occupation numbers will change the distance between the relative centers of
gravity of the occupied E
g
and T
2g
states because of U, while the unoccupied
states are not inuenced. The ligand eld splitting of the occupied states is not
so clear now, as can be seen in Figure 5.8. There is a high resonant E
g
peak
at an energy of -5.5 eV. This peak shifts the center of gravity of the E
g
states
signicantly. But it is not from the symmetry reason. This is the reason that
we use the distance between the unoccupied E
g
and T
2g
states to estimate the
ligand eld splitting in the forthcoming discussions.
Under ambient pressure, the electrons will take the occupation favored by
Hunds rst rule, because the ligand eld splitting
l
between the E
g
and T
2g
levels is small ( 1 eV). If we put the energy of the high spin states to zero,
the ip of one spin in E
g
state to T
2g
state gives an energy contribution of
E =
ex

l
= 2J
l
. If
l
< 2J, the system will stay in the high spin
state, otherwise, it will transform into a state with lower spin. Under pressure,
the ligand eld splitting
l
is enlarged because of stronger hybridization with
the ligands, and the E
g
and T
2g
states are pushed further apart. The PDOS
under pressure is shown in Figure 5.10. The model DOS of this situation is
shown in Figure 5.9(b).
There are all together seven electrons in the atomic Co-3d shell. According
to Hunds rst rule, the spin up shell is fully occupied. The spin down channel
is occupied with two electrons. The spin moment in the atomic ground state
is 3
B
. In LSDA+U, the DOS shows that the majority spin states are fully
occupied as well. For the minority spins, we can see from Figure 5.8 that the E
g
states are empty and the T
2g
states are partially occupied. When switched into
the IR of the rhombohedral lattice, for the minority spin state, the a
g
(m
l
= 0)
states are unoccupied and one of the e
g
bands is unoccupied, while the other
is fully occupied. This occupation pattern results in a DOS as shown in Figure
5.11.
As shown by Korotin [93], the initial electronic occupation of the localized
orbitals will inuence the nal occupation of the orbitals because of the orbital-
74 5.6 LSDA+U pictures
E
g
T
2g
E
g
T
2g
E=0
l


ex
LSDA+U

l
U-J
E
g
T
2g
E
g
T
2g
l


ex
LSDA+U under pressure

l
U-J
(a) (b)





Figure 5.9: The LSDA+U model DOS (a) and its variation under pressure (b).
-10 -8 -6 -4 -2 0 2 4 6
E (eV)
2
1
0
1
2
D
O
S

(
1
/
e
V
.
C
o
)
Co: 3d-T
2g
Co: 3d-E
g
Figure 5.10: The LSDA+U PDOS at a volume of 90 a.u./f.u., about 70% of the
theoretical equilibrium volume.
5 Magnetic transitions in CoO under high pressure 75
-6 -4 -2 0 2 4
E (eV)
3
2
1
0
1
2
D
O
S

(
1
/
e
V
.
C
o
)
m
l
=0
|m
l
|=2
|m
l
|=1
Figure 5.11: The Co-3d PDOS at the equilibrium lattice constant (the same
calculation as Figure 5.8). It is resolved into dierent angular quantum number
with respect to the rhombohedral lattice.
dependent potential in LSDA+U. We obtain the possible solutions by giving
the initial occupations at the very beginning. In order to explore the possible
low spin solutions, we tried all the initial congurations. There are 3 3 = 9
possibilities if we ip one of the electrons in e
g,1(2)
or a
g
state of the majority
channel to the e
g,1(2)
or a
g
state of the minority channel. After self-consistent
iterations, the initial populations converge into the following two cases:
A ip between m
l
= 0 (a
g
) (LS-I). If we just ip one a
g
electron, then
the original occupied a
g
state in the up-spin channel will be un-occupied.
Because just the a
g
spin ips, the insulating state is unaected. The
energy is higher than the high spin states in all the volume region we
considered as will be shown in Figure 5.15, which means that this kind of
spin ip is energetically unfavorable. This is because that the exchange
interaction is increased by an amount of 2J, but there are no signicant
contributions to the decrease of the electronic static energy, for example,
by decreasing the ligand eld energy.
A ip from m
l
,= 0 to m
l
= 0 (LS-II). In this case, the singlet a
g
spin
down state will be totally occupied. The four e
g,1(2)
spin down states
will be partially occupied by two electrons. Therefore, a metallic state is
obtained and the Fermi level situates inside the admixture of the e
g,1
and
e
g,2
states.
The Co-3d DOS of the LS-I and LS-II low spin solutions are shown in Figure
5.12 (a) and (b), respectively. It can be seen that the ip of the e
g,1(2)
state to
the a
g
state gives a metallic solution. After decomposing the PDOS according
to the cubic symmetry, it turns out that the Fermi level crosses the E
g
orbital
and the T
2g
states are totally occupied as shown in Figure 5.13. In this case
we obtain the model DOS for the low spin solution as in Figure 5.14.
The magnetic moment and the energy versus the relative volume of dierent
solutions are shown in Figure 5.15. It can be seen that with the increase of
76 5.6 LSDA+U pictures
-12 -10 -8 -6 -4 -2 0 2 4 6
E (eV)
1
0
1
2
D
O
S

(
1
/
e
V
.
C
o
)
m
l
=0
|m
l
|=2
|m
l
|=1
(a) The DOS of a
g
a
g
ips (LS-I).
-10 -8 -6 -4 -2 0 2 4
E (eV)
2
1
0
1
D
O
S

