Born effective charge tensors, dielectric permitivity tensors, the phonon frequencies at the Bril- louin zone center and mode oscillator strengths are calculated using density func tional perturbation theory.
Born effective charge tensors, dielectric permitivity tensors, the phonon frequencies at the Bril- louin zone center and mode oscillator strengths are calculated using density func tional perturbation theory.
Born effective charge tensors, dielectric permitivity tensors, the phonon frequencies at the Bril- louin zone center and mode oscillator strengths are calculated using density func tional perturbation theory.
AB INITIO CALCULATIONS ON ELECTRONIC AND LATTICE
DYNAMICAL PROPERTIES OF COPPER INDIUM DISELENIDE
AND COPPER INDIUM DISULPHIDE
by
Cihan Parlak
(234992
‘THE GRADUATE SCHOOL OF NATURAL AND APPLIED SCIENCES
OF
THE ABANT IZZET BAYSAL UNIVERSITY
‘THESIS SUBMITTED TO
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE
OF
MASTER OF SCIENCE
IN
THE DEPARTMENT OF PHYSICS
ve viusexoe ca
Bolu, Turkey
July 2002Approval of the Graduate School of Natural Sciences —
2 1
Lv
Prof. Dr. Nihet Celebi
Director
certify that this thesis satisfies all the requirements as a thesis for the degree of
Prof. Dr. Ahmet Turan Alan )
Head of Physics Department
Master of Science.
‘This is to certify that we have read this thesis and that in our opinion it is fully
adequate, in scope and quality as a thesis for the degree of Master of Science.
Assistant Prof. at
Supervisor
‘Examining Committee Members
1. Assist. Prof. Dr. Resul Eryigit ba Leja
2. Assoc. Prof. Dr. Ayge Morkan
3. Assist. Prof. Dr. Cabir Terziogiu
iiTo My Family ...ABSTRACT
Ab initio Calculations on Electronic and Lattice Dynamical
Properties of CulnSe, and CulnS,
Cihan Parlak
Master of Science, Department of Physics
Supervisor: Assist. Prof. Dr. Resul Eryigit
July 2002, 68 pages
We have performed a first-principles study of structural, dynamical, dielec-
tric and electronic properties of chalcopyrite semiconductors CulnSe, and CulnS».
‘The calculations have been carried out within the local density functional approx-
imation using norm-conserving pseudopotentials and a plane-wave basis. Elec-
tronic band structure of both compounds found to be similar. Born effective
charge tensors, dielectric permitivity tensors, the phonon frequencies at the Bril-
louin zone center and mode oscillator strengths are calculated using density func-
tional perturbation theory. The calculated properties agree with some measure-
ments.
Keywords: Chalcopyrite semiconductors, electronic band structures, Born ef-
fective charges, phonon frequencies, dielectric permitivity tensors.
iv6ZET
CulnSe; ve CulnS,’iin Elektronik ve Orgii Dinamigi Ozellikleri
Uzerine Ab initio Hesaplamalar
Cihan Parlak
Yiiksek Lisans, Fizik Béliimit
Tez Danigmam: Yrd. Dog. Dr. Resul Eryigit,
‘Temmuz 2002, 68 sayfa
Bu teade chalcopyrite yar iletkenler CulnSe, ve CulnS, ‘iin yapisal, dinamik,
dielektirik ve elektronik zelliklerini inceledik. Oncelikle hesaplarda yerel yogunluk
yaklagum, norm-korunumlu pseudopotential ve diizlem dalga taban) kullanild
Elekronik band yapilan ikki madde iginde benzer bulundu. Born etkin yiik tensérleri,
dielektirik gecirgenlik tensdrleri, Brillouin bélgesi merkezindeki fonon frekanslart
ve mod titresim giicleri hesapland:. Bu degerler literatiirdeki deerlerle karglastirilarak
baz verilerle uyumlu oldugu gorildi
Anahtar Kelimeler: Chalcopyrite yaniiletkenler, elektronik band yapilan, Born
ektin yiik tensérleri, fonon frekanslan, dilektrik gegirgenlik tensérleri,ACKNOWLEDGEMENTS
I would like to thank my advisor, Dr. Resul Eryigit for his support, guidance,
patience and encouragement during years of my work as a graduate student at
AIBU. I am obliged for his continuous guidance and support and thank to him
for his help in the revision of scientific content and style of this dissertation
And I want to thank him not only for guidance for scientific content’ but also his
guidance for all contents of life
I would like to thank my friends, colleagues and teachers in physics department
for their friendship and encouragement all trough my years at AIBU.
