Sei sulla pagina 1di 65

MODELING AND SIMULATIONS OF DIELECTRIC MATERIALS

A Thesis

Presented to

The Graduate Faculty of The University of Akron

In Partial Fulfillment

of the Requirement for the Degree

Master of Science

Yuan Zhou

August, 2007
MODELING AND SIMULATIONS OF DIELECTRIC MATERIALS

Yuan Zhou

Thesis

Approved: Accepted:

__________________________ __________________________
Advisor Department Chair
Dr. Alper Buldum Dr. Robert R. Mallik

__________________________ ___________________________
Co- Advisor Dean of the College
Dr. Ang Chen Dr. Ronald F. Levant

___________________________ ___________________________
Committee Member Dean of the Graduate School
Dr. Jutta Luettmer-Strathmann Dr. George R. Newkome

__________________________
Date

ii
ABSTRACT

The perovskites ABO3 are the most important class of dielectric materials. The

perovskite SrTiO3 and the perovskite-related material CaCu3Ti4O12 are studied in this

work.

In the first part of this work, we find the structural temperature of SrTiO3 by

performing molecular dynamics simulations and investigate electronic and structural

properties of SrTiO3 by performing ab initio calculations. A strong chemical bonding

nature between Ti and O is found. This is responsible for the TiO6 octahedron behavior

throughout the phase transition.

In the second part, ab initio calculations on CaCu3Ti4O12 are performed. We

investigate the electronic properties of this material. An antiferromagnetic character of

CaCu3Ti4O12 is observed. Furthermore, we investigate the electronic properties of new

materials of different Ca, Cu ratios. This is the first time to perform ab initio calculations

to study Ca1+xCu3-xTi4O12. We find that with the increase of with increase of Ca in the

material, the optimized lattice constant and band gap increase, and insulator character

becomes much more pronounced in the material. This is in good agreement with the

experimental results.

iii
ACKNOWLEDGEMENTS

Many thanks to my advisors; to Dr. Alper Buldum, for offering valuable advice and

guidance in my research; to Dr. Ang Chen, for his patience, direction and help not only in

my research but also in my life. I will never forget the days working with them.

Many thanks to the professors in the Department of Physics at the University of

Akron, for their excellent lectures and generous assistance through these past few years. I

continue to benefit from their knowledge in living life.

Many thanks to my colleagues and friends in Akron, for their help and assistance.

Many thanks to everyone I met in Akron, for their warmness and hospitality which

makes me feel at home.

Special thanks to my parents for their encouragement and loving support.

iv
TABLE OF CONTENTS

Page

LIST OF TABLES………………………………………………………………….…...viii

LIST OF FIGURES……………………………………………………………………....ix

CHAPTER

І INTRODUCTION ........................................................................................................ 1

1.1 Background of SrTiO3 ........................................................................................... 1

1.2 Background of CaCu3Ti4O12 .................................................................................. 9

1.3 Outline of the thesis ............................................................................................ 12

ІI MODELING AND SIMULATION METHODS ...................................................... 13

2.1 Molecular dynamics ............................................................................................ 13

2.2 Ab initio (First-principles) calculations .............................................................. 17

III MOLECULAR DYNAMICS SIMULATIONS ON SrTiO3 ..................................... 21

3.1 Structure .............................................................................................................. 21

3.1.1 Cubic ......................................................................................................... 21

3.1.2 Tetragonal .................................................................................................. 22

3.1.3 Force field ................................................................................................. 23

3.2 Energy minimization ........................................................................................... 24

3.2.1 One unit cell and its superlattice ............................................................... 24

3.2.2 5*5*5 cell and its superlattice ................................................................... 25

v
3.3 Dynamics simulations ................................................................................................ 26

IV FIRST-PRINCIPLES CALCULATIONS ON SrTiO3 .............................................. 27

4.1 Structure of SrTiO3 .............................................................................................. 27

4.2 Cambridge Sequential Total Energy Package ..................................................... 28

4.3 Electronic structures and optical properties of SrTiO3 ........................................ 28

4.3.1 Band structure of SrTiO3............................................................................ 28

4.3.2 Total density of states of SrTiO3 ................................................................ 30

4.3.3 Charge density of SrTiO3 ........................................................................... 31

4.3.4 Optical properties of SrTiO3 ...................................................................... 32

4.4 Summary of the work on SrTiO3 ......................................................................... 33

V FIRST-PRINCIPLES CALCULATIONS ON CaCu3Ti4O12 ...................................... 35

5.1 Structure of CaCu3Ti4O12 ..................................................................................... 35

5.2 Energy minimization of CaCu3Ti4O12 .................................................................. 36

5.3 Geometry Optimization of CaCu3Ti4O12 ............................................................. 38

5.4 Band structure of CaCu3Ti4O12 ............................................................................ 39

5.5 Density of states of CaCu3Ti4O12 ......................................................................... 40

5.6 Charge densities of CaCu3Ti4O12 ......................................................................... 41

VI FIRST-PRINCIPLES CALCULATIONS ON Ca1+xCu3-xTi4O12 .............................. 42

6.1 Structure .............................................................................................................. 42

6.2 Geometry optimization ....................................................................................... 44

6.3 Band structure ..................................................................................................... 46

6.4 Density of states .................................................................................................. 48

6.5 Charge density .................................................................................................... 50

vi
6.6 Summary of the work on Ca1+xCu3-xTi4O12 ........................................................ 51

REFERENCES ................................................................................................................ 53

vii
LIST OF TABLES

Table Page

1.1 Summary of SrTiO3 x-ray data...............................................................................3

3.1 Coordinates and charges for SrTiO3 in tetragonal phase.......................................22

3.2 Atomic data for Sr, Ti and O………………………….........................................23

5.1 The Wyckoff positions for CaCu3Ti4O12................................................................35

6.1 The Wyckoff positions for Ca2Cu2Ti4O12…………………………………..…....44

6.2 Summary of optimized lattice constant for different Ca, Cu ratio.........................46

6.3 Summary of direct band gap for different Ca, Cu ratio………………………….48

viii
LIST OF FIGURES

Figure Page

1.1 The cubic ABO3 perovskite structure .................................................................. 2

1.2 Low temperature phase of SrTiO3 ( tetragonal).................................................... 6

1.3 The MEM charge density distribution of STO at room temperature (a) (001)
(b) (002) and (c) (110) plane................................................................................. 7

1.4 The MEM charge density distribution map of STO at 70 K (a) (001) (b) (002)
and (c) (110) plane ................................................................................................ 8

1.5 Structure of CCTO shown as TiO6 octahedra, Cu atoms bonded to four


oxygen atoms, and large Ca atoms without bonds.............................................. 10

2.1 Simulation as a bridge between microscopic and macroscopic properties......... 14

2.2 Kohn-Sham approach.......................................................................................... 18

2.3 Local density approximation............................................................................... 19

3.1 Cubic structure of SrTiO3 ................................................................................... 22

3.2 Total potential energy vs lattice constant for SrTiO3 .......................................... 24

3.3 Total potential energy vs lattice constant for 5*5*5 superlattice........................ 26

4.1 The structure of SrTiO3 built by MS Modeling .................................................. 27

4.2 The calculated energy band structure of SrTiO3 ................................................. 29

4.3 Total density of states of SrTiO3 ......................................................................... 30

ix
4.4 Charge densities of SrTiO3 (a) Sr-O plane (b) Ti-O plane .................................. 31

4.5 Refractive index and extinction coefficient of SrTiO3 ........................................ 33

5.1 The structure of CaCu3Ti4O12 built by MS Modeling......................................... 36

5.2 Energy minimization of CaCu3Ti4O12 ................................................................. 37

5.3 Bond lengths and angles of CaCu3Ti4O12 after Geometry Optimization............ 38

5.4 Band structure of CaCu3Ti4O12 ........................................................................... 39

5.5 Density of states of CaCu3Ti4O12 ........................................................................ 40

5.6 Spin-up and spin-down change densities of CaCu3Ti4O12 .................................. 41

6.1 The structure of CuTiO3 ...................................................................................... 43

6.2 The structure of Ca2Cu2Ti4O12 ............................................................................ 43

6.3 The structure of Ca3CuTi4O12 ............................................................................. 43

6.4 The structure of CaTiO3 ...................................................................................... 43

6.5 Band structure of CuTiO3 ................................................................................... 47

6.6 Band structure of Ca2Cu2Ti4O12 .......................................................................... 47

6.7 Band structure of Ca3CuTi4O12 ........................................................................... 47

6.8 Band structure of CaTiO3 .................................................................................... 47

6.9 Density of states of CuTiO3 ................................................................................ 49

6.10 Density of states of Ca2Cu2Ti4O12 .................................................................... 49

6.11 Density of states of Ca3CuTi4O12 ...................................................................... 50

6.12 Density of states of CaTiO3 .............................................................................. 50

6.13 Charge densities of Ca2Cu2Ti4O12 (a) Ca-Cu-O (b) Ti-O ................................. 51

x
CHAPTER І

INTRODUCTION

Many experimental and theoretical methods are performed to study the structural

phase transitions of the perovskites ABO3. SrTiO3 is an important member of the

perovskites. The purpose of this work is to find the structural temperature of SrTiO3 by

molecular dynamics simulations and to examine the ground state properties of SrTiO3

using ab initio calculations in order to understand the chemical bonding nature of SrTiO3,

which is very important to explain the TiO6 octahedron behavior throughout the phase

transition.

