University of the www.mscnano.

eu
Basque Country

MASTER THESIS
Optical properties of
pentacene and picene

by
Jeiran Jokar

Supervisors:
Prof. Angel Rubio Secades
Dr. Matteo Gatti
Dr. Pierluigi Cudazzo

14 September 2011
Acknowledgement

I would like to express my gratitude to all those who gave me the possibility
to complete this thesis. This work would not have been possible without
the support from Prof. Dr. Angel Rubio under whose guidance I chose this
topic. I would like to gratefully acknowledge the supervision of my advisor,
Dr. Matteo Gatti, who has been abundantly helpful and has assisted me in
numerous ways. I specially thank him for his infinite patience. The discus-
sions I had with him were invaluable. I would like to say a big thanks to
Dr. Pierluigi Cudazzo, who helped me a lot.
I would like to thank Donostia International Physics Center and the Uni-
versity of the Basque Country for financial supports.

3
Contents

1 Introduction 7

2 Theory 9
2.1 DFT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.1 The Hohenberg-Kohn theorems . . . . . . . . . . . . . 11
2.1.2 The Kohn-Sham method . . . . . . . . . . . . . . . . . 12
2.1.3 Local Density Approximation . . . . . . . . . . . . . . 13
2.1.4 Pseudopotential approximation . . . . . . . . . . . . . 14
2.2 Time Dependent Density Functional Theory . . . . . . . . . . 15
2.2.1 The Runge-Gross theorem . . . . . . . . . . . . . . . . 15
2.2.2 Time-Dependent Kohn-Sham Theory . . . . . . . . . . 15
2.2.3 xc functionals . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.4 DFT vs TDDFT . . . . . . . . . . . . . . . . . . . . . 16
2.2.5 Linear response theory . . . . . . . . . . . . . . . . . . 17
2.2.6 Casida’s equations . . . . . . . . . . . . . . . . . . . . 20
2.2.7 Electronic excitations by means of time-propagation . 20

3 Results 23
3.1 Ground state properties . . . . . . . . . . . . . . . . . . . . . 23
3.1.1 Convergence study . . . . . . . . . . . . . . . . . . . . 23
3.1.2 Structural optimisation . . . . . . . . . . . . . . . . . 26
3.1.3 Kohn-Sham orbitals and energy levels . . . . . . . . . 28
3.2 Absorption spectra from time propagation . . . . . . . . . . . 33
3.2.1 Convergence study . . . . . . . . . . . . . . . . . . . . 33
3.3 Absorption spectra from Casida equation . . . . . . . . . . . 38

4 Conclusion 45

bibliography 47

List of Figures 49

List of Tables 51

5
Chapter 1

Introduction

Organic molecular solids have attracted considerable interest for both funda-
mental reasons and possible technological applications. In fact they are very
promising materials for a wide range of applications [1], as their properties
can be efficiently tailored for their use in (opto)electronic devices [2] such as
field-effect transistor [3], light-emitting diodes [4][5], photovoltaic cells [6][7],
etc. Among the different classes of organic molecular solids, the most stud-
ied one in the last years is probably that of the aromatic molecules. The
interest for this class of molecular crystals recently increased after the dis-
covery of superconductivity in potassium doped picene and potassium doped
coronene at critical temperature Tc =18 K [8] and 15 K respectively [9].
Since the constituents of molecular solids are molecular units interacting
through weak Van der Waals forces their physical properties, as for example
dynamical, electronic and optical properties, are to a large extent governed
by the individual molecular units [10]. Thus, in principle, by studying the
isolated molecule we have access also to the corresponding condensed states.
In this work we present a first-principles study based on density func-
tional theory of the electronic and optical properties of two aromatic molecules
namely picene and pentacene. This will be a good starting point for the un-
derstanding of the electronic and optical properties of the related molecular
crystals.
Aromatic molecules are obtained by combining benzene rings in differ-
ent ways. In particular picene and pentacene both with molecular formula
C22 H14 consist of five benzene rings in two different configurations. In pen-
tacene the benzene rings are arranged in a linear manner (Fig. 3.7(a)) while
in picene they are arranged in a zigzag manner (Fig. 3.7(b)). Since C atoms
present 4 valence electrons (2 in the 2s state and 2 in the 2p state) they form
4 covalent bonds while H atoms only one bond involving their 1s electron.
Due to the planar structure in picene and pentacene as well as in benzene, for
each carbon atom the s orbital combines with two p (px , py ) orbitals. The
resulting three hybrid orbitals are oriented at 120◦ angles to each other.

7
8 CHAPTER 1. INTRODUCTION

These orbitals are called sp2 hybrid orbitals because they are composed of
one s and two p orbitals. If the sp2 orbitals of different C atoms overlap
in phase between them or with the H s orbital they interact constructively
to form a bonding molecular orbitals. The corresponding electron density
results centred along the line connecting the atomic nuclei. This type of
bond is called a σ bond. When they interact out of phase they give rise to
anti-bonding states called σ ∗ state. The remain pz orbitals (one for each C
atoms) combine among them giving the so called π bond. The π bond is
parallel to the line connecting the C nuclei and the related electron density is
localized above and below the line connecting the nuclei. The corresponding
antibonding states are called π ∗ states. Since the overlap between in-plane
atomic orbitals is larger than that between out of plane atomic orbitals, σ
bonds result stronger than the π ones [11]. Thus the electronic structure of
picene and pentacene can be described in terms of σ states responsible for
in plane C-C and C-H bonds and π states responsible for out of plane C-C
bonds. However being the π bonds the weakest, the chemical and optical
properties of these molecules are set mainly by π and π ∗ states.
The present work is organized as follows: in chapter one we discuss
the theoretical tools used in this thesis, namely DFT and TDDFT while
in the second chapter we present our results. Here, first we address the
structural and electronic properties of picene and pentacene then we focus on
their optical properties evaluating the absorption spectra with two different
method: performing time propagation and solving the Casida equation. The
last chapter is devoted to our conclusions.
Chapter 2

Theory

2.1 DFT
The first goal is to find the ground state of the molecules. In fact, by knowing
the ground state we can obtain the atomic structure of the molecules, and the
electronic properties (the orbital wavefunctions and energies). Moreover, the
ground state of the system will be the starting point to employ perturbation
theory (see sec. (2.2.5)) to study the optical response.
In order to find the ground state of a system of M atoms (with N elec-
trons), one has to solve the Schrödinger equation:

HΨ(r1 , . . . , rN , R1 , . . . , RM ) = EΨ(r1 , . . . , rN , R1 , . . . , RM ), (2.1)

and obtain the eigenstate Ψ0 corresponding to the lowest eigenvalue E0 . In

atomic units (e = m = ~ = 1), the non-relativistic Hamiltonian for the
electrons and the nuclei of mass MI , is:

N N N
1 X 2 1 X ∇2I 1 X 1
H=− ∇i − +
2 2 MI 2 |ri − rj |
i=1 I=1 i6=j=1
X ZI 1X ZI ZJ
− + , (2.2)
|ri − RI | 2 | RI − RJ |
iI I6=J

where the index i runs over electrons and I runs over nuclei. These are just
the kinetic terms of nuclei and electrons, plus the instantaneous Coulomb
interaction between all pairs of bodies.
One of the first proposal to solve Eq. (2.1) was published in 1927 by
Max Born and Julius Robert Oppenheimer. They observed that the nuclei
in the system move much slower than the electrons because they are more
massive [12]. From this, they assumed that the electrons are following the
slow motion of the nuclei. For each fixed ionic configuration, they are in
their ground state. The total wavefunction Ψ(r1 , . . . , rN , R1 , . . . , RM ) is

9
10 CHAPTER 2. THEORY

decoupled in the product of an electronic wavefunction and a ionic wave-

function. The ionic coordinates appear as given parameters in the electronic
Hamiltonian.
Thus, the electronic Hamiltonian reduces to [13]
N N
1X 2 1 X 1 X ZI
H=− ∇i + − . (2.3)
2 2 |ri − rj | |ri − RI |
i=1 i6=j=1 iI

Even in the Born-Oppenheimer approximation, solving the electronic prob-

lem (2.3) remains a formidable task. A possible strategy of solution is to
formulate a trial wavefunction and solve the Hamiltonian (2.3) with this
trial wavefunction.

In 1928, the british physicist Douglas Rayner Hartree proposed a prod-

uct of the single-electron wavefunctions as the solution of the many-body
Hamiltonian in a self-consistent way. This method is known as Hartree ap-
proximation [14].
The wave function in Hartree approximation is simply:

ΨHartree (r1 , . . . , rN ) = ψ1 (r1 )ψ2 (r2 ) . . . ψN (rN ) (2.4)

Two years later the russian physicist Vladimir Aleksandrovich Fock mod-
ified the Hartree’s method by allowing the wavefunction to be represented
by a Slater determinant [15]. This approach is commonly known as the
Hartree-Fock method (HF). The trial wavefunction ΨHF is antisymmetric
with respect to the interchange of both spin and space coordinates of any
two electrons.
 
 ψ1 (r1 ) ψ2 (r1 ) · · · ψN (r1 ) 
 
1  ψ1 (r2 ) ψ2 (r2 ) · · · ψN (r2 ) 
ΨHF = √  . .. .. (2.5)
N !  ..