Tam grateful to Dr. Ayge Morkan and Dr. Cabir Terzioglu along with my
advisor for serving on my dissertation.
Tam grateful to "Teknik Bilgisayar ve Hizmetleri” for their computer support.
Tam thankful to my family, for their support, patience, understanding and
financial support.
TD
upekOGRETIN KOETL
pac MARTASYON MERITTABLE OF CONTENTS
ABSTRACT
OZET ee eee eee
ACKNOWLEDGEMENTS
1 INTRODUCTION .
2 DENSITY FUNCTIONAL THEORY .. .
2.1. The Plane-Wave Pseudopotential Method
2.2 Plane-Wave Solution of Kohn-Sham Equations:
3 REVIEW OF EXPERIMENTAL STUDIES
4 ELECTRONIC STRUCTURE OF CHALCOPYRITES .. . . .
4.1 Pseudopotentials
4.2 Convergence Issues
4.3 Atomic Structure -
4.4 Electronic Structure
5 LATTICE DYNAMICS RESULTS
5.1 Born Effective Charge Tensors .
5.2 Phonons .. .
5.3 Lattice Dielectric Tensors
6 CONCLUSION .
REFERENCES .
iv
vi3.1
3.2
33
41
42
43
44
45
46
47
48
49
4.10
411
4.12
413
ald
415
4.16
5.1
5.2
53
54
LIST OF FIGURES
Conventional unit cell of chalcopyrite crystal structure. 15
Conventional unit cell of zinc-blende crystal structure 15
‘Tetragonal chalcopyrite primitive unit-cell 16
Energy Levels of Cu, In and Se atoms. 24
Energy Levels of Cu, In and $ atoms. : sees BB
‘The ‘Troulliers Martin pseudopotentials for Cu... +... 26
‘The Troulliers Martin pseudopotentials for In 27
The Troulliers Martin pseudopotentials for Se 28
‘The Troulliers Martin pseudopotentials for 8 29
‘Wave functions of Copper valance orbitals. 30
‘The band structure of CulnSe,. .. . . . 32
‘The band structure of CulnS2. . . BB
Details of the high valance and low conduction states of CulnSe,. 34
Details of the high valance and low conduction states of CulnS; . 35
‘Modulus square of the wave functions of CulnSe,-first band 36
Modulus square of the wave functions of CulnSey-sixth band... 37
Modulus square of the wave functions of CulnSey-10 band 38
Modulus square of the wave functions of CulnSe,-28" band 39
Modulus square of the wave functions of CulnSe,-29" band 40
Reflectance spectra obtained by using sum form of the model di-
electric function for El}c polarization . 47
Reflectance spectra obtained by using sum form of the model di-
electric function for E-Le polarization . . 8
Reflectance spectra obtained by using product form of the model
dielectric function for Elle polarization ............... 48
Reflectance spectra obtained by using product form of the model
dielectric function for Elle polarization .... 2.2.0.2... 50
wa ze, VouSEKGGRETIM KURULDLIST OF TABLES
3.1 Coordinates of the atoms of chalcopyrite primitive unit cell .
4.1 Convergence study on total energy of CulnSe, and Culn$a.
4.2 Calculated structural parameters of CulnSea :
4.3 Band-Gaps of CulnSe, and CulnS, . .
Calculated Born effective charges of CulnSe,
Frequencies of phonons at the I’ point .
