Applications of density functional theory for modeling...
Transcript of Applications of density functional theory for modeling...
Applications of Density functional theory for
modeling metal-semiconductor contacts, reaction
pathways, and calculating oxidation states
Sergei Posysaev
DFT
University of Oulu Graduate School
Reports series in Physical Sciences
Report No. 123 (2018)
APPLICATIONS
APPLICATIONS OF DENSITY FUNCTIONAL THEORY
FOR MODELING METAL-SEMICONDUCTOR CONTACTS,
REACTION PATHWAYS, AND CALCULATING OXIDATION
STATES
SERGEI POSYSAEV
University of Oulu Graduate School
University of Oulu
Faculty of Science
Nano and Molecular Systems Research Unit
Academic Dissertation to be presented with the assent of the Faculty of Science,
University of Oulu, for public discussion in the Auditorium IT 116, on December
14th, 2018 at 12 o’clock noon.
REPORTS SERIES IN PHYSICAL SCIENCES Report No. 123
OULU 2018 UNIVERSITY OF OULU
Opponent
Prof. Marcel Sluiter, Delft University of Technology, Netherlands
Reviewers
Prof. Bernardo Barbiellini, Lappeenranta University of Technology, Finland
Prof. Kalevi Kokko, University of Turku, Finland
Custos
Prof. Matti Alatalo, University of Oulu, Finland
ISBN 978-952-62-2131-1
ISBN 978-952-62-2132-8 (PDF)
ISSN 1239-4327
Juvenes print
Oulu 2018
Posysaev, Sergei, Applications of Density functional theory for modeling
metal-semiconductor contacts, reaction pathways, and calculating oxidation
states
University of Oulu Graduate School; University of Oulu, Faculty of Science,
Nano and Molecular Systems Research Unit
Acta Univ. Oul. X XXX, 2018
University of Oulu, P.O. Box 8000, FI-90014 University of Oulu, Finland
Abstract
Density functional theory (DFT) is a well-established tool for calculating the
properties of materials. The volume of DFT-related publications doubles every 5–
6 years, which has resulted in the appearance of continuously growing open
material databases, containing information on millions of compounds. Furthermore,
the results of DFT computations are frequently coupled with experimental ones to
strengthen the computational findings.
In this thesis, several applications of DFT related to physics and chemistry are
discussed. The conductivity between MoS2 and transition metal nanoparticles is
evaluated by calculating the electronic structure of two different models for the
nanoparticles. Chemical bonding of Ni to the MoS2 host is proven by the system’s
band alignment. To meet the demand for cleaner fuel, the applicability of the (103)
edge surface of molybdenum disulfide in relation to the early stages of the
hydrodesulfurization (HDS) reaction is considered. The occurrence of the (103)
edge surface of molybdenum disulfide in the XRD patterns is explained. A method
for calculating oxidation states based on partial charges using open materials
databases is suggested. We estimate the applicability of the method in the case of
mixed valence compounds and surfaces, showing that DFT calculations can be used
for the estimation of oxidation states.
Keywords: Bader charge, density functional theory, molybdenum disulfide,
oxidation state, reaction pathway
Posysaev, Sergei, Tiheysfunktionaaliteorian sovelluksia metalli-
puolijohdeliitosten, reaktiopolkujen ja hapetustilojen mallinnuksessa
Oulun yliopiston tutkijakoulu; Oulun yliopisto, Luonnontieteellinen tiedekunta
Nano- ja molekyylisysteemien tutkimusyksikkö
Acta Univ. Oul. X XXX, 2018
Oulun yliopisto, PL 8000, 90014 Oulun yliopisto
Tiivistelmä
Tiheysfunktionaaliteoria (density functional theory, DFT) on yleisesti käytetty
työkalu laskennallisessa materiaalitutkimuksessa. DFT:llä tuotettujen julkaisujen
määrä kaksinkertaistuu 5-6 vuoden välein, minkä johdosta käytettävissä on
jatkuvasti kasvava määrä avoimia materiaalitietokantoja, joihin on talletettu
miljoonien yhdisteiden ominaisuuksia. DFT-laskujen tuloksia täydennetään myös
usein kokeellisilla tuloksilla.
Tässä työssä tarkastellaan tiheysfunktionaaliteorian sovelluksia fysiikassa ja
kemiassa. MoS2:n ja metallisten nanopartikkelien välistä johtavuutta on tutkittu
mallintamalla erilaisia nanopartikkeleita. Nikkelin ja MoS2:n välinen kemiallinen
sidos selittyy systeemin energiavöiden kohdistumisella. MoS2:n (103)-pinnan
soveltuvuutta rikinpoistoreaktion varhaisissa vaiheissa on tutkittu tarkoituksena
löytää uusia menetelmiä puhtaan polttoaineen tuottamiseksi. Myös (103)-pinnan
esiintyminen röntgendiffraktiokuvissa selitetään. Työssä on myös esitetty
menetelmä hapetustilojen laskemiseksi tietokannoista löytyvien laskettujen
varausjakaumien avulla. Menetelmän soveltuvuutta on tarkasteltu erilaisille
yhdisteille ja pinnoille. Tämä tarkastelu osoittaa, että DFT-tuloksia voidaan käyttää
hapetustilojen laskemiseen.
Asiasanat: Bader-varaus, tiheysfunktionaaliteoria, molybdeenidisulfidi,
hapetustila, reaktiopolku
1
Acknowledgements
The present work was carried out in the Nano and Molecular Systems Research
Unit (NANOMO), University of Oulu. I would like to thank my supervisor Prof.
Matti Alatalo, who saw in me a possible Ph.D. student, guided all the way, taught
me the Finnish culture and became a friend. Special thanks to Prof. Marko Huttula
for being a good boss and providing me with the chance to finish my doctoral study.
This work involved many close collaborations and I would like to acknowledge
Dr. Wei Cao and Xinying Shi for not giving up on our project and not just finish it,
but hugely improve it, which led to the publication in a good journal. I thank Prof.
Talat Rahman for providing an opportunity for my research visit and for an amazing
time in Florida, Dr. Duy Le for productive collaboration. Also, big thanks go to
Lyman Baker, before knowing him I was sure that combining beer and own library
is impossible.
I want to express my gratitude to my follow-up group members: Prof. Erkki
Thuneberg, Prof. Juha Vaara, Doc. Saana-Maija Huttula for their guidance and
understanding.
I am grateful to NANOMO for the friendly and productive environment.
Particularly to Jack Lin, Nønne Prisle and especially to Georgia Michailoudi for
being a good friend and for tasty cakes.
I especially thank my beloved Olga Miroshnichenko, who brought me to Oulu
and helped me during the whole Ph.D. Thanks to Irina Ignatova and to Roman
Frolov who taught me how to stand up for oneself and that one can have both a
career and a lot of children. Anton Artamonov thank you for being Anton. A special
thanks to the Poluyanov family, for the numerous evenings filled with joy
The last thanks are for my family. Big thank to my Mother, she is the main
reason why I went to university, even though it was difficult for her to let 16 years
old son go to another city to study. All my journey wouldn’t be possible without
Father’s support. Sister, who inspired me, and whom I always want to catch up but
have not succeeded yet…Thanks to my son for all the joy he brought. He taught
me that four hours of uninterrupted sleep is one of the best things in this world.
Oulu, December 2018 Sergei Posysaev
2
Abbreviations
API application programming interface
DFT density functional theory
etc. et cetera
i.e. id est
ILC inorganic layered crystals
e.g. exempli gratia
GIGO garbage in, garbage out
GGA generalized gradient approximation
HDS hydrodesulfurization
LDA Local Density Approximation
OS oxidation state
PBC periodic boundary conditions
REST–API The REpresentational State Transfer–Application Programming
Interface
TMD transition metal dichalcogenides
XRD X-ray diffraction
3
Original publications
This thesis is based on the following publications, which are referred throughout
the text by their Roman numerals:
I Shi, X., Posysaev, S., Huttula, M., Pankratov, V., Hoszowska, J., Dousse, J.-Cl.,
Zeeshan, F., Niu, Y., Zakharov, A., Li, T., Miroshnichenko, O., Zhang, M.,
Wang, X., Huang, Z., Saukko, S., González, D. L., Dijken, S., Alatalo, M., Cao.
W., Metallic contact between MoS2 and Ni via Au nanoglue. Small, 14,
1704526 (2018). DOI: 10.1002/smll.201704526
II Posysaev, S., Alatalo, M., Surface morphology and sulfur reduction pathways
of MoS2 Mo-edges of monolayer, (100) and (103) surfaces by molecular
hydrogen: A DFT study. ACS Omega, submitted for publication
III Posysaev, S., Miroshnichenko, O., Alatalo, M., Le, D., Rahman, T., Oxidation
states of binary oxides from data analytics of the electronic structure.
Computational Materials Science, submitted for publication
All papers in the thesis are a result of teamwork. The author has been the main
writer of the Papers II and III. Theoretical calculations and data management were
done by the author for Papers I and II and partially for Paper III.
4
Papers not included in the thesis
The author contributed to the publications which are not included in the thesis
publication
I Miroshnichenko, O., Posysaev, S., Alatalo, M., A DFT study of the effect of SO4
groups on the properties of TiO2 nanoparticles. Phys. Chem. Chem. Phys.,
2016, 18, 33068-33076. DOI: 10.1039/C6CP05681D
II X. Shi, M. Huttula, V. Pankratov, J. Hoszowska, J.-Cl. Dousse, F. Zeeshan, Y.