. . 

ψ1 (rN ) ψ2 (rN ) · · · ψN (rN )

In the Slater determinant, exchanging the coordinates of two electrons is the

same as exchanging two rows, with a resulting change in the sign. Because
of this property of a Slater determinant, the basic antisymmetric property
of the wavefunction related with its fermionic nature is conserved.
The introduction of the antisymmetric principle in the wavefunction al-
lowed an improved description of the total electronic energy, because the
exchange energy effect is considered. However, in many problems in physics
this term is not enough to predict the properties of a electronic system.
This method does not take in account the fact that the motion of a single
electron affects the motion of all other electrons in the system. The missed
term in the total energy is called as correlation energy. A way of taking
into account correlation effects is called configuration interaction, in which
2.1. DFT 11

the many body wavefunction is expressed as a linear combination of Slater

determinants.
These approaches are all based upon improving the trial wavefunctions.
In the configuration interaction method, the combination of wavefunctions
soon becomes very complicated, resulting in severe limiting of the size of
problems, that can be treated.
Another and very different approach called the density functional theory
(DFT) has been proposed. DFT is the practical first-principles scheme that
is employed in the calculation of the ground-state properties. DFT is a
minimum-information theory. Instead of dealing with the many-body wave-
function Ψ(r1 , ..., rN ), DFT leads to calculate directly the simplest quantity,
i.e. the electronic density, that one needs in order to have access to the
ground-state properties. The correlation effect is taken into account, and
the size of the system which can be handled is far larger. Since 1980, this
method has established a position as one of the main tools of calculating
the properties of solid and molecules.

2.1.1 The Hohenberg-Kohn theorems

Density-functional theory is based upon two theorems first proved by Ho-
henburg and Kohn in 1964. The Hohenberg and Kohn theorems prove that,
for a system with a nondegenerate ground state [16], the ground-state en-
ergy of electrons is a unique functional of the electron density. Furthermore,
given an external potential, it is shown that the ground-state energy can
be obtained by minimizing the energy functional, with respect to the elec-
tron density. When the density is the true ground-state electron density,
this minimizes the energy functional. Their results can be summarized as
follows:

• The first theorem states that, for a system with a nondegenerate

ground state, there exists a one-to-one correspondence, up to an addi-
tive constant, between the ground-state density and the static exter-
nal potentialVext (r). Since there is also a one-to-one correspondence
between the external potential and the ground-state many-body wave-
function Ψ, Ψ can be written as a function of the density Ψ = Ψ[n(r)].

• The second theorem states that a universal functional for energy E[n]
in terms of the density n(r) can be defined, valid for any external
potential Vext (r). For any particular Vext (r), the exact ground state
energy of the system is the global minimum value of this functional,
and the density n(r) that minimizes the functional is the exact ground
state density n0 (r).
The functional E[n] alone is sufficient to determine the exact ground
state energy and density. In general, excited states of the electrons
must be determined by other means.
12 CHAPTER 2. THEORY

The Hohenberg-Kohn theorem is valid only for the densities that are the
ground state of some potential. These densities are called V -representable.
While the V -representability condition assesses the existence of the one-
to-one mapping between density and Vext , the Hohenberg-Kohn theorem
establishes its uniqueness. In the worst case, the potential has degenerate
ground states such that the given density is representable as a linear com-
bination of the degenerate ground-state densities (the density is then called
ensemble-V -representable) [17].
The Hohenberg-Kohn theorem has been then generalized in many ways:
to cases where the ground states is degenerate, or where the system is spin-
polarized,to relativistic DFT [18], to multicomponent DFT,to DFT for su-
perconductors [19], and so on.

2.1.2 The Kohn-Sham method

In a subsequent paper by Kohn and Sham [20], it is shown that the energy
functional can be recast by using one particle orbitals as EKS [{Φi }]. The
one-electron wavefunctions Φi (r) are moreover subjected to the orthogonal-
ization condition.
Z Z
1X
EKS ({Φi }) = − ki Φi (r)∇ Φi (r)d r + n(r)Vion-electron d3 r
2 3
2
i
n(r)n(r0 ) 3 3 0
Z
1
+ d rd r + Exc [n(r)] (2.6)
2 |r − r0 |
where the i- summation takes over all one-electron orbital, ki is the number
of occupations in i-state, Exc the exchange energy, and n(r) is the electronic
density. It is given by

occ
X
n(r) = |Φi (r)|2 . (2.7)
i

The wavefunctions Φi which minimize the Kohn-Sham functional energy in

Eq. (2.6) satisfy the following eigenvalue equations,

HKS Φi (r) = εi Φi (r), (2.8)

where HKS is Kohn-Sham Hamiltonian,

1
HKS = − ∇2 + Vion-electron (r) + VH (r) + Vxc (r). (2.9)
2
Here, VH (r) is the Hartree potential

n(r0 ) 3 0
Z
VH (r) = d r, (2.10)
|r − r0 |
2.1. DFT 13

Vxc (r) is the exchange correlation potential

δExc [n]
Vxc (r) = , (2.11)
δn(r)
and
X ZI
Velectron−ion (r) = (2.12)
|r − RI |
I

is the electron-ion interaction (see Eq. (2.3)). εi and Φi denote the eigen-
values and eigenfunctions of the Kohn-Sham equation, respectively. The
wavefunctions calculated by Eq. (2.8) yield the charge density by Eq.( 2.7).
Hence, the Kohn-Sham equation must be solved self-consistently. It seems
that Eq. (2.8) plays a role of Schrödinger equation of one-electron wave-
function, but the idea underlying these equation is quite different. While
in Hartree, Hartree-Fock, or configuration interaction the key variable is
the many body wavefunction Ψ(r1 , ..., rN ) and the ground-state is found by
solving the many body Hamiltonian (2.3), in DFT the key variable is the
much simpler electronic density n(r) and the ground-state energy is found
by minimizing the energy functional or, equivalently, by solving the Kohn-
Sham equations. However, the Kohn-Sham orbitals and energies do not have
a physical meaning. The Slater determinant obtained using Kohn-Sham or-
bitals is different from the Slater determinant that is the solution of the
Hartree-Fock equations.

2.1.3 Local Density Approximation

A price of mathematical simplification of the density functional method,
which replaces the many-electronic problem by one-electron problem is paid
by introducing an unknown functional of exchange and correlation Exc of the
charge density. Fortunately, there is a simple approximation for Exc . The
most widely used form of Exc is the so-called local density approximation
(LDA). The true system is locally (that is for every point r) approximated
by a homogeneous electron gas of density n = n(r). Thus the exchange-
correlation energy becomes:
Z
Exc [n] = n(r)εHEG
LDA
xc (n(r))d3 r, (2.13)

where εHEG
xc (n(r)) is the exchange-correlation energy per electron in a ho-
mogeneous electron gas of density n [20]. This equation is, by construction,
exact for the homogeneous electron gas and we can expect it to work well
for systems where the density has small spatial variations, or where the
electron-electron interaction is well-screened.
The domain of applicability of LDA has been unexpectedly found to
go much beyond the nearly-free electron gas and accurate results can be
obtained for inhomogeneous systems like atoms or molecules.
14 CHAPTER 2. THEORY

Figure 2.1: Schematic representation of a pseudopotential (left, dark curve)

and a pseudo-wavefunction (right, dark curve) along with the all-electron
potential (with the 1/r tail) and wavefunction (indigo curves). Notice that
the all-electron and pseudofunctions are identical beyond the radial cutoff
Rc and the pseudo-functions are smooth outside the core region.

2.1.4 Pseudopotential approximation

In order to solve the Kohn-Sham equations on a computer, a possibility is
to represent them on a discrete grid of points in space. A limitation to this
approach is given by the fact that the electron-ion interaction Eq. (2.12)
is rapidly varying on a very short scale length. These rapid variations are
mostly due to the interactions with core electrons, which are more tightly
bound to the ionic cores than valence electrons.
Fortunately, the physical and chemical characteristics of many materials are
governed by the valence electrons which extend to more wide region, and the
core states are insensitive to those properties. Then, we can make an approx-
imation by using valence electrons solely in describing the chemical combin-
ing characters of materials. Therefore, needed potentials have relatively
slowly varying characters. The electron-ion interactionis then expressed
by means of slowly varying “pseudopotentials”. In the norm-conserving
pseudopotential formalism, the corresponding valence wavefunctions (called
“pseudowavefunctions”) match the true all electron ones outside a cutoff
radius Rc . The process is sketched in Fig. (2.1).
2.2. TIME DEPENDENT DENSITY FUNCTIONAL THEORY 15

2.2 Time Dependent Density Functional Theory

As I said before, with DFT we can calculate the ground-state properties
of a system. However, to simulate absorption, we need to deal with time-
dependent external potentials. The description of time-dependent phenom-
ena was incorporated properly into a density functional framework by Runge
and Gross, who generalized the Hohenberg-Kohn theorem to time-dependent
densities and potentials, establishing the time dependent density functional
theory (TDDFT) [21].
Now we should solve the time-dependent Schrödinger equation
∂
i Ψ(r1 , . . . , rN , t) = H(r1 , . . . , rN , t)Ψ(r1 , . . . , rN , t). (2.14)
∂t
The only difference between the Hamiltonian in Eq. (2.3) is that we have
time-dependent external potential Vext (r, t) in the Hamiltonian.
The time-dependent Schrödinger equation is a first order differential equa-
tion in time, therefore, the initial value Ψ(r, t0 ) must be given. In the present
case, it will always be the ground-state.

