Karl Michael Ziems

Photophysics and
Photochemistry
of a BODIPY-Based
Photosensitizer
Quantumchemical Simulations
BestMasters
Mit „BestMasters“ zeichnet Springer die besten Masterarbeiten aus, die an renom­
mierten Hochschulen in Deutschland, Österreich und der Schweiz entstanden sind.
Die mit Höchstnote ausgezeichneten Arbeiten wurden durch Gutachter zur Veröf­
fentlichung empfohlen und behandeln aktuelle Themen aus unterschiedlichen
Fachgebieten der Naturwissenschaften, Psychologie, Technik und Wirtschaftswis­
senschaften. Die Reihe wendet sich an Praktiker und Wissenschaftler gleicherma­
ßen und soll insbesondere auch Nachwuchswissenschaftlern Orientierung geben.

Springer awards “BestMasters” to the best master’s theses which have been com­
pleted at renowned Universities in Germany, Austria, and Switzerland. The ­studies
received highest marks and were recommended for publication by supervisors.
They address current issues from various fields of research in natural sciences,
psychology, technology, and economics. The series addresses practitioners as well
as scientists and, in particular, offers guidance for early stage researchers.

More information about this series at http://www.springer.com/series/13198

Karl Michael Ziems

Photophysics and
Photochemistry
of a BODIPY‐Based
Photosensitizer
Quantumchemical Simulations
Karl Michael Ziems
Selb, Germany

ISSN 2625-3577 ISSN 2625-3615 (electronic)

BestMasters
ISBN 978-3-658-26187-0 ISBN 978-3-658-26188-7 (eBook)
https://doi.org/10.1007/978-3-658-26188-7

Springer Spektrum
© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2019
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or
dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, expressed or implied, with respect to the material contained
herein or for any errors or omissions that may have been made. The publisher remains neutral with
regard to jurisdictional claims in published maps and institutional affiliations.

This Springer Spektrum imprint is published by the registered company Springer Fachmedien Wiesbaden
GmbH part of Springer Nature
The registered company address is: Abraham-Lincoln-Str. 46, 65189 Wiesbaden, Germany
Contents
List of Abbreviations vii

Abstract xi

Abstract (Deutsch) xiii

1 Introduction 1

2 Theory 9
2.1 Hamiltonian operator . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Born-Oppenheimer Approximation . . . . . . . . . . . . . . . 10
2.3 Beyond Born-Oppenheimer . . . . . . . . . . . . . . . . . . . 12
2.4 Hartree-Fock theory . . . . . . . . . . . . . . . . . . . . . . . 14
2.4.1 Ansatz of the Hartree-Fock theory . . . . . . . . . . . 14
2.4.2 The Hartree-Fock equations . . . . . . . . . . . . . . . 15
2.4.3 Restricted and Unrestricted Hartree-Fock . . . . . . . 18
2.4.4 The Basis Set Approximation . . . . . . . . . . . . . . 18
2.4.5 The Roothaan-Hall equations . . . . . . . . . . . . . . 19
2.4.6 The SCF Procedure . . . . . . . . . . . . . . . . . . . 20
2.5 Electron Correlation Methods . . . . . . . . . . . . . . . . . . 21
2.5.1 Configuration Interaction . . . . . . . . . . . . . . . . 22
2.5.2 Multi-Configuration Self-Consistent Field . . . . . . . 24
2.6 Density Functional Theory . . . . . . . . . . . . . . . . . . . 32
2.6.1 Hohenberg-Kohn Theorems . . . . . . . . . . . . . . . 33
2.6.2 Kohn-Sham Theory . . . . . . . . . . . . . . . . . . . 33
2.6.3 Exchange-Correlation Functionals . . . . . . . . . . . 35
2.6.4 Time-Dependent Density Functional Theory . . . . . . 37
2.6.5 (TD)DFT Problems . . . . . . . . . . . . . . . . . . . 41
2.7 Spin-Orbit Coupling . . . . . . . . . . . . . . . . . . . . . . . 41

3 Computational Details 45
3.1 Used Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.2 Active Spaces for RASSCF/RASPT2 calculations . . . . . . . 48
3.2.1 Active spaces for dissociation and charge transfer . . . 49
vi Contents

3.2.2 Level shift considerations . . . . . . . . . . . . . . . . 51

4 Results for non-reduced singlet system 53

4.1 TDDFT and RASPT2 benchmark . . . . . . . . . . . . . . . 53
4.2 Spin-Orbit coupling at ground state geometry . . . . . . . . . 56
4.3 TDDFT geometry-states correlation diagram . . . . . . . . . 57
4.4 Dissociation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.4.1 Minimum energy pathway . . . . . . . . . . . . . . . . 60
4.4.2 Dissociation coordinate . . . . . . . . . . . . . . . . . 61
4.4.3 Spin-Orbit Kinetics . . . . . . . . . . . . . . . . . . . 68
4.5 Charge-Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . 71

5 Results for reduced doublet system 75

5.1 Comparison of TDDFT and RASPT2 results . . . . . . . . . 75
5.2 TDDFT geometry-states correlation diagram . . . . . . . . . 78
5.3 Dissociation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.4 Charge Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . 82