Another perovskite-related material, CaCu3Ti4O12 has drawn much interest due to its

unusual dielectric behavior. Another part of this work is to examine the ground state

properties of this material using ab initio calculations. We also investigate the electronic

structures of Ca1+xCu3-xTi4O12 (x=-1, 1, 2, 3). This is the first time to study these

materials using ab initio calculations.

1.1 Background of SrTiO3

The perovskites ABO3 are the most important class of ferroelectric materials. At

high temperatures, they all share the simple cubic structure, with monovalent or divalent

cation A at the cube corners, penta- or tetravalent metal B at the body centers, and O

atoms at the cube face centers (see Figure 1.1). This structure can also be thought of as a
1
set of oxygen octahedra arranged in a simple cubic pattern and connected together by

shared O atoms, with the A atoms occupying the spaces in between. The latter picture of

corner-shared octahedra has great structural utility, because most structural

transformations observed in perovskites can be characterized as collective rotations of the

octahedra.

Figure 1.1 The cubic ABO3 perovskite

The most interesting aspect of the cubic perovskite structure is that as the

temperature is reduced, it can display a variety of structural phase transitions, ranging


[1]
from ferroelectric and antiferroelectric to non-polar antiferrodistortive in nature.

SrTiO3 (STO) is an important member of the perovskites, which possesses


[1]
non-ferroelectric structural phase transitions at about 105 K. Different experimental
[2] [3, 4]
methods involving electron paramagnetic resonance , ultrasonic measurements ,
[5] [6, 7]
nuclear magnetic resonance , and the polarizing microscope were performed to
[8]
show that there was a phase transition around 110 K. Lytle performed x-ray

diffractometry on single-crystal STO over a large temperature range, from 4.2 K to 300 K.

2
The lattice parameters were determined as a function of temperature. The data was

summarized to show that as the temperature goes down, STO changes to lower symmetry

structures (as shown in Table 1.1).

Table 1.1 Summary of SrTiO3 x-ray data [8]

Besides a great deal of experimental research, a wide range of theoretical studies

were performed to find and explain the structural phase transitions in STO. Pytte and
[9]
Feder studied phase transitions in perovskites with the help of a model Hamiltonian.

They found that three degrees of freedom (15 degrees of freedom per unit cell in the

perovskite structure) are directly connected with the structural phase transitions. For STO,

the phase transition from the cubic to the tetragonal phase involves the rotation of the

octahedron about a cube axis. The predictions of their Hamiltonian model qualitatively

agree well with the experimental results except for the magnitude of the angular

distortion. This disagreement suggests that additional anharmonic interactions are needed
[10]
to be taken into account. One year later, Feder and Pytte included the interaction

between soft-mode coordinates and the strains into their Hamiltonian model. Good

3
agreement was obtained with all the experimental results this time.
[11]
Zhong and Vanderbilt claimed that the previous phenomenological model
[9, 10]
Hamiltonian approach had been limited by oversimplification and ambiguities in

interpretation of experiment. They performed first-principle density-functional

calculations (a more accurate method) to show that the coexistence of both

antiferrodistortive (AFD) and ferroelectric (FE) instabilities is very common in cubic

perovskite compounds. Monte Carlo (MC) simulations were performed on the STO cubic

lattice with periodic boundary conditions. The system was found to adopt the cubic

structure at high temperature. As the temperature is reduced, a transition to an AFD

tetragonal structure occurs at 130 K, a second transition occurs at 70 K to a FE tetragonal

structure, and the system transforms to the low-symmetry monoclinic structure at 10 K.

They also found that increased pressure enhances the AFD instability while suppressing

the FE one.

As we know, a first-principles approach can study the zero-temperature properties of

the materials. However, it is important to see whether one can understand such features as

the phase transition sequence and transition temperatures on a material-specific basis.


[12]
Zhong and Vanderbilt extended the ab initio effective Hamiltonian treatment of first

principles to include quantum fluctuations. Path-integral MC quantum simulations were

applied to STO to find that the AFD phase transition temperature decreases from 130 K
[11]
to 110 K, in excellent agreement with the experimental result of 105 K. They also

found that the quantum fluctuations completely suppress the FE phase transition, i.e. the

FE phases disappear. This implies that quantum effects on the FE instability are much

stronger than on the AFD instability in STO. Since the structural differences and energy

4
barriers between the cubic structure and the distorted structures are very small, a rough

estimate of the importance of quantum fluctuations can be obtained from the

Heisenberg uncertainty principle Δp ⋅ Δq ≥ h / 2 , or equivalently,

ΔE ≥ h 2 /(8mΔq 2 ) (1.1)

Where,

Δq is the structural difference between phases, and ΔE is the energy uncertainty,

which may prevent the occurrence of the distorted phase if it is larger than the classical

free-energy reduction. It is known that the AFD instability involves only the motion of

oxygen atoms, while the FE instability involves mainly Ti atoms. The structural change

involved in the FE distortion (0.1 a.u. for Ti in SrTiO3) is much smaller than for the AFD

distortion (0.3 a.u. for O). Therefore, mΔq2 turns out to be three times larger for the AFD

case, even though Ti atoms are three times heavier than the oxygen atoms. According to

Equation (1.1), the effect of the quantum fluctuations will be more significant for the FE

case.

The structural phase transition of STO at ~110 K is represented by the rotation of the

TiO6 octahedron. [13] The tetragonal structure of STO at low temperature is shown in

Figure 1.2. The oxygen atoms are all equivalent in the cubic phase, while they can be

distinguished as two kinds O(1) and O(2) in the tetragonal phase. It should be noted that

O(2) atoms are no longer located at the face centers. The mechanism of the phase

transition is understood by both neutron [15] and X-ray [16] diffraction, however why the

rotational mode of the TiO6 octahedron is unstable at 110 K is still unclear.

5
Figure 1.2 Low temperature phase of SrTiO3 (tetragonal) [14]

[14]
Therefore Ikeda et al. applied the Maximum Entropy Method (MEM) to X-ray

diffraction data to understand the chemical bonding nature of STO in both phases and to

reveal the structural changes at the electron level which occur throughout the phase

transition. MEM charge density maps of STO at room temperature are shown in Figure

1.3.

6
Figure 1.3 The MEM charge density distribution of STO at room temperature
(a) (001) (b) (002) and (c) (110) plane.
Contour lines are drawn from 0.4 at 0.2 eÅ3 intervals.
Four unit cells are shown. [14]

From Figure 1.3, both Sr and O atoms are found to be isotropic and no covalence

electrons are found between them. A rather strong covalent bond is found to exist

between the Ti and O atoms. MEM charge density maps of STO at 70 K are shown in

Figure 1.4 to compare with Figure 1.3.

7
Figure 1.4 The MEM charge density distribution map of STO at 70 K.
(a) (001) (b) (002) and (c) (110) plane. Contour lines are the same as Figure 1.3.
Two tetragonal cells are shown. [14]

From Figure 1.4 (a), both the Sr and O(1) atoms are fairly isotropic and no

covalence electrons are found between Sr and O(1). From Figure 1.4 (b), there is a strong

covalent bond between Ti and O(2). The charge density distribution around the oxygen

atoms is not symmetrical with respect to the Ti-Ti lines. There is a little positional shift of

8
the O(2) atom in the direction of the arrow. In Figure 1.4 (c), the covalent bond between

Ti and O(1) is observed to be similar to that in Figure 1.3 (c). Therefore, it shows that

there is no significant change of chemical bonding through the phase transition. The

phase transition of STO at 110 K is purely due to the TiO6 octahedron, which seems to be

maintained by the strong Ti-O covalent bond.