Niu, A. Zakharov, Z. Huang, G. Wang, S. Posysaev, O. Miroshnichenko, M.
Alatalo, W. Cao. Quantification of bonded Ni atoms for Ni-MoS2 metallic
contact through X-ray photoemission electron microscopy. Microscopy and
Microanalysis, 24, 458−459 (2018). DOI: 10.1017/S1431927618014526
5
Contents
Abstract
Tiivistelmä
Acknowledgements 1
Abbreviations 2
Original publications 3
Papers not included in the thesis 4
Contents 5
1 Introduction 7
2 Quantum mechanics 9
2.1 Postulates of quantum mechanics ................................................................ 9
2.1.1 The wave function............................................................................ 10
2.1.2 Observables ...................................................................................... 10
2.1.3 Measurements .................................................................................. 11
2.1.4 Expectation Values ........................................................................... 11
2.2 Schrödinger Equation ................................................................................. 11
3 Computational methods 13
3.1 Density functional theory ........................................................................... 13
3.1.1 Born–Oppenheimer approximation ................................................ 13
3.1.2 Hohenberg–Kohn theorems ............................................................. 14
3.1.3 Kohn-Sham equations ...................................................................... 14
3.1.4 Exchange-correlation functional ..................................................... 15
3.1.5 Computational details of DFT ......................................................... 17
3.2 Application of DFT ..................................................................................... 18
3.2.1 Surface energy .................................................................................. 18
3.2.2 Calculation of reaction pathways .................................................... 20
3.2.3 Bader Charges .................................................................................. 21
3.3 Data analysis ................................................................................................ 22
3.3.1 Materials databases .......................................................................... 22
3.3.2 Data collection ................................................................................. 22
3.3.3 Data processing ................................................................................ 23
3.3.4 Data cleaning and data exploring .................................................... 23
4 Summary of articles 24
4.1 Metallic Contact between MoS2 and Ni via Au Nanoglue ....................... 24
6
4.2 Surface morphology and sulfur reduction pathways of MoS2 Mo-
edges of monolayer, (100) and (103) surfaces by molecular
hydrogen ....................................................................................................... 27
4.3 Oxidation states of binary oxides from data analytics of the
electronic structure ...................................................................................... 29
5 Conclusions 32
List of references 33
Original publications 40
7
1 Introduction
One of the greatest things a human being can do is to ask ‘Why is this so?’. Greater
yet is to search for the answer, by studying nature. A lot of time has passed and a
lot of ‘why’s’ have found answers. Thus, a scheme has been formed for justifying
new answers consisting of the following steps: observe a phenomenon, form a
hypothesis describing the phenomenon, conduct a controllable experiment to verify
the hypothesis.
To conduct an experiment is not a trivial task. Many aspects can affect the
results of experiments and one must ensure that all variables are correctly
controlled.
Nowadays with modern techniques and increasing computer power, сomputers
have enabled us to conduct experiments by computer simulation, which saves both
time and money. We can model processes from interacting quantum particles [1]
up to the formation of galaxies [2]. Even though one can in principle control
everything in modeling, it is possible to forget/skip something important and end
up with a model that will not reflect reality.
Today many computational models are available for the calculations of
material properties, on different scales of the phenomena studied. When one wants
to know the behavior of a material at the scale of electrons, Density Functional
Theory (DFT) is one of the methods that he/she should use.
Since its formulation in the 1960’s and up to nowadays of DFT has proven to
be a reliable tool [3,4]. It has been applied to chemistry, physics, and biology [5].
The importance of DFT was evidenced by the Nobel prize awarded to one of the
creators of DFT, Walter Kohn [6].
DFT is widely applied in physics for calculating properties of materials. MoS2
is one of the transition metal dichalcogenides (TMDs), which falls in the category
of inorganic layered crystals (ILC). It is a cheap material and abundant on Earth.
MoS2 exhibits semiconducting band structure and it has also been proposed as a
promising alternative to traditional silicon-based semiconductors [7,8]. However,
the lack of dangling bonds in pristine ILC material surface makes it difficult to
bond with metals, resulting in contact instability and Schottky barriers [9,10] with
high contact resistance, which restrains the performance of TMDs based devices.
One of the aims of this thesis was to prove theoretically the existence of an effective
metallic contact between MoS2 and Ni nanoparticles.
The edge planes of MoS2 are used as a catalyst for hydrodesulfurization (HDS)
reaction during which the emission of sulfur dioxide is reduced. Sulfur dioxide can
8
provoke acid rains and affects human health [11]. This is why the demand for
cleaner fuel rises each year within stricter environmental regulations [12].
Extensive studies have been performed in order to improve the processes of fuel
purification such as doping [13], novel supports [14,15], and changing the
morphology of the catalyst [16,17]. However, desired properties have not yet been
achieved. A way of possible improvement of the catalytic activity of MoS2 was the
second aim of this thesis. We studied reaction pathways of one of the surfaces which
most frequently occurs in X-ray diffraction (XRD).
Currently, the volume of DFT-related publications doubles every 5–6 years
[18]. This growth has given rise to continuously growing open materials databases
of millions of compounds. Such databases have been used for merging together
physics and machine learning techniques to show a unique possibility to predict
materials properties by intelligent algorithms trained at a large number of known
data [19] as well as for the inverse task - to find a material with desired properties
[20].
Oxidation state (OS) indicates the number of electrons an atom has lost or
gained. Properties of each material depend on OS of its atoms. DFT has not been
designed to calculate OS directly, although DFT data is routinely used for the
extraction of atomic charges [21–24]. We used materials databases to resolve the
long-standing question about the applicability of partial charges to the
determination of the OS. Empirical rules for establishing the OS in binary oxides
are discussed later in this work, based on a large number of data.
This thesis is structured as follows: Chapter 2 introduces the basics of quantum
mechanics. In Chapter 3 the foundations and applications of density functional
theory are discussed along with data analysis and materials databases. In a summary
of articles, Chapter 4, the main findings are included along with additional material
which was not included in the Papers. The last chapter consists of conclusions
based on the original publications.
9
2 Quantum mechanics
This chapter gives a brief overview of quantum mechanics. Before explaining the
theory, which describes the nature of subatomic particles and energies at the
smallest scale and to understand that quantum mechanics is not pure magic, it is
important to provide an example of quantization, as many people find it puzzling
that some physical quantities can have only discrete values.
Quantization appears only when some restriction is present in a system. An
analogous example in classical mechanics would be the potential energy of a
human in a skyscraper. The human can only be on a certain floor with a certain
potential energy, analogous to electron levels in an atom (Fig. 2.1), whereas a
falling human experiences a continuous range of potential energy. The same goes
for a free particle, it can have a continuous range of energies.
2.1 Postulates of quantum mechanics
Postulates of quantum mechanics were built taking classical mechanics as a starting
point and do not have a “real” justification. Even though for a scientist, a theory
which should be accepted by postulates usually sows a grain of doubt, quantum
mechanics has been with us about 100 years and is “the most precisely tested and
most successful theory in the history of science” [25].
Fig. 2.1 Example of discrete and continuous range of energy in quantum and classical
mechanics.
10
2.1.1 The wave function
The first postulate says that each state of the system is fully defined by a wave
function, 𝛹(𝑟, 𝑡) . The probability of finding a particle (or particles in a many
particle system) in a certain volume 𝑑3𝑟 at time 𝑡 is given by the square of the wave
function, i.e. taking the complex conjugate of the wave function multiplied by the
wave function itself and then by the volume element:
𝑃(𝑟, 𝑡) = 𝛹∗(𝑟, 𝑡)𝛹(𝑟, 𝑡)𝑑3𝑟. (2.1)
For example, if an electron is present in an infinite vacuum, then we need to
integrate the right-hand side of Eq. (2.1) over infinity to get one as our answer.
2.1.2 Observables
The second postulate says that for each observable quantity there is a linear, self-
adjoint operator.
In classical mechanics, one can define observables as quantities that can be
measured in an experiment. Given the trajectory of the system, x(t), or the phase
space trajectory, p(x), we can determine all observables of the system. For instance,
we could find the momentum, the energy, the angular momentum, etc.
Table 1 Physical observables and the corresponding quantum operators (denoted by
hat).
In quantum mechanics, the observables x(t), p, etc. are replaced with operators.
Because the state of the system is uniquely determined by a wave function instead
of the trajectory, the operators act on the wave function. The position operator acts
Classical Quantum
x(t) �̂�𝛹(�⃗⃗�, 𝑡) = 𝑥𝛹(�⃗⃗�, 𝑡)
p(t) �̂�𝛹(�⃗⃗�, 𝑡) = −𝑖ℎ𝑑
𝑑𝑥𝛹(�⃗⃗�, 𝑡)
E �̂�𝛹(�⃗⃗�, 𝑡) = [�̂�2
2𝑚+ �̂�(𝑥)] 𝛹(�⃗⃗�, 𝑡)
11
on the wave function by simple multiplication. The momentum operator acts
through differentiation etc. (Table 1).
2.1.3 Measurements
The third postulate concerns observables. It says that in any measurement, the only
possible values one can obtain are the eigenvalues 𝜆 of the operator 𝛰 , that
correspond to the observable to be measured. The operator and its eigenvalue
satisfy the eigenvalue equation:
𝛰𝛹(𝑟, 𝑡) = 𝜆𝛹(𝑟, 𝑡). (2.2)
An important point is that after the measurement, the wavefunction 𝛹(𝑟, 𝑡)
immediately collapses into the corresponding eigenstate 𝜆. Thus, the measurement
affects the state of the system.