(
1
/
e
V
.
C
o
)
m
l
=0
|m
l
|=2
|m
l
|=1
(b) The DOS of e
g
ips (LS-II).
Figure 5.12: The Co-3d PDOS for the two possible low spin solutions. The lled
curve is the PDOS with a
g
symmetry.
-10 -8 -6 -4 -2 0 2 4
E (eV)
4
3
2
1
0
1
2
3
D
O
S

(
1
/
e
V
.
C
o
)
Co-3d: T
2g
Co-3d: E
g
Figure 5.13: The PDOS after the electron with e
g
symmetry ips its spin. It is
the same as Figure 5.12(b), but decomposed according to the cubic symmetry.
5 Magnetic transitions in CoO under high pressure 77
E
g
T
2g
T
2g
E
F
l
LS
~3eV

ex
E
g
l
+(U-J)


Figure 5.14: The LSDA+U model DOS after the magnetic transition (LS-II).
compression, the magnetic moment of the high spin state decreases. The total
energy of the HS state is the lowest until the volume reaches 60% of the equi-
librium volume. Under further compression the low spin solution LS-II takes
over. Because the low spin solution is metallic, the magnetic moment decreases
to zero much faster than in the insulator. The transition pressure determined
by the common tangent of the energy versus volume curves of the two cases is
about 320 GPa. This pressure is much too high compared with the experiments.
The pressure, however, is at least U dependent and we will discuss this later.
The magnetic transition between HS and LS-II is a rst order transition. The
magnetic moment of Co jumps from 2.5
B
to 0.5
B
with a the volume collapse
of about 6.5%.
5.7 The reason for the magnetic transition
Based on our calculation, the reason for the spin ip in CoO can be traced back
to competitions between ligand eld splitting and exchange interactions.
In the ligand eld model, the spin state changes at the crossing point, i.e., the
point at which Hunds rst rule for the electronic ground state of the transition
metal ion breaks down. The crossing point is given by

l
(V
c
) = 2J, (5.18)
where
l
(V
c
) is the ligand eld splitting between E
g
and T
2g
states at the
critical volume V
c
and J is the spin pair energy given by the Slater integrals.
The ligand eld splitting
l
increases as the volume is reduced because of the
increasing interaction with the ligand anion, but J is less aected as it is an
intra-atomic interaction.
Our self-consistent electronic structure calculation conrms the simple ar-
guments above. The center of gravity (E
c
) of a band is dened as
E
c
=
_
ED(E)dE
_
D(E)dE
, (5.19)
78 5.8 Discussions
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
V/V
0
-0.2
0
0.2
0.4
0.6
0.8
1
E
+
2
9
3
1
.
5

(
H
a
r
t
r
e
e
/
f
.
u
.
)
HS
LS-I
LS-II
0.4 0.6 0.8 1
V/V
0
0
0.5
1
1.5
2
2.5
3
S
p
i
n

m
o
m
e
n
t

(

/
C
o
)
Figure 5.15: The total energy and the magnetic moment versus the relative
volume with respect to V
0
=129.4 a.u.. It can be seen that the low spin solution
(LS-II) is energetic favorable under high pressures. The other low spin solution
(LS-I) where the ip of the a
g
electrons takes place is always above the curve
of the high spin state.
where E is the energy and D(E) is the related DOS. The center of gravity of
the unoccupied E
g
state (E
E
g
c
diamond lines) and T
2g
state (E
T
2g
c
circle black
line) are shown in Figure 5.16. It can be clearly observed that the splitting of
the E
g
and T
2g
(cross red line) states is enlarged from 1 eV to 4 eV when
the volume is shrunk to 50% while the average (the star red line) of the two
is only slightly decreased. If we take J = 1 eV for Co, the 60% compression
of the volume where the cross point appears in Figure 5.15 gives
l
= 3.1 eV,
satisfying the simple consideration
l
2J for the magnetic transition.
We notice that the PDOS in the range of -3-1 eV in our trigonal phase
in Figure 5.11 is roughly split into two, namely [m
l
[ = 2 and [m
l
[ = 1. These
states are the mixture of the cubic T
2g
and E
g
states. We note that the e
g,1
states have a larger portion of E
g
, being
2
3
E
g
and only
1
3
T
2g
. The E
g
states
in the cubic system have the strongest (dp) overlap with O ions, and thus are
most aected by the pressure. So under pressure, the e
g,1
state ips its spin
rst. After self-consistency, the electrons are redistributed among the 3d shells
resulting in the full occupation of T
2g
state and the partial occupancy of the
E
g
state. This is favorable by the ligand eld because the T
2g
has the lower
electrostatic interaction with the ligand O anion. The metallic solution comes
from the strong overlapping of the orbitals, thus a wide band width under this
large volume compression, although the unoccupied states and the occupied
states are separated by U J.
5.8 Discussions
In our calculation, the LSDA+U is used, with U=5 eV given ad hoc. We choose
U so as to reproduce the experimental results under ambient pressure, especially
the lattice constant, the bulk modulus and the optical gap, reasonably well. This
5 Magnetic transitions in CoO under high pressure 79
0.5 0.6 0.7 0.8 0.9 1 1.1
V/V
0
0
1
2
3
4
5
E
n
e
r
g
y