6 Conclusion 85

7 Summary 89

8 Zusammenfassung 93

Bibliography 97

A TDDFT and RASPT2 states 111

B Reference weight 125

C Spin-Orbit couplings of the non-reduced singlet system 129

D Further diabatic potentials 133

Danksagung 139
List of Abbreviations
AMFI Atomic-Mean-Field Integral

AS Active Space

BOA Born-Oppenheimer Approximation

BODIPY Boron Dipyrromethene

BP Breit-Pauli

CASPT2 Complete Active Space Perturbation Theory through

Second-Order

CASSCF Complete Active Space Self-Consistent Field

CI Configuration Interaction

CSF Configurational State Function

CT Charge Transfer

DFT Density Functional Theory

DKH Douglas-Kroll-Hess

EC Electron Correlation

ECP Effective Core Potential

ER Electron Relay

GASSCF Generalized Active Space Self-Consistent Field

GGA Generalized Gradient Approximation

GS Ground State
GTO Gaussian Type Orbital
viii List of Abbreviations

HF Hartree-Fock
HOMO Highest Occupied Molecular Orbital
H-L HOMO → LUMO
H-1-L HOMO-1 → LUMO
IC Internal Conversion
ISC Intersystem Crossing
LCAO Linear Combination of Atomic Orbitals
LDA Local Density Approximation
LICC Linear-Interpolated Cartesian Coordinate
LSDA Local Spin Density Approximation
LUMO Lowest Unoccupied Molecular Orbital
MCSCF Multi-Configuration Self-Consistent Field
MEP Minimum Energy Path
MO Molecular Orbital
MP Møller-Plesset Perturbation Theory
MS-CASPT2 Multi-State Complete Active Space Perturbation Theory
through Second-Order
MS-RASPT2 Multi-State Restricted Active Space Perturbation Theory
through Second-Order
PCM Polarizable Continuum Model
PES Potential Energy Surface
PMD Photochemical Molecular Devices
PS Photosensitizer
RASPT2 Restricted Active Space Perturbation Theory through
Second-Order
RASSCF Restricted Active Space Self-Consistent Field
List of Abbreviations ix

RASSI Restricted Active Space State Interaction

RHF Restricted Hartree-Fock
SCF Self-Consistent Field

SD Slater Determinant
SOC Spin-Orbit Coupling
SOMF Spin-Orbit Mean Field

SOMO Singly Occupied Molecular Orbital

SOO Spin-Other Orbit
SSO Spin-Same Orbit
STO Slater Type Orbital
TDDFT Time-Dependent Density Functional Theory

TDSE Time-Dependent Schrödinger Equation

TEA Triethylamine
THF Tetrahydrofurane

TISE Time-Independent Schrödinger Equation

UHF Unrestricted Hartree-Fock
UV Ultraviolet
Vis Visible
Abstract
A meso-mesityl-2,6-Iodine substituted Boron Dipyrromethene (BODIPY)
dye is investigated computationally regarding its functionality in a two-
component light-driven hydrogen evolution with [Pd(PPh3 )Cl2 ]2 as precursor
for catalytically-active Pd nanoparticles and triethylamine as sacrificial
electron donor. Therefore, quantum chemical calculations are performed at
the time-dependent density functional (TDDFT) and multi-state restricted
active space perturbation theory through second-order (MS-RASPT2) level
of theory. Earlier reports showed a significantly higher hydrogen production
activity of the present dye compared to its unsubstituted parent molecule and
suggested a heavy atom effect with ISC to long-lived charge transfer states
located at the mesityl moiety as an explanation. Furthermore, a dissociation
of the substituted Iodines was reported. The present computational work is
applied to the non-reduced and singly reduced BODIPY dye and aims to
elucidate several dissociation and charge transfer channels along selected
coordinates of both systems. It could be revealed that the photoactivation
mechanism relies predominantly on a reduction of the BODIPY dye prior
to the photoexcitation, which does not involve any ISC. However, the
dissociation of the Iodines, i.e. the photodegradation, is closely coupled to
the population of charge transfer states.
Abstract (Deutsch)
In dieser Arbeit wurde ein meso-Mesityl-2,6-Iod substituierter Bor-dipyrro-
methen (BODIPY) Farbstoff bezüglich seiner Funktionalität in einem licht-
induzierten Zwei-Komponenten Wasserspaltungssystem, zusammen mit dem
Komplex [Pd(PPh3 )Cl2 ]2 zur Erzeugung von katalytisch aktiven Pd-Nano-
partikeln und Triethylamin als Elektronen-Opferdonor, mittels quanten-
chemischer Simulationen untersucht. Als Methoden kamen zeitabhängige
Dichtefunktionaltheorie (TDDFT) und multi-state restricted active space
Störungstheorie zweiter Ordnung (MS-RASPT2) zum Einsatz. In früheren
Veröffentlichungen konnte für diesen Farbstoff eine signifikant gesteigerte
Wasserstoff-Produktion im Vergleich zum unsubstituierten BODIPY Farb-
stoff festgestellt werden. Diese wurde einem Schweratomeffekt mit ISC zu
langlebigen, ladungsseparierten Zuständen am Mesitly-Rest zugeschrieben.
Des Weiteren wurde von einem Bruch der C-I Bindungen, d.h. einer Pho-
tozersetzung, berichtet. Die Berechnungen dieser Arbeit werden auf das
nicht reduzierte und einfach reduzierte BODIPY angewendet. Dabei wer-
den für beide Systeme mehrere dissoziative und Ladungstransfer-Übergänge
entlang ausgewählter Koordinaten untersucht. Abschließend konnte fest-
gestellt werden, dass der Mechanismus der Photoaktivierung überwiegend
auf einer vorangestellten chemischen Reduktion des BODIPY, ohne Betei-
ligung eines ISC, beruht und die Photozersetzung mit der Population von
ladungsseparierten Zuständen gekoppelt ist.
1 Introduction
Renewable and sustainable power sources are a key scientific and technologi-
cal challenge in time of climate change and limitations of fossil fuels. [1–5]
A strong research focus is hereby on the use of solar energy. The sun
provides approximately 3 × 1024 J of usable energy per year, exceeding our
current energy demands by thousands of joules. Thus, it represents a viable
energy resource for solving our future economic and ecological challenges. [5]
Beside the development of solar cells, solar energy conversion is utilized to
produce chemical fuels from carbon dioxide or water. [5–7]
The latter one - light-driven water-splitting - aims to gain molecular
hydrogen to use as an environmentally clean and renewable fuel. Currently,
most of the hydrogen used in industry is derived from natural gas, coal,
petroleum or water electrolysis at the expense of carbon dioxide emissions
or high levels of electric power consumption. Therefore, a carbon-free and
power-efficient generation of hydrogen is a highly desirable prospect. [8–13]
The water-splitting mechanism is divided into two half-reactions. The
water oxidation to gain oxygen and the water reduction to produce hydrogen.