2.1.4 Expectation Values
The fourth postulate says that if the system is characterized by a state 𝛹 , the
average of the measurements of the observable 𝛰 at time 𝑡 is given by the
expectation value of 𝛰:
⟨𝛰⟩𝑡 = ∫ 𝑑3𝑟𝐷
𝛹∗(𝑟, 𝑡)𝛰𝛹(𝑟, 𝑡). (2.3)
Expectation values are important in an experiment because an experiment is
not measuring just one copy of the system. For example, when measuring the
spectrum of a water molecule, one is measuring the spectrum of many molecules
in the sample.
2.2 Schrödinger Equation
Schrödinger equation summarizes all postulates together. The state and the time
evolution of the system can be obtained by solving the time-dependent Schrödinger
equation:
𝑖ħ𝜕
𝜕𝑡𝛹(𝑟, 𝑡) = �̂�𝛹(𝑟, 𝑡), (2.4)
12
where ħ is Planck's constant divided by 2𝜋 and �̂� is the Hamiltonian operator,
which corresponds to the total energy of the system.
The significance of the Schrödinger equation in quantum mechanics is the
same as that of Newton’s second law in classical mechanics. They are both
equations of motion. The goal here is to find 𝛹(𝑟, 𝑡) knowing an initial state
𝛹(𝑟, 0) and the solution to the time dependent Schrödinger equation describes the
dynamical behavior of the particle. In contrast to Newton’s equation, where we can
determine the exact position of the particle as a function of time, we get a wave
function, which tells us the probability of finding the particle in some region in
space.
13
3 Computational methods
This chapter gives an overview of the computational methods used in this thesis.
The Schrödinger equation can be solved analytically only for simple systems. For
resolving the many-body problem, approximations should be made. They are
discussed further along with the application of DFT. The use of materials databases
and principles of data analysis end this chapter.
3.1 Density functional theory
DFT is an ab initio method for solving the Schrödinger equation, where the
properties of the many-electron system are fully defined by the electron density that
depends only on 3 spatial coordinates.
Electrons are present as a set of n one-electron Kohn-Sham (Schrödinger like)
equations, which are evaluated in a potential (external + effective, see section 3.1.4).
Before going to the basics of DFT, first, the term “functional” needs to be defined.
A functional is related to the concept of a function, as one can guess based on the
name. A function takes a value of a variable (variables) and defines a single number
from those variables. A simple example of a function is 𝑓(𝑥) = ax + b . A
functional, denoted by 𝐹, is similar, for example 𝐹(𝑓) = ∫ 𝑓(𝑥)𝑑𝑥, but it takes a
function instead of the value of a variable and yields a single number from the
function.
3.1.1 Born–Oppenheimer approximation
The Born-Oppenheimer approximation [26] is the assumption that the nuclear
motion and the electronic motion can be separated. It leads to a molecular wave
function in terms of the electron coordinates 𝑟𝑖 and nuclear coordinates �⃗⃗�𝑗.
𝛹𝑚𝑜𝑙𝑒𝑐𝑢𝑙𝑒 (𝑟𝑖 , �⃗⃗�𝑗) = 𝛹𝑒𝑙𝑒𝑐𝑡𝑟𝑜𝑛𝑠(𝑟𝑖 , �⃗⃗�𝑗)𝛹𝑛𝑢𝑐𝑙𝑒𝑖(�⃗⃗�𝑗). (3.1)
Further assumptions are: 1) Nuclei are present as fixed points, creating a static
external potential V. 2) Electrons respond much faster to changes in their
surroundings than nuclei due to the greater mass of nucleus which is at least 1800
times heavier than the electrons. Hence we can assume that electrons instantly
follow the motion of the nuclei. 3) The nuclear motion sees a smeared out potential
arising from the electrons.
14
3.1.2 Hohenberg–Kohn theorems
The roots of DFT stem from two theorems proved by Hohenberg and Kohn [27].
The first theorem states: ‘The ground-state energy from Schrödinger’s equation is
a unique functional of the electron density’. This means that there is an exact
dependence between the ground-state wave function and the electron density. It
allows us to reduce complexity from 3N dimensions to just 3 dimensions (Fig. 3.1).
The first Hohenberg–Kohn theorem does not yield information on how to find
the functional. The second theorem states: ’ The electron density that minimizes the
energy of the overall functional is the true electron density corresponding to the
full solution of the Schrödinger equation’. Taking an assumption that the functional
is known, one needs to vary the electron density until the energy from the functional
is minimized.
3.1.3 Kohn-Sham equations
Solving the Schrödinger equation using just the Hohenberg–Kohn theorems is
still an intractable many-body problem of interacting electrons in a static external
potential. Kohn and Sham reduced this problem to a tractable problem of
noninteracting electrons moving in an effective potential. This led to a set of self-
consistent, single particle equations known as the Kohn-Sham equations [28]:
Fig. 3.1 Schematic illustration of transforming many-electron system to electron density
15
(−1
2𝛻2 + 𝜗𝑒𝑓𝑓(𝑟) − 𝜖𝑗) 𝜑𝑗(𝑟) = 0, (3.2)
where 𝜗𝑒𝑓𝑓 is the sum of the external, Hartree and exchange-correlation potentials,
𝜑𝑗(𝑟) is the jth Kohn-Sham orbital of energy 𝜖𝑗 . The solutions of the Kohn–Sham
equations are single-electron wave functions that depend on three spatial
coordinates. Thus, to find the effective potential, we need to define the electron
density, which is given by:
𝑛(𝑟) = ∑|𝜑𝑗(𝑟)|2
𝑁
𝑗=1
. (3.3)
The following self-consistent cycle should be used to find the ground state
energy: an initial guess for 𝜑𝑗(𝑟) which is applied to Eq. (3.3) and later on used for
finding the effective potential, which is subsequently used in Eq. (2.4). Then Kohn-
Sham equation is solved for 𝜑𝑗(𝑟) iteratively until the initial and new densities are
the same, then the ground state density has been found. Otherwise, one must select
a new trial density through minimization of the total energy and continue to repeat
the iterative procedure.
3.1.4 Exchange-correlation functional
The Hohenberg-Kohn theorems refer to a functional, which therefore needs to
be defined more accurately. As discussed briefly above, a functional is a function
of another function; in the case of DFT, it is the spatially dependent electron density.
Kohn and Sham separated the effective potential into three terms:
𝜗𝑒𝑓𝑓(𝑟) = 𝜗(𝑟) + ∫𝑛(𝑟′)
|𝑟 − 𝑟′|𝑑𝑟′ + 𝜗𝑥𝑐(𝑟), (3.4)
where the first and second terms are the external and classical electrostatic
(Hartree) potentials, which can be found exactly. The third term is a definition of
the exchange-correlation potential which contains the non-classical electrostatic
interaction energy and the difference between the kinetic energies of the interacting
and non-interacting systems.
Results of DFT calculations depend on the choice of the exchange-correlation
functional. The simplest approximation is the Local Density Approximation (LDA)
[28], where the exchange energy and correlation energy are based on uniform
16
electron gas. In this situation, the electron density is constant at all points in space;
𝑛(𝑟) = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡.
LDA is not the only functional, there are several others. One of the most
popular classes of functionals uses information about the local gradient in the
electron density. This approach defines the generalized gradient approximation
(GGA), which include information from the gradient of the electron density [29–
32]. More accurate calculations became possible with including second derivatives
(meta-GGA functionals) [33–35] of the electron density. Popularity and accuracy
of DFT increased significantly after the development of hybrid functionals [36–38].
Range-separation allowed application of hybrid functionals to solids and surfaces
[39–43]. Accuracy was improved further by local hybrids [44–46], and even more
after coupling it with range-separation [47,48].
Some functionals give good results for one system and some for another, so
when reporting results of DFT calculation it is important to indicate which
functional has been used. More accurate functionals usually consume more
computer resources and nowadays it is more a trade-off between accuracy and the
speed of calculation. Perdew and co-workers [49] provide a classification of
functionals using a reference to the Biblical account of Jacob’s ladder (Fig. 3.2).
Scientists get more and more physical information, climbing by ladder and in the
top, the Schrödinger equation is solved without approximations.
Fig. 3.2 Illustration of Perdew’s classification of DFT functionals using “Jacob’s ladder”.
17
3.1.5 Computational details of DFT
There are many important parameters involved in a DFT calculation such as:
whether to perform spin-polarized or non-spin-polarized calculations, the use of
symmetry, the choice of eigensolver, structural optimization algorithms etc. These
parameters are sometimes a matter of taste, sometimes the choice depends on the
system studied. The following concepts are the most crucial ones and they need to
be discussed before diving in the applications of DFT: unit cell, periodic boundary
conditions, Brillouin zone [50] and methods for expanding the wave functions. This
is a minimum for understanding the principles of DFT calculations.
A unit cell is the most basic and least volume consuming repeating structure of
any solid. The lattice structure is created if the unit cell is repeated in all three
directions. To avoid using large systems in computations, save computational time
and prevent molecules/atoms from flying off into space, periodic boundary
conditions (PBC) are normally employed (Fig. 3.3). For example, if a molecule
leaves the red cell through the right wall, its image will enter the cell through the
left wall.