Reanalyzed reflectivity data
5.4 Static and high frequency dielectric tensor components of CulnSes.
5.5 Oscillator strength tensor, magnitude of mode-effective charge ve
tors and LO frequencies of CulnSez .........- 20.005CHAPTER 1
INTRODUCTION
‘The increasing energy demand, dwindling fossil-fuel resources and a growing trend
towards so called clean-energy have been fuelling the research for high-efficiency
solar materials for last three decades. Among the promising solar-cell materials,
Cu-based chalcopyrite (CH) semiconductors have been shown to have the highest
conversion efficiency of 18 % (1, 2]
As experimental work on these compounds continues, a better theoretical
understanding of their electronic properties need to be provided. ‘This thesis is
about the electronic and lattice properties of two of these group of semiconductors,
namely CulnSez and CulnS,
Nowadays the standard tool of theoretical condensed matter physics to deal
with the electronic and lattice properties of solids is density functional theory
(DFT) of Kohn-Sham [3] and Hohenberg-Kohn [4]. This extremely successful
and simple theory, which was established almost 40 years ago has been immensely
popular, first in physics and chemistry and for last 10 years also for biology.
Density functional theory, in its many forms helped in understanding many
different. properties of systems as simple as atoms to very complicated surface
structures. In condensed matter applications, it is generally used within so called
local density approximation (LDA) and pseudopotentials to account: for the inter-
action between the valance electrons and the nuclei. In Section 2, we will review
the basics of this theory and its implementation with a plane-wave basis.
One should also mention the dramatic increase in computing power and de-
velopment of very efficient algorithms and programs to solve the differential equa-
1tions of density functional theory. These two factors made it possible to solve
problems, that could not be solved on a supercomputer 10 years ago, on a per-
sonal computer. This trend, if continues as it is, will make the first principles
calculations possible for much more complicated structures, within next decade.
We have set the stage for the possibility of obtaining the quantum mechanical
properties of moderately complicated chalcopyrite semiconductors. For solar en-
ergy applications, the most important property is of course, the band structure
of the material. Band-structure of material determines the allowed energy values
for the electrons of the crystal. For semiconducting systems, the band structure
has a typical forbidden band between the valance bands which are completely
full at 0 °K, and conduction bands, which are completely empty at 0 °K. For a
good absorber of sun-light, one needs a material with a spectrum of sun-light.
Of course, the band-gap photon energy match is not enough for high-efficiency
absorption. Absorption /emission process involves an energy and a momentum
transfer and so called direct band-gap materials would have higher efficiencies.
All Cu-based ternary semiconductors have direct band-gaps at the Brillouin zone
center (I point). In this thesis we investigate the electronic band structure of
CulnSe, and CulnS; with density functional theory within local density approx-
imation using norm-conserving pseudopotentials and a plane-wave basis set for
expansion of the wave function.
Actually, these settings for the calculations are not the best ones possible. It
is well-known that DFT-LDA underestimates the band-gap of semiconductors.
This property has been thoroughly investigated over the last 15 years and it is
well understood. A different theory, so called GW approximation is needed in
solids{5]. Unfortunately, GW approach is computationally extremely demanding.
But over the years, it has been well established that the GW band structureand DFT-LDA band structure can be brought to agreement by a k-independent
shifting (so called "scissor approximation") of DFT-LDA valance bands.
‘As the electronic structure is important for the way photons absorbed and
emitted, the lattice dynamical properties of solids are very important for the
thermal properties such as thermal expansion coefficient, heat capacity and ther-
modynamical Griineisen parameters. The phonon frequencies for the wave vector
at the center of the Brillouin zone are also very important diagnostic properties.
Optical diagnostic tools, such as Raman and Infrared reflectivity spectroscopies
can be used to measure symmetry allowed modes at the zone center. This infor-
mation can be used as a guide for the quality of the crystal.
Lattice dynamical studies, experimental as well as theoretical have been the
standard tools of solid state physics for last 60 years. On the experimental scene
the main improvement came from the development of lasers, 40 years ago, and
better detectors, such as charge coupled devices (CCD) in 1980's and 90’s. The
major theoretical development in this field came in last 15 years like almost
every other development in computational condensed matter physics; impressive
developments in computing power was one of driving forces. Theoretical studies in
this field can be broadly divided into two classes; the so called phenomenological
models and ab initio methods. Phenomenological models try to guess the most
important interactions between the constituent atoms compromising the system
and tries to find the parameters of the interaction by fitting the frequencies of the
model with the experimentally measured ones. Because of this empirical nature,
the predictions of phenomenological models are not always good.