2 H2 O −−→ 2 H2 + O2 (1.1)
+ −
2 H2 O −−→ 4 H + 4 e + O2 (1.2)
+ −
4 H + 4 e −−→ 2 H2 (1.3)

If researchers are interested solely in hydrogen production, the water oxida-

tion is replaced by a suitable sacrificial electron donor. [5]
In recent years, the field of solar hydrogen production has grown. In
general, semiconductor-based devices [14–18] , photoactive dyes in homoge-
neous systems, and photoactive dyes in multi-component systems [19–23]
are differentiated. Research about the latter two topics increased after
Lehn and Sauvage observed hydrogen production in a multicomponent
system of [Ru(bpy)3 ]2+ (bpy = 2,2’-bipyridine) as photosensitizer (PS),
[Rh(bpy)3 ]2+ as electron relay (ER), colloidal Platin as reduction catalyst
and triethanolamine as sacrificial electron donor. [23,24] This and other three-
component solar hydrogen-generation systems were a focus of research in the
1970s and 1980s. [24–28] The excited PS transfers an electron to an ER, which
© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2019
K. M. Ziems, Photophysics and Photochemistry of a BODIPY-Based
Photosensitizer, BestMasters, https://doi.org/10.1007/978-3-658-26188-7_1
2 1 Introduction

functions as electron mediator and donates the electron to the catalyst. Here,
the water splitting to hydrogen occurs. The two-component system omits
the electron relay and the excited electron from the PS is transferred directly
to the catalyst. [29–33] A major problem with these systems is controlling and
monitoring the electron transfer processes, which depend on many factors in
heterogeneous systems. Thus, supramolecular, homogeneous, one-component
approaches have been the focus of research in recent years. They couple a PS
as light-harvesting unit with a catalytic center in a supramolecular approach
- commonly referred to as photochemical molecular devices (PMD). [34–39]
One prominent example is [(tbbpy)2 Ru(tpphz)PdCl2 ](PF6 )2 (tbbpy = 4,4’-
Di-tert-butyl-2,2’-bipyridin, tpphz = Tetrapyrido-[3,2-a:2’,3’-c:3”,2”’,3”’-
j]phenanzin) shown in Figure 1.1, [40,41] where the Ruthenium moiety acts as
PS and transfers one of several excess charges via MLCT excitation to the
tpphz bridging ligand and from there to the catalytic active Palladium cen-
ter. [34,41] The design concept underlying all heterogeneous systems discussed
so far is depicted in Figure 1.2. [5]
The preliminary step, the energy supply for these processes, is achieved
via light harvesting by the PS through population of an electronic excited
state. Such a PS is defined as a molecule with a chromophor absorbing
in the ultraviolet (UV) and visible (Vis) range of the solar spectrum, and
capable of transferring electrons to an electronic acceptor. An efficient
PS is distinguished by a high absorbency in a broad range of the UV/vis
part of the solar light, long-term stability, its electron-transfer efficiency
and excited state lifetime, all with respect to cost, productivity, toxicity
and feasibility of hydrogen generation. [42] The most prominent example
of a PS in nature is chlorophyll, used for photosynthesis in photosystem I
and II. [34,35] Thus, man-made light-induced hydrogen-producing systems

N
N N N N Cl
Ru2+ Pd
N N N N Cl
N

Figure 1.1: Molecular structure of [(tbbpy)2 Ru(tpphz)PdCl2 ]2+ , used together

with PF6 – as PMD.
1 Introduction 3

SD H2
(1) PS + ER + Cat
SD+ H2O

SD H2
(2) PS + Cat
SD+ H2O

SD H2
(3) PS Cat
SD+ H2O

Figure 1.2: Three different photocatalytic hydrogen production systems. The

heterogeneous systems: (1) three-component system and (2) two-
component system, as well as the homogeneous approach: (3) a
one-component system, PMD.