2.2.1 The Runge-Gross theorem

The theorem was published by Erich Runge and Eberhard K. U. Gross in
1984 [22]. It states that two densities n(r, t) and n0 (r0 , t0 ) evolving from a
common initial state Ψ(r, t0 ) under the influence of two potentials Vext (r, t)
and Vext0 (r, t), both Taylor-expandable about the initial time t , eventually
0
differ if the potentials differ by more than a purely time-dependent function,
0 (r, t) 6= C(t). So, there is a one to one correspondence between
Vext (r, t)−Vext
the time-dependent external potential, Vext (r, t), and the time-dependent
electron density, n(r, t) for a fixed initial state [22] [21].

2.2.2 Time-Dependent Kohn-Sham Theory

According to the Runge-Gross theorem there is a one to one correspondence
between the potential and the density. This can be seen as a generalization
of the Hohenberg- Kohn theorem for electronic ground states. The density
of the interacting system can be reproduced by the non-interacting time-
dependent Kohn-Sham orbitals
occ
X
n(r, t) = |Φi (r, t)|2 , (2.15)
i

where the orbitals Φi (r) satisfy the time-dependent Kohn-Sham equations

∇2
 
∂
i Φi (r, t) = − + VKS [n](r, t) Φi (r, t). (2.16)
∂t 2
16 CHAPTER 2. THEORY

The Kohn-Sham potential is conventionally separated in the following way

VKS [n](r, t) = Vext (r, t) + VHartree [n](r, t) + Vxc [n](r, t). (2.17)

With respect to Kohn-Sham DFT, all the potentials become here time-
dependent.
The potential Vxc is a functional of the the density n(r, t). This means that
in principle to obtain Vxc at given r̄ and t̄ one needs to know the density at
all the points r and times t < t̄. The exact expression of Vxc as a functional
of the density is unknown. At this point we are obliged to perform an
approximation [21][23][24].

2.2.3 xc functionals
Adiabatic approximations In adiabatic approximations, we neglect the
dependence on the past. The approximate Vxc (r, t) only depend on the
static [n](r) evaluated at n = n(r, t) to obtain
density at time t. We can use Vxc
adiabatic (r, t).
Vxc
By inserting the LDA functional in Vxc adiabatic (r, t) we obtain the so-called

adiabatic local density approximation (ALDA)

ALDA HEG
Vxc (r, t) = Vxc (n)|n=n(r,t) . (2.18)

The ALDA assumes that the xc potential at the point r, and time t is equal
to the xc potential of a (static) homogeneous-electron gas (HEG) of density
n(r, t).

2.2.4 DFT vs TDDFT

We now compare DFT with TDDFT:

1. DFT is based on the Hohenberg-Kohn theorem while, TDDFT is based

on the Runge-Gross theorem. They both establish a one-to-one corre-
spondence between densities and external potentials. The key variable
is in both cases the density. Additionally, in TDDFT the initial con-
ditions must be given.

2. In DFT, the ground-state of the system can be determined through

the minimization of the total energy functional

hΦ|Ĥ|Φi = E[n]. (2.19)

In TDDFT, there is no variational principle on the basis of the total

energy for it is not a conserved quantity. There exists the quantum
2.2. TIME DEPENDENT DENSITY FUNCTIONAL THEORY 17

mechanical action
Z t1
dthΦ(t)|i∂t − Ĥ(t)|Φ(t)i = A[n]. (2.20)
t0

Equating the functional derivative of Eq. (2.20) in terms of n(r, t)

to zero we arrive at the time-dependent Schrödinger equation. We
can therefore solve the time-dependent problem by calculating the
stationary point of the functional A[n].
In DFT, equating the functional derivative of Eq. (2.19) in terms
of n(r) to zero we arrive the minimum of E[n]. In TDDFT, is no
“minimum principle”, but only a “stationary principle”.

3. In
PNDFT,KSthe density just depends on the space coordinate n(r) =
2
i=1 |Φi (r)| . In TDDFT, the density depends on the time and
space coordinates n(r, t) = N KS 2
P
i=1 |Φi (r, t)| .
The same difference holds for the potentials. VKS in DFT
n(r0 )
Z
δExc [n]
VKS (r) = Vext (r) + dr0 0
+ , (2.21)
|r − r | δn(r)
while in TDDFT
n(r0 , t) δAxc [n]
Z
VKS (r, t) = Vext (r, t) + dr0 + . (2.22)
|r − r0 | δn(r, t)

4. In DFT, the static Schrödinger equation is a second order differen-

tial equation in the space coordinates. It can be formulated as an
eigenvalue problem.

HKS (r)ΦKS KS KS
i (r) = εi Φi (r). (2.23)

In contrast, in TDDFT, the time-dependent Schrödinger equation is

a first-order differential equation in the time coordinate. The wave-
function (or the density) thus depends on the initial state, which im-
plies that the Runge-Gross theorem can only hold for a fixed initial
state (and that the xc potential depends on that state) [25][23]. In our
case, we will neglect this initial state dependence, because the initial
state will always be the ground state.

2.2.5 Linear response theory

We are interested in calculating electronic excitations of our system. Two
regimes can be observed: If the time-dependent potential in respect to the
internal interactions is weak, it is sufficient to resort to linear-response theory
to study the system. In this way it is possible to calculate, e.g. optical
absorption spectra. On the other hand, if the time-dependent potential is
18 CHAPTER 2. THEORY

strong, a full solution of the Kohn-Sham equations is required. A canonical

example of this regime is the treatment of atoms or molecules in strong
laser fields. In this case, TDDFT is able to describe nonlinear phenomena
like high-harmonic generation, or multi-photon ionization. Here I will limit
the discussion to the linear response aspect, the particular focus being on
electronic excitation [23]. Where the external time-dependent potential is
weak, perturbation theory may prove sufficient to determine the behaviour
of the system. We will focus on the linear change of the density, that allows
us to calculate, e.g., the optical absorption spectrum.
We start with a system in its ground state and assume that for t < t0 the
time-dependent potential is zero. At t0 add a small perturbation δVext . Due
to this perturbation the density of the system evolves differently than in the
unperturbed system which would remain unchanged. Since the perturbation
is small the change in the density is also small

n(r, t) = ngs (r) + δn(r, t) (2.24)

The first order variation of the density in frequency space is

Z
δn(r, ω)
δn(r, ω) = d3 r0 |n δVext (r0 , ω). (2.25)
δVext (r0 , ω) gs
| {z }
χ(r,r0 ,ω)

The quantity χ is the linear density-density response function of the system.

We recall that in the time-dependent Kohn-Sham framework, the density
of the interacting system of electrons is obtained from a fictitious system of
non-interacting electrons. The Kohn-Sham system has to produce the same
change in the density by changing the Kohn-Sham potential, i.e.
Z
δn(r, ω)
δn(r, ω) = d3 r0 |n δVKS (r0 , ω). (2.26)
δVKS (r0 , ω) gs
| {z }
χKS (r,r0 ,ω)

Note that the response function that enters Eq. (2.26), χKS , is the density
response function of a system of non-interacting electrons and is, conse-
quently, much easier to calculate than the full interacting χ. In terms of the
unperturbed stationary Kohn-Sham orbitals it reads :
∞
X Φj (r)Φ∗j (r0 )Φk (r0 )Φ∗k (r)
χKS (r, r0 , ω) = lim (fk − fj ) , (2.27)
η→0+ ω − (j − k ) + iη
jk

where fm is the occupation number of the m orbital in the Kohn-Sham

ground-state. The δVKS that enters Eq. (2.26) can be calculated explicitly
from the definition of the Kohn-Sham potential

δVKS (r, t) = δVext (r, t) + δVHartree (r, t) + δVxc (r, t). (2.28)
2.2. TIME DEPENDENT DENSITY FUNCTIONAL THEORY 19

It is useful to introduce the exchange-correlation kernel, fxc , defined by

δVxc (r, ω)
fxc (r, r0 , ω) = , (2.29)
δn(r0 , ω)

and to remind that

Z
1
δVHartree (r, ω) = dr0 δn(r0 , ω). (2.30)
|r − r0 |

Combining the previous results, in frequency space we arrive at:

Z
δn(r, ω) = d3 r0 χKS (r, r0 , ω)δVext (r0 , ω)
Z  
3 0 1
3
+ d r1 d r χKS (r, r1 , ω) + fxc (r1 , r , ω) δn(r0 , ω).
0
|r1 − r0 |
(2.31)

From Eq. (2.25) and Eq. (2.31) follows the relation

χ(r, r0 , ω) = χKS (r, r0 , ω)+

Z Z  
1
3 3
d r1 d r2 χKS (r, r1 , ω) + fxc (r1 , r2 , ω) χ(r2 , r0 , ω).
|r1 − r2 |
(2.32)