1.2 Background of CaCu3Ti4O12

CaCu3Ti4O12 (CCTO) has drawn much recent interest due to its unusual dielectric

properties. Subramanian et al. [17] first observed the unusually high dielectric constant of

CCTO. This compound possesses a large dielectric constant about 12,000 at 1 kHz, which

is nearly constant over the temperature range 100-380 K. [18] It is also found that the

dielectric constant of this material is frequency dependent, decreasing with increasing

frequency. One curious result is that the dielectric constant drops as low as 100 when

cooling below 100 K.

The structure of CCTO was first determined from neutron powder diffraction

data. [19] The crystal structure is shown in Figure 1.5. The structure of CCTO can be

viewed as a perovskite-related structure, in which four ATiO3 units comprise the primitive

cell (where A is either Ca or Cu). Because there are 5 atoms in an ATiO3 unit cell, the

structure in Figure 1.5 contains 40 atoms. This is a doubled conventional cell, which is

also the primitive cell of the spin structure. The CuO4 plaquettes are nearly square,

because the bonds between Cu and O are all the same length, while the angles between

bonds differ slightly from 90 degrees. In a similar way, the distances between Ti and its

six neighbors are the same, but the angles deviate slightly from 90 degrees. Every oxygen

9
atom belongs to a single planar CuO4 plaquette and to the two connected tilted TiO6

octahedra.

Figure 1.5 Structure of CCTO shown as TiO6 octahedra, Cu atoms


bonded to four oxygen atoms, and large Ca atoms without bonds. [17]

A material with giant dielectric constant is always classified as a relaxor or a

ferroelectric, but empirical evidence tends to exclude CCTO from either category.
[18]
Ramirez et al. excluded ferroelectricity as a cause of the unusual high dielectric

response by X-ray diffraction and thermodynamic data. Instead they explained their

results by a relaxor model. However, nanodomains or disorder effects, which are

common to relaxor materials, are not observed. Neither superstructure peaks nor strong

diffuse scattering are present in diffraction experiments. [20]

Many researchers have presented theoretical insights into intrinsic lattice of CCTO.

10
[20]
He et al. performed first-principles calculations within the local spin-density

approximation (LSDA) on CCTO to study the structural and electronic properties. Their

calculations appear to limit certain intrinsic mechanisms to explain the enormous,

low-frequency dielectric response. Furthermore, they suggested that increased attention

should be given to extrinsic effects. An extrinsic mechanism which is associated with

defects, domain boundaries, or other crystalline deficiencies, was initially proposed by

Subramanian et al. [17] He et al. also found CCTO to be stable in a centrosymmetric

crystal structure with space group Im3, arguing against the possibility that CCTO is a

conventional ferroelectric or relaxor.

He et al. [21] extended their previous work on CCTO by carrying out a parallel study

on the closely related material CdCu3Ti4O12 (CdCTO). Replacing Ca with Cd is found to

leave many calculated quantities largely unaltered. This indicates that both CCTO and

CdCTO possess similar intrinsic structural, vibrational and dielectric properties. As for

the lattice contributions to the static dielectric constant, the discrepancy between their

computations and those measured at frequencies below the Debye cutoff range, reinforces

the conclusion that some extrinsic mechanism is likely to be the source of the large

dielectric constant present in both materials.


[22]
Johannes et al. performed first-principles calculations with a full-potential

linearized augmented plane-wave method (FLAPW) within the local spin-density

approximation (LSDA). Their results reproduce the observed antiferromagnetic (AFM)

insulating character of CCTO, and show a similar density of states as that presented by
[20]
He et al. They applied the superexchange picture of magnetism to CCTO, and

suggested that extremely long-range interactions are important and perhaps even

11
responsible for the AFM order.

Li et al. [23] applied FLAPW method within the generalized gradient approximation

(GGA) to investigate the basic electronic and magnetic properties of CCTO. Their

calculated indirect band gap of 0.51 eV and direct band gap of 0.58 eV within GGA are

much closer to the experimental optical gap (≥ 1.5 eV) [24] than those in He et al.’s work
[20]
, but still underestimate the experimental data. A metastable state is found to be

ferromagnetic (FM) and semiconducting, indicating semiconducting behavior and an

AFM-FM transition in CCTO. Besides the Ti cation expected to be involved in the


[22]
magnetic coupling , a possible oxygen path is proposed in the superexchange

interaction between Cu spins.

1.3 Outline of the thesis

The background knowledge of molecular dynamics simulations and ab initio

calculations is discussed in Chapter II. The molecular dynamics simulations and ab initio

calculations on SrTiO3 are discussed in Chapter III and IV, respectively. The electronic

properties are provided. The strong covalent bonding between Ti and O is found to

explain the TiO6 octahedron behavior throughout the structural phase transition. In

Chapter V and VI, we present the results of ab initio calculations on Ca1+xCu3-xTi4O12, i.e.

CuTiO3, CaCu3Ti4O12, Ca2Cu2Ti4O12, Ca3CuTi4O12 and CaTiO3. With the increase of Ca

in the material, the optimized lattice constant and band gap increase, and insulator

character becomes much more pronounced.

12
CHAPTER ІI

MODELING AND SIMULATION METHODS

2.1 Molecular dynamics

In general, molecular systems, which consist of a great number of particles, are so

large that it is impossible to find their properties analytically. Molecular dynamics (MD)

simulation serves as a good method to solve the problem numerically. Nowadays MD

simulations, first introduced by Alder and Wainwright in 1950, are a widely used method

to study time-dependent properties of a molecular system in various material-related

fields. [25] In physics, MD is used to observe the dynamics of atomic-level phenomena,

such as thin film growth and ion-subplantation. In biology, the MD method can provide

much detailed information on the fluctuations and conformation changes of proteins and

nucleic acids. In chemistry, MD simulation is a useful approach to help in the efficient

synthesis of compounds and in drug design.

In a MD simulation, the macroscopic properties of a system, such as pressure,

energy and heat capacities, are deduced from microscopic information. [26] Therefore MD

simulations act as a bridge between the macroscopic world and microscopic length and

time scales (shown in Figure 2.1). The connection between macroscopic properties and

microscopic information is made by mathematical expressions which relate the

macroscopic properties to the distribution and motion of the atoms. MD solves the

dynamics of a classical interacting many-particle system by solving the equations of


13
motion of the particles.

Figure 2.1 Simulation as a bridge between microscopic

and macroscopic properties [26]


r r
Newton’s equation ( Fi = mi ai ) is used to simulate atomic motion. For each atom i in
r
a system of N atoms, Fi is the total force acting on it by all the other particles, mi is the

r
atom mass, and ai is its acceleration. This is the so-called N-body problem. Since the

r
r r d 2 ri
acceleration a can be expressed as ai = 2 , Newton’s equation is expressed in the
dt

following form,

r r
d 2 ri
Fi = mi 2 (2.1)
dt
r
The force F can be considered as the derivative of the potential energy with

respect to the change in the atom’s position.


r r
Fi = −∇iU (2.2)

14
r
We need to find the potential energy U to calculate the force Fi from Equation 2.2.
r
If Fi is known, with the help of Equation 2.1, we can compute trajectories. (The

coordinates and velocities of all the particles are called the trajectory of the system.)

In MD simulations, a force field is used to describe the time evolution of bond

lengths, bond angles and torsions, also the non-bonding van der Waals and electrostatic

interactions between atoms. [27] A simple force field functional is shown in Equation 2.3.

r ki k
U (r ) = ∑
bonds 2
(li − lo ) 2 + ∑ i (θ i − θ o ) 2
angles 2

Vn
+ ∑
torsions 2
(1 + cos(nω − γ )) (2.3) [27]

N N ⎛ σ ij σ ij qi q j ⎞
+∑ ∑ ⎜⎜ 4ε [( )12 − ( )6 ] + ⎟
4πε 0 rij ⎟⎠
ij
i =1 j = i +1 ⎝ rij rij

r
U (r ) denotes the potential energy which is a function of the positions of N

particles. The first two terms in Equation 2.3 model the interactions between pairs of

bonded atoms, modeled by a harmonic potential. The third term is a torsional interaction

potential. The last term is the non-bonded term. The non-bonded term is modeled using a

Coulomb potential term for electrostatic interactions and a Lennard-Jones potential for

van der Waals interactions. [27]

From a potential energy forcefield, MD can solve the classical equations of motion

for a system of N interacting atoms. Combining Equation 2.1 and 2.2, we can rewrite

Newton’s equation,

r r
d 2 ri
−∇iU = mi 2 (2.4)
dt

Given the initial coordinates and velocities at time t , the positions and velocities at

15
time t + Δ t are calculated. The initial coordinates are determined in the input file or

from a previous operation such as energy minimization, while the initial velocities are

randomly generated at the beginning of each run based on the desired temperature.