The concept of the Brillouin zone is of major importance in the band theory of
materials. The Brillouin zone is the unit cell of the reciprocal lattice. We will not
go into the details of this construction here but simply state that many parts of the
mathematical problems posed by DFT are much more convenient to solve in terms
of reciprocal space and the accuracy of a calculation depends on the number of k
points in each direction. One of the most popular solutions to split the reciprocal
space was suggested by Monkhorst and Pack [51]. Together with the number of k
Fig. 3.3 Two-dimensional schematic of periodic boundary conditions. The unit cell of
system is in the lower left corner (red).
18
points, the accuracy of a calculation depends on the kinetic energy cutoff, which is
proportional to the number of plane waves being utilized as basis functions to
represent the wavefunction.
It is worth mentioning that nowadays there are many computer programs for
atomic scale materials modelling from first principles. Examples of well-known
packages include ABINIT [52], GPAW [53], Quantum ESPRESSO [54], and VASP
[55,56].
3.2 Application of DFT
DFT is used in physics, chemistry, materials science, and biology. Due to the
increase in computer power, DFT constantly finds new applications. Results can be
obtained for a wide range of physical quantities, starting from the simple lattice
constant of a BCC metal up to the prediction of complex system behavior.
Here we concentrate on the techniques, which have been used in the original
publications.
3.2.1 Surface energy
Before discussing the surface energy, one needs to explain how to model surfaces.
The best model for a surface would be a slice of a material which is infinite in two
dimensions, which can be done using PBC. The common idea for modeling
surfaces is presented in Fig. 3.4, where four-unit cells are used along the y direction
and a vacuum layer is added on top of the five-layer slab. It is important to ensure
that the vacuum layer is thick enough to avoid interaction between the bottom and
the top layer of the slab in the z direction. Molecular surface coverage, percentages
of defects/adatoms on the surface and other structural features can be controlled by
varying the number of unit cells in the x and y direction. A number of layers in the
z direction usually depend on the material and the properties studied. A thicker slab
gives more accurate results but consumes more computational power; in other
words, there is usually a trade-off between accuracy and computational time.
19
Surface energy is the energy needed to create the surface by a cleavage along a
plane in a bulk material. This is a difficult quantity to determine experimentally
because it usually requires measuring the surface tension at the melting temperature
of the metal [57]. On the other hand, a theoretical determination is relatively easy.
The surface energy (𝐸𝑠) is calculated as
𝐸𝑠(𝑛) =1
2𝐴(𝐸𝑛 − 𝑛𝐸𝑏), (3.5)
where 𝑛 is the number of unit cells, 𝐴 is the surface area, 𝐸𝑛 is the total energy of
the relaxed surface slab and 𝐸𝑏 is the total energy of a bulk unit cell. Also, the
cleavage energy can be calculated using the same equation; in that case, 𝐸𝑛 is the
total energy of the unrelaxed slab.
One can cleavage different surfaces from the same bulk material and the surface
with the lowest surface energy is the most stable surface. It means that this surface
is the most likely one to be created under mechanical impact and the one which
most frequently occurs in nature.
Surface energy calculated by Eq. (3.5) does not properly converge with
increasing number of layers, due to numerical errors arising from differences in the
Brillouin-zone sampling [58]. There are other techniques for calculating the surface
energy which avoid these errors [58–61]. Usually, however, the surface energy
calculated by Eq. (3.4) gives an acceptable result if care is used when choosing the
k-point mesh [60,62].
Fig. 3.4 Surface slab model.
20
3.2.2 Calculation of reaction pathways
Knowledge about reaction steps and reaction barriers is crucial for the commercial
application of catalysts. Modeling of reaction pathways helps to increase the rates
of catalytical reaction and adjust conditions which leads to a cheaper product for
customers.
Before starting to model a chemical reaction, it is important to find the proper
initial and final states of the reaction with a minimum energy of the system studied.
Chemical reactions usually take place on the surface of the catalyst. Due to the
complexity of the surface structure, finding the initial and final states is not a trivial
task. If we take even a simple adsorption of the CO molecule, there are several
adsorption sites on the ideal surface of any solid (Fig. 3.5). A number of possible
adsorption sites significantly increases on real surfaces due to adatoms, steps, and
vacancies. Picking a wrong site results in a higher barrier for the reaction.
Once the initial and final states are known, the reaction/adsorption barrier can
be found. There are several methods for finding the barrier.
The nudged elastic band (NEB) method [63] uses several images between the
initial and final state during the search for the minimum reaction pathway. It
optimizes the intermediate images by adding spring forces along the band between
the images and by projecting the force perpendicular to the band. A variation of the
NEB method is the climbing image NEB (cNEB) [64], where the highest energy
image can be adjusted to the top of the saddle point (Fig. 3.6). The dimer method
[65] is another method for finding saddle points. It allows starting from any initial
21
configuration near a saddle point. The dimer method needs only one image and
searches for the saddle points in random directions. Sometimes switching from a
pre-converged NEB calculation to the dimer method helps to save computational
time and provide a possibility to use finer parameters in the calculation due to the
usage of only one image.
3.2.3 Bader Charges
There are several techniques for the estimation of the total electronic charge of an
atom [21,22,24,66]. Bader analysis [21] is one of them. It utilizes electron density
for estimating the net charge on an atom. A volume of a cell is separated in Bader
volumes when:
∇𝜌(𝑟) = 0, (3.6)
where 𝜌(𝑟) is the electron density, and the gradient is taken with respect to the
three-dimensional coordinate 𝑟. This zero flux condition effectively yields distinct
regions (Bader volumes) for each atom in a system and the integral of the density
inside them is assigned as a net charge of an atom.
22
3.3 Data analysis
Nowadays, the amount of data is rapidly growing. This growth helps to find new
patterns in all areas. In our information age, data analysis of multidimensional data
has become an essential skill for researchers in all areas. Data analysis includes
processes such as cleaning, mining, retrieving, modeling, transforming of data with
the purpose of finding useful information, patterns and to support decision-making.
3.3.1 Materials databases
Quantum mechanical calculations consume a lot of computational power of modern
computers. Not all researchers have the possibility to repeat a calculation from a
scientific article. Providing the results of your calculation to an open database
allows everyone to see full results of the calculation.
Materials databases such as AFLOW [67,68], Materials Project [69],
Engineering Virtual Organization for Cyber Design [70], and NOMAD [71]
provide easy data mining. AFLOW requires strict parameters for calculation [72],
which simplifies the comparison between the results of different calculations.
3.3.2 Data collection
Most of the databases allow a user to download data. With the data available, there
are two different ways to proceed. The first one is to find and download the
compounds needed manually through the web interface of a database. It is an easy
and straightforward way to proceed if one needs only a few compounds. The second
way is to use an application programming interface (API), which enables users to
search and download desired data for post-processing, through appropriate queries.
AFLOW provides The REpresentational State Transfer–Application Programming
Interface (REST–API) [73] For example, using REST–API a query to get all links
for titanium oxides in the database looks like this:
“http://aflowlib.duke.edu/search/API/?species('Ti', 'O'),$nspecies(2)”. It is a simple
query where you indicate the species and number of species in a compound. The
filter can be refined further by various values such as band gap, formation energy
etc.
23
3.3.3 Data processing
Downloading data from multiple sources results in files in different formats, which
contain different quantities. One needs to prepare all the collected data in one
format with the same features. In Paper III, for faster processing and analyzing of
data, a database of compounds was transformed to a database of cations, meaning
that each record in the new data set represents an atom rather than a compound (Fig.
3.7).
3.3.4 Data cleaning and data exploring
The open databases can include duplicates, errors, and missing or unnecessary
information. Before analyzing data, one needs to clean it. Cleaning is one of the
crucial components of data science. It is supported by the famous statement
‘garbage in, garbage out (GIGO)’. GIGO means that nonsense input data produces
nonsense output.
During data exploring one uses visual exploration to understand what is in a
dataset and the characteristics of the data. Usually, cleaning goes together with data
exploring. Examples of data cleaning and exploring are presented in the summary
of Paper III, where the dataset used in Paper III is used as an example dataset.
24
4 Summary of articles
4.1 Metallic Contact between MoS2 and Ni via Au Nanoglue
It was noticed that the edge plane of MoS2 is more chemically active than basal
plane [74]. A majority of computational studies only focused on the contact
between metal and monolayer TMDs [75], where resistance is high owing to the
larger physical separation between contacts. A comprehensive computational study
of metal-mTMD contacts [76] has demonstrated the high advantage of edge
contacts in terms of electron injection efficiency.
Our aim was to prove conduction between the edge surface of MoS2 and a
metallic contact. The experimental results showed two types of interfaces. In type-
I, the MoS2 and Ni are connected via Au nanoparticles, which features bulk-like
properties. In contrast, the type-II interface is nanocrystallized Au nanoparticles,
with sizes typically below 5 nm. Both types of interfaces were modeled. Each one
contains a 3×1×4 unit cell of the MoS2 host. To reach a reasonable computational
time for model-I (type-I contact), the nanocrystallized Au nanoglue was simplified
to 4 layers of Au atoms between the Ni and MoS2 sides. Two contact layers are
relaxed for structural optimization at each side, and another two fixed as Au (111).
The second model (type-II) employs the interfacial character where the Au and Ni
atoms are randomly placed to reach possible alloying. The Ni (111) surface is
connected to the interface, opposite to the MoS2.