This equation is a formally exact representation of the linear density re-

sponse in the sense that, if we possessed the exact fxc a solution of (2.32)
would yield the response function, χ, of the interacting system. If ω is equal
to a transition frequency of our system the change in the density is dramatic
and χ has a pole. χKS has its instead poles at the transition frequencies of
the KS system.
The main ingredient in linear response theory is the xc kernel. fxc ,
as expected, is a very complex quantity that includes - or, in other words,
hides - all non-trivial many-body effects. Many approximate xc kernels have
been proposed in the literature over the past years. The most ancient, and
certainly the simplest is the ALDA kernel
ALDA
fxc (rt, r0 t0 ) = δ(r − r0 )δ(t − t0 )fxc
HEG
(n)|n=n(r) , (2.33)

where

HEG d HEG
fxc (n) = V (n) (2.34)
dn xc
is just the derivative of the xc potential of the homogeneous electron gas.
The ALDA kernel is local both in the space and time coordinates [23].
20 CHAPTER 2. THEORY

2.2.6 Casida’s equations

Casida has reformulated the calculation of the poles (namely the transition
energies) of the response function χ into a generalized Hermitian eigenvalue
problem [26].
Casida showed that, for real orbitals, finding the poles of χ is equivalent to
solving the eigenvalue problem:
X
Ω̃qq0 (ω)Vq0 = ω̃q2 Vq , (2.35)
q0

where q is a double index, representing a transition from occupied KS orbital

j to unoccupied KS orbital k, ωq = εk − εj , and ϕq (r) = Φj (r)Φk (r). the
matrix is
√
Ω̃qq0 (ω) = δqq0 ωq2 + 2 ωq ωq0 hq|fHxc (ω)|q 0 i, (2.36)

where
Z Z
0
hq|fHxc (ω)|q i = 3
d r d3 r0 ϕq (r)fHxc (r, r0 , ω)ϕq0 (r0 ). (2.37)

In this equation, fHxc is the Hartree-exchange-correlation kernel,

1/|r − r0 | + fxc (r, r0 , ω). In the special case of an adiabatic approximation,
these equations are a straightforward matrix equation [27].
The eigenvalues ω̃q of Eq. (2.35) yields the excitation energies and the
eigenvectors of the Casida’s equation Vq can be used to extract the strength
of the response to the external field and to obtain the cross-section tensor,
which is proportional to imaginary part of polarizability:
4πω
σ(ω) = Imα(ω). (2.38)
c
And finally the optical absorption spectrum can be given as the “strength
function”, which is [28]:

cσ(ω)
S(ω) = . (2.39)
2π 2

2.2.7 Electronic excitations by means of time-propagation

Equivalently, the calculation of the electronic excitations such as the dy-
namical polarizability can be performed by propagating in real time the
Kohn-Sham equation. This methodology does not require the calculation of
unoccupied KS states, and scales well with the size of the system, and is thus
our preferred scheme for large systems. Let us recall the essentials of this
formulation. We will restrict hereafter to electrical dipole perturbations:

δVext (r, ω) = −xj k(ω). (2.40)

2.2. TIME DEPENDENT DENSITY FUNCTIONAL THEORY 21

This defines an electrical perturbation polarized in the direction j : δE(ω) =

k(ω)êj . The response of the system dipole moment in the i direction
Z
δhX̂i i(ω) = d3 rxi δn(r, ω) (2.41)

is then given by:

Z Z
δhX̂i i(ω) = −k(ω) 3
d r d3 r0 xi χ(r, r0 , ω)x0j . (2.42)

We may define the dynamical dipole polarizability αij (ω) as the quotient
of the induced dipole moment in the direction i with the applied external
electrical field in the direction j , which yields:
Z Z
αij = − d3 r d3 r0 xi χ(r, r0 , ω)x0j . (2.43)

The dynamical polarizability elements may then be arranged to form a

second-rank symmetric tensor, α(ω). The cross-section tensor is propor-
tional to its imaginary part:
4πω
σ(ω) = Imα(ω). (2.44)
c
We consider a sudden external perturbation at t = 0 (delta function in
time), which means k(ω) = k, equal for all frequencies. This perturbation
is applied along a given polarization direction, say êj . By propagating the
time-dependent Kohn-Sham equations, we obtain the density δn(r, ω) and
δhX̂i i(ω) through Eq. (2.41). The polarizability element αij may then be
calculated via [29]:

δhX̂i i(ω)
Z
1
αij = − =− d3 rxi δn(r, ω). (2.45)
k k
Chapter 3

Results

3.1 Ground state properties

3.1.1 Convergence study
We want to solve Kohn-Sham equation (2.8)
 
1 2
− ∇ + Vion-electron (r) + VH (r) + Vxc (r) Φi (r) = εi Φi (r). (3.1)
2

In principle the quantities that appear in Eq. (3.1) are defined for a contin-
uous set of infinite points. However, we cannot deal with an infinite number
of points on a computer. To implement Eq. (3.1) on a computer we have two
possibilities. We can either expand the functions on a basis and truncate
the expansion in a finite number of basis elements.
Or we can represent the wavefunction on a discrete and finite grid of points.
In both cases we introduce an error which we can control by doing a conver-
gence analysis. Octopus uses the grid representation. So we need to check
that the grid is accurate with respect to its spacing and its extension, and
at the same time we should balance the accuracy with the time that the
calculation takes to run.
So we have to study the convergence of the results with respect to two pa-
rameters: first the grid spacing (i.e. the distance between mesh points) and
second the radius of the simulation box that contains the molecule.

Convergence with respect to grid spacing

As I said before, we need to check that the grid is accurate for the spacing.
And here spacing means the spacing between the points in the mesh.
We run the Octopus for different spacing and for each of them we calculate
the total energy of the molecules.
For pentacene we have the results in Fig. (3.1). We can see that for
spacings smaller than 0.12 Å, we have very small error. For example the

23
24 CHAPTER 3. RESULTS

Convergence with spacing for pentacene

−3638
Convergence with spacing
−3640
−3642
−3644

Energy [eV]
−3646
−3648
−3650
−3652
−3654
−3656
−3658
0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0.26 0.28
Spacing [Å]

Figure 3.1: Convergence with spacing for pentacene

difference between 0.12 Å and 0.10 Å spacing is about 0.18 eV. It is very
small with respect the energy: only 0.005% of the energy. And in addition,
the time of the calculation is important. So, we choose 0.12 Å spacing for
calculation of pentacene.
We do the same calculation for picene. We can see the result in Fig. (3.2).
We can see that the suitable spacing for picene is the same as for pentacene.

Convergence with spacing for picene

−3639
Convergence with spacing

−3640

−3641
Energy [eV]

−3642

−3643

−3644

−3645
0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22
Spacing [Å]

Figure 3.2: Convergence with spacing for picene

We can compare the results for picene and pentacene in Fig. (3.3). The
spacing just depends on a type of atoms that compose the molecules. In fact
it depends on the pseudopotential. So for picene and pentacene we have the
same suitable spacing that is equal to 0.12 Å.
3.1. GROUND STATE PROPERTIES 25

Convergence with spacing for picene and pentacene

−3638
picene
−3640 pentacene
−3642
−3644
Energy [eV]

−3646
−3648
−3650
−3652
−3654
−3656
−3658
0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0.26 0.28
Spacing [Å]

Figure 3.3: Convergence with spacing for picene and pentacene

Convergence with respect to box radius

For our calculation we need a finite space around the molecule, for this we
choose a box and we should sure that the size of the box is suitable. So, we
should try to do the calculation and try to converge with respect to radius
of the simulation box.
We run the Octopus for different size of the simulation box and for each of
them we calculate the total energy of the molecules.
For pentacene we have the results in Fig. (3.4). We can see that for radius

Convergence with radius of box for pentacene

−3644
Convergence with radius
−3646

−3648
Energy [eV]

−3650

−3652

−3654

−3656

−3658
2 2.5 3 3.5 4 4.5 5
Radius [Å]

Figure 3.4: Convergence with radius of box for pentacene

bigger than 4 Å, we have very small error. For example the difference
between 4 Å and 5 Å radius is about 0.01 eV. It is very small with respect
the energy: only 3 × 10−4 % of the energy. And in addition, the time of the
calculation is important. So, we choose 4 Å radius for the calculation on
pentacene.
26 CHAPTER 3. RESULTS

We do the same calculation for picene. We can see the result in Fig. (3.5).
We can see that the suitable radius for picene is the same as for pentacene.

Convergence with radius of box for picene

−3624
Convergence with radius
−3626

−3628

−3630
Energy [eV]

−3632

−3634

−3636

−3638

−3640
2 2.5 3 3.5 4 4.5 5
Radius [Å]

Figure 3.5: Convergence with radius of box for picene

We can compare the results for picene and pentacene in Fig. (3.6). The

Convergence with radius of box for picene and pentacene

−3620
picene
pentacene
−3625

−3630

−3635
Energy [eV]

−3640

−3645

−3650

−3655

−3660
2 2.5 3 3.5 4 4.5 5
Radius [Å]

Figure 3.6: Convergence with radius of box for picene and pentacene

radius of the simulation box just depends on the type of atoms that compose
the molecules. In fact it depends on the pseudopotential. So for picene and
pentacene we have the same suitable radius that is equal to 4 Å.