Therefore, dynamics runs can not be repeated exactly except if the same speed is used for

the random number generator.

Numerous numerical algorithms have been developed for integrating the equations

of motion (Equation 2.4), such as the the Verlet algorithm, the Leapfrog algorithm, the

Verlet velocity algorithm and Beeman’s algorithm. Verlet algorithm is used to our MD

simulations. All the integration algorithms assume the positions, velocities and

accelerations can be approximated by a Taylor series expansion, [25]

1
r (t + δ t ) = r (t ) + v(t )δ t + a (t )δ t 2 + K ,
2
1
v(t + δ t ) = v(t ) + a(t )δ t + b(t )δ t 2 + K , (2.5)
2
a(t + δ t ) = a(t ) + b(t )δ t + K ,

where r , v and a represent the position, velocity and acceleration respectively.

When choosing which algorithm to use, we should consider whether the algorithm

satisfies the conservation of energy and momentum, whether it is computationally

efficient and whether it permits the use of a relatively long time step for integration.

The MD simulation acts not only as a bridge between macroscopic and microscopic

as I mentioned above, but also as a bridge between theory and experiment. Sometimes we

can conduct a simulation and compare the theoretical results with experimental results to

test a theory or a model. A better advantage of MD simulation is that we can carry out

simulations on the computer that are difficult or impossible to conduct in the laboratory.

16
2.2 Ab initio (First-principles) calculations

As we know, a material is composed of atoms bound by chemical bonds, which are

simply interactions between electrons. These interactions can be described by the laws of

quantum physics. This means that all material properties (chemical, mechanical, electrical,

magnetic, optical, thermal…) can, in principle, be predicted from the atomic number and

mass of the atomic species involved, with the aid of quantum physics. This is precisely

the basic idea of first principles calculations.

An ab initio calculation is performed from first principle. The Latin term ab initio

means “from the beginning”. It implies that the calculation starts directly at the level of

established laws of physics and does not depend on empirically derived parameters.

Density functional theory (DFT), for which Prof. Walter Kohn was awarded the

1998 Nobel prize in chemistry, is considered to be an ab initio method to determine the

molecular electronic structure. The fundamental idea of DFT is that any property of a

system of many interacting particles can be viewed as a functional of the ground state

charge density. DFT, which maps the original many-electron problem into an equivalent

single-electron problem, is an extremely successful approach for the description of

ground state properties of metals, semiconductors, and insulators.

The Kohn-Sham method, the most common implementation of DFT, has provided a

way to make useful approximate ground state functions for real systems of electrons. The

Kohn-Sham ansatz replaces the intractable many-body problem of interacting electrons in

a static external potential to a tractable problem of non-interacting electrons moving in an

effective potential. (shown in Figure 2.2)

17
Figure 2.2 Kohn-Sham approach

The Kohn-Sham method assumes that the ground state density of the original

interacting system is equal to that of some chosen non-interacting system.

nks (r ) = n( r ) (2.6)

The ground state energy functional can be expressed in the form

EKS (n) = TKS (n) + EHartree (n) + Eext (n) + Exc (n)
where
1 n
TKS (n) = − ∑
2 i =1
ψ i ∇2 ψ i (2.7)

n(r )n(r ')


EHartree (n) = ∫ dr d r '
r −r'
Eext (n) = ∫ Vext (r )n(r )dr

Exc (n) is the so-called exchange-correlation energy.

The Kohn-Sham variational equation, shown in Equation 2.8, is similar in form to

the time-independent Schrödinger’s equation, except that the potential experienced by the

electrons is formally expressed as a functional of the electron density.

1
[− ∇ 2 + VHartree ( nGS ; r ) + Vext (nGS ; r ) + Vxc ( nGS ; r )]ψ i ( r ) = ε iψ i ( r ) (2.8)
2

In practice, the exact functional of the electron density is unknown. Practical


18
applications of DFT are based on approximations for the so-called exchange-correlation

potential, which describes the effects of the Pauli principle and the Coulomb potential

beyond a pure electrostatic interaction of the electrons.

The most widely used and simplest approximation of the exchange-correlation

energy functional is the local density approximation (LDA) which assumes that the

density can be treated locally as an non-interacting homogeneous electron gas (shown in

Figure 2.3), and the exchange-correlation energy at each point in the system is the same

as that of an uniform electron gas of the same density.

inhomogeneous system at point r homogeneous electron gas


with local density n(r) with same density n(r)

Figure 2.3 Local density approximation

The exchange-correlation energy can be expressed in the form

Exc ( n( r )) = ∫ n( r )ε xc ( n( r )) dr (2.9)

where ε xc (n(r )) is an energy per electron at point r.

The local, energy-independent exchange-correlation potential Vxc (n(r )) is the

functional derivative of Exc (n(r )) ,

∂Exc (n(r ))
Vxc (n(r )) = (2.10)
∂n(r )

The LDA functional within the Kohn-Sham approach has been proved to be an

accurate, practical way to investigate the solid-state properties of materials.

There are two methods to deal with the external potential Vext , which is complicated

in real circumstances. One is called the all-electron calculation method, the other is the

pseudopotential method. The primary application of the pseudopotential is to replace the


19
strong Coulomb potential of the atomic nucleus and the effects of the tightly bound core

electrons by an effective ionic potential acting on the valence electrons. Most

pseudopotential calculations are based on norm-conserving potentials, in which

pseudo-wavefunctions are normalized and the total pseudocharge inside the core matches

that of the all-electron wave function. This method provides a way to construct

potentials that are successfully applicable to calculations on molecules and solids.

However, for many cases, such as O 2p or Ni 3d orbitals, a very large plane-wave

basis-set size is needed to satisfy the norm-conserving condition. This leads to more
[28]
calculations than those in other cases. Therefore, Vanderbilt developed ultrasoft

pseudopotentials which create much smoother pseudofunctions and use considerably

fewer plane-waves for calculations of the same accuracy. This method is always used in

the study of ferroelectrics.

20
CHAPTER III

MOLECULAR DYNAMICS SIMULATIONS ON SrTiO3

Molecular dynamics simulations were carried out using Cerius2 Version 4.10.

Cerius2 provides a wealth of tools for applications in life and materials science modeling

and simulation. It offers capabilities for modeling materials structure, properties, and

processes, with applications in catalysis, crystallization, and polymer science.

3.1 Structure

Both cubic and tetragonal structures were built. The cubic structure was used to get

energy minimization and the tetragonal structure was used in dynamics simulations.

3.1.1 Cubic

The space group of STO is pm3m. The Wyckoff positions are Sr (0, 0, 0), Ti (0.5,

0.5, 0.5) and O (0.5, 0.5, 0). The charges are fixed to be +2, +2.2, -1.4 for Sr, Ti and O,

respectively, due to the bond covalency— the Ti-O bond in real crystals has both ionic

and covalent character. [29] The crystal structure is shown in Figure 3.1.

The lattice constant is chosen to be a=3.905 Å

21
Figure 3.1 Cubic structure of SrTiO3

3.1.2 Tetragonal

The space group is p4. The Wyckoff positions and charges are shown in Table 3.1.

The charges are chosen to be the same as in the cubic structure.

Table 3.1 Coordinates and charges for SrTiO3 in the tetragonal phase

Coordinate Charge

Sr (0, 0, 0) 2.0

Ti (0.5, 0.5, 0.5) 2.2

O (0.5, 0.5, 0) (0, 0.5, 0.5) -1.4

According to x-ray diffraction measurements of the lattice parameters of


[8]
single-crystal SrTiO3 by F. W. Lytle , there is a small tetragonal distortion c/a=1.00056

below 110 K but the unit-cell volume remains unchanged through the transition. At 300 K,

a=3.905 Å. The volume is a3= 59.54744 Å3. In the tetragonal phase, the volume is
22
a2*c=59.54744 Å3.