l
e
v
e
l

(
e
V
)
E
c
T
2g
E
c
E
g

l
center
Figure 5.16: The evolution of the center of gravity of the unoccupied E
g
state (diamonds) and T
2g
state (circles) under relative volumes with respect
to V
0
=129.4 a.u./f.u.. The split of the two states
l
is shown by the crosses,
while the average of them are shown by the stars. The lines are for guiding eyes.
U is also used when we apply high pressures. Although under high pressure,
the occupation uctuation eect will be enhanced, so that the U should not be
the same, but the pressure dependence of U is not readily available. Our main
results, the spin ip process, or the low spin conguration is not changed for
dierent values of U. For example, we have results for U=3, 5, 7.8 eV. The self-
consistent calculations give the same low spin state T
6
2g
E
1
g
, but U is essential to
stabilize the low spin solution. If we switch o the U, the solutions can only be
reduced to LSDA nonmagnetic case, no matter what the initial density is. The
transition pressure will be changed for dierent values of U of course because
on one hand the energy versus volume curve is largely dependent on U. On
the other hand also the position of the unoccupied states is largely dependent
on U, so that the relative positions of the highest occupied states of the up
spins and the lowest unoccupied down spin state is dependent on U. The gap
is approximately (U J +
ex

l
), cf. Figure 5.9. At least from these two
aspects, the value of U will inuence the transition pressure. In Figure 5.17 the
enthalpy variations with pressure of dierent states obtained by dierent values
of U are shown. The transition pressure is reduced from 680 GPa to 130 GPa
if U is decreased from 7.8 eV to 3 eV. We see that U=3 eV tunes the transition
pressure to the vicinity of the experimental one (140 GPa in Figure 5.1).
In our calculation, no lattice distortion is taken into account. Experimen-
tally, distortions are found both below the Neel temperature and under high
pressures. Our calculations with the experimental distortions but without U
does not give the insulating solutions which means that CoO is indeed not a
band insulator. This distortion does not alter the electronic structure qualita-
tively if we compare the DOS from both. So we neglect the distortions when
we discuss the spin ip processes.
80 5.8 Discussions
600 640 680
0
0.1
0.2
E
n
t
h
a
l
p
y

(
H
a
r
t
r
e
e
/
f
.
u
.
)
U=7.8 eV HS
U=7.8 eV LS-II
300 320 340
U=5 eV HS
U=5 eV LS-II
40 80 120 160
Pressure (GPa)
0.2
0.4
E
n
t
h
a
l
p
y

(
H
a
r
t
r
e
e
/
f
.
u
.
)
U=3 eV HS
U=3 eV LS-II
Pressure (GPa)
P
c
= 680 GPa
P
c
=320 GPa
P
c
=130 GPa
Figure 5.17: The enthalpy of CoO under pressure with dierent values of U.
The point where the two lines cross is denoted by P
c
. This is the pressure at
which the magnetic transition happens.
Chapter 6
Summary and Outlook
In this thesis three kinds of compounds are investigated by density functional
calculations in local spin density approximation.
The rst one is the Heusler compound Rh
2
MnGe. It is found that the cubic
phase is not stable at T = 0. It may undergo tetragonal distortions by either
extending or compressing the lattice in one direction. The electronic reason
for this distortion is identied to be the band Jahn-Teller eect. The Jahn-
Teller active states are found to be the 4d states of Rh placed at the Fermi
level due to magnetic polarization. The tetragonal lattice distortion is clearly
observed by recent low temperature XRD measurements. The phase transition
from the high temperature cubic phase to lower temperature tetragonal phase
is of second order, where the order parameter (c/a-