3.1.2 Structural optimisation

First of all I obtained the molecular structures of picene and pentacene by
assembling 5 benzene rings, in a zigzag and a linear configuration, respec-
tively. I tried to optimise these structures to find the ground state.
As I said before in section (2.1), in order to find the ground state of a
system we should solve the Schrödinger equation (2.1). According to the
3.1. GROUND STATE PROPERTIES 27

Born-Oppenheimer approximation, for each fixed ionic configuration, elec-

trons are in their ground state. Within the DFT framework we can calculate
the ground-state electronic energy E(R1 , . . . , RM ), where the ionic positions
appear as parameters. We want to minimize E as a function of ionic posi-
tions in order to find the ground-state geometry of the molecules. Different
algorithms exist to find the minimum of E(R1 , . . . , RM ).

I found that the minimisation was very slow. For this reason, we decided
to use the structures that appear in the (American) National Institute of
Standard and Technology (NIST) website.
From (http://webbook.nist.gov/cgi/cbook.cgi?Str3File=C135488) for pen-
tacene and for picene from (http://webbook.nist.gov/cgi/cbook.cgi?ID=213-
46-7).
In Fig. (3.7(a)) and Fig. (3.7(b)) we show the optimisation structures of
pentacene and picene.
As can be inferred from Fig. (3.7(a)), pentacene is symmetric with respect

Figure 3.7: Optimisation structure (a) optimisation structure of pentacene

(b) optimisation structure of picene (c) structure of benzene
28 CHAPTER 3. RESULTS

to the xy, xz, and yz planes. While picene (see Fig. 3.7(b)) is symmetric
only with respect to the yz and xy plane. Thus picene has fewer symmetries
than pentacene.

Now, we compare C–C and C–H bond lengths in pentacene, picene, and
benzene.
All C–H bond lengths in picene, pentacene, and benzene are equal to 1.10
Å. All C–C bond lengths in benzene are equal to 1.40 Å, but in picene and
pentacene we can see different bond lengths. For instance, we compare the
C–C bond lengths in the middle rings for pentacene and picene. One could
expect that the bond lengths of C–C atoms 6–7 in picene is equal to 7–16
in pentacene, we can see the bond length of 6–7 in picene is equal to 1.36 Å
but 7–16 in pentacene is equal to 1.43 Å. Another example is bond length
of C atoms 5–6 in picene and 7–6 in pentacene. We can see the bond length
of 5–6 in picene is equal to 1.42 Å and 7–6 in pentacene is equal to 1.40 Å.
So, we can see that we have different bond lengths in pentacene and picene.

3.1.3 Kohn-Sham orbitals and energy levels

We Know that we have 22 Carbon atoms and 14 Hydrogen atoms in both
picene and pentacene. Each Carbon atom has 4 valence electrons and each
Hydrogen atom has one valence electron. So we have 102 valence electrons.
According to the Pauli exclusion principle, in each state we have 2 electrons.
So, we have 51 occupied orbitals.
Since π states are linear combinations of pz atomic orbitals, their number is
set by the number of carbon atoms. So that for the two molecules we have
11 π states. The other 40 states present σ character. We know that σ states
have lower energy than π states. We write the results of the energy of some
states in Table (3.1). From i = 1 to i = 40 we have the σ states, from i = 41
to i = 51 we have π states, from i = 52 to i = 62 we have π ∗ states, and the
others are σ ∗ states.

In Table (3.2) we write the energy width of the Kohn-Sham levels that
belong to each group of states. While the σ states have the different energy
width (13.58 eV in pentacene and 13.92 eV in picene), the σ ∗ states present
smaller and different energy width (4.25 eV in pentacene and 4.42 eV in
picene). The energy width of π states is larger for pentacene than for picene.
It is 3.71 eV for π states and 3.95 eV for π ∗ states in pentacene, while it is
2.76 eV for π states and 4.42 eV for π ∗ states in picene.
In Table (3.3) we compare the energy gap between the highest occu-
pied molecular orbital (HOMO) and lowest unoccupied molecular orbital
(LUMO) in picene and pentacene. In our calculation the HOMO–LUMO
energy gap for pentacene is 1.13 eV. We compare it with the data that we
find in literature.
3.1. GROUND STATE PROPERTIES 29

States Energy of pentacene (eV) Energy of picene (ev)

i=1 -22.20 -22.32
i=40 -8.62 -8.40
i=41 -8.34 -8.30
i=50 -5.75 -5.73
i=51 -4.63 -5.54
i=52 -3.51 -2.45
i=53 -2.28 -2.41
i=61 0.13 0.52
i=62 0.44 0.53
i=63 0.49 0.53
i=64 0.88 0.56
i=102 4.74 4.95

Table 3.1: The Energy of some states of pentacene and picene

Pentacene Picene
σ band(eV) 13.58 13.92
π band(eV) 3.71 2.76
π ∗ band(eV) 3.95 2.98
σ ∗ band(eV) 4.25 4.42

Table 3.2: The Energy band of σ, σ ∗ , π, and π ∗ state

Pentacene Picene
HOMO–LUMO energy gap (eV) 1.13 3.09

Table 3.3: HOMO–LUMO energy gap for pentacene and picene

30 CHAPTER 3. RESULTS

The agreement with other calculations is very good: in Ref. [30] the authors
find 1.1 eV and in Ref. [31] they find 1.12 eV. From this comparison, we
understand that our calculation is correct.
But when we compare this value with experimental data, we see that this
value is about 80% smaller than the experimental value, 5.22 eV [30]. The
underestimation of the energy gap is due to two reasons. First, because the
Kohn-Sham orbitals have no physical meaning. Second, because when we
model Exc using a local density approximation where the system is locally
approximated by the homogeneous electron gas of density n = n(r).
The same observation holds for picene. For picene in our calculation the
HOMO–LUMO energy gap is 3.09 eV. In Ref. [31] the authors report a value
of 2.96 eV. So, again, we conclude that our calculation is correct.

Even though the Kohn-Sham HOMO–LUMO energy gaps underestimate

the experimental ones, we can use these results to compare picene with
pentacene. Here we can see that the picene HOMO–LUMO gap is much
larger than in pentacene.
We show the Kohn-Sham HOMO and LUMO orbital isosurfaces for
picene and pentacene in Fig. (3.8(a),3.8(b),3.9(a),3.9(b)), where we take
only the positive region of the wavefunctions. From the shape of the wave-
functions we note that the corresponding charge is localized out of the plane
of the molecules. This means that these states present π character and
are responsible for the out-of-plane C–C bonds. In particular comparing
Figs. (3.8(a)) and (3.9(a)) with Figs. (3.8(b)) and (3.9(b)) we clearly iden-
tified the HOMO and LUMO orbitals as bonding π states and anti-bonding
π ∗ states respectively.
The higher symmetry of the pentacene molecule clearly reflects on the
shape of the wave functions. In fact, comparing Fig. (3.9(a)) with Fig. (3.9(b))
we find that in pentacene both HOMO and LUMO orbitals are symmet-
ric respect to the yz plane while respect to the xz plane they are anti-
symmetric and symmetric respectively. On the other hand, for picene (see
Fig. (3.8(a)) and (3.8(b))) the HOMO and LUMO orbitals are symmetric
and anti-symmetric respect to the yz plane respectively but they do not
have any symmetry respect to the xz plane. Moreover being the HOMO
(LUMO) orbitals as well as all the π (π ∗ ) states linear combinations of pz
atomic orbitals they are anti-symmetric respect to the plane of the molecule
(xy plane) in both picene and pentacene.
We show in Figs. ((3.9(c)), (3.8(c)), (3.9(d)), (3.8(d))) the orbital iso-
surface (positive region) for the highest occupied σ state and the lowest
unoccupied σ ∗ state for both picene and pentacene. Being these states re-
lated to the s, px and py atomic orbitals, their charge distribution is strongly
localized on the plane of the molecules resulting in the formation of strong
C-H bonds and in-plane C-C bonds. A comparison of Fig. (3.8(c)) and
(3.9(c)) with Fig. (3.9(d)) and (3.8(d)) clearly reveal the bonding (anti-
3.1. GROUND STATE PROPERTIES 31

bonding) character of the σ (σ ∗ ) states. Finally, contrarily to the π (π ∗ )

states, σ (σ ∗ ) states are symmetric respect the molecular plane.

Figure 3.8: Kohn-Sham orbital for picene (a) Kohn-Sham orbital of the
HOMO (b) Kohn-Sham orbital of the LUMO (c) Kohn-Sham orbital of the
last σ state (d) Kohn-Sham orbital of the first σ ∗ state
32 CHAPTER 3. RESULTS

Figure 3.9: Kohn-Sham orbital for pentacene (a) Kohn-Sham orbital of the
HOMO (b) Kohn-Sham orbital of the LUMO (c) Kohn-Sham orbital of the
last σ state (d) Kohn-Sham orbital of the first σ ∗ state
3.2. ABSORPTION SPECTRA FROM TIME PROPAGATION 33

3.2 Absorption spectra from time propagation

According to the section (2.2.5), we know that if the time-dependent poten-

tial is weak with respect to the internal interactions, we are in the regime
of the linear-response approximation.
As we said before in section (2.2.7), our time-dependent external potential
is a sudden external perturbation at t = 0 (delta function in time). So,
first of all we should check the strength of the delta perturbation to be sure
that is not too strong and we can use the linear-response approximation.
We call the strength of the delta perturbation TDDeltaStrength. We do the
calculation for the TDDeltaStrength is equal to 0.005 (1/Å) and 0.01 (1/Å)
for pentacene in one direction. We show the results in Fig. (3.10). We see
that both of them have the same absorption spectrum, so we decided to use
0.01 (1/Å) strength.