In the present calculations, DFT structural optimization for the edge surface
slab MoS2 (Fig. 4.1a) has been carried out. Two views of the unit cell of bulk MoS2
are shown in Fig. 4.1b, c. First-principles calculations were performed, using the
projector augmented wave method as implemented in Vienna Ab initio Simulation
Package (VASP) [55,56]. The plane wave cut-off energy was set to 400 eV. The
Perdew-Burke-Ernzerhof GGA functional was used to treat the electron exchange
and correlations with van der Waals (DFT-D3) correction [32,77]. The convergence
criterion was chosen so that the maximum force on one atom is smaller than 0.02
eV/Å. The calculations were performed with periodic boundary conditions. The
vacuum region in the z-direction was set to 15 Å to avoid mirror interactions
between nearby images. During relaxation all atoms were allowed to relax except:
in model-I, two middle layers of gold, which represent the bulk of gold; and in
model-II, two bottom layers of MoS2, which represent the bulk of MoS2.
25
The Brillouin zone was sampled using a Monkhorst-Pack k-point set [51]
containing 4×1×2 k-points during the structural optimization and PDOS calculation.
30 k-points were used along with special k-point path in the Brillouin zone for band
structure calculation. To obtaining magnetic properties spin-polarized calculations
were carried out. The VESTA3 code [78] was used to plot 3D isosurfaces for the
𝛥𝜌 function. The calculations are consistent with those using other software
packages [79,80], as well as results crosschecked with experimental determinations
[81].
The partial density of states (PDOS) of Mo and S atoms subjected to the Au
adsorption is shown in Fig. 4.2a. After adsorbing gold atoms, structural relaxation
is small at the host side. The guest gold buffer tends to be adsorbed by the ideal
MoS2 edge, forming a contact with the host. The charge density difference clarifies
the charge transfer between the gold layer and the MoS2 side, as shown in Fig. 4.2b.
The non-zero PDOS (gray line in Fig. 4.2a) of the host disperses crossing the Fermi
level, denoting a total metallization via side contact after the introduction of the
gold slab. This is different from the bare MoS2 side whose PDOS (red line in Fig.
4.2a) suffers from discontinuity around the Fermi-level, as pointed out by the green
regions in the figure.
Fig. 4.1 Structure of MoS2. (a) Representation of edge (blue) and top (red) MoS2 surfaces,
(b) and (c) unit cell of bulk MoS2. The pink and yellow spheres represent the Mo and S
atoms respectively.
26
Separate spin-polarized calculations were performed for bulk Ni and with
different percentage of Ni atoms substitutions (0, 8, 16%) for Au-Ni and MoS2-Ni
interfaces. Our theoretical result for the magnetic moment of bulk nickel is 0.67 µB,
in good agreement with experimental results. Reduction of the magnetic moment
of nickel was observed on atoms in contact with Au (111) down to 0.44 µB and with
MoS2 (100) surface down to 0.2 µB. Small magnetic moments were found on Mo,
S and Au atoms as well.
Introduction of the Au-Ni alloy onto the MoS2 side offers new electron
migration channels between the metal and the semiconductor side, leading to a
metallic contact mechanism. Such TMDs possess unique properties with numerous
promising applications in the semiconductor industry.
27
4.2 Surface morphology and sulfur reduction pathways of MoS2
Mo-edges of monolayer, (100) and (103) surfaces by molecular
hydrogen
In addition to the above-mentioned applications, molybdenum disulfide is a well-
known catalyst applied in fuel purification. Various methods for improving the
catalytic activity of MoS2 have been tried [13–17]. However, chemically activity
of the (103) edge surface of MoS2, which has the highest XRD peak in several
experimental works [82–85], has not been studied.
Our aim was to shed light on the catalytic properties of the (103) edge surface
of MoS2 and discover reaction pathways. Another motivation was to find the reason
for the occurrence of the (103) surface of MoS2 in XRD patterns.
Ab initio DFT calculations were performed in spin-polarized mode, with VASP
[55,56], using the PBE functional to treat the electron exchange and correlations,
together with van der Waals (DFT-D3) correction [32,77]. The electronic density
was calculated with a 350 eV cut-off energy. The convergence criterion was chosen
so that the maximum force on one atom is smaller than 0.02 eV/Å. The calculations
were performed with periodic boundary conditions. A Monkhorst-Pack k-point set
containing 2×2×1 k-points was used to sample the Brillouin zone during the
structural optimization [51].
Transition state (TS) search was performed using two methods in the following
order: first, climbing image nudged elastic band [64] and then the dimer method
[65]. To ensure that the system is in the TS, for each TS the dynamical matrix was
calculated and verified that it contains exactly one imaginary frequency.
According to our findings, the (100) surface is less energetically favorable than
the (103) surface, which explains the higher peak of the (103) surface in the XRD
patterns, but contradicts with Refs. [86–88], where the (100) peak is higher.
Apparently, the morphology of the samples depends on the preparation route of
MoS2.
The reaction pathway for the formation of sulfur vacancy consists of the same
steps as proposed by Prodhomme et al. [89] and presented in Fig. 4.3. We described
the pathway for the (100) monolayer, (100) surface, and (103) surface. According
to our results the barriers among the three models are similar.
28
To summarize, we demonstrated that the reason for the occurrence of the (103)
surface of MoS2 in XRD patterns is its low surface energy. The difference in the
barriers between the (100) monolayer, (100) surface, and (103) surface is less than
0.07eV. It allows us to conclude that the monolayer and bulk MoS2 possess similar
catalytic properties in the HDS reaction. In addition, a new reaction pathway for
the formation of sulfur a vacancy was found.
Fig. 4.3 Reaction pathway for the formation of a sulfur vacancy. Black, green and blue
paths relate to (100) monolayer, (100) surface, and (103) surface, respectively. Paths are
shifted vertically for better readability. The energies of the S1 state are taken as
references for each reaction path. Mo atoms in turquoise, S atoms in yellow, and H
atoms are in green and in red (the two hydrogen atoms are depicted with different colors
for clarity). Geometrical structures are depicted for the (100) monolayer surface.
29
4.3 Oxidation states of binary oxides from data analytics of the
electronic structure
Several studies have tried to establish a relationship between the OS and Bader
charge, and it has proven to be a nontrivial task. According to Reeves and Kanai,
the charge difference between two consecutive OS (between Ru(II) and Ru(III))
[90] is rather marginal and it is much smaller than unity. It is natural to look for a
method which gives an integer number for OS, but is it true in nature? Can we say
that all 8 valence electrons of osmium in OsO4 go entirely to oxygen atoms? It is
well known that all ionic bonds have some features of a covalent bond. Jansen and
Wedig [91] are confident that a connection between OS and Bader charge exists
and they highlight “the need for factors other than calculated charges to be taken
into consideration in order to obtain a meaningful connection with oxidation states”.
To our knowledge, all previous work tried to connect OS with Bader charges only
in a few compounds, and the main topic in those articles was different.
Nowadays the popularity of open and big data gives rise to continuously
growing open materials databases with DFT results for millions of compounds. Our
aim was to find a dependency between OS and Bader charge, using compounds
from an open database, and to provide a possibility to determine the oxidation state
from ordinary DFT calculations. Our focus was on binary oxides, i.e. compounds
of two elements, one of which is oxygen. They were chosen for simplicity as they
contain only two species. They are among the most calculated compound types in
materials databases. We show that a correlation indeed exists between OSs and
Bader charges and discuss the applicability of determining OS by Bader charges in
surfaces and in mixed-valence compounds (MV), where an element in the
compound can be in several OSs. In addition, the applicability of results with other
exchange-correlation functionals was examined.
The data set was downloaded from the AFLOW database [67,68], a materials
properties repository which has a standard for ab initio calculations [72]. To
classify materials by their OS we need to formulate an accurate definition of the
cation’s OS. By stating that the sum of the OSs in a compound has to be neutral,
that oxygen has an OS of -2, and that the OS of the anion is a known quantity, we
calculate the OS per compound and the OS of the cation as follows:
OS per compound = −OS of anion ∗ number of anions, (4.1)
30
OS of cation =OS per compound
number of cations. (4.2)
Databases can include duplicates, errors etc., as was mentioned in chapter 3.3.4.
Also, a result for OS of cation obtained from Eq. (4.2) can be wrong in some cases.
For example, in Fig. 4.4 a unit cell is represented, where molecules of oxygen are
confined between layers of copper atoms. OS of copper atoms is -4 by Eq. (4.2),
which is wrong because copper does not have a bond with oxygen.
To ensure that all oxygen atoms present in -2 state we disregard compounds
using several rules such as: compounds were disregarded if they have the shortest
bond between cation and anion larger than 130% of the sum of covalent radii [92]
to ensure that each atom in a compound has a covalent bond. Compounds were
disregarded if there is an O-O bond length smaller than 1.65 Å. The OS of oxygens
in an O-O bond can be 0 or -1 depending on the multiplicity of the bond (double or
single). The length of the double O=O bond is 1.21 Å, the length of a single O-O
bond is 1.48 Å [93]. To take into account that DFT gives errors [94] we had to
increase the threshold bond length, thus compounds with the O-O bond smaller
than 1.65 Å were disregarded. The resulting data set includes 697 compounds and
consists of 10743 atoms.
We carried out analysis of the resulting data set and demonstrated a correlation
between OS and Bader charge for a large set of single and mixed valence systems.