Using the relation c/a=1.00056, we can find a=3.90427 Å and c=3.90646 Å. Therefore,

the lattice constants are chosen to be a=b=3.904 Å and c=3.906 Å.

3.1.3 Force field

Universal Force field 1.02 [29] was used. Sr6+2, Ti3+4 and O_2 were set as the types

for Sr, Ti and O, respectively. A five-character mnemonic label is used to describe the

atom types. The first two characters correspond to the chemical symbol; an underscore

appears in the second column if the symbol has one letter. The third column describes the

hybridization or geometry: 1 =linear, 2 = trigonal, R =resonant, 3 = tetrahedral, 4 =

square planar, 5 = trigonal bipyramidal, 6 = octahedral. The forth and fifth columns are
[29]
used as indicators of alternate parameters such as formal oxidation state. Thus Sr6+2

indicates an octahedral Sr in the +2 oxidation state. The atomic data are shown in Table

3.2.

Table 3.2 Atomic data for Sr, Ti and O [29]

Atom Angle Distance Energy Effective


Bond(Å) scale
type (Degree) (Å) (kcal/mol) charge

Sr6+2 2.052 90.0 3.641 0.235 12.0 2.449

Ti3+4 1.412 109.47 3.175 0.017 12.0 2.659

O_2 0.634 120.0 3.500 0.060 14.085 2.300

23
3.2 Energy minimization

Two different methods were used to determine the energy minimum. (1) In the first

method, the minimization algorithm of cerius2 searches the lattice constant and atomic

positions for the local minimum energy. (2) In the second method, the lattice constants

were changed systematically and the total potential energy for each lattice constant was

calculated. Thus we obtain a potential energy vs lattice constant curve. This allows us to

find the lattice constant from the condition that the energy is at a minimum.

3.2.1 One unit cell and its superlattice

Initially, the cubic structure which is shown in Figure 3.1 was used. Keeping the

fractional coordinates (relative distance and orientation) fixed, the current energy is

calculated for every given lattice constant. (see Figure 3.2) From the graph, we find that

when a = 3.619 Å, the energy is the minimum. The first minimization method also

confirms the result of 3.619 Å.

0.00
3.40 3.50 3.60 3.70 3.80 3.90 4.00 4.10
-500.00
Energy/ cal

-1000.00

-1500.00

-2000.00

-2500.00
Lattice constant/ Å

Figure 3.2 Total potential energy vs lattice constant for SrTiO3


[29]
The simulation result from Katsumata et al is a = 3.9056 Å, b = 3.9054 Å, c =

3.9055 Å. The difference between our result and their result may due to the different
24
simulation temperatures. Their calculation was fixed at 300 K, while our energy

minimization program is based on 0 K. Our result is also 7.3 % less than the experimental

data of 3.9051 Å at 300 K.

The superlattice was created from one unit cell. After creation of the superlattice, the

symmetry changes to p1. In the first minimization method, we allow all atoms to move

while keeping the fractional coordinates fixed. Not only the lattice constant a, b, and c,

but also the angle α, β and γ will change after minimization. The final structure gives that

a = b = c = 3.618 Å

α = β = γ =90.01o

Sr atoms are still located at the vertex of the cubic cell, Ti atoms are located at the

body center as before, but O atoms are no longer at the face center positions.

3.2.2 5*5*5 cell and its superlattice

The same procedure as in 3.2.1 was performed except the 5*5*5 cell was used for

the calculation. The simulation cell contains 625 particles. It turns out that the

energy-lattice constant graph is the same as Figure 3.2.

Another superlattice was built based on the 5*5*5 cell. The final lattice constant

from the first minimization method is 18.094 Å, which should be divided by 5 to get the

lattice constant for one unit cell. It gives that a = b = c = 3.619 Å, α = β = γ =90o.

To confirm the result, the second minimization method was performed. The

energy-lattice constant graph is shown in Figure 3.3. This gives the same lattice constant

3.619 Å.

25
Figure 3.3 Total potential energy vs lattice constant for 5*5*5 superlattice

3.3 Dynamics simulations

The tetragonal structure was used as the initial structure. The simulation was based

on the 5*5*5 supercell. The initial lattice constants are chosen to be a = b = 3.904 Å and

c = 3.906 Å. (see section 3.1.2) The temperature ranges from 60 K to 300 K. Constant

NVT was chosen at 60 K, and then constant NPT were chosen for the other temperatures.

The temperature was fixed for every run, while the structure was allowed to change to

reach the equilibrium at that temperature. T_Damping Thermostat was chosen to keep the

temperature constant. The relaxation time is 0.1 ps and the dynamics time step is 0.001 ps.

Number of steps was chosen large enough to make the system reach equilibrium.

The transition temperature is found to be around 130 K. The average value is 122.5

K, which is 16.7 % higher than the experimental value of 105 K [1].

26
CHAPTER IV

FIRST-PRINCIPLES CALCULATIONS ON SrTiO3

4.1 Structure of SrTiO3

Accelrys MS Modeling 4.0 is applied to build the structure. MS Modeling is a

flexible client-server software environment that provides advanced materials simulation

and modeling technology.

The space group of STO is pm3m. The Wyckoff positions are Sr (0, 0, 0), Ti (0.5,

0.5, 0.5) and O (0.5, 0.5, 0). The charges are chosen to be +2, +4, -2 for Sr, Ti and O,

respectively. The charges are used differently from those in molecular dynamics

simulation. Normal charges are always used in the ab initio calculations. The crystal

structure is shown in Figure 4.1.

Figure 4.1 The structure of SrTiO3 built by MS Modeling

27
4.2 Cambridge Sequential Total Energy Package

The Cambridge Sequential Total Energy Package (CASTEP) [31] is a program that

employs DFT to simulate the properties of solids, interfaces, and surfaces for a wide

range of materials classes. Based on total energy plane-wave pseudopotential methods,

CASTEP takes the number and types of atoms in a system and predicts properties

including lattice constant, molecular geometry, structural properties, band structures,

density of states, charge densities and wave functions, and optical properties.

4.3 Electronic structures and optical properties of SrTiO3

The crystal structure of STO shown in Figure 4.1 is loaded. Computations are

performed via the LDA with ultra-soft pseudopotentials. The initial lattice constant is

chosen to be the experimental value of 3.905 Å. [33] Geometry optimization is performed

at a plane-wave cutoff energy of 340 eV, with a 1×1×1 Monkhorst-Pack k-point mesh.
[34]
The relaxations are performed using the BFGS algorithm while the fractional

coordinates are kept fixed during changes to the lattice.

The optimized lattice constant is 3.867 Å, slightly less than the value of 3.905 Å by

less than 1 %, as is typical in LDA calculations.

4.3.1 Band structure of SrTiO3

The calculated band structure of STO is shown in Figure 4.2. The G point in Figure

4.2 represents the Gamma point, Γ. Our band structure is similar to that in Samantaray et

al.’s work.[32]

28
Figure 4.2 The calculated energy band structure of SrTiO3

The valence band maximum is located at the R point, i.e., at the corner of the cubic

Brillouin zone. The lowest conduction band state is located at the zone center Γ, which

introduces an indirect band gap. There is a second local valence band maximum at Γ,

which gives rise to a direct band edge at Γ. The calculated direct band gap at the Γ point is
[32]
1.92 eV in Samantaray et al.’s work. From Figure 4.2, our direct band gap at the Γ
[35]
point is 2.3 eV, which is closer to the experimental value of 3.2 eV. But both

theoretical band gaps are smaller than experimental result. As is well known, the LDA

method underestimates the band gap in both semiconductors and insulators.

The band structure of STO has energetically separated low lying bands, which are

derived from the O 2s states. These bands are separated from a group of relatively narrow

bands arising from the Sr 4p states. Near the Fermi level in the valence band region, there

is a manifold of several bands, nominally derived from the O 2p states. The conduction

29
bands near the Fermi level have a strong Ti 3d character.

4.3.2 Total density of states of SrTiO3

The total density of states of STO is shown in Figure 4.3. We have obtained a similar

picture for the total density of states of STO as in Samantaray et al.’s work.[32]

Figure 4.3 Total density of states of SrTiO3

The lowest bands are located around -18 eV with double peaks and belong to O 2s

states. O 2s band states are -17.5 eV and Sr 4p band state is -15 eV. The Sr 4p state is

very close to O 2s states, which is in good agreement with the XPS measurement. [36, 37]

The upper valence band is made up predominately of the O 2p components. The width of

the O 2p valence band is 5.5 eV, close to 5 eV in Zollner et al.’s work. [38] The peak

nearest to the valence band maximum originates from the O 2p nonbonding components,

while the other two peaks correspond to the O 2p antibonding combinations of

oxygen-oxygen interactions. The lowest conduction bands are made up of Ti 3d bands.