are commonly summarized by the term "artificial photosynthesis". Since

the 1970s, mostly transition-metal-based complexes have been successfully
employed as photosensitizer; more specific, complexes on the basis of Ruthe-
nium, Platinum, Iridium, and Rhenium. [40,42–51] However, their high price
and toxicity limited their applications and opened the research field towards
metal-free approaches. Several systems containing classical organic dyes,
such as Fluorescein, porphyrins or Eosin Y have been reported as PS. [52–59]
Moreover, Boron Dipyrromethene (BODIPY) (IUPAC name: 4,4-difluoro-
4-bora-3a,4a-diaza-s-indacene) dyes were used as light-harvesting unit in
inter- [60] and intramolecular [61–65] photocatalytic hydrogen systems.
BODIPYs show excellent photophysical properties like high extinction co-
efficients and quantum yields while having a small and simple structure. [66,67]
Moreover, their properties, such as absorption and emission wavelength or
photostability, can easily be tuned via chemical modification. Thus, a big
variety of BODIPY dyes has been reported and investigated with extensive
experimental and theoretical efforts. [66,68–75] Due to their excellent and mod-
ifiable properties, BODIPYs find use in a wide range of applications other
than as photosensitizers in multicomponent catalytic systems. Research has
been done in the field of photodynamic therapy, [76,77] electrochemistry and
electro-generated chemiluminescence, [78] laser dyes, [79,80] labeling and fluo-
rescent indicators and probes, [81,82] imaging and sensing, [83,84] and singlet
oxygen generation. [85]
4 1 Introduction

R
Cl PPh3

Pd
X1 X2 Cl Cl
N N N
Pd
B
Ph3P Cl
F F

PS Pd source TEA

Figure 1.3: Two-component system, investigated by T. Beweries et al. [86,87]

Various BODIPY dyes for two-component systems were studied by the

group of T. Beweries. [86,87] Their experimental results outlined the basis of
this thesis. In their multicomponent approach they used different BODIPYs
as PS, triethylamine (TEA) as sacrificial electron donor and [Pd(PPh3 )Cl2 ]2
as a precursor for catalytically-active Palladium nanoparticles in tetrahydro-
furane (THF)/water (see Figure 1.3). Two major observations were made:
Firstly, a sterical demanding group at the meso-position (R) of the BODIPY
enhances the dyes’ stability. Using a mesityl-group exceeded previously
reported long-term stability of 20 h by a factor of over ten. Secondly, the
incorporation of Iodine at the 2,6 position of the BODIPY (X1 and X2 )
resulted in a significant increase of hydrogen production, surpassing its
unsubstituted equivalent (X1 = X2 = H) by a factor of seven.
Two different mechanisms were discussed by T. Beweries and co-workers
that might provide an explanation for the observation of an increased hydro-
gen production with Iodine-substituted BODIPYs. [86] The mechanisms are
described below and depicted in Figure 1.4.
The first hypothesis postulates that the high hydrogen production is
due to a relativistic effect originating from the Iodines, which results in a
pronounced coupling between singlet and triplet states and a population of
charge separated triplet states. An electron transfer to the catalyst starting
from these long-lived triplet charge transfer (CT) states would be very likely.
Proof for intersystem crossing (ISC) from the initially populated excited
singlet state to excited triplet states was experimentally found through low
emission quantum yields and low lifetimes for the excited singlet state in
comparison with unsubstituted BODIPYs. The population of long-lived
1 Introduction 5

Heavy atom effect with Chemical reduction with

non-reduced BODIPY system singly reduced BODIPY system
Figure 1.4: Two possible mechanisms described by T. Beweries et al. for a
two-component light-induced water reduction with BODIPY as PS,
Pd as catalyst and TEA as sacrificial electron donor.

states1 in transient absorption spectroscopy hinted towards an ISC to triplet

states, as well. Furthermore, the excited state absorptions were identified as
triplet CT state by quantum chemical calculations with Time-Dependent
Density Functional Theory (TDDFT) as level of theory. This first mechanism
is referred to as non-reduced system and implies that the electron is first
separated within the BODIPY in form of a triplet charge transfer state,
then transferred to the catalyst and as last step the sacrificial electron donor
restores the original electronic state of the PS.
A second approach turns the sequence of the first mechanism hypothesis
around. Here, the sacrificial electron donor firstly reduces the BODIPY,
then a CT state is generated upon irradiation and, finally, an electron is
transferred to the catalyst. Due to the reduction of the BODIPY as initial
step, this is called the reduced mechanism. Experimental evidence for this
pathway are quenching experiments via electron transfer with TEA and
BODIPY. However, triplet quenching experiments were not conducted and
the bright, first excited singlet state exposed only slow quenching properties
(Stern-Volmer constant of KSV = 1.08 ± 0.03).
Therefore, the mechanism with a non-reduced BODIPY as PS, generating
CT triplet states after irradiation to transfer the electron to the catalyst,
1 They were interpreted as triplet states, but no experimental research was done proving
that. The assignment is based on literature values
6 1 Introduction

was postulated to be the dominant mechanism by the group of T. Beweries

and co-workers.
Another photophysical observation in the two-component system, con-
sisting of Pd-catalyst, the BODIPY with two Iodines (X1 = X2 = I) and
TEA, was the photodegradation of the dye via cleavage of the C-I bonds
after ∼20 h under catalytic conditions, yielding the less hydrogen-productive
unsubstituted BODIPY (X1 = X2 = H). This represents a big disadvantage
of the otherwise promising Iodine-substituted BODIPY.

The aim of this thesis is to investigate the BODIPY dye shown in Fig-
ure 1.5. The structure was slightly modified compared to the Reference [86] by
replacing the mesityl group with a phenyl group to reduce the computational
costs of the quantum chemical simulations. Several tasks are outlined, based
on the experimental observations and mechanistic thoughts described above.
First of all, a comparison of the results obtained by different levels of theory
for the BODIPY has to be conducted. This is especially important since the
calculations from the previous work were mainly done by TDDFT, which is
known to perform poorly for boron-containing systems2 . As state-of-the-art
computational reference method Multi-State Restricted Active Space Per-
turbation Theory through Second-Order (MS-RASPT2) has been chosen.
The main goal is to unravel the photo-induced mechanism (photoactivation)
leading to hydrogen production, i.e. heavy atom effect (non-reduced system)
vs. chemical reduction (singly reduced system), and to study the C-I cleavage
(photodegradation) in order to prevent such cleavage and to enhance the
dye’s photostability. Therefore, pathways and kinetics for dissociative and
CT processes have to be identified and an investigation of the role of the
Iodines and a possible relativistic heavy atom effect has to be incorporated.