The results can be summarized as below:
• Bader charge is a good parameter for predicting OS in most of the binary
oxides
31
• There is a unique linear relationship between Bader charges and OS for
each binary oxide
• The results in Figure 2 (Paper III) may be used for predicting OS
• Our approach fails to predict OS of iron in MV compounds
• Bader charge intervals are not equal between OSs and tend to shrink with
higher OS
• Oxygen in -1 or 0 OS can have Bader charge down to -0.96 e, oxygens
with lower Bader charge are in -2 OS
• With a high probability cations of the same element in binaries are in
different OS, if the maximal Bader charge difference is more than 0.1 e
• The results can be compared with those for other exchange-correlational
functionals, but an estimation of the discrepancy is needed.
We present a technique which can give at least a hint about the OS of the
cations and predict OS in MV compounds. For surfaces, clusters and more complex
systems than binary, the charges can differ.
32
5 Conclusions
Density functional theory was used to study an effective metallic contact between
MoS2 and Ni nanoparticles, the catalytic activity of the (103) edge surface of MoS2
and the dependency between Bader charges and oxidation states.
We demonstrated by investigation of the band structure of the contact region
that a metallic contact to the edge surface of molybdenum disulfide can be created
with Au-Ni alloy.
The (103) edge surface of MoS2 was proven to have similar catalytic properties
to the (100) edge surface of MoS2 and the monolayer in relation to the
hydrodesulfurization reaction. Moreover, a more energetically favorable pathway
was found.
The long-standing issue about the applicability of Bader charges to oxidation
states has been solved for binary oxides. In addition, a new method for obtaining
oxidation states in mixed-valence compounds has been suggested.
DFT took root in physics and now flourishes almost in every field of science
where materials are involved. We have shown here only a fraction of possible
applications. They continue to grow and DFT finds new ways to be applied in
various scientific problems.
Future considerations in the use of online open databases for computed
properties of materials enable the utilization of big data to find connections between
materials and their properties. For example, doping of materials has led to only a
limited set of materials with new properties, whereas creation of an artificial
intelligence framework for the automated recommendation of doping candidates
can provide a huge speed increase in finding new materials with desired properties.
33
List of references
[1] P. Kościk, T. Sowiński, Exactly solvable model of two trapped quantum particles
interacting via finite-range soft-core interactions, Sci. Rep. 8 (2018) 48.
doi:10.1038/s41598-017-18505-5.
[2] R.S. Somerville, R. Davé, Physical Models of Galaxy Formation in a Cosmological
Framework, Annu. Rev. Astron. Astrophys. 53 (2015) 51–113.
doi:10.1146/annurev-astro-082812-140951.
[3] R.O. Jones, Density functional theory: Its origins, rise to prominence, and future,
Rev. Mod. Phys. 87 (2015). doi:10.1103/RevModPhys.87.897.
[4] A.D. Becke, Perspective: Fifty years of density-functional theory in chemical
physics, J. Chem. Phys. 140 (2014) 18A301. doi:10.1063/1.4869598.
[5] T. van Mourik, M. Bühl, M.-P. Gaigeot, Density functional theory across chemistry,
physics and biology., Philos. Trans. A. Math. Phys. Eng. Sci. 372 (2014) 20120488.
doi:10.1098/rsta.2012.0488.
[6] W. Kohn, Nobel Lecture: Electronic structure of matter—wave functions and
density functionals, Rev. Mod. Phys. 71 (1999) 1253–1266.
doi:10.1103/RevModPhys.71.1253.
[7] R. Ganatra, Q. Zhang, Few-layer MoS2: A promising layered semiconductor, ACS
Nano. (2014). doi:10.1021/nn405938z.
[8] K.F. Mak, C. Lee, J. Hone, J. Shan, T.F. Heinz, Atomically thin MoS2: A new direct-
gap semiconductor, Phys. Rev. Lett. 105 (2010) 136805.
doi:10.1103/PhysRevLett.105.136805.
[9] W. Wu, L. Wang, Y. Li, F. Zhang, L. Lin, S. Niu, D. Chenet, X. Zhang, Y. Hao,
T.F. Heinz, J. Hone, Z.L. Wang, Piezoelectricity of single-atomic-layer MoS2 for
energy conversion and piezotronics, Nature. (2014). doi:10.1038/nature13792.
[10] A.T. Neal, H. Liu, J.J. Gu, P.D. Ye, Metal contacts to MoS2: A two-dimensional
semiconductor, in: Device Res. Conf. - Conf. Dig. DRC, 2012.
doi:10.1109/DRC.2012.6256928.
[11] M. Kampa, E. Castanas, Human health effects of air pollution, Environ. Pollut. 151
(2008) 362–367. doi:10.1016/j.envpol.2007.06.012.
[12] A. Stanislaus, A. Marafi, M.S. Rana, Recent advances in the science and technology
of ultra low sulfur diesel (ULSD) production, Catal. Today. 153 (2010) 1–68.
doi:10.1016/j.cattod.2010.05.011.
[13] G. Liu, A.W. Robertson, M.M.J. Li, W.C.H. Kuo, M.T. Darby, M.H. Muhieddine,
Y.C. Lin, K. Suenaga, M. Stamatakis, J.H. Warner, S.C.E. Tsang, MoS2 monolayer
catalyst doped with isolated Co atoms for the hydrodeoxygenation reaction, Nat.
Chem. 9 (2017) 810–816. doi:10.1038/NCHEM.2740.
[14] F. Cui, G. Li, X. Li, M. Lu, M. Li, Enhancement of hydrodesulfurization of 4,6-
dimethyldibenzothiophene catalyzed by CoMo catalysts supported on carbon-
covered γ-Al2O3, Catal. Sci. Technol. 5 (2015) 549–555. doi:10.1039/C4CY00814F.
[15] O.Y. Gutierrez, S. Singh, E. Schachtl, J. Kim, E. Kondratieva, J. Hein, J.A. Lercher,
Effects of the support on the performance and promotion of (Ni)MoS2 catalysts for
34
simultaneous hydrodenitrogenation and hydrodesulfurization, ACS Catal. 4 (2014)
1487–1499. doi:10.1021/cs500034d.
[16] J. Chen, E. Dominguez Garcia, E. Oliviero, L. Oliviero, F. Maugé, Effect of high
pressure sulfidation on the morphology and reactivity of MoS2 slabs on
MoS2/Al2O3 catalyst prepared with citric acid Dedicated to the 60th birthday of J.-
P. Gilson, J. Catal. 339 (2016) 153–162. doi:10.1016/j.jcat.2016.04.010.
[17] E.J.M. Hensen, P.J. Kooyman, Y. Van der Meer, A.M. Van der Kraan, V.H.J. De
Beer, J.A.R. Van Veen, R.A. Van Santen, E.J.M. Hensen, P.J. Kooyman, Y. Van
der Meer, A.M. Van der Kraan, The relation between morphology and hydrotreating
activity for supported MoS2 particles, J. Catal. 199 (2001) 224–235.
doi:10.1006/jcat.2000.3158.
[18] R. Haunschild, A. Barth, W. Marx, Evolution of DFT studies in view of a
scientometric perspective, J. Cheminform. (2016). doi:10.1186/s13321-016-0166-
y.
[19] O. Isayev, C. Oses, C. Toher, E. Gossett, S. Curtarolo, A. Tropsha, Universal
fragment descriptors for predicting properties of inorganic crystals, Nat. Commun.
8 (2017) 15679. doi:10.1038/ncomms15679.
[20] O. Isayev, D. Fourches, E.N. Muratov, C. Oses, K. Rasch, A. Tropsha, S. Curtarolo,
Materials Cartography: Representing and Mining Materials Space Using Structural
and Electronic Fingerprints, Chem. Mater. 27 (2015) 735–743.
doi:10.1021/cm503507h.
[21] R.F.W. Bader, Atoms in Molecules: A Quantum Theory, 1990.
[22] R.S. Mulliken, Electronic Population Analysis on LCAO[Single Bond]MO
Molecular Wave Functions. I, J. Chem. Phys. 23 (1955) 1833.
doi:10.1063/1.1740588.
[23] C.I. Bayly, P. Cieplak, W.D. Cornell, P.A. Kollman, A well-behaved electrostatic
potential based method using charge restraints for deriving atomic charges: The
RESP model, J. Phys. Chem. 97 (1993) 10269–10280. doi:10.1021/j100142a004.
[24] P.O. Löwdin, On the non-orthogonality problem connected with the use of atomic
wave functions in the theory of molecules and crystals, J. Chem. Phys. 18 (1950)
365–375. doi:10.1063/1.1747632.
[25] D. Kleppner, R. Jackiw, One hundred years of quantum physics., Science. 289
(2000) 893–8. doi:10.1126/science.289.5481.893.
[26] M. Born, R. Oppenheimer, Zur Quantentheorie der Molekeln, Ann. Phys. (1927).
doi:10.1002/andp.19273892002.
[27] P. Hohenberg, W. Kohn, Inhomogeneous Electron Gas, Phys. Rev. 136 (1964)
B864–B871. doi:10.1103/PhysRev.136.B864.
[28] W. Kohn, L.J. Sham, Self-consistent equations including exchange and correlation
effects, Phys. Rev. 140 (1965) A1133–A1138. doi:10.1103/PhysRev.140.A1133.