Absorption spectrum of pentacene

8
TDDeltaStrength=0.01 [1/Å2]
7 TDDeltaStrength=0.005 [1/Å2]

5
Absorption [Å ]
2

−1
0 5 10 15 20
Energy [eV]

Figure 3.10: The absorption spectrum for pentacene in one direction for
TDDeltaStrength = 0.005 (1/Å) and TDDeltaStrength = 0.01 (1/Å)

3.2.1 Convergence study

We try to converge the absorption spectrum with respect to the calculation

parameters.
Here we have two important parameters: the first is the time-step for the
time propagation (TDTimeStep) while the second one is the number of time-
propagation steps (TDMaximumIter). TDMaximumIter is equal to the ratio
between the total propagation time (T) and TDTimeStep.
So, we need to converge the absorption spectrum with respect to the TD-
TimeStep and total propagation time.
34 CHAPTER 3. RESULTS

Convergence with respect to time step

We should choose the suitable length of each time step. Longer time steps
are desirable to assure a shorten computational propagation time, but they
limits to the maximum frequency we will able to observe. So, we should find
the largest value that it is possible without the evolution becomes unstable.
For this purpose, we should run the calculation just with the time-evolution
of the system, departing from the ground-state, under the influence of no
external perturbation. As a consequence if the time step is correct, the
electronic system evolve in such a way that the total energy does not change.
We do this calculation with the total time propagation equal to 0.1 (}/eV )
and time step equal to 0.002 (}/eV ), 0.0015 (}/eV ), and 0.001 (}/eV ). We
found that the total energy of the system is conserved just for the time step
equal to 0.001 (}/eV ). So, we chose the time step equal to 0.001 (}/eV ) for
our calculation.
For checking that this time step is correct, we did the calculation also
in presence of an external perturbation. We did the calculation with ex-
ternal perturbation in one direction with total time propagation equal to
15 (}/eV ), and compare the result for time step equal to 0.001 (}/eV ) with
that for time step equal to 0.0005 (}/eV ).
In Fig. (3.11) we can see the absorption spectrum for two time steps for
pentacene and in Fig. (3.12) for picene. In these two figures, we plot the
strength function Eq. (2.39).
We can see that we have the same results for both time steps. From these

Absorpthin spectrum of pentacene

16000
TDTimeStep=0.0005 [−h/eV]
14000 TDTimeStep=0.001 [−h/eV]

12000
StrengthFunction [1/eV]

10000

8000

6000

4000

2000

−2000
0 5 10 15 20
Energy [eV]

Figure 3.11: The absorption spectrum for two time step

(TDTimeStep=0.001 (}/eV ) and TDTimeStep=0.0005 (}/eV )) for
pentacene

results, we understand that our time step is correct.

3.2. ABSORPTION SPECTRA FROM TIME PROPAGATION 35

Absorpthin spectrum of picene

8000
TDTimeStep=0.0005 [−h/eV]
7000 TDTimeStep=0.001 [−h/eV]

StrengthFunction [1/eV] 6000

5000

4000

3000

2000

1000

−1000
0 5 10 15 20
Energy [eV]

Figure 3.12: The absorption spectrum for two time step

(TDTimeStep=0.001 (}/eV ) and TDTimeStep=0.0005 (}/eV )) for
picene

Convergence with respect to total propagation time

Increasing the total propagation time (T) reduces the width of the peaks.
We do the calculation for time step equal to 0.001 (}/eV ) and T=10 (}/eV ),
T=20 (}/eV ), T=25 (}/eV ), T=30 (}/eV ), T=35 (}/eV ), and T=40 (}/eV ).
In Fig. (3.13) for pentacene and in Fig. (3.14) for picene, we plot the average
absorption coefficient T r σ/3 for these different total propagation time. As
we can see, when we increase the total propagation time, the width of the
peaks reduces. Moreover the increase of total propagation time causes the
appearance of new structures in the spectra indicating that we are not in
convergence up to about T=40 (}/eV ). We checked that above this value
the spectra does not change any more.
Thus, we find the suitable total propagation time equal to 40 (}/eV ).

Analysis of the absorption spectra from time propagation

We do the calculation with total propagation time equal to 40 (}/eV ) and

the time step equal to 0.001 (}/eV ) and we applied the external pertur-
bation in three directions (x, y, and z). According to the section (2.2.7)
by propagating the time-dependent Kohn-Sham equations, we obtained the
density δn(r, ω) and δhX̂i i(ω) through Eq. (2.41) and finally we obtained
the spectra.

First we start with the analysis of the pentacene spectrum. In Fig. (3.15(a))
we show the absorption spectrum for pentacene obtained plotting the av-
erage absorption coefficient while in Fig. (3.15(b)) we show the absorption
spectrum for three different direction of polarizations obtained from the
36 CHAPTER 3. RESULTS

Absorption spectrum of pentacene

20
T=10 [−h/eV]
18 T=20 [−h/eV]
T=25 [−h/eV]
16 T=30 [−h/eV]
14 T=35 [−h/eV]
T=40 [−h/eV]

Absorption [Å ]
12

2
10
8
6
4
2
0
−2
0 2 4 6 8 10 12
Energy [eV]

Figure 3.13: The absorption spectrum for (T=10 (}/eV ), T=20 (}/eV ),
T=25 (}/eV ), T=30 (}/eV ), T=35 (}/eV ), and T=40 (}/eV )) for pen-
tacene

strength function.
According with the distribution of the Kohn-Sham energy levels shown in
Table (3.2) and Table (3.3) we can distinguish two main energy windows
in the absorption spectra: a low energy region between 0 eV and 4.84 eV
where only π−π ∗ transitions are allowed by energy conservation and a higher
energy one (above 4.84 eV) involving both π − π ∗ and π − σ ∗ (σ − π ∗ ) transi-
tions. As can be inferred from Fig. (3.15(b)) the lower energy region of the
spectrum is dominated by electronic transitions induced by in-plane dipole
moment, in particular in the x direction. While the out-of-plane component
(z component) of the dipole moment gives a remarkably contribution only at
energy higher than 10 eV where only π −σ ∗ (σ −π ∗ ) transitions are involved.
This behaviour can be explained in terms of the symmetry properties of the
electronic wave functions discussed in the previous section. In fact as we
have seen previously π (π ∗ ) and σ (σ ∗ ) states are respectively antisymmetric
and symmetric respect to the xy plane. This means that the only allowed
electronic transitions in the dipole approximation are π − π ∗ and σ − σ ∗ for
polarization direction belonging to the plane of the molecule and π − σ ∗ and
σ − π ∗ for the direction perpendicular to the plane. Interestingly, we note
that, although the main contribution to the lower energy region of the spec-
trum arises from the x component of the polarization (see for example the
feature at 3.98 eV), the onset of the absorption (1.58 eV) involving mainly
HOMO-LUMO transition is related only to the y component. In fact being
HOMO and LUMO orbitals both symmetric respect to the yz plane, the
HOMO-LUMO transition is forbidden for dipole moment along the x axis.
On the other hand it is allowed for dipole along the y axis being HOMO and
LUMO orbitals respectively antisymmetric and symmetric respect to the xz
plane.
3.2. ABSORPTION SPECTRA FROM TIME PROPAGATION 37

Absorption spectrum of picene

9
T=10 [−h/eV]
8 T=20 [−h/eV]
T=25 [−h/eV]
7 T=30 [−h/eV]
T=35 [−h/eV]
6 T=40 [−h/eV]
Absorption [Å ]
2

5
4
3
2
1
0
−1
0 2 4 6 8 10 12
Energy [eV]

Figure 3.14: The absorption spectrum for (T=10 (}/eV ), T=20 (}/eV ),
T=25 (}/eV ), T=30 (}/eV ), T=35 (}/eV ), and T=40 (}/eV )) for picene

In the following we focus on picene.