This thesis is constructed as follows: First of all, an overview of the used

methods is given in chapter 2, followed by the computational details with
a special emphasis on the construction of the active spaces in chapter 3.
Subsequently, all obtained results are introduced and discussed for both
systems, complemented by a conclusion in chapter 6 connecting all results.
The summary will give a brief overview of the entire work and an outlook to
future investigations.

Parts of the results have been published as an article in the journal

Catalysts with similar figures and schemes. For reference see [88] .

2 This (TD)DFT problem for boron-systems is further deepened in subsection 2.6.5.

1 Introduction 7

Figure 1.5: Structure of T. Beweries et al. most promising BODIPY dye re-
garding hydrogen production in a two-component system. The
main heteroaromatic ring and the side ring are perpendicular to
each other and the structure has been slightly modified substitut-
ing the mesityl with a phenyl side ring to reduce computational
costs.
2 Theory
The following chapter presents an overview of the used methods, as well as
the underlying physical principles and approximations.

2.1 Hamiltonian operator

The main equation to describe (molecular) systems and their properties
within the framework of quantum mechanics is the Time-Dependent Schrö-
dinger Equation (TDSE)

∂  = Ĥ |Ψ(t, r, R)

 .
ih̄ |Ψ(t, r, R) (2.1)
∂t

Ĥ is the Hamilton operator corresponding to the total energy Etot of the

 is
system by incorporating all particles and their interaction, and |Ψ(t, r, R)
the corresponding wave function, depending on time t, electron coordinates

r and nuclear coordinates R.
A Hamiltonian with no time dependency enables a separation Ansatz for
the TDSE into the Time-Independent Schrödinger Equation (TISE) with
Etot being the total energy of the stationary system.
 = Etot |Ψ(r, R)
Ĥ |Ψ(r, R)  (2.2)

The total (non-relativistic) Hermitian Hamiltonian can be written as a

sum of kinetic (T̂ ) and potential operators (V̂ ) of K nuclei and N electrons:

Ĥ =T̂n + T̂e + V̂ne + V̂ee + V̂nn

K
h̄2 2  h̄2 2
N
=− ∇α − ∇
α=1
2mα i=1
2me i
⎡ ⎤
e2 ⎣   Zα  1 
K N N K
Zα Zβ ⎦
+ − + + , (2.3)
4π0 α i
riα i<j rij Rαβ
α<β

© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2019

K. M. Ziems, Photophysics and Photochemistry of a BODIPY-Based
Photosensitizer, BestMasters, https://doi.org/10.1007/978-3-658-26188-7_2
10 2 Theory

here T̂n is the nuclear kinetic energy operator, T̂e the electronic kinetic energy
operator and V̂ = V̂ne + V̂ee + V̂nn the corresponding potential energy operator
of the interactions of electron and nucleus, V̂ne , electron and electron, V̂ee ,
and nucleus and nucleus, V̂nn . The parameter me and mα describe the
electron- and nuclear-mass, respectively, e corresponds to the elementary
charge, Zα to the nuclear charge and 0 is the vacuum permittivity. The
distances between the elementary particles are indicated by r and R. [89,90]

2.2 Born-Oppenheimer Approximation

The first step in the Born-Oppenheimer Approximation (BOA) is the Ansatz
of transforming the Hamiltonian to the center of mass system to write the
Hamiltonian as follows:
Ĥ = Ĥe + T̂n (+Ĥmp ) (2.4)
Ĥe = T̂e + V̂ne + V̂ee + V̂nn . (2.5)

In this case, Ĥe is the electronic Hermitian Hamiltonian and Ĥmp the
so-called mass-polarization. The latter denotes that it is impossible to
strictly separate the center of mass motion from the internal motion. This
contribution to the total Hamiltonian will be neglected from now on.1
The Born-Oppenheimer Ansatz (without introducing any approximations
yet) is an expansion of the total, exact wave function in a complete set of

electronic wave functions |ψi (r, R), with the expansion coefficient being

functions of the nuclear coordinates |ψni (R).
∞

 =
|Ψ(r, R)  |ψi (r, R)
|ψni (R)  (2.6)
i=1


This means that the total wave function is separated in a term, |ψni (R),
depending on the nuclear coordinates (nuclear wave function), and a term,

|ψi (r, R), depending on the electronic coordinates and the current nuclear
position R (electronic wave function).2
Equation 2.5 and 2.6 allow to define an electronic Schrödinger equation:
 = Ei (R)
Ĥe |ψi (r, R)  |ψi (r, R)
 ; i = 1, 2, ..., ∞ (2.7)
1 The adiabatic and BOA both neglect the Ĥmp term. For the sake of clarity, it will
already be neglected in the derivation of these approximations.
2 Thus, the index i sums over the electronic eigenfunctions |ψ ( 
i r , R), whereas n indicates
the nuclear eigenfunctions |ψni (R) for a given electronic eigenfunction |ψi (
 
r, R).
2.2 Born-Oppenheimer Approximation 11

 being the solution at a given nuclear geometry R for the electronic

with Ei (R)
Schrödinger equation. The fact that the electronic Hamiltonian is Hermitian
allows the solutions to be chosen orthonormal.