2) increases continuously
from zero when the temperature is reduced below 180 K. However, the c/a
ratio is much smaller from the experiment than our prediction. The reason may
be a small LSDA error or disorder eects in experiments. In the proposed band
Jahn-Teller scenario, the amount of energy that can be gained by the distortion
is largely dependent on the height of the van Hove singularity near the Fermi
level. Disorder will suppress the singularity, as well as the nite temperature in
the experiments. Thus, the distortion is reduced. Improvement of the quality
of samples of Rh
2
MnGe is favorable to decrease the gap between theory and
experiments. Up to now, we have only the frozen phonon calculation. What
does the phonon dispersions in this compound look like? This will be done
and after that, a more complete discussion about thermodynamics of structural
phase transitions can be made.
The second kind consists of series of isostructural compounds in the cubic
Laves phase: YFe
2
, ZrFe
2
, HfFe
2
, and LuFe
2
. The magnetic moment is found to
decrease when the lattice constant is decreased, and nally disappears in all of
these four compounds. Particularly, there exists a lattice constant range where
the magnetic moment is rapidly reduced by decreasing the lattice constant. The
magnetic moment behavior when approaching zero is categorized into rst order
(in ZrFe
2
) and second order (in YFe
2
, HfFe
2
, and LuFe
2
) phase transition. The
dierence is explained by the dierence of the topology of DOS near the Fermi
level and Landaus expansion of magnetic free energy. As shown by our calcula-
tion, the magnetism is suppressed under high pressures. The Invar anomaly in
two of these compounds (ZrFe
2
and LuFe
2
) is discussed, which show a high spin-
low spin transition close to ambient pressure. As one of the major properties of
82
Invar alloys, the spontaneous volume magnetostriction (
s
) is calculated and it
is found to be comparable with the experiments. The compounds show dierent
moment-volume behavior as compared with typical Invar alloy. In typical Invar
alloy such as Fe
3
Ni, there are double wells of HS and LS state in the binding
curves, but in our Laves phase compounds a continuous transition form HS to LS
is obtained. The lattice constant at which the transition takes place excludes
the other two (YFe
2
and HfFe
2
) from being Invar alloy. In the Laves phase
compounds, the magnetic structure may be complicated due to the geometri-
cally highly frustrated Fe sublattice. Especially when the magnetic moment is
not ferromagnetically coupled, noncollinear magnetic moment arrangement is
expected. Simulation of possible noncollinear magnetism in these compounds
may improve our understanding of the properties, especially the Invar anomaly
in these compounds.
Substitution of Zr by Fe in ZrFe
2
results in the decrease of the lattice constant
and increase of the average magnetic moment per iron atom. The doped Fe takes
the Zr site, and shows strong ferromagnetism, which is the reason for the increase
of the average magnetic moment. The original problem of the homogenous
region of Fe substitutions are going to be investigated theoretically.
At last the HS-LS transition under pressure in CoO is investigated by LSDA+U.
The collapse of the magnetic moment is discerned to be caused by the increase
of the ligand eld splitting under pressure. This increase of the ligand eld split-
ting not only suppresses the intraatomic exchange, but also pushes the upper
Hubbard band downwards (the band gap is decreased). This transition is of no
surprise as it results from the competition between the ligand eld split and the
spin pairing energy. The results give a microscopic picture of the magnetic tran-
sition under the approximation of LSDA+U. We obtained the metallic low spin
state, which is in quite good agreement with recent experiments. The LSDA+U,
however, shows some uncertainty in the transition region from a strongly cor-
related system to a weakly correlated system under ultrahigh pressure, because
the pressure dependent U is not available ab initio. This transition region is no-
toriously dicult in theory. Another parameter J, which enters the LSDA+U,
will have inuence on the exchange energy. This parameter will surely inuence
the transition and the low spin solution. The eect will be investigated. How
will other treatment of the correlation eect aect the HS-LS transitions? This
point can be checked.
The interplay between structure and magnetism produces various states in
itinerant systems as well as in strongly correlated systems. DFT calculations
provide a way to understand the physics underlying them. In this thesis, only
three interesting systems were investigated. Preliminary results show that the
behavior can be explained by LSDA+(U). Investigations can be extended to
the problems as mentioned above.
Bibliography
[1] O. Madelung. Introduction to Solid-State Theory. Springer-Verlag,