Contrary to phenomenological models, ad initio methods do not assume any
form for the interactions of the atoms of the system and tries to calculate ev-
erything from first-principles quantum mechanical quantities. There are numberof different ad initio approaches with respective advantage and disadvantages.
Partly because of its computational ease, the so called frozen-phonon method (6]
was the first ab initio phonon method. This method tried to find the phonon dis-
placement pattern and calculate the energy difference between the perfect crystal
and the crystal where the atoms are displaced according to particular phonon
pattern, This technique is good for phonons at high symmetry points of the Bril-
louin zone. For a phonon at an arbitrary point in Brillouin zone, the displacement
pattern would require a large unit cell and calculation becomes prohibitively large
A completely different approach to ab initio phonon problem is so called linear
response approach {7]. In this approach, one considers phonon as a perturbation
an the electronic density of the crystal and phonon frequencies are computed from
the linear response of the crystal. This method has the added advantage that
one can consider other perturbations, such as electrical field, which is important
for the longitudinal optical modes of polar materials.
Lattice dynamical properties of CulnSe have been investigated experimen-
tally by many groups over the last thirty years (8, 9, 10, 11, 12, 13, 14, 15, 16,
17, 18). Phonon frequencies at the I point have been measured by all possi-
ble spectroscopies, Gan et al. (9], Bodnar et al. [8], Rincon et al. (14), Bacewicz
et al. (15), Tanino et al. {18] report Raman measurements at 100 °K and 300
°K while Bodnar et al. [8], Gan et al. [9], Riede et al. [10], Neuman et al. (11),
report infrared measurements at ~ 4°K and Derollez et al. (17] reported Neutron
measurements. Unfortunately, the picture that arises from these reports is one
of confusion. Even the frequencies measured by different groups using the same
technique show considerable discrepancies for some of modes. Same confusion
persist for the dielectric properties that can be measured by infrared reflectivity
methods.
4 ec vunst: uM KCURUS
‘DOKUMANIASYONSo we combine the predictive power of ab initio methods with linear response
approach to try to shed some light on the question of the zone center phonon
frequencies and dielectric tensors of chalcopyrite CulnSey semiconductor.
The plan of the thesis is as follows; in Section 2, we give a brief overview of
the main theory of density functional theory in pseudopotential plane-wave im-
plementation. Section 3 review previous experimental as well as theoretical work
on lattice dynamical and electronic properties of CulnSe, and CulnS,, The main
ingredients of the calculation and electronic band structure and wave function of
the system are displayed and discussed in Section 4. Section 5 is about the calcu-
lated phonon, Born effective charges and dielectric properties. Thesis concludes
with discussion and further research suggestion in Section 6.CHAPTER 2
DENSITY FUNCTIONAL THEORY
From the standpoint of quantum mechanics, a crystal can be considered as system
of nuclei and electrons interacting through Coulombic forces. The standard way
of entangling the nuclear degrees of freedom is called "adiabatic approximation”
which is based on the fact that the mass of electrons is much smaller than nucleic
mass.
With this approximation, one transforms the problem into two parts; the
problem of electrons as if the nuclei are stationary and problems of nuclei moving
on a potential energy surface determined by the electronic distribution.
‘The solution of the first problem would give the electronic energy band struc-
ture and electronic wave functions of the crystal, while the solution to second
problem are the vibrational properties of the crystal.
Let us start with the first problem, here we have N interacting electrons in
an extemal potential, created by stationary nuclei. This system is described
quantum mechanically by the solutions of Schrédinger many body equation;
HY(F, Fv) = BVA, Fn) (2.1)
where H in the Hamiltonian:
(2.2)
os Veal) + & oy
A 2 = 7]
where #{ is the position of the i electron and Vize(fi) is the external potential
that act on the electron i. In these formulas we are using atomic units, where
61; ifm, = } and e* = 2 then the energies are in Rydberg (Ryd) or if m, = 1,
1, energies are in Hartree (Ha)
‘Unfortunately, the solution of Equation (2.1) cannot be found analytically or
numerically, because Coulombic form couples all the electronic degrees of freedom.