0, for i = j
ψi |ψj  = (2.8)
1, for i = j

Combining the TISE in Equation 2.2 with the separation of the total wave
function in Equation 2.6 gives:
∞
 ∞

 |ψi (r, R)
(Ĥe + T̂n ) |ψni (R)  = Etot  |ψi (r, R)
|ψni (R)  . (2.9)
i=1 i=1

For reasons of clarity, from now on the dependencies of the wave functions
will be left out and atomic units will be used for the operators. Moreover,
the electronic wave function is written as ψi and the nuclear wave function
as ψni .3

K
h̄2 2
∇2n = T̂n = − ∇
α=1
2mα α

ψni = |ψni (R)

ψi = |ψi (r, R) (2.10)

Inserting the exact expression for the nuclear operator and using Equation 2.7,
results after some conversions in
∞
  ∞
 ψi (∇2n ψni ) + 2(∇n ψi )(∇n ψni )+ 
= E tot ψni ψi . (2.11)
i=1
ψni (∇2n ψi ) + ψni Ei ψi i=1

Multiplying the equation from the left by a specific electronic wave function
ψj∗ and integrating over the electronic coordinates yields, with the use of the
orthonormality of the electronic wave function (Equation 2.8), the following
equation:
∞
  ∞
 2 ψj |∇n |ψi  (∇n ψni )+ 
2
∇n ψnj + Ej ψnj + = E tot ψnj . (2.12)
i=1
ψj |∇2n |ψi  ψni i=1
3 The notation in bras and kets will also be neglected. The braket notation will only
be used to indicate scalar products (integration over electronic coordinates) in this
section.
12 2 Theory

The terms in the curly bracket couple different electronic states and are called
first- and second-order non-adiabatic coupling elements. These become very
important at the (near-)degeneracy of electronic states and for describing
interactions between them, e.g. for photochemical reactions.
In the adiabatic approximation, the total wave function is restricted to
one electronic surface. This means that in Equation 2.12 only the coupling
elements i = j survive and the first-order non-adiabatic elements are zero
(within the assumption of no spatial degeneracy).

∇2n + Ej + ψj |∇2n |ψi  ψnj = Etot ψnj

 + U (R)
T̂n + Ej (R)  ψnj (R)
 = Etot ψnj (R)
 (2.13)

 term is the so-called diagonal correction and is significantly

The U (R))
 [91] In the BOA the diagonal correction is neglected,
smaller than Ej (R).
which results in a nuclear Schrödinger equation with the electronic energy
in the role of a potential energy.

 ψnj (R)
T̂n + Ej (R)  = Etot ψnj (R)
 (2.14)

In the Born-Oppenheimer picture, the nuclei move on a Potential Energy

Surface (PES), which is the solution of the electronic Schrödinger equation.
This means that the electrons react instantly (adiabatically) to any change
in the nuclear motion (∇2n ). The PES can be constructed by solving the
electronic Schrödinger equation, which yields the energy for this surface, in
a parametric dependence on the nuclear coordinate. Thus, for each position
R̄ of the nuclei, a corresponding electronic energy Ej (R̄) is obtained.
Solving Equation 2.14 for the nuclear wave function on one PES results
in energy levels of molecular vibrations and rotations, whereas Equation 2.7,
the electronic Schrödinger equation, leads to electron distributions and
PES’s. These are the two fundamental quantum chemical equations in the
framework of the BOA. [89,90,92,93]

2.3 Beyond Born-Oppenheimer

The BOA is usually a good approximation, but breaks down when two (or
more) electronic states (solutions of the electronic Schrödinger equation)
come close in energy. In this case vibronic coupling between states occurs,
which results in crossings and radiationless decays (e.g. internal conversion);
processes that are especially important in photochemical studies.
2.3 Beyond Born-Oppenheimer 13

Also, because two (or more) states with different electronic distributions
become energetically equal, the electronic wave function cannot adjust
instantly on any change in the nuclear motion and therefore, an adiabatic
approach is not reasonable. As a result of this, concepts such as molecular
geometries become blurred and energy surfaces no longer exist.
Calculating such PES with crossing states in the framework of the BOA
would result in adiabatic potentials which are described poorly in the crossing
region and does not allow any switch of state. This is based on the break-
down of the main approximation within the BOA, the neglection of the
non-adiabatic coupling elements:
(1)
Tij = ψj |∇n |ψi 
(2)
Tij = ψj |∇2n |ψi  , (2.15)

which are essential to describe a mix of different electronic states as a result of

small vibrations. These couplings can become large in the region of crossings;
therefore, a non-adiabatic treatment of the TISE is favorable. This means -
for an example of a two-level system - that Equation 2.16 is no longer a good
description and the non-adiabatic Equation 2.17 should be used instead.
Another approach is using diabatic potentials (Equation 2.18). These can be
obtained via a basis transformation that diagonalizes the kinetic matrix in
the non-adiabatic equation and introduces non-diagonal potentials between
mixing states. Diabatic potentials differ from adiabatic ones in terms of
following a specific electron distribution, even in the case of crossing other
potentials. [89,91,94]
       
T 0 E1 0 ψn1 ψn1
+ = Etot (2.16)
0 T 0 E2 ψn2 ψn2

       
(1)
T T12 E1 0 ψn1 ψn1
(1) + = Etot (2.17)
T12 T 0 E2 ψn2 ψn2

       
(d) (d)
T 0 E1 E12 ψn1 ψn1
+ (d) (d) = Etot (2.18)
0 T E12 E2 ψn2 ψn2
14 2 Theory

2.4 Hartree-Fock theory

Taking the BOA into consideration allows only to solve the electronic
Schrödinger equation for the H2 + molecule and similar one-electron systems
exactly. Thus, further (numerical) approximations are needed to treat larger
systems.
From now on Ψ will be used interchangeably as total electronic wave
function ψi , as well as E as electronic energy Ei , since the following sections
deal only with the electronic part of the wave function.