Reprinted by Beijing World Publishing Corporation, 2003.
[2] H. Eschrig. The Fundamentals of Density Functional Theory (2nd Edition).
Edition am Gutenbergplatz Leipzig, 2003.
[3] P. Hohenberg and W. Kohn. Inhomogeneous Electron Gas. Phys. Rev.,
136(3B):B864B871, Nov. 1964.
[4] E. H. Lieb. Density Functionals for Coulomb Systems. Int. J. of Quant.
Chem., 24(3):243277, Mar. 1983.
[5] W. Kohn and L. J. Sham. Self-Consistent Equations Including Exchange
and Correlation Eects. Phys. Rev., 140(4A):A1133A1138, Nov 1965.
[6] D. M. Ceperley and B. J. Alder. Ground State of the Electron Gas by a
Stochastic Method. Phys. Rev. Lett., 45(7):566569, Aug. 1980.
[7] John P. Perdew and Y. Wang. Accurate and Simple Analytic Representa-
tion of the Electron-gas Correlation Energy. Phys. Rev. B, 45(23):13244
13249, Jun. 1992.
[8] K. Koepernik and H. Eschrig. Full-potential Nonorthogonal Local-orbital
Minimum-basis Band-structure Scheme. Phys. Rev. B, 59(3):17431757,
Jan. 1999.
[9] N. W. Ashcroft and N. D. Mermin. Solid State Physics. Thomson Learning,
Inc., reprinted by Beijing World Publishing Corporation, 1976.
[10] E. C. Stoner. Collective Electron Ferromagnetism. Proc. Roy. Soc. Lond.
Series A, 165(922):372414, Apr. 1938.
[11] H. Eschrig. Lecture notes on electronic structure of solid. unpublished,
2006.
[12] V. L. Moruzzi. Singular Volume Dependence of Transition-Metal Mag-
netism. Phys. Rev. Lett., 57(17):22112214, Oct. 1986.
[13] D.-K. Seo. Density Functional Perturbational Orbital Theory of Spin
Polarization in Electronic Systems. I. Formalism. J. Chem. Phys.,
125(15):154105, 2006.
[14] G. Stollho, Andrzej M. Oles, and V. Heine. Stoner Exchange Interaction
in Transition Metals. Phys. Rev. B, 41(10):70287041, Apr. 1990.
84 BIBLIOGRAPHY
[15] H. A. Jahn and E. Teller. Stability of Polyatomic Molecules in Degener-
ate Electronic States. I. Orbital Degeneracy . Proc. R. Soc. London A,
161(905):220235, Jul. 1937.
[16] Sir R. Peierls. Selected scientic Papers of Sir Rudolf Peierls with Com-
mentary, volume 19. World Scientic, 1997.
[17] S. Lee and R. Homann. Bcc and Fcc Transition Metals and Alloys: A Cen-
tral Role for the Jahn-Teller Eect in Explaining Their Ideal and Distorted
Structures. J. Am. Chem. Soc., 124(17):48114823, Apr. 2002.
[18] A. Ayuela, J. Enkvaara, K. Ullakko, and R.M. Nieminen. Structural prop-
erties of magnetic Heusler alloys. J.Phys.: Cond. Matt., 11:20172026,
1999.
[19] C. Bungaro, K. M. Rabe, and A. Dal Corso. First-Principles Study of Lat-
tice Instabilities in Ferromagnetic Ni
2
MnGa . Phys. Rev. B, 68(13):134104,
Oct. 2003.
[20] P. Dowben. Half Metallic Ferromagnets. J. Phys.: Cond. Matt.,
19(31):310301(1p), 2007.
[21] F. A. Hames and J. Crangle. Ferromagnetism in Heusler-Type Alloys
Based on Platinum-Group or Palladium-Group Metals. J. Appl. Phys.,
42(4):13361338, 1971.
[22] J. C. Suits. New magnetic compounds with Heusler and Heusler-related
structures. Phys. Rev. B, 14(9):41314135, 1976.
[23] R.A. Dunlap, S. Jha, and G.M. Julian. Sn hyperne eld systematics in
Rh
2
MnZ Heusler alloys. Can. J. Phys., 62:396399, 1984.
[24] M. Pugacheva and A. Jezierski. Dependence of the Magnetic Moment on
the Local Atomic Order in Rh
2
MnX Heusler Alloys . J. Magn. Magn. Mat.,
151:202206, 1995.
[25] I. Galanakis, P. H. Dederichs, and N. Papanikolaou. Slater-Pauling Behav-
ior and Origin of the Half-metallicity of the Full-Heusler Alloys. Phys. Rev.
B, 66(17):174429, Nov 2002.
[26] Y. Adachi, H. Morita, T. Kanomata, A. Sato, H. Yoshida, T. Kaneko, and
H. Nishihara. Pressure Eect on the Curie Temperature of the Heusler
Alloys Rh
2
MnZ(Z=Sn,Ge) . J. Alloy. Comp., 383:3739, 2004.
[27] A. T. Zayak, P. Entel, K. M. Rabe, W. A. Adeagbo, and M. Acet. Anoma-
lous Vibrational Eects in Nonmagnetic and Magnetic Heusler Alloys.
Phys. Rev. B, 72(5):054113, 2005.
[28] J. K ubler, A. R. Williams, and C.B. Sommers. Formation and Coupling of
Magnetic Moments in Heusler Alloys. Phys. Rev.B, 28:17451755, 1983.
[29] S. Fujii, S. Ishida, and S. Asano. Electronic Structure and Lattice Trans-
formation in Ni
2
MnGa and Co
2
MnSn. J. Phys. Soc. Jap., 58:36573665,
1989.
BIBLIOGRAPHY 85
[30] Z. Manuel. Aufklarung magnetisch und thermisch induzierter Struk-
turanderungen mittels Rontgenmethoden. Masters thesis, Technische Uni-
versitat Dresden, 2007.
[31] M. Shiga, Y. Muraoka, and Y. Nakamura. Invar Eects of Laves Phase
Intermetallic Compounds . J. Magn. Magn. Mat., 10:280287, 1979.
[32] M. P. Pasternak, R. D. Taylor, R. Jeanloz, X. Li, J. H. Nguyen, and C. A.
McCammon. High Pressure Collapse of Magnetism in Fe
0.94
O: Mossbauer
Spectroscopy Beyond 100 GPa. Phys. Rev. Lett., 79(25):50465049, Dec.
1997.
[33] E. Duman, M. Acet, and E.F. Wassermann et al. Magnetic Instablities in
Fe
3
C Cementite Particles Observed with Fe K-edge X-ray Circular Dichro-
ism Under Pressure . Phys. Rev. Lett., 94:075502, 2005.
[34] I. A. Troyan, A. G. Gavrilyuk, and V. A. Sarkisyan et al. Transition from
the Antiferromagnetic to a Nonmangetic State in FeBO
3
under High Pres-