The usual practice, for last four decades, to overcome this problem has been to
forget about the many body wave function to find the ground-state charge density,
which can also be used to calculate the observable ground-state properties of the
crystal.
In practice, Equation (2.2) is replaced by equation derived from density func-
tional theory (DFT) of Hohenberg and Kohn (4]. This theory shows that the
external potential acting on electrons is a functional of the ground-state charge
density. This potential, in turn, determines the ground-state wave-function and
therefore all the ground-state properties of the electronic system are functionals
of the electronic density.
An important ground-state observable is the total energy of the system. Charge
density can be found as the function that minimizes the total energy functional
as demonstrated by Kohn and Sham (KS) (3]
Formally, one can write the ground-state energy of the many-body system as
the expectation value of the Hamiltonian as;
Gtr] =< voll > (23)
Let us specialize the situation to properties of periodic solids, where nuclear
position are fixed by the introduction of Bravais lattice vectors K,, and by the
choice of basis vectors 7, which characterize the positions of the nuclei inside the
unit cell so that the total energy of the electrons and the nuclei can be written
as;Eo = Ex + G[n) (2.4)
where B,.; , the Coulombic potential energy of the ionic subsystem which can be
computed as;
where Z,, the charge of nucleus at r, and prime on summation indicates that
=v and s = $1 term is omitted from the sum.
Kohn and Sham introduced a new functional G{n] with help of wave-functions
of independent electron gas with the same density as the interacting system.
Within this approach, let |y > be the single particle wave functions of non-
interacting system, then the kinetic energy would be
T
EEDA calves > 26)
where f, is the Fermi function, which is equal to 2 (taking into account spin
degeneracy) if the energy of the i* level is less than Fermi energy and equals 0 if
it is higher. The density of same non-interacting system would be;
nf) =D Ala? (2.7)
this density, by construction, is same as the density of the interacting system. So
Ey would include a Coulombic term of the form;
Bun f ie rgd, MEE (2.8)
External interaction would give;Eze =D) fi < VilVeatlvi > (2.9)
In practice Veg is the effect of nuclei. If these terms are extracted from the
total energy of the interacting system one would get a functional that includes all
the many-particle effects. This part of the total energy is known as the exchange-
correlation energy of the system
Ezeln = G{n| = Bee —T — En (2.10)
If E,c{n] was known, the ground state density could be computed by min-
imizing total ground state energy. This minimization is alone with respect to
single-particle wave function ¥;(r) with the orthogonality constraint;
< Pils >= 5 (2.11)
which leads to Kohn-Sham equations;
[aac + al) + Vel) + Veal] LAD = ewieCA) (212)
The solutions of this equation are used to build up charge density n(F) in
Equation (2.7) in Equation (2.12);
2 iE an 2D, (2.13)
and
Vae(7) = oat (2.14)
‘The physical observables of KS are total ground-state energy and ground-statedensity.
2.1 The Plane-Wave Pseudopotential Method
Ina periodic solid, Equation (2.12) can be solved by expanding ¥i(7) in a complete
set of basis functions, such as plane waves (PW), gaussian functions or atomic
orbitals. We will discuss only the plane wave expansion because of its conceptual
simplicity and computational advantages. Because of lattice periodicity, plane
‘waves are natural basis and they don’t suffer from basis towards atomic positions.
Hartree potential is easier to compute with PW, and one can systematically
increase the resolution of the computation by increasing the number of plane
waves used to expand ¥(7).
Since V.zs(F) is the muclear potential, the wave functions if electrons close to
nuclei would have highly oscillatory structure, which is very difficult to reproduce
with a limited number of plane waves. One possible solution of this problem
causes from the realization that the energies of care states are orders of magnitude
larger than the valence states and the electrons that are important in bonding
are valence electrons. So, core electrons can be eliminated from Equation (2.12)
This approximation amounts to substitution of nuclear potential with a pseu-
dopotential whose lowest energies coincides with the valence all-electron energies
and whose wave-functions coincide with all-electron wave functions in regions far
away from the nucleus.
Pseudopotential theory has been developed to great accuracies in last three
decades In practice, the pseudopotential of nuclei s is written as
loa
VA) = V+ YVR (2.15)
Ss
where P; is the projector
10