[29] D.C. Langreth, M.J. Mehl, Easily Implementable Nonlocal Exchange-Correlation
Energy Functional, Phys. Rev. Lett. 47 (1981) 446–450.
doi:10.1103/PhysRevLett.47.446.
[30] A.D. Becke, Density-functional exchange-energy approximation with correct
35
asymptotic behavior, Phys. Rev. A. 38 (1988) 3098–3100.
doi:10.1103/PhysRevA.38.3098.
[31] J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R. Pederson, D.J. Singh,
C. Fiolhais, Atoms, molecules, solids, and surfaces: Applications of the generalized
gradient approximation for exchange and correlation, Phys. Rev. B. 46 (1992)
6671–6687. doi:10.1103/PhysRevB.46.6671.
[32] J.P. Perdew, K. Burke, M. Ernzerhof, Generalized gradient approximation made
simple, Phys. Rev. Lett. 77 (1996) 3865–3868. doi:10.1103/PhysRevLett.77.3865.
[33] J. Tao, J.P. Perdew, V.N. Staroverov, G.E. Scuseria, Climbing the density functional
ladder: Nonempirical meta–generalized gradient approximation designed for
molecules and solids, Phys. Rev. Lett. (2003). doi:10.1103/PhysRevLett.91.146401.
[34] J.P. Perdew, S. Kurth, A. Zupan, P. Blaha, Accurate density functional with correct
formal properties: A step beyond the generalized gradient approximation, Phys. Rev.
Lett. (1999). doi:10.1103/PhysRevLett.82.2544.
[35] Y. Zhao, D.G. Truhlar, A new local density functional for main-group
thermochemistry, transition metal bonding, thermochemical kinetics, and
noncovalent interactions, J. Chem. Phys. (2006). doi:10.1063/1.2370993.
[36] A.D. Becke, Density-functional thermochemistry. III. The role of exact exchange,
J. Chem. Phys. (1993). doi:10.1063/1.464913.
[37] C. Adamo, V. Barone, Toward reliable density functional methods without
adjustable parameters: The PBE0 model, J. Chem. Phys. (1999).
doi:10.1063/1.478522.
[38] Y. Zhao, N.E. Schultz, D.G. Truhlar, Exchange-correlation functional with broad
accuracy for metallic and nonmetallic compounds, kinetics, and noncovalent
interactions, J. Chem. Phys. (2005). doi:10.1063/1.2126975.
[39] H. Iikura, T. Tsuneda, T. Yanai, K. Hirao, A long-range correction scheme for
generalized-gradient-approximation exchange functionals, J. Chem. Phys. (2001).
doi:10.1063/1.1383587.
[40] O.A. Vydrov, G.E. Scuseria, Assessment of a long-range corrected hybrid
functional, J. Chem. Phys. (2006). doi:10.1063/1.2409292.
[41] J. Heyd, G.E. Scuseria, M. Ernzerhof, Hybrid functionals based on a screened
Coulomb potential, J. Chem. Phys. (2003). doi:10.1063/1.1564060.
[42] T. Yanai, D.P. Tew, N.C. Handy, A new hybrid exchange-correlation functional
using the Coulomb-attenuating method (CAM-B3LYP), Chem. Phys. Lett. (2004).
doi:10.1016/j.cplett.2004.06.011.
[43] J. Da Chai, M. Head-Gordon, Systematic optimization of long-range corrected
hybrid density functionals, J. Chem. Phys. (2008). doi:10.1063/1.2834918.
[44] J. Jaramillo, G.E. Scuseria, M. Ernzerhof, Local hybrid functionals, J. Chem. Phys.
(2003). doi:10.1063/1.1528936.
[45] E.R. Johnson, Local-hybrid functional based on the correlation length, J. Chem.
Phys. (2014). doi:10.1063/1.4896302.
[46] H. Bahmann, A. Rodenberg, A. V. Arbuznikov, M. Kaupp, A thermochemically
competitive local hybrid functional without gradient corrections, J. Chem. Phys.
36
(2007). doi:10.1063/1.2429058.
[47] R. Haunschild, G.E. Scuseria, Range-separated local hybrids, J. Chem. Phys. (2010).
doi:10.1063/1.3451078.
[48] A. V. Arbuznikov, M. Kaupp, Importance of the correlation contribution for local
hybrid functionals: Range separation and self-interaction corrections, J. Chem. Phys.
(2012). doi:10.1063/1.3672080.
[49] J.P. Perdew, A. Ruzsinszky, J. Tao, V.N. Staroverov, G.E. Scuseria, G.I. Csonka,
Prescription for the design and selection of density functional approximations: More
constraint satisfaction with fewer fits, J. Chem. Phys. 123 (2005) 062201.
doi:10.1063/1.1904565.
[50] L. Brillouin, Wave Propagation in Periodic Structures, J. Sound Vib. (1953).
doi:10.1007/978-3-319-28179-7.
[51] H. Monkhorst, J. Pack, Special points for Brillouin zone integrations, Phys. Rev. B.
13 (1976) 5188–5192. doi:10.1103/PhysRevB.13.5188.
[52] X. Gonze, J.M. Beuken, R. Caracas, F. Detraux, M. Fuchs, G.M. Rignanese, L.
Sindic, M. Verstraete, G. Zerah, F. Jollet, M. Torrent, A. Roy, M. Mikami, P.
Ghosez, J.Y. Raty, D.C. Allan, First-principles computation of material properties:
The ABINIT software project, Comput. Mater. Sci. (2002). doi:10.1016/S0927-
0256(02)00325-7.
[53] J.J. Mortensen, L.B. Hansen, K.W. Jacobsen, Real-space grid implementation of the
projector augmented wave method, Phys. Rev. B - Condens. Matter Mater. Phys.
(2005). doi:10.1103/PhysRevB.71.035109.
[54] P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli,
G.L. Chiarotti, M. Cococcioni, I. Dabo, A. Dal Corso, S. De Gironcoli, S. Fabris,
G. Fratesi, R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri, L.
Martin-Samos, N. Marzari, F. Mauri, R. Mazzarello, S. Paolini, A. Pasquarello, L.
Paulatto, C. Sbraccia, S. Scandolo, G. Sclauzero, A.P. Seitsonen, A. Smogunov, P.
Umari, R.M. Wentzcovitch, QUANTUM ESPRESSO: A modular and open-source
software project for quantum simulations of materials, J. Phys. Condens. Matter.
(2009). doi:10.1088/0953-8984/21/39/395502.
[55] G. Kresse, J. Furthmüller, Efficient iterative schemes for ab initio total-energy
calculations using a plane-wave basis set, Phys. Rev. B - Condens. Matter Mater.
Phys. 54 (1996) 11169–11186. doi:10.1103/PhysRevB.54.11169.
[56] D. Joubert, From ultrasoft pseudopotentials to the projector augmented-wave
method, Phys. Rev. B - Condens. Matter Mater. Phys. 59 (1999) 1758–1775.
doi:10.1103/PhysRevB.59.1758.
[57] W.R. Tyson, W.A. Miller, Surface free energies of solid metals: Estimation from
liquid surface tension measurements, Surf. Sci. 62 (1977) 267–276.
doi:10.1016/0039-6028(77)90442-3.
[58] J.C. Boettger, Nonconvergence of surface energies obtained from thin-film
calculations, Phys. Rev. B. (1994). doi:10.1103/PhysRevB.49.16798.
[59] M. Methfessel, D. Hennig, M. Scheffler, Trends of the surface relaxations, surface
energies, and work functions of the 4d transition metals, Phys. Rev. B. (1992).
37
doi:10.1103/PhysRevB.46.4816.
[60] J.L.F. Da Silva, C. Stampfl, M. Scheffler, Converged properties of clean metal
surfaces by all-electron first-principles calculations, Surf. Sci. 600 (2006) 703–715.
doi:10.1016/J.SUSC.2005.12.008.
[61] V. Fiorentini, M. Methfessel, Extracting convergent surface energies from slab
calculations, J. Phys. Condens. Matter. (1996). doi:10.1088/0953-8984/8/36/005.
[62] N.E. Singh-Miller, N. Marzari, Surface energies, work functions, and surface
relaxations of low-index metallic surfaces from first principles, Phys. Rev. B -
Condens. Matter Mater. Phys. (2009). doi:10.1103/PhysRevB.80.235407.
[63] H. Jónsson, G. Mills, K.W. Jacobsen, Nudged elastic band method for finding
minimum energy paths of transitions, in: Class. Quantum Dyn. Condens. Phase
Simulations, 1998. doi:10.1142/9789812839664_0016.
[64] G. Henkelman, B.P. Uberuaga, H. Jónsson, Climbing image nudged elastic band
method for finding saddle points and minimum energy paths, J. Chem. Phys. 113
(2000) 9901–9904. doi:10.1063/1.1329672.
[65] G. Henkelman, H. Jónsson, A dimer method for finding saddle points on high
dimensional potential surfaces using only first derivatives, J. Chem. Phys. 111
(1999) 7010–7022. doi:10.1063/1.480097.
[66] F.L. Hirshfeld, Bonded-atom fragments for describing molecular charge densities,
Theor. Chim. Acta. (1977). doi:10.1007/BF00549096.
[67] S. Curtarolo, W. Setyawan, G.L.W. Hart, M. Jahnatek, R. V. Chepulskii, R.H.
Taylor, S. Wang, J. Xue, K. Yang, O. Levy, M.J. Mehl, H.T. Stokes, D.O.