sure. JETP Letters, 74:2427, 2001.
[35] G. Wortmann, K. Rupprecht, and H. Giefers. High-pressure Studies of
Magnetism and Lattice Dynamics by Nuclear Resonant Scattering of Syn-
chrotron Radiation. Hyperne interactions, 114/115:103117, 2002.
[36] T. Dumelow, P. C. Riedi, P. Mohn, K. Schwarz, and Y. Yamada. Pressure
Dependence of the Hyperne Field of YFe
2
and ZrFe
2
. J. Magn. Magn.
Mater., 54-57:10811082, 1986.
[37] P. C. Riedi and G. D. Webber. Pressure and Temperature Dependence of
the Hyperne Fields of YFe
2
, YFe
3
and YIG. J. Phys. F: Metal Phys.,
13(5):10571069, 1983.
[38] K. Kai and T. Nakamichi. Magnetic Moment of the Iron Atom in the Laves
Phase Compounds. J.Phys. Soc. Japan, 30:1755, 1971.
[39] J. G. M. Armitage, T. Dumelow, R. H. Mitchell, P. C. Riedi, J. S. Abell,
P. Mohn, and K. Schwarz. Pressure Dependence of the Magnetisation of
YFe
2
and ZrFe
2
: Computation and Experiment. J. Phys. F: Metal Phys.,
16(7):L141L144, 1986.
[40] M. Brouha and K. H. J. Buschow. Pressure Dependence of the Curie
Temperautre of Intermetallic Compounds of Iron and Rare-earth Elements,
Th, and Zr. J. Appl. Phys., 44(4):18131816, 1973.
[41] S. Asano and S. Ishida. Competition of Magnetism in Iron Laves Phase
Compounds. J. Phys.: Cond. Matt., 1(44):85018508, 1989.
[42] H. Yamada and M Shimizu. Electronic Structure and Magnetic Proper-
ties of the Cubic Laves Phase Transition Metal Compounds . Physica B,
149:390402, 1988.
[43] B.M. Klein, W.E. Pickett, D.A. Papaconstantopoulos, and L.L. Boyer.
Electronic Structure, Superconductivity, and Magnetism in the C15 Com-
pounds ZrV
2
, ZrFe
2
, and ZrCo
2
. Phys. Rev. B, 27:67216731, 1982.
86 BIBLIOGRAPHY
[44] P. Mohn and K. Schwarz. Binding Mechanism and Itinerant Magnetism of
ZrFe
2
and YFe
2
. Physica, 130B:2628, 1985.
[45] M. S. S. Brooks, O. Eriksson, and B. Johansson. 3d-5d Band Magnetism in
Rare Earth Transition Metal Intermetallics: LuFe
2
. J. Phys.: Cond. Matt.,
1(34):58615874, 1989.
[46] C. Ritter. Polarised Neutron Study of Magnetic Ordering in the Simple
Alloy YF
2
. J.Phys.:Cond. Matt., 01:27652769, 1989.
[47] V. Heine. s d Interaction in Transition Metals. Phys. Rev., 153(3):673
682, Jan. 1967.
[48] K Schwarz and P Mohn. Itinerant Metamagnetism in YCo
2
. J. Phys. F:
Metal Physics, 14(7):L129L134, 1984.
[49] J. K ubler. Theory of Itinerant Electron Magnetism. Oxford science publi-
cations, 2000.
[50] Roy L. Johnston and R. Homann. Structure-bonding Relationships in the
Laves Phases. Z. Anorg. Allg. Chem., 616:105120, 1992.
[51] M. Isoda and S. Mori. Magnetic Correlation and Breakdown of Fermi
Liquid Picture in Frustrated Intinerant Electron Magnet. J. Phys. Soc.
Japan, 69(5):15091516, 2000.
[52] Walter. A. Harrison. Elementary Electronic Structure (Revised Edition).
World Scientic, 2004.
[53] David R. Lide (editor-in chief). CRC handbook of chemistry and physics,
84th edition. CRC Press, 2003.
[54] J. Mathon. Pressure Dependence of the Magnetization in the Itinerant
Electron Model of Ferromagnetism. J. Phys. F: Metal Phys., 2(1):159168,
1972.
[55] R. Coehoorn. Calculated Electronic Structure And Magnetic Properties of
Y-Fe Compounds. Phys. Rev. B, 39(18):1307213085, Jun 1989.
[56] Wasserman. Ferromagnetic Materials, a Handbook on the Properties of
Magnetically Ordered Substances, edited by K.H.J. Buschow and E.P.
Wohlfarth, volume 5. Elsevier Science Publishers, 1990.
[57] Y. Amako, H. Nagai, Yokoyama T., Yamada H., and Naka T. et al. Pressure
Dependence of the Curie Temperature and the Spontanous Magnetization
in Ferromagnetic Hf(Fe,Co)
2
. Physica B, 382(1-2):189192, June 2006.
[58] C. Peiderer. Why First Order Quantum Phase Transitions Are Interesting.
J. Phys.: Cond. Matt., 17(11):S987S997, 2005.
[59] M. Shimizu. On the Conditions of Ferromagnetism by the Band Model: II.
Proc. Phys. Soc., 86:147157, 1965.
[60] R. J. Weiss. The Origin of the Invar Eect. Proc. Phys. Soc., 82(2):281
288, 1963.
BIBLIOGRAPHY 87
[61] V. L. Moruzzi. High-spin and Low-spin States in Invar and Related Alloys.
Phys. Rev. B, 41(10):69396946, Apr. 1990.
[62] P. Entel, E. Homann, P. Mohn, K. Schwarz, and V. L. Moruzzi. First-
principles Calculations of the Instability Leading to the Invar Eect. Phys.
Rev. B, 47(14):87068720, Apr. 1993.
[63] M. van Schilfgaarde, I.A. Abrikosov, and B. Johansson. Origin of the Invar
Eect in Iron-nickel Alloys. Nature, 400:4649, Jul. 1999.
[64] V. Crisan, P. Entel, H. Ebert, H. Akai, D. D. Johnson, and J. B. Staunton.
Magnetochemical Origin for Invar Anomalies in Iron-nickel Alloys. Phys.
Rev. B, 66(1):014416, Jul. 2002.
[65] S. Khmelevskyi, I. Turek, and P. Mohn. Large Negative Magnetic Contri-
bution to the Thermal Expansion in Iron-Platinum Alloys: Quantitative
Theory of the Invar Eect. Phys. Rev. Lett., 91(3):037201, Jul. 2003.
[66] N. Mattern, W. X. Zhang, S. Roth, H. Reuther, C. Baehtz, and M. Richter.
Structural and magnetic properties of non-stoichiometric Fe
2
Zr. J. Phys.:
Cond. Matt., 19(37):376202 (8pp), 2007.