30
The Ti 3d contribution is zero at the valence band maximum but rises strongly with

increasing binding energy, while the O 2p contribution rises from zero at the conduction

band minimum with increasing energy. These reflect the Ti 3d to O 2p covalency, i.e. the

O 2p orbitals are hybridized with Ti 3d orbitals. Following Ti 3d above the conduction

band, Sr 4d and O 3p contribute to the high energy region.

4.3.3 Charge density of SrTiO3

(a) (b)

Figure 4.4 Charge densities of SrTiO3 (atoms are colored as in Figure 4.1)
(a) Sr-O plane (b) Ti-O plane

Figure 4.4 (a) and (b) display the calculated charge densities of STO on the Sr-O and

Ti-O planes. The high charge density region is around the nuclear sites and is nearly

spherically symmetric. There is very weak chemical bonding between Sr and O atoms,

which reflects the presence of an ionic bond between Sr and O atoms. From the density

of states of STO, it has been shown that there exists a significant hybridization of Ti 3d
31
with O 2p states. This means that the bonding in Ti-O system cannot be purely ionic but

must exhibit a large covalent part. Figure 4.4 (b) presents a clearer picture about the

nature of chemical bonding between Ti and O atoms. It confirms that the Ti-O bond has a

strong covalent character. This is apparent from the noticeable overlapping charge

distribution at the middle of the Ti-O bond. However there is not much bonding charge to

link the Sr and O atoms, it indicates that the bonding between Sr and O atoms is mainly

ionic.

4.3.4 Optical properties of SrTiO3

The optical properties can be derived from the complex dielectric function

ε (ω ) = ε1 (ω ) + iε 2 (ω ) , where ε1 (ω ) and ε 2 (ω ) are the real part and the imaginary part

of the dielectric constant, respectively. The refractive n(ω ) and extinction coefficient

k (ω ) follow from Equations 4.1 and 4.2.

1
n(ω ) = ( )[ ε1 (ω ) 2 + ε 2 (ω ) 2 + ε1 (ω )]1/ 2 (4.1)
2

1
k (ω ) = ( )[ ε1 (ω ) 2 + ε 2 (ω ) 2 − ε1 (ω )]1/ 2 (4.2)
2

Figure 4.5 shows the calculated optical properties of STO. Comparing our result

with Samantaray et al.’s work [32], we find that we have obtained similar results.

32
Figure 4.5 Refractive index and extinction coefficient of SrTiO3

4.4 Summary of the work on SrTiO3

The optimized lattice constant a=3.619 Å and a=3.867 Å from molecular dynamics

simulations and ab initio calculations, respectively. The result from ab initio calculations

is in good agreement with the experimental value of 3.905 Å, based on the consideration

that underestimation is the typical limitation is LDA calculations. The difference between

the result from molecular dynamics simulations and the experimental value is due to the

different temperatures. The experimental value is measured at 300 K while our

minimization result is calculated at zero temperature.

The transition temperature is found to be 122.5 K, which is 16.7 % higher than the

experimental value of 105 K. Molecular dynamics simulations are not sufficient to find

the structural transition temperatures.

33
The calculated electronic properties show that there is a strong covalent bonding

between Ti and O, which explains why TiO6 octahedron behaves as a rigid body

throughout the phase transition.

34
CHAPTER V

FIRST-PRINCIPLES CALCULATIONS ON CaCu3Ti4O12

5.1 Structure of CaCu3Ti4O12

Accelrys MS Modeling 4.0 is applied to build the structure. The space group of

CCTO is Im3. The Wyckoff positions are shown in Table 5.1. The charges are chosen to

be +2, +2, +4, -2 for Ca, Cu, Ti and O, respectively. The crystal structure of CCTO is

shown in Figure 5.1.

Table 5.1 The Wyckoff positions for CaCu3Ti4O12

Atom Position

Ca (0, 0, 0)

Cu (1/2, 0, 0)

Ti (1/4, 1/4, 1/4)

O (0.303, 0.175, 0)*

*The structural parameters for the O atom are those of Ref [20]

35
Figure 5.1 The structure of CaCu3Ti4O12 built by MS Modeling

This structure can be obtained from the ideal structure of CaTiO3 by replacing three

out of every four Ca site ions with a Cu ion, which quadruples the cell to CaCu3Ti4O12,

and then performing a correlated rotation of the four octahedron until the Cu ion is

fourfold coordinated with O ions in a nearly square arrangement-the bonds connecting Cu

and O are all the same length, though the angles between bonds deviate slightly from 90

degrees. All the following calculations are performed using this 40-atom cell in Figure

5.1.

5.2 Energy minimization of CaCu3Ti4O12

The crystal structure of CCTO shown in Figure 5.1 is loaded. Computations are

performed via the LDA with ultra-soft pseudopotentials. The highest occupied p shell

36
electrons, for Ca and Ti, are treated as valence. The potential considers electrons in the 3d

and 4s shells of Cu as valence electrons. Energy minimization is performed at a


[20]
plane-wave cutoff energy of 500 eV (equivalent to 37 Ry ), with a 2×2×2

Monkhorst-Pack k-point mesh.

The total energy is computed while the lattice constant and the fractional coordinates

are kept fixed. The lattice constant is changed in steps and each time the energy is

determined. The total energy vs lattice constant is shown in Figure 5.2. Therefore, we

find the equilibrium lattice constant when the total energy is minimum.

-33390
6.3 6.5 6.7 6.9 7.1 7.3 7.5 7.7 7.9 8.1 8.3 8.5
-33400
Total energy/ eV

-33410

-33420

-33430

-33440

-33450
Lattice constant/ Å

Figure 5.2 Energy minimization of CaCu3Ti4O12

The lattice constant ranges from 6.50 Å to 8.10 Å. a = 7.29 Å is found to be the

lattice constant for the minimum energy. It is the same result as in He et al.’s work. [20]

The experimental value is 7.384 Å measured at 35 K [20], which is 1 % more than our

LDA result. This is the common limitation in LDA calculation.

37
5.3 Geometry Optimization of CaCu3Ti4O12

The crystal structure of CCTO shown in Figure 5.1 is loaded. Computations are

performed via the LDA with ultrasoft pseudopotentials. The initial lattice constant is

chosen to be our calculated lattice constant 7.29 Å. Geometry optimization is performed

at a plane-wave cutoff energy of 500 eV which results in convergence of the total energy

to 1 meV/atom, with a 2×2×2 Monkhorst-Pack k-point mesh. The relaxations are

performed using the BFGS [34] algorithm. The internal parameters of the structure are

relaxed at our calculated lattice constant.

The optimized lattice constant is 7.29 Å, which confirms the result from energy

minimization. The Wyckoff position for O is (0.305, 0.174, 0), which is somewhat

different from that in He et al.’s work. [20] while all the other positions for Ca, Cu and Ti

are the same.

Figure 5.3 Bond lengths and angles of CaCu3Ti4O12 after Geometry Optimization
(bond lengths are in blue and angles are in black.)
38
Bond lengths and angles after geometry optimization are shown in Figure 5.3. The

Cu cation along with its four nearest neighboring O atoms forms a slightly distorted

square of CuO4 with Cu-O distance of 1.907 Å, 0.68% smaller than 1.92 Å in He et al.’s

work [20] and O-Cu-O angle of 96.498o. The four next nearest neighboring oxygens are

2.785 Å away from Cu, with O-Cu-O angle of 118.153o. The third nearest neighboring

oxygens are 3.252 Å away from Cu, with O-Cu-O angle of 93.808o. These three kinds of

CuO4 complexes are orthogonal to one another.

5.4 Band structure of CaCu3Ti4O12

Figure 5.4 Band structure of CaCu3Ti4O12

Figure 5.4 shows the calculated band structure of CCTO. The unoccupied

conduction bands above 1 eV have predominant Ti 3d character. The bands below 1 eV

are composed of the Cu 3d-O 2p manifold. The Fermi energy is 0.48 eV. The direct band
39
gap is 0.57 eV, which is greater than 0.27 eV in He et al.’s work [20] and closer to the

experimental optical gap of 1.5 eV as a lower limit.[see ref. 20 in (20)] The underestimation of

the gap is common in LDA calculations and results from the underestimation of

correlation effects.