2.4.1 Ansatz of the Hartree-Fock theory

To get approximate solutions to the electronic wave function the Hartree-
Fock (HF) theory employs the variational principle, which states that any
approximate wave function has an energy above or equal to the exact energy.
This is implemented by constructing a trial wave function containing a
number of parameters. In the next step, the energy is minimized as a
function of these parameters. In general, the energy of an approximate wave
function is calculated as the expectation value of the Hamiltonian, divided
by the norm of the wave function:

Ψ|Ĥe |Ψ
E= . (2.19)
Ψ|Ψ
In the HF method, the trial wave function is built by one-electron functions
- Molecular Orbitals (MO). Since the electronic wave function must be
antisymmetric with respect to interchange of any two electron coordinates
(Pauli principle), these orbitals are organized in a Slater Determinant (SD).
As another Ansatz in the HF theory, the electron spin is introduced as
an ad hoc quantum effect. This results in orthonormal spin orbitals χa (xi )
as product of a spatial orbital ϕa (ri ) and a spin function s(ωi ) with two
possible arrangements (α(ωi ) or β(ωi )).

α(ωi )
χa (xi ) = ϕa (ri ) · (2.20)
β(ωi )

Here, the index a specifies the MO and i the electron. The spin functions
are orthogonal to each other:

α|α = β|β = 1
α|β = β|α = 0 . (2.21)
2.4 Hartree-Fock theory 15

These two approximations within HF result in a SD of spin orbitals as

approximate electronic wave function. For the general case of N electrons
in N spin orbitals, the SD is given as:
 
 χ1 (x1 ) χ2 (x1 ) · · · χN (x1 ) 
 
1  χ1 (x2 ) χ2 (x2 ) · · · χN (x2 ) 
Ψ= √  . .. .. .. ; with χa |χb  = δab .
N !  .. . . . 

χ1 (xN ) χ2 (xN ) · · · χN (xN )
(2.22)
The columns in the SD are single-electron wave functions, spin orbitals χa ,
while each row represents an electron coordinate xi . [89,90,95]

2.4.2 The Hartree-Fock equations

In order to derive the HF equations, an expression for the energy of a single
SD is needed. For this, the operators are collected according to the number
of electron indices.

N 
N
Ĥe = ĥ(xi ) + ĝ(xi , xj ) + V̂nn (2.23)
i j>i

1  Zα N
with ĥ(xi ) = − ∇2i − (2.24)
2 α
riα
1
and ĝ(xi , xj ) = (2.25)
rij

The one-electron operator ĥ(xi ) describes the kinetic energy of the electron
i in the field of all nuclei, and ĝ(xi , xj ), the two-electron operator, gives the
electron-electron repulsion. Applying these operators to the HF-Ansatz for
the electronic energy as expectation value of a N electron, N orbital SD,
results in Equation 2.26
16 2 Theory


N
E= χa (x1 )|ĥ(x1 )|χa (x1 )
a

1 
N N
+ χa (x1 )|Jˆb (x1 )|χa (x1 ) − χa (x1 )|K̂b (x1 )|χa (x1 ) + V̂nn
2 a
b=a


N
1 
N N
= haa + (Jab − Kab ) (2.26)
2 a
a b=a

with the Coulomb operator Jˆb (x1 ) and the exchange operator K̂b (x1 )

Jˆb (x1 )χa (x1 ) = χb (x2 )|ĝ(x1 , x2 )|χb (x2 ) χa (x1 ) (2.27)
K̂b (x1 )χa (x1 ) = χb (x2 )|ĝ(x1 , x2 )|χa (x2 ) χb (x1 ) (2.28)

representing the two-electron contributions.4 The corresponding Coulomb

integral
Jab = χa (x1 )χb (x2 )|ĝ(x1 , x2 )|χa (x1 )χb (x2 ) (2.29)
represents the classical repulsion between two charge distributions, whereas
the exchange integral
Kab = χa (x1 )χb (x2 )|ĝ(x1 , x2 )|χb (x1 )χa (x2 ) (2.30)
has no classical analogy. The electron-electron interactions are only ac-
counted for in an average fashion in both terms. By summing over all b =  a
one obtains the total averaged potential acting on the electron in χa , coming
from N − 1 electrons in the other spin orbitals.
The next step is minimizing this energy in such a way, that the orbitals
remain orthonormal. This constrained optimization is done by Lagrange
multipliers and under the condition that a small change in the orbitals should
not change the Lagrange function.