Demchenko, D. Morgan, AFLOW: An automatic framework for high-throughput
materials discovery, Comput. Mater. Sci. 58 (2012) 218–226.
doi:10.1016/j.commatsci.2012.02.005.
[68] S. Curtarolo, W. Setyawan, S. Wang, J. Xue, K. Yang, R.H. Taylor, L.J. Nelson,
G.L.W. Hart, S. Sanvito, M. Buongiorno-Nardelli, N. Mingo, O. Levy,
AFLOWLIB.ORG: A distributed materials properties repository from high-
throughput ab initio calculations, Comput. Mater. Sci. 58 (2012) 227–235.
doi:10.1016/j.commatsci.2012.02.002.
[69] A. Jain, S.P. Ong, G. Hautier, W. Chen, W.D. Richards, S. Dacek, S. Cholia, D.
Gunter, D. Skinner, G. Ceder, K.A. Persson, Commentary: The materials project: A
materials genome approach to accelerating materials innovation, APL Mater. 1
(2013). doi:10.1063/1.4812323.
[70] T. Haupt, N. Sukhija, M.F. Horstemeyer, Cyberinfrastructure support for
engineering virtual organization for CyberDesign, in: Lect. Notes Comput. Sci.
(Including Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinformatics), 2012: pp.
161–170. doi:10.1007/978-3-642-31500-8_17.
[71] NOMAD Repository, (n.d.). http://nomad-repository.eu/ (accessed May 22, 2018).
[72] C.E. Calderon, J.J. Plata, C. Toher, C. Oses, O. Levy, M. Fornari, A. Natan, M.J.
Mehl, G. Hart, M. Buongiorno Nardelli, S. Curtarolo, The AFLOW standard for
high-throughput materials science calculations, Comput. Mater. Sci. 108 (2015)
233–238. doi:10.1016/j.commatsci.2015.07.019.
38
[73] R.H. Taylor, F. Rose, C. Toher, O. Levy, K. Yang, M. Buongiorno Nardelli, S.
Curtarolo, A RESTful API for exchanging materials data in the AFLOWLIB.org
consortium, Comput. Mater. Sci. 93 (2014) 178–192.
doi:10.1016/j.commatsci.2014.05.014.
[74] S.M. Tan, A. Ambrosi, Z. Sofer, Š. Huber, D. Sedmidubský, M. Pumera, Pristine
basal- and edge-plane-oriented molybdenite MoS2 exhibiting highly anisotropic
properties, Chem. - A Eur. J. (2015). doi:10.1002/chem.201500435.
[75] L.Y. Gan, Y.J. Zhao, D. Huang, U. Schwingenschlögl, First-principles analysis of
MoS2/Ti2C and MoS2/Ti2CY2 (Y=F and OH) all-2D semiconductor/metal contacts,
Phys. Rev. B - Condens. Matter Mater. Phys. (2013).
doi:10.1103/PhysRevB.87.245307.
[76] J. Kang, W. Liu, D. Sarkar, D. Jena, K. Banerjee, Computational study of metal
contacts to monolayer transition-metal dichalcogenide semiconductors, Phys. Rev.
X. (2014). doi:10.1103/PhysRevX.4.031005.
[77] S. Grimme, J. Antony, S. Ehrlich, H. Krieg, A consistent and accurate ab initio
parametrization of density functional dispersion correction (DFT-D) for the 94
elements H-Pu, J. Chem. Phys. 132 (2010) 154104. doi:10.1063/1.3382344.
[78] K. Momma, F. Izumi, VESTA 3 for three-dimensional visualisation of crystal,
voluemtric and morphology data, J. Appl. Crystallogr. (2011).
doi:doi:10.1107/S0021889811038970.
[79] J. Vähäkangas, P. Lantto, J. Vaara, M. Huttula, W. Cao, Orienting spins in dually
doped monolayer MoS2: From one-sided to double-sided doping, Chem. Commun.
(2017). doi:10.1039/c7cc01560g.
[80] M. Zhang, Z. Huang, X. Wang, H. Zhang, T. Li, Z. Wu, Y. Luo, W. Cao, Magnetic
MoS2 pizzas and sandwiches with Mnn(n = 1–4) cluster toppings and fillings: A
first-principles investigation, Sci. Rep. 6 (2016) 19504. doi:10.1038/srep19504.
[81] V. Pankratov, J. Hoszowska, J.C. Dousse, M. Huttula, A. Kis, D. Krasnozhon, M.
Zhang, W. Cao, Vacuum ultraviolet excitation luminescence spectroscopy of few-
layered MoS2, J. Phys. Condens. Matter. (2015). doi:10.1088/0953-
8984/28/1/015301.
[82] J. Li, W. Tang, H. Yang, Z. Dong, J. Huang, S. Li, J. Wang, J. Jin, J. Ma, Enhanced-
electrocatalytic activity of Ni(1-x) Fex alloy supported on polyethyleneimine
functionalized MoS2 nanosheets for hydrazine oxidation, RSC Adv. 4 (2014) 1988–
1995. doi:10.1039/C3RA42757A.
[83] Y. Liu, Y. Zhao, L. Jiao, J. Chen, A graphene-like MoS2/graphene nanocomposite
as a highperformance anode for lithium ion batteries, J. Mater. Chem. A. 2 (2014)
13109–13115. doi:10.1039/C4TA01644K.
[84] P. Cheng, K. Sun, Y.H. Hu, Mechanically-induced reverse phase transformation of
MoS2 from stable 2H to metastable 1T and its memristive behavior, RSC Adv. 6
(2016) 65691–65697. doi:10.1039/C6RA12735E.
[85] S. Ravula, J.B. Essner, G.A. Baker, Kitchen-Inspired Nanochemistry: Dispersion,
Exfoliation, and Hybridization of Functional MoS2 Nanosheets Using Culinary
Hydrocolloids, ChemNanoMat. 1 (2015) 167–177. doi:10.1002/cnma.201500022.
39
[86] J. Hu, B. Huang, C. Zhang, Z. Wang, Y. An, D. Zhou, H. Lin, M.K.H. Leung, S.
Yang, Engineering stepped edge surface structures of MoS2 sheet stacks to
accelerate the hydrogen evolution reaction, Energy Environ. Sci. 10 (2017) 593–
603. doi:10.1039/C6EE03629E.
[87] X. Zhang, H. Tang, M. Xue, C. Li, Facile synthesis and characterization of ultrathin
MoS2 nanosheets, Mater. Lett. 130 (2014) 83–86. doi:10.1016/j.matlet.2014.05.078.
[88] L. Yang, X. Cui, J. Zhang, K. Wang, M. Shen, S. Zeng, S.A. Dayeh, L. Feng, B.
Xiang, Lattice strain effects on the optical properties of MoS2 nanosheets, Sci. Rep.
4 (2015) 5649. doi:10.1038/srep05649.
[89] P.Y. Prodhomme, P. Raybaud, H. Toulhoat, Free-energy profiles along reduction
pathways of MoS2 M-edge and S-edge by dihydrogen: A first-principles study, J.
Catal. 280 (2011) 178–195. doi:10.1016/j.jcat.2011.03.017.
[90] K.G. Reeves, Y. Kanai, Theoretical oxidation state analysis of Ru-(bpy)3: Influence
of water solvation and Hubbard correction in first-principles calculations, J. Chem.
Phys. 141 (2014). doi:10.1063/1.4886406.
[91] M. Jansen, U. Wedig, A piece of the picture - Misunderstanding of chemical
concepts, Angew. Chemie - Int. Ed. 47 (2008) 10026–10029.
doi:10.1002/anie.200803605.
[92] B. Cordero, V. Gómez, A.E. Platero-Prats, M. Revés, J. Echeverría, E. Cremades,
F. Barragán, S. Alvarez, Covalent radii revisited, Dalt. Trans. (2008) 2832.
doi:10.1039/b801115j.
[93] D.R. Lide, CRC Handbook of Chemistry and Physics, EBook. (2003) 3485.
doi:978-1466571143.
[94] K. Lejaeghere, V. Van Speybroeck, G. Van Oost, S. Cottenier, Error estimates for
solid-state density-functional theory predictions: An overview by means of the
ground-state elemental crystals, Crit. Rev. Solid State Mater. Sci. 39 (2014) 1–24.
doi:10.1080/10408436.2013.772503.
40
Original publications
I Reprinted with permission from Small. Shi, X., Posysaev, S., Huttula, M.,
Pankratov, V., Hoszowska, J., Dousse, J.-Cl., Zeeshan, F., Niu, Y., Zakharov,
A., Li, T., Miroshnichenko, O., Zhang, M., Wang, X., Huang, Z., Saukko, S.,
González, D. L., Dijken, S., Alatalo, M., Cao. W., Metallic contact between
MoS2 and Ni via Au nanoglue. Small, 14, 1704526 (2018). DOI:
10.1002/smll.201704526. Copyright Wiley-VCH Verlag GmbH & Co. KGaA.
Reproduced with permission.
II Posysaev, S., Alatalo, M., Surface morphology and sulfur reduction pathways
of MoS2 Mo-edges of monolayer, (100) and (103) surfaces by molecular
hydrogen: A DFT study. ACS Omega, submitted for publication
III Posysaev, S., Miroshnichenko, O., Alatalo, M., Le, D., Rahman, T., Oxidation
states of binary oxides from data analytics of the electronic structure.
Computational Materials Science, submitted for publication
Original publications are not included in the electronic version of the dissertation.