5.5 Density of states of CaCu3Ti4O12

Figure 5.5 Density of states of CaCu3Ti4O12


(Red line indicates the Fermi level.)

Our density of states shown in Figure 5.5 is similar to that presented by He et al.[20]

The bands shown in the energy range in Figure 5.4 are mainly O 2p and Cu 3d orbitals.

The bands from -7.25 eV to -0.5 eV consist mainly of Cu 3d orbitals that hybridize

weakly with O 2p orbitals that point toward their Ti neighbors. In the range from -0.5 eV

to 3 eV around the Fermi level, the band structure reflects two narrow bands each of

40
which contains three bands. These six isolated bands consist of antibonding interactions

of Cu 3d orbitals and O 2p orbitals pointing to the Cu from its four near oxygen

neighbors. These are responsible for the magnetism of the compound.

5.6 Charge densities of CaCu3Ti4O12

The total charge density is symmetric and the spin density is antisymmetric, under

the fractional lattice translation (½, ½, ½), consistent with an AFM state. Spin-up and

spin-down charge densities are shown in Figure 5.6. It indicates antibonding interactions

between Cu 3d and O 2p, which extend over the central cluster composed of a Cu ion and

its four nearest O neighbors.

Figure 5.6 Spin-up (left panel) and spin-down (right panel)


charge densities of CaCu3Ti4O12
41
CHAPTER VI

FIRST-PRINCIPLES CALCULATIONS ON Ca1+xCu3-xTi4O12

We performed first-principles calculations on Ca1+xCu3-xTi4O12 (x= -1, 1, 2, 3).

Different Ca and Cu ratios give different materials, i.e. CuTiO3, Ca2Cu2Ti4O12,

Ca3CuTi4O12, CaTiO3. Electronic properties are calculated.

6.1 Structure

Accelrys MS Modeling 4.0 is applied to build the structure. The structures are

shown in Figure 6.1-6.4. The structures of CuTiO3, Ca3CuTi4O12 and CaTiO3 are built

based on the structure of CaCu3Ti4O12 shown in Figure 5.1. CuTiO3 is built by

substituting Ca by Cu. The crystal structure of CuTiO3 is shown in Figure 6.1.

Ca3CuTi4O12 is built by exchanging the Ca and Cu positions. The crystal structure of

Ca3CuTi4O12 is shown in Figure 6.3. CaTiO3 is built by substituting Cu by Ca. The crystal

structure of CaTiO3 is shown in Figure 6.4.

However, the structure of Ca2Cu2Ti4O12 can be built from the structure of

CaCu3Ti4O12, so this structure is built separately from the others. The space group of

Ca2Cu2Ti4O12 is Pm3. The Wyckoff positions are shown in Table 6.1. The charges are

chosen to be +2, +2, +4, -2 for Ca, Cu, Ti and O, respectively. The crystal structure of

Ca2Cu2Ti4O12 is shown in Figure 6.2.

42
Figure 6.1 The structure of CuTiO3 Figure 6.2 The structure of Ca2Cu2Ti4O12

Figure 6.3 The structure of Ca3CuTi4O12 Figure 6.4 The structure of CaTiO3

43
Table 6.1 The Wyckoff positions for Ca2Cu2Ti4O12

Atom Position

Ca (0, 0, 0) (1/2, 1/2, 0)

Cu (1/2, 0, 0) (1/2, 1/2, 1/2)

Ti (1/4, 1/4, 1/4)

O (1/4, 1/4, 0) (1/4, 1/4, 1/2)

6.2 Geometry optimization

Computations are performed via the LDA with ultra-soft pseudopotentials. The

initial lattice constant is 7.29 Å which is the equilibrium lattice constant from our former

single energy calculation for CCTO. Geometry optimization is performed at a plane-wave

cutoff energy of 500 eV, with a 2×2×2 Monkhorst-Pack k-point mesh. The relaxations are

performed using the BFGS [34] algorithm and all the atoms are allowed to relax during

changes to the lattice.

For CuTiO3, the optimized lattice constant is found to be 7.25 Å. The optimization

result shows that Cu and Ti atoms are still in the original positions, which agrees well

with the fact that the CuTiO3 structure is perovskite structure. The O atoms move from

their original positions, but the result doesn’t show obvious tendency of moving towards

to the center of each side of the primitive cell CuTiO3.

For Ca2Cu2Ti4O12, the optimized lattice constant is found to be 7.58 Å.

44
For Ca3CuTi4O12, the optimized lattice constant is found to be 7.55 Å. From the

optimized structure, Ca, Cu and Ti atoms are still in the original positions. The O atoms

move towards the center of each side of the primitive cell, which shows a tendency

towards pure perovskite structure. Ca3CuTi4O12 is a composite of CaCu3Ti4O12 and

CaTiO3. (CaCu3Ti4O12 : CaTiO3=1:8) Since it is rich in CaTiO3, its properties resemble

CaTiO3 which is a perovskite structure.

For CaTiO3, the optimized lattice constant is found to be 7.57 Å. Wang et al. [39]

performed generalized gradient approximation (GGA) calculations using the plane-wave

pseudopotential method to find that the theoretical lattice constant is 3.88 Å. Since we

use 40-atom cell in our calculations, the lattice constant is 7.57/ 2= 3.785 Å in a 20-atom

cell, which is 2.5 % smaller than the theoretical one in Wang et al.’s work. [39] Compared

with the experimental one of 3.8950 Å [40], our theoretical lattice constant is 2.8 % smaller.

The optimized structure shows that Ca and Ti atoms are still in the original positions. The

O atoms move towards to the center of each side of the primitive cell CaTiO3. CaTiO3 has

perovskite structure. Therefore, during optimization the structure changes from that of

CaCu3Ti4O12 structure to that of CaTiO3 as we expected.

The summary of optimized lattice constants for different Ca, Cu ratio is shown in

Table 6.2. As the Ca, Cu ratio increases from 0 to ∞, the optimized lattice constant

increases as well except for Ca2Cu2Ti4O12, which shows a divergence. As we know, the

pauling ionic radii of Ca2+ is 99 pauling radius/pm and that of Cu2+ is 96 pauling

radius/pm. Since Ca2+ > Cu2+, the lattice constant should increase if more calcium is

added to the material. This agrees with our result.

As for the abnormal behavior of Ca2Cu2Ti4O12, one reason may be the simulation

45
structure of Ca2Cu2Ti4O12 is not based on the structure of CCTO, while the structures of

the others are all based on the structure of CCTO.

Table 6.2 Summary of optimized lattice constant for different Ca, Cu ratio

Optimized lattice
Material type Ca, Cu ratio
constant

CuTiO3 Ca: Cu=0 7.25 Å

CaCu3Ti4O12 Ca: Cu=1:3 7.29 Å

Ca2Cu2Ti4O12 Ca: Cu=2:2 (1:1) 7.58 Å

Ca3CuTi4O12 Ca: Cu=3:1 7.55 Å

CaTiO3 Ca: Cu=∞ 7.57 Å

6.3 Band structure

The band structures of CuTiO3, Ca2Cu2Ti4O12, Ca3CuTi4O12, CaTiO3 are shown in

Figure 6.5-6.8.

For CuTiO3, the direct band gap at the Γ point is 0.53 eV. The indirect band gap,

which is between the valence band maximum Γ point and the conduction band minimum

X point, is 0.65 eV. The Fermi level is 0.32 eV.

For Ca2Cu2Ti4O12, the direct band gap at the Γ point is 1.22 eV. The indirect band

gap, which is between the valence band maximum Γ point and the conduction band

minimum X point, is 1.99 eV. The Fermi energy is 0.19 eV.

For Ca3CuTi4O12, the direct band gap at the Γ point is 2.30 eV. The indirect band gap,

which is between the valence band maximum Γ point and the conduction band minimum

X point, is 2.36 eV. The Fermi energy is 0.19 eV.


46
For CaTiO3, the direct band gap at the Γ point is 2.20 eV. The indirect band gap,

which is between the valence band maximum Γ point and the conduction band minimum

X point, is 2.30 eV. The Fermi energy is 0.27 eV.