[67] K. Koepernik, B. Velick y, R. Hayn, and H. Eschrig. Self-consistent LCAO-
CPA Method for Disordered Alloys. Phys. Rev. B, 55(9):57175729, Mar.
1997.
[68] V. I. Anisimov, F. Aryasetiawan, and A. I. Lichtenstein. First-principles
Calculations of the Electronic Structure and Spectra of Strongly Correlated
Systems: the LDA+U method. J. Phys.: Cond. Matt., 9(4):767808, 1997.
[69] Z. Fang, I. V. Solovyev, H. Sawada, and K. Terakura. First-principles Study
on Electronic Structures and Phase Stability of MnO and FeO under High
Pressure. Phys. Rev. B, 59(2):762774, Jan. 1999.
[70] D. Kasinathan, K. Koepernik, and W. E. Pickett. Pressure-driven magnetic
moment collapse in the ground state of MnO. New J. Phys., 9(7):235, 2007.
[71] D. Kasinathan, J. Kunes, K. Koepernik, C. V. Diaconu, R. L. Martin,
I. D. Prodan, G. E. Scuseria, N. Spaldin, L. Petit, T. C. Schulthess, and
W. E. Pickett. Mott Transition of MnO under Pressure: A Comparison of
Correlated Band Theories. Phys. Rev. B, 74(19):195110, 2006.
[72] D. Yang, R. Yang, C. Paul, J. Zhang, S. C. Vogel, B. Winkler, J. Xu,
Y. Zhao, and H.-K. Mao. Pressure-induced Long-range Magnetic Ordering
in Cobalt Oxide. Phys. Rev. B, 74(14):144101, 2006.
[73] Q.Z. Guo, H.-K. Mao, J.Z. Hu, J.F. Shu, and R. J. Hemley. The Phase
Transitions of CoO under Static Pressure to 104 GPa. J. Phys.: Cond.
Matt., 14(44):1136911374, 2002.
[74] J-P Rue, A. Mattila, J. Badro, G. Vanko, and A. Shukla. Electronic
Properties of Transition-metal Oxides under High Pressure Revealed by
X-ray Emission Spectroscopy. J. Phys.: Cond. Matt., 17(11):S717S726,
2005.
88 BIBLIOGRAPHY
[75] T. Atou, M. Kawasaki, and S. Nakajima. Electronic Transition of Cobalt
Monoxide under High-Pressure. Jap. J. Appl. Phys., 43(10A):L1281L1283,
2004.
[76] K. Tomiyasu, T. Inami, and N. Ikeda. Magnetic Structure of CoO
Studied by Neutron and Synchrotron X-Ray Diraction. Phys. Rev. B,
70(18):184411, 2004.
[77] V. I. Anisimov, J. Zaanen, and O. K. Andersen. Band Theory and Mott
Insulators: Hubbard U Instead of Stoner I. Phys. Rev. B, 44(3):943954,
Jul. 1991.
[78] U. D. Wdowik and K. Parlinski. Lattice Dynamics of CoO from First
Principles. Phys. Rev. B, 75(10):104306, 2007.
[79] F. Tran, P. Blaha, K. Schwarz, and P. Novak. Hybrid Exchange-Correlation
Energy Functionals for Strongly Correlated Electrons: Applications to
Transition-Metal Monoxides. Phys. Rev. B, 74(15):155108, 2006.
[80] R. E. Cohen, I. I. Mazin, and D. G. Isaak. Magnetic Collapse in Transi-
tion Metal Oxides at High Pressure: Implications for the Earth. Science,
275:654657, 1997.
[81] V. I. Anisimov, I. V. Solovyev, M. A. Korotin, M. T. Czy zyk, and G. A.
Sawatzky. Density-functional Theory and NiO Photoemission Spectra.
Phys. Rev. B, 48(23):1692916934, Dec. 1993.
[82] M. T. Czy zyk and G. A. Sawatzky. Local-density Functional and On-Site
Correlations: the Electronic Structure of La
2
CuO
4
and LaCuO
3
. Phys.
Rev. B, 49(20):1421114228, May 1994.
[83] H. Eschrig, K. Koepernik, and I. Chaplyginne. Density Functional Appli-
cation to Strongly Correlated Electron Systems. J. of Solid State Chem.,
176(2):482495, Dec. 2003.
[84] H. Gebhard. The Mott Metal-Insulator Transition: Models and Methods.
Springer-Verlag Berlin Heidelberg, 1997.
[85] M. Imada, A. Fujimori, and Y. Tokura. Metal-insulator transitions. Rev.
Mod. Phys., 70(4):10391263, Oct. 1998.
[86] K. Terakura, A. R. Williams, T. Oguchi, and J. K ubler. Transition-Metal
Monoxides: Band or Mott Insulators. Phys. Rev. Lett., 52(20):18301833,
May. 1984.
[87] M. R. Norman. Orbital Polarization and the Insulating Gap in the
Transition-Metal Oxides. Phys. Rev. Lett., 64(10):11621165, Mar. 1990.
[88] S. Ohnishi. A Theory of the Pressure-Induced High-Spin-Low-Spin
Transtion of Transition Metal Oxides. Phys. Earth and Plan. Inter.,
17(2):130139, Sep. 1978.
[89] P. Werner and A. J. Millis. High-spin to Low-Spin And Orbital Polarization
Transitions in Mutiorbital Mott Systems. Phys. Rev. Lett., 99(12):126405,
Sept. 2007.
BIBLIOGRAPHY 89
[90] F. Birch. Finite Strain Isotherm and Velocities for Single-crystal and
Polycrystalline NaCl at High Pressures and 300 K. J. Geophys. Res.,
83(B3):12571268, 1978.
[91] Z.-X. Shen, C. K. Shih, O. Jepsen, W. E. Spicer, I. Lindau, and J. W.
Allen. Aspects of the Correlation Eects, Antiferromagnetic Order, and
Translational Symmetry of the Electronic Structure of NiO and CoO. Phys.
Rev. Lett., 64(20):24422445, May. 1990.
[92] M. Magnuson, S. M. Butorin, J.-H. Guo, and J. Nordgren. Electronic
Structure Investigation of CoO By Means of Soft X-Ray Scattering. Phys.
Rev. B, 65(20):205106, Apr. 2002.
[93] M. A. Korotin, S. Yu. Ezhov, I. V. Solovyev, V. I. Anisimov, D. I. Khomskii,
and G. A. Sawatzky. Intermediate-spin State and Properties of LaCoO
3
.
Phys. Rev. B, 54(8):53095316, Aug 1996.
90 BIBLIOGRAPHY
Appendix
The real spherical harmonics
1
The spherical harmonics of various authors dier by constant phase factors. We
use the denition of
}
lm
(, ) = i
l
_
2l + 1
4
(1)
(|m|+m)/2
e
im
_
(l +[m[)!(l [m[)!
(1
2
)
|m|/2