In Fig. (3.16(a)) we see the absorption spectrum for picene obtained eval-
uating the average absorption coefficient while in Fig. (3.16(b)) we see the
absorption spectrum for three different directions of polarizations obtained
from the strength function.
As for pentacene, also for picene, according to the results shown in Ta-
ble (3.2) and Table (3.3), we can identify two main energy windows in the
absorption spectrum: a lower energy region between 0 eV and 5.85 eV re-
lated to π − π ∗ transitions and a higher energy region (above 5.85 eV)
involving both π − π ∗ and π − σ ∗ (σ − π ∗ ) transitions. As for pentacene
the symmetry of the electronic wave functions reflects on the structure of
the absorption spectrum, so that the lower energy region of the spectrum
is mainly related to the in plane component of the polarization while the
higher energy one to the z component. Contrarily to what happens in pen-
tacene, we find that for picene the onset of the absorption (3.4 eV) presents
contributions from both the x component and the y component of the dipole
moment. In fact, in the case of picene the HOMO and LUMO orbitals are
respectively symmetric and antisymmetric respect to the yz plane so that
HOMO-LUMO transitions are allowed in dipole approximation for dipole
moment along the x axis. Moreover since for picene the xz plane is not a
symmetry plane we do not have any restriction imposed by the symmetry
on the transitions induced by the dipole moment oriented along the y axis.
38 CHAPTER 3. RESULTS

20
(a)
15

Absorption [Å ]
2
10

0
0 5 10
Energy [eV]
40000

StrengthFunction [1/eV]
x direction
30000 (b) y direction
z direction
20000

10000

0
0 5 10
Energy [eV]

Figure 3.15: The absorption spectrum for pentacene (a)The average absorp-
tion (b)The strength function for three direction (x,y,z)

3.3 Absorption spectra from Casida equation

Convergence study

We have to converge the absorption spectrum with respect to the calculation

parameters.
According to the section (2.2.6) we know that the most important parameter
that we have here is the number of unoccupied states. So, we have to
increase the number of unoccupied states until our absorption spectrum is
in convergence. We performed the Casida calculation for pentacene with
the number of unoccupied states equal to 51, 100, 150, and 200.
We show the results for pentacene in Fig. (3.17). As we can see the spectra
for the number of unoccupied states equal to 150 and 200 are exactly the
same. So in order to have a shorter calculation we chose to do the calculation
with a number of unoccupied states equal to 150.
For picene we performed the calculation for the number of unoccupied
states equal to 51, 100, and 150.
We show the results for picene in Fig. (3.18). As for pentacene we decided
to do the calculation with a number of unoccupied states equal to 150.

Analysis of the casida spectra

As discussed in section (2.2.6), we can obtain the absorption spectra by

solving the eigenvalue problem in Eq. (2.35) (Casida equation). As we know
the eigenvalues of the Casida equation yield the excitation energies while the
corresponding eigenvectors yield the weight of the different electron-hole
pairs contributing to that excitations. In an independent particle picture
(i.e. when fHxc = 0) the Casida equation is already in the diagonal form
3.3. ABSORPTION SPECTRA FROM CASIDA EQUATION 39

8 (a)

Absorption [Å ]
2
6

0
0 5 10
Energy [eV]
20000
StrengthFunction [1/eV]

(b) x direction
15000 y direction
z direction
10000

5000

0
0 5 10
Energy [eV]

Figure 3.16: The absorption spectrum for picene (a)The average absorption
(b)The strength function for three direction (x,y,z)

in the basis of electron-hole Kohn-Sham wave functions so that it exists a

one to one correspondence between excitation energies and the energies of
the electron-hole pairs. However when fHxc 6= 0 the Casida equation is no
more diagonal in this basis since the Hartree and the exchange-correlation
terms cause a mixing of electron-hole transitions. Thus, solving Eq. (2.35)
not only gives access to the absorption spectra but also allows to understand
the nature of the different excitations.
We can see the absorption spectrum with the number of unoccupied
states equal to 150 for pentacene in Fig. (3.19(a)) and picene in Fig. (3.20(a)).
In Fig. (3.19) and Fig. (3.20) we compare the spectra from Casida with
that obtained from time propagation and we find that they are in good
agreement between them.
To understand the role played by the exchange-correlation effects we
compare in Fig.(3.22) and Fig.(3.21) the spectra evaluated with and without
fxc . We see that the effect of fxc is negligible in the first region of the spectra
and causes only a small red shift of the spectra of about 0.1 eV in both picene
(Fig. (3.22)) and pentacene (Fig.(3.21)) at high energy. This means that
in these systems the mixing of electron-hole transitions is mainly related to
the Hartree term.
Thus in the following we focus on the analysis of the main features of the
spectra looking to the eigenstates of the Casida equation solved neglecting
the fxc term. First we analyse the results for pentacene.
As I said before the onset of the absorption is at about 1.58 eV. We find
from the solution of Casida equation that this excitation is related to the
mixing of two transitions: the dominant contribution arises from HOMO-
LUMO transition with energy difference of about 1.13 eV while the other
one is related to the transition from the level 50 to the level 53 with energy
40 CHAPTER 3. RESULTS

Absorption spectrum of pentacene

6000
NumberUnoccStates=51
NumberUnoccStates=100
5000 NumberUnoccStates=150
NumberUnoccStates=200

StrengthFunction [1/eV]
4000

3000

2000

1000

0
0 2 4 6 8 10 12
Energy [eV]

Figure 3.17: The absorption spectrum for Casida equation with the number
of unoccupied states are equal to 51, 100, 150, and 200 for pentacene

Absorption spectrum of picene

4500
NumberUnoccStates=51
4000 NumberUnoccStates=100
NumberUnoccStates=150
3500
StrengthFunction [1/eV]

3000

2500

2000

1500

1000

500

0
0 2 4 6 8 10 12
Energy [eV]

Figure 3.18: The absorption spectrum for Casida equation with the number
of unoccupied states are equal to 51, 100, and 150 for picene

difference of about 3.47 eV. Thus the Hartree term causes a red shift of the
onset of about 0.45 eV respect to the value set by the HOMO-LUMO gap.
The dominant peak at 3.98 eV is related to the mixing of transitions with
energy from 2.899 eV to 4.98 eV revealing a strong effects of the Hartree
term.
Now we focus on picene.
The onset of the absorption is at about 3.4 eV. We find from the solution of
Casida equation that this excitation is related to the mixing of transitions
with energy from 3.09 eV to 4.184 eV. Thus the Hartree term causes a red
shift of the onset of about 0.31 eV respect to the value set by the HOMO-
LUMO gap (3.09 eV).
The dominant peak at 4.39 eV is related to the mixing of transitions with
3.3. ABSORPTION SPECTRA FROM CASIDA EQUATION 41

6000

StrengthFunction [1/eV]
5000
4000
(a)
3000
2000
1000
0
0 5 10
Energy [eV]
20

(b)
15
Absorption [Å ]
2

10

0
0 5 10
Energy [eV]

Figure 3.19: The absorption spectrum for pentacene (a) absorption spec-
trum from Casida (b) the absorption spectrum from time propagation

energy from 3.09 eV to 6.06 eV. Even in this case the mixing of transitions
due to the Hartree term is very strong.
42 CHAPTER 3. RESULTS

5000

StrengthFunction [1/eV]
(a)
4000

3000

2000

1000

0
0 5 10
Energy [eV]
8
(b)
Absorption [Å ]
2

0
0 5 10
Energy [eV]

Figure 3.20: The absorption spectrum for picene (a) absorption spectrum
from Casida (b) the absorption spectrum from time propagation

Absorption spectrum of pentacene

5000
Without fxc
4500 With fxc

4000
StrengthFunction [1/eV]

3500
3000
2500
2000
1500
1000
500
0
0 2 4 6 8 10 12
Energy [eV]

Figure 3.21: The absorption spectrum for Casida equation with fxc and
without fxc of pentacene
3.3. ABSORPTION SPECTRA FROM CASIDA EQUATION 43

Absorption spectrum of picene

3500
Without fxc
With fxc
3000
StrengthFunction [1/eV]

2500

2000

1500

1000

500

0
0 2 4 6 8 10 12
Energy [eV]

Figure 3.22: The absorption spectrum for Casida equation with fxc and
without fxc of picene
Chapter 4

Conclusion

In this work I studied the optical properties of two aromatic molecules:

pentacene and picene. They are both made of 5 benzene rings. While in
pentacene they are in a linear configuration, in picene they have a zigzag
configuration, as we can see in Fig. (3.7(a) and 3.7(b)). Picene has fewer
symmetries than pentacene.
I used the Octopus code in the framework of density functional theory
for the ground state and time-dependent density functional theory for the
absorption spectra.
The first step has been the calculation of the Kohn-Sham orbitals and their
energy levels.
We have 11 π (π ∗ ) states and 40 σ (σ ∗ ) states. According to the Table (3.2)
we see the σ states have the different energy width (13.58 eV in pentacene
and 13.92 eV in picene), the σ ∗ states present smaller and different energy
width (4.25 eV in pentacene and 4.42 eV in picene). The energy width of π
states is larger for pentacene than for picene. It is 3.71 eV for π states and
3.95 eV for π ∗ states in pentacene, while it is 2.76 eV for π states and 4.42
eV for π ∗ states in picene.
In our calculation the HOMO–LUMO energy gap for pentacene is 1.13 eV,
and when compare it with the data that we find in literature we under-
stand that our calculation is correct. But when we compare this value with
experimental data, we see that this value is about 80% smaller than the ex-
perimental value, 5.22 eV. For picene in our calculation the HOMO–LUMO
energy gap is 3.09 eV. And when compare it with the data of literature,
again, we conclude that our calculation is correct.
We see that the picene HOMO–LUMO gap is much larger than in pen-
tacene.
In Fig. (3.8(a),3.8(b),3.9(a),3.9(b)) we show the Kohn-Sham HOMO and
LUMO orbital isosurfaces for picene and pentacene. The higher symmetry
of the pentacene molecule clearly reflects on the shape of the wave functions.
In fact, comparing Fig. (3.9(a)) with Fig. (3.9(b)) we find that in pentacene