Figure 6.5 Band structure of CuTiO3 Figure 6.6 Band structure of Ca2Cu2Ti4O12

Figure 6.7 Band structure of Ca3CuTi4O12 Figure 6.8 Band structure of CaTiO3

47
The direct band gaps are summarized in Table 6.3. With the increase of Ca in the

material, the band gap increases. The increasing band gap indicates the material is more

like an insulator. Therefore, with the increase of Ca in the material, insulator character

becomes much more pronounced. This is in good agreement with the experimental

results.

Table 6.3 Summary of direct band gap for different Ca, Cu ratio

Material type Direct band gap

CuTiO3 0.53 eV

CaCu3Ti4O12 0.57 eV

Ca2Cu2Ti4O12 1.22 eV

Ca3CuTi4O12 2.30 eV

CaTiO3 2.20 eV

6.4 Density of states

The densities of states of CuTiO3, Ca2Cu2Ti4O12, Ca3CuTi4O12, CaTiO3 are shown in

Figure 6.9-6.12. The dashed line does not indicate the Fermi level in every figure. The

Fermi level is located at 0.32 eV, 0.19 eV, 0.19 eV and 0.27 eV for CuTiO3, Ca2Cu2Ti4O12,

Ca3CuTi4O12, CaTiO3, respectively.

The bands below the valence band maximum are mainly O 2p and Cu 3d in

character. Cu 3d orbitals hybridize weakly with O 2p orbitals. The lowest conduction


48
bands are made up of Ti 3d bands, which hybridize with O 2p orbitals. Ca contribution is

quite small in this energy range.

Figure 6.9 Density of states of CuTiO3

Figure 6.10 Density of states of Ca2Cu2Ti4O12

49
Figure 6.11 Density of states of Ca3CuTi4O12

Figure 6.12 Density of states of CaTiO3

6.5 Charge density

The charge densities of CuTiO3, Ca2Cu2Ti4O12, Ca3CuTi4O12, CaTiO3 are calculated.

They show similar results. In this thesis, the charge densities of Ca2Cu2Ti4O12 are shown

as an example. Figure 6.13 (a) and (b) shows the charge densities of Ca2Cu2Ti4O12 on the

50
Ca-Cu-O and Ti-O planes. The high charge density region is around the nuclear sites and

is nearly spherically symmetric. The Ti-O bond has a strong covalent character which

results from the noticeable overlapping charge distribution at the middle of the Ti-O bond.

There is no chemical bonding between any other types of atoms, e.g., Ca-O or Cu-O.

(b)
(a)
Figure 6.13 Charge densities of Ca2Cu2Ti4O12(atoms are colored as in Figure 6.2)

(a) Ca-Cu-O (b) Ti-O

6.6 Summary of the work on Ca1+xCu3-xTi4O12

Our calculated ground state properties of CCTO are in good agreement with other

people’s work.[20] The optimized lattice constant is 7.29 Å, which is 1 % less than the

experimental value of 7.384 Å. The density of states and charge densities confirm the

AFM character of CCTO.

For Ca1+xCu3-xTi4O12, we find that with increase of Ca in the material, the optimized
51
lattice constant and band gap increase, and the insulator character becomes much more

pronounced in the material. This is in good agreement with the experimental results.

52
REFERENCES

[1] M.E. Lines and A.M. Glass, Principles and Applications of Ferroelectrics and Related
Materials, Clarendon Press, Oxford, 1977

[2] L. Rimai and G. deMars, Phys. Rev. 127, 702 (1962)

[3] R. O. Bell and G. Rupprecht, Phys. Rev. 129, 90 (1963)

[4] R. S. Krogstad and R. W. Moss, Bull. Am. Phys. Soc. 7, 192 (1962)

[5] M. J. Weber and R. R. Allen, J. Chem. Phys. 38, 726 (1963)

[6] T. Mitsui and W. B. Westphal, Phys. Rev. 124, 1354 (1961)

[7] E. Sawaguchi, A. Kikuchi and Y. Kodera, J. Phys. Soc. Japan 18, 459 (1963)

[8] F. W. Lytle, J. Appl. Phys. 35, 2212 (1964)

[9] E. Pytte and J. Feder, Phys. Rev. 187, 1077 (1969)

[10] J. Feder and E. Pytte, Phys. Rev. B 1, 4803 (1970)

[11] W. Zhong and D. Vanderbilt, Phys. Rev. Lett. 74, 2587 (1995)

[12] W. Zhong and D. Vanderbilt, Phys. Rev. B 53, 5047 (1996)

[13] P. A. Fleury, J. F. Scott and J. M. Worlock, Phys. Rev. Lett. 21, 16 (1968)

[14] T. Ikeda, T. Kobayashi, M. Takata, T. Takayama and M. Sakata, Solid State Ionics,
108, 157 (1998)

[15] Y. Shirane, Y. Yamada, Phys. Rev. 177, 858 (1969)

[16] H. Fujishita, Y. Shiozaki and E. Sawaguchi, J. Phys. Soc. Jpn. 46, 581 (1979)

[17] M. A. Subramanian, D. Li, N. Duan, B. A. Reisner and A. W. Sleight, J. Solid State


Chem. 151, 323 (2000)

53
[18] A. P. Ramirez, M. A. Subramanian, M. Gardel, G. Blumberg, D. Li, T. Vogt and S. M.
Shapiro, Solid State Commun. 115, 217 (2000)

[19] B. Bochu, M. N. Deschizeaux and J. C. Joubert, J. Solid State Chem. 29, 291 (1979)

[20] L. He, J. B. Neaton, M. H. Cohen, D. Vanderbilt and C. C. Homes, Phys. Rev. B 65,
214112 (2002)

[21] L. He, J. B. Neaton, D. Vanderbilt and M. H. Cohen, Phys. Rev. B 67, 012103 (2003)

[22] M. D. Johannes, W. E. Pickett and R. Weht, Mat. Res. Soc. Symp. Proc. 718, 25
(2002)

[23] G. Li, Z. Yin and M. Zhang, Phys. Lett. A 344, 238 (2005)

[24] C. C. Homes, T. Vogt, S. M. Shapiro, S. Wakimoto, M. A. Subramanian and A. P.


Ramirez, Phys. Rev. B 67, 092106 (2003)

[25] G. Sutmann, Quantum Simulations of Complex Many-Body Systems: From Theory


to Algorithms, Vol. 10, 211-254 (2002)

[26] M. P. Allen, Computational Soft Matter, Vol 23, 1-28 (2004)

[27] A. R. Leach, Molecular Modeling: Principles and Applications, Longman Singapore


Publishers Ltd. (1996)

[28] D. Vanderbilt, Phys. Rev. B 41, 7892 (1990)

[29] T. Katsumata, Y. Inaguma, M. Itoh, K. kawamura, Solid State Ionics, 108(1998)


175-178

[30] A. K Rappe, C. J. Caeswit, K. S. Colwell, W. A. Goddard III and W. M. Skiff, J. Am.


Chem. Soc., 114, 10024 (1992)

[31] Accelrys Inc., Materials Studio CASTEP, Accelrys Inc., San Diego, 2006

[32] C. B. Samantaray, H. Sim and H. Hwang, Microelectronics Journal 36, 725 (2005)

[33] A. D. Polli, T. Wagner, T. Gemming, M. Ruhle, Surf. Sci. 448, 279 (2000)

[34] J. Nocendal and S. J. Wright, Numerical Optimization, Springer, 193-201(1999)

[35] M. Cardona, Phys. Rev. 140, A651 (1965)

[36] S. P. Kowalczyk, F. R. McFeely, L. Levy, V. T. Gritsyna and D. A. Shirely, Solid


State Commun. 23, 161 (1977)
54
[37] M. Fujita, R. Sekine and S. Sugihara, Jpn J. Appl. Phys. 38, 5664 (1999)

[38] S. Zollner, A. A. Demkov, R. Liu, P. L. Fejes, R. B. Gregory, J. A. Curless, Z. Yu, J.


Ramdani, R. Droopad, T. E. Tiwald, J. N. Hilfiker and J. A. Woollam, J. Vac. Sci.
Technol. B 18(4), 2242 (2000)

[39] Y. X. Wang, M. Arai, T. Sasaki and C. L. Wang, Phys. Rev. B 73, 035411 (2006)

[40] B. J. Kennedy, C. J. Howard and B. Chakoumakos, J. Phys.: Condens. Matter 11,


1479 (1999)

55

Potrebbero piacerti anche