1|m|

l|m|

p=0,2,...
(1)
p/2
(l [m[)!(2l 1 p)!!
(l [m[ p)!p!!

p
,
(6.1)
with = z/r, = arctan(y/x), 0 2. The real spherical harmonics is
dened by
Y
lm
=
_
_
_
i

2
(}
l|m|
(1)
m
}
l|m|
) <
}
l0
for m = 0
1

2
(}
l|m|
+ (1)
m
}
l|m|
) >
(6.2)
The rst harmonics in Equ. (6.2) are
Y
00
=
1

4
,
Y
11
= i
_
3
4
y
r
, Y
10
= i
_
3
4
z
r
, Y
11
= i
_
3
4
x
r
,
Y
22
=
_
15
4
xy
r
2
, Y
21
=
_
15
4
yz
r
2
, Y
20
=
_
5
16
(3
z
2
r
2
1),
Y
21
=
_
15
4
xz
r
2
, Y
22
=
_
15
16
(x
2
y
2
)
r
2
,
1
H. Eschrig, Optimized LCAO Method, Akademie-Verlag Berlin 1989
92
Acknowledgment
At this point, I want to thank everyone who accompanied me during the last
three years. During these three years, I have experienced a second order
transition from doing experiments to doing theory. The driving force behind
this transition is my personal interests in understanding something theoretically.
A lot of people help me to survive this process and I would like to express my
thanks to these loving people.
First I would like to thank my supervisor, Prof. Dr. Helmut Eschrig. I rst
thank him for setting up for me a model how a scientist should be like. In his
words, A scientist lives on his reputation. His lectures are always enlightening.
He carefully explained how a formula is understood by physical arguments,
and at the same time, how physical ideas are formulated with mathematical
precision.
The second thanks are given to my second supervisor PD. Dr. Manuel
Richter. I thank him for all the helps, both scientic and nonscientic, during
my stay in Dresden. I thank him for his very careful correction of my thesis,
and a lot of critical suggestions of the physical ideas. He also corrected me
some naive, or sometimes error explanations of some results. His suggestions
can always improve the quality of the thesis. His instructive, and impressive
drawings always help me to understand complicate phenomena.
Special thanks are given to my cooperators: Mr. Z. Manuel at TU Dresden,
and Mr. S. Nils at IFW for the experimental work on Rh
2
MnGe, and Dr. N.
Mattern at IFW for introducing me the Zr-Fe compounds and cooperations on
the nonstoichiometric compounds.
Many thanks go to my dear colleagues in our theoretical group. Dr. Klaus
Koepernick is always ready to answer any questions about FPLO. You are the
fastest person in thinking and computing I have ever met. Ulrike Nitzsche is
always willing to help on any problems about our workhorses, the clusters. Of
course, discussions with Dr. Michael Kuzmin, Dr. Ingo Opahle, Dr. Ferenc
Tasnadi, and Dr. Mahdi Sargolzaei are always helpful. I will also thank my
German colleagues, Casten Neise, Stephan Schonecker and Mathias Beyer for
providing my a lot of fun during the work. All my Chinese friends and colleagues
in Dresden, especially Hongbin Zhang, Prof. Dr. Weimin Wang, Prof. Dr.
Chongde Cao are acknowledged. It is you who make the study and work here
much more harmonic. I cannot make a full list of the people should be thanked.
Here I just mentioned several of them: PD. Dr. Anulf Mobius, Dr. Saad El
Gazar, Maengsuk Kim, etc.
I would also like to take this opportunity to acknowledge my colleagues in
China: Associate Prof. Wanli Zhang, Prof. Dr. Huaiwu Zhang, Prof. Dr.
Shiqing Yang, Associate Prof. Dr. Hongchuan Jiang, Dr. Bin Peng, and Mr.
94
Rujun Tang, et al. for all their supports when I was in Dresden. Without your
continuous support and help, I do not think that I can nish my thesis.
The last but not the least thanks go to my family. My wife, and dear parents
take good care of the family when I am abroad. They always favor my decisions
and encourage me. I hope my little lovely daughter, Lele, will excuse me for not
being able to accompany her when she was young. As a father, I shall say that
I am sorry. I will spend all my free time with you when I am back, my dear
daughter.
Two years of nancial support from Deutscher Akademischer Austausch Di-
enst (DAAD) is greatly acknowledged.
Versicherung
Hiermit versichere ich, dass ich die vorliegende Arbeit ohne unzulassige Hilfe
Dritter und ohne Benutzung anderer als der angegebenen Hilfsmittel angefertigt
habe; die aus fremden Quellen direkt oder indirekt ubernommenen Gedanken
sind als solche kenntlich gemacht. Die Arbeit wurde bisher weder im Inland
noch im Ausland in gleicher oder ahnlicher Form einer anderen Pr ufungsbehorde
vorgelegt. Die Arbeit entstand mit wissenschaftlicher Betreuung durch Prof.
Helmut Eschrig am Leibniz-Institut f ur Festkorper- und Werkstoforschung
Dresden.
Ich habe vorher kein anderes Promotionsverfahren eronet. Ich erkenne
die Promotionsordnung der Fakultat Mathematik und Naturwissenschften der
Technischen Universitat Dresden in der aktuell g ultigen Fassung an.

Wenxu Zhang
Dresden, den 15. Nov. 2007
96