45
46 CHAPTER 4. CONCLUSION

both HOMO and LUMO orbitals are symmetric respect to the yz plane
while respect to the xz plane they are anti-symmetric and symmetric re-
spectively. On the other hand, for picene (see Fig. (3.8(a)) and (3.8(b)))
the HOMO and LUMO orbitals are symmetric and anti-symmetric respect
to the yz plane respectively but they do not have any symmetry respect to
the xz plane. Moreover being the HOMO (LUMO) orbitals as well as all
the π (π ∗ ) states linear combinations of pz atomic orbitals they are anti-
symmetric respect to the plane of the molecule (xy plane) in both picene
and pentacene. In Fig. (3.8(c)) and (3.9(c)) with Fig. (3.9(d)) and (3.8(d))
we see σ and σ ∗ states that they are symmetric respect the molecular plane.
Then we find the absorption spectrum with two ways: first from time
propagation and second from Casida equation. In the spectra that we ob-
tained from time propagation for pentacene (Fig. (3.15)) and for picene
(Fig. (3.16)) we have two main energy windows: a low energy region be-
tween 0 eV and 4.84 eV for pentacene and for picene between 0 eV and 5.85
eV where only π − π ∗ transitions are allowed by energy conservation and a
higher energy one (above 4.84 eV for pentacene and for picene above 5.85
eV) involving both π − π ∗ and π − σ ∗ (σ − π ∗ ) transitions. The lower energy
region of the spectrum is dominated by electronic transitions induced by
in-plane dipole moment. The main contribution to the lower energy region
of the spectrum arises from the x component of the polarization for picene
and pentacene. The onset of the absorption for pentacene involving mainly
HOMO-LUMO transition is related only to the y component, but for picene
the onset of the absorption presents contributions from both the x compo-
nent and the y component of the dipole moment.
We see the spectra from Casida with that obtained from time propagation
are in good agreement. We see that the effect of fxc is negligible in the
first region of the spectra and causes only a small red shift of the spectra
of about 0.1 eV in both picene (Fig. (3.22)) and pentacene (Fig.(3.21)) at
high energy. This means that in these systems the mixing of electron-hole
transitions is mainly related to the Hartree term.
Bibliography

[1] S. R. Forrest, Nature 428, 911 (2004).

[2] M. L. Tiago, J. E. Northrup, and S. G. Louie, Phys. Rev. B 67, 115212
(2003).
[3] M. E. Gershenson, V. Podzorov, and A. F. Morpurgo, Rev. Mod. Phys.
78, 973 (2006).
[4] M. A. Baldo et al., Nature 395, 151 (1998).
[5] R. H. Friend et al., Nature 397, 121 (1999).
[6] P. Peumans, S. Uchida, and S. R. Forrest, Nature 425, 158 (2003).
[7] G. Li et al., Nature Materials 4, 864 (2005).
[8] R. Mitsuhashi et al., nature 464, 08859 (2010).
[9] Y. Kubozono et al., Phys. Chem. Chem. Phys. (2001).
[10] M. Grobosch et al., Phys. Rev. B 74, 155202 (2006).
[11] L. G. Wade, in Organic chemistry, sixth ed., edited by J. Challice and
R. Mullaney (Pearson Prentice Hall, Upper Saddle River, NJ 07458,
United States of America, 2006).
[12] M. Born and J. R. Oppenheimer, Ann. Physik. (1927).
[13] F. Bruneval, Ph.D. thesis, Ecole Polytechnique, 2005.
[14] D. R. Hartree, Proc. Camb. Phil. Soc. (1928).
[15] V. A. Fock, Z. Phys. A 126 (1930).
[16] P. Hohenberg and W. Kohn, Phys. Rev. 136, B864 (1964).
[17] J. T. Chayes, L. Chayes, and M. B. Ruskai, J. Stat. Phys. 38, (1985).
[18] R. Dreizler, in A Primer in Density Functional Theory, edited by M.
Marques, C. Fiolhais, and F. Nogueira (Springer-Verlag, Berlin Heidel-
berg, 2003).

47
48 BIBLIOGRAPHY

[19] M. Gatti, Ph.D. thesis, Ecole Polytechnique, 2007.

[20] W. Kohn and L. J. Sham, Phys. Rev. 140, A1133 (1965).

[21] N. L. Doltsinis, in Computational Nanoscience: Do it yourself ! (PUB-

LISHER, Ruhr-Universität Bochum 44780 Bochum, Germany, 2006),
Chap. Time-dependent density functional theory, pp. 357–373.

[22] E. Runge and E. K. U. Gross, Phy. Rev. Lett. 52, 997 (1984).

[23] M. A. L. Marques, E. K. U. Gross, C. Fiolhais, and F. Nogueira, in

A Primer in Density Functional Theory, edited by C. Marques, M. A.
L. Fiolhais and F. Nogueira (Springer-Verlag, Berlin Heidelberg, 2003),
Vol. 620, Chap. Time-dependent density functional theory, pp. 144–184.

[24] R. van Leeuwen, Phys. Rev. Lett. 82, 3863 (1999).

[25] V. Olevano, Introduction to TDDFT Linear-Response TDDFT in

Frequency-Reciprocal space on a Plane-Waves basis: the DP (Dielectric
Properties) code.

[26] D. Rocca, Ph.D. thesis, Internazionale Superiore di Studi Avanzati,

2007.

[27] K. Burke, The ABC of DFT (PUBLISHER, Department of Chemistry,

University of California, Irvine, CA 92697, 2007).

[28] M. Gatti and G. Onida, Phys. Rev. B 72, 045442 (2005).

[29] A. Castro et al., Phys. Stat. Sol. (b) 243, 2465 (2006).

[30] R. G. Endres et al., Computational Materials Science 29, 362 (2004).

[31] T. Kosugi et al., Journal of the Physical Society of Japan 78, 113704
(2009).
List of Figures

2.1 Schematic representation of a pseudopotential (left, dark curve)

and a pseudo-wavefunction (right, dark curve) along with the
all-electron potential (with the 1/r tail) and wavefunction
(indigo curves). Notice that the all-electron and pseudofunc-
tions are identical beyond the radial cutoff Rc and the pseudo-
functions are smooth outside the core region. . . . . . . . . . 14

3.1 Convergence with spacing for pentacene . . . . . . . . . . . . 24

3.2 Convergence with spacing for picene . . . . . . . . . . . . . . 24
3.3 Convergence with spacing for picene and pentacene . . . . . . 25
3.4 Convergence with radius of box for pentacene . . . . . . . . . 25
3.5 Convergence with radius of box for picene . . . . . . . . . . . 26
3.6 Convergence with radius of box for picene and pentacene . . 26
3.7 Optimisation structure (a) optimisation structure of pentacene
(b) optimisation structure of picene (c) structure of benzene . 27
3.8 Kohn-Sham orbital for picene (a) Kohn-Sham orbital of the
HOMO (b) Kohn-Sham orbital of the LUMO (c) Kohn-Sham
orbital of the last σ state (d) Kohn-Sham orbital of the first
σ ∗ state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.9 Kohn-Sham orbital for pentacene (a) Kohn-Sham orbital of
the HOMO (b) Kohn-Sham orbital of the LUMO (c) Kohn-
Sham orbital of the last σ state (d) Kohn-Sham orbital of the
first σ ∗ state . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.10 The absorption spectrum for pentacene in one direction for
TDDeltaStrength = 0.005 (1/Å) and TDDeltaStrength = 0.01
(1/Å) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.11 The absorption spectrum for two time step (TDTimeStep=0.001 (}/eV )
and TDTimeStep=0.0005 (}/eV )) for pentacene . . . . . . . 34
3.12 The absorption spectrum for two time step (TDTimeStep=0.001 (}/eV )
and TDTimeStep=0.0005 (}/eV )) for picene . . . . . . . . . 35
3.13 The absorption spectrum for (T=10 (}/eV ), T=20 (}/eV ),
T=25 (}/eV ), T=30 (}/eV ), T=35 (}/eV ), and T=40 (}/eV ))
for pentacene . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

49
50 LIST OF FIGURES

3.14 The absorption spectrum for (T=10 (}/eV ), T=20 (}/eV ),

T=25 (}/eV ), T=30 (}/eV ), T=35 (}/eV ), and T=40 (}/eV ))
for picene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.15 The absorption spectrum for pentacene (a)The average ab-
sorption (b)The strength function for three direction (x,y,z) . 38
3.16 The absorption spectrum for picene (a)The average absorp-
tion (b)The strength function for three direction (x,y,z) . . . 39
3.17 The absorption spectrum for Casida equation with the num-
ber of unoccupied states are equal to 51, 100, 150, and 200
for pentacene . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.18 The absorption spectrum for Casida equation with the num-
ber of unoccupied states are equal to 51, 100, and 150 for
picene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.19 The absorption spectrum for pentacene (a) absorption spec-
trum from Casida (b) the absorption spectrum from time
propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.20 The absorption spectrum for picene (a) absorption spectrum
from Casida (b) the absorption spectrum from time propagation 42
3.21 The absorption spectrum for Casida equation with fxc and
without fxc of pentacene . . . . . . . . . . . . . . . . . . . . . 42
3.22 The absorption spectrum for Casida equation with fxc and
without fxc of picene . . . . . . . . . . . . . . . . . . . . . . . 43
List of Tables

3.1 The Energy of some states of pentacene and picene . . . . . . 29

3.2 The Energy band of σ, σ ∗ , π, and π ∗ state . . . . . . . . . . . 29
3.3 HOMO–LUMO energy gap for pentacene and picene . . . . . 29

51