N
L=E− λab (χa |χb  − δab )
a,b


N
δL = δE − λab (δχa |χb  − χa |δχb ) = 0 (2.31)
a,b
4 The integrations are over a single or two electrons. The dummy variables of integration
are, by convention, chosen to be the coordinates of electron one, and electron one and
two, respectively.
2.4 Hartree-Fock theory 17

Solving Equation 2.31 results in a final set of N Hartree-Fock equations:


N
fˆ(x1 )χa (x1 ) = λab χb (x1 ) (2.32)
b

with an effective one-electron energy operator, the Fock operator:


N
fˆ(x1 ) = ĥ(x1 ) + Jˆb (x1 ) − K̂b (x1 ) , (2.33)
b

describing the kinetic energy of an electron and the attraction to all nuclei
via the one-electron operator ĥ(x1 ), as well as the repulsion to a (average)
sum of all other electrons via the Coulomb and exchange operator.5 After a
unitary transformation of the Hartree-Fock equation that diagonalizes the
matrix of Lagrange multipliers λab , a new HF equation with a special set of
so-called canonical orbitals χ and their orbital energies a is obtained:

fˆ(x1 )χa (x1 ) = a χa (x1 ) . (2.34)

Because a specific orbital can only be determined if all other occupied

orbitals are known6 , it is called pseudo-eigenvalue equation7 . Thus, iterative
methods must be used to solve these equations and to generate so-called
Self-Consistent Field (SCF) orbitals.
The total electronic energy of the system can be obtained as in Equa-
tion 2.26 or in terms of MO energies. [89,90,95]


N
1
N
E= a − (Jab − Kab ) + V̂nn (2.35)
a
2
a,b

a = χa |fˆ|χa  (2.36)

5 Note that the Fock operator is an operator for the variational HF method and therefore,
associated with the variation of the energy. It is not associated with the energy itself.
Thus, the Hamilton operator is not a sum of Fock operators.
6 The Coulomb and exchange operator sum over all orbitals.
7 The prime in the HF solution (Equation 2.34) will be neglected from now on.
18 2 Theory

2.4.3 Restricted and Unrestricted Hartree-Fock

So far, the SD and HF equations have been discussed in terms of a general
set of spin orbitals χa (xi ) (Equation 2.20) without any restrictions. In
this restriction-free case, the trial wave function is an unrestricted Hartree-
Fock (UHF) wave function, which means that different spatial orbitals for
each of the two electrons (α and β) in a MO are allowed. [96]
For systems with an even number of electrons and interests in a singlet
type wave function, the restriction that each spatial orbital has two electrons,
one with α and one with β spin, can be introduced. In other words, spin
orbitals are constrained to have the same spatial orbital for α and β spin.

ϕa (ri )α(ωi ) = ϕa (ri )
χa (xi ) = (2.37)
ϕa (ri )β(ωi ) = ϕa (ri )

This method for closed-shell systems is called restricted Hartree-Fock (RHF).

The resulting wave function (SD) changes to:

|Ψ = |χ1 χ2 · · · χN −1 χN  = |ϕ1ϕ1 · · · ϕaϕa · · · ϕN/2 ϕN/2  . (2.38)

Inserting Equation 2.37 into the general spin orbital Hartree-Fock equa-
tion (Equation 2.34) and integrating out the spin functions gives the RHF
equation:
fˆ(r1 )ϕa (r1 ) = a ϕa (r1 ) (2.39)
and operators not depending on the spin coordinate. Furthermore, one
contribution of the exchange operator is lost due to the fact that it only
occurs between electrons of the same spin:


N/2
fˆ(r1 ) = ĥ(r1 ) + 2Jˆb (r1 ) − K̂b (r1 ) . (2.40)
b

Even though the RHF means a restriction in the degrees of freedom,

it overcomes the big problem of spin contamination in UHF wave func-
tions. [89,90,95,96]

2.4.4 The Basis Set Approximation

The Hartree-Fock Equation 2.34 can only be solved for atoms and very small
molecules numerically. [97] Hence, a basis set expansion is used to express the
unknown MOs with a set of known functions for "larger" systems. Usually,
2.4 Hartree-Fock theory 19

a Linear Combination of Atomic Orbitals (LCAO) is applied to project the

spatial orbitals.

K
ϕa (ri ) = caμ φμ (ri ) (2.41)
μ

The parameter K corresponds to the number of basis functions (atomic

orbitals) φμ (ri ) used for each MO.
In principle, any type of basis function may be used, as long as their
behavior agrees with the physics of the problem and the chosen functions
facilitate the calculations of all required integrals. Slater Type Orbitals (STO)
reflect the exact functions of the hydrogen atom but are computational
demanding.
φSTO (rr ) ∝ exp−αr (2.42)
Therefore, the use of Gaussian Type Orbitals (GTO) is common nowadays.
2
φGTO (rr ) ∝ exp−αr (2.43)

Although they are poorer in describing the electronic behavior compared to

the solutions of the hydrogen atom (STO), the computational advantages
make more than up for this. Especially the fact that the product of two
GTOs is again a GTO, offers a very efficient way to calculate the two-electron
integrals. A way to compensate for the poor physics of GTOs is a linear
combination of various GTOs to contracted GTOs to mimic the behavior of
STOs.
As the number of basis functions increases, the accuracy of the MOs
improves (assured by the variational principle), but also do the computational
costs. In the limit of a complete basis set (K → ∞) the results are identical to
numerical HF; this is called the Hartree-Fock limit. By now, a vast number of
basis sets exists, different in amount and kind of its used functions; common
basis sets are e.g. by Pople, Dunning-Huzinaga or Ahlrich. [98–100]

2.4.5 The Roothaan-Hall equations

Expanding the RHF Equation 2.39 with the LCAO Equation 2.41 for a
closed-shell system (RHF) gives:


K 
K
fˆ(r1 ) caμ φμ (r1 ) = a caμ φμ (r1 ). (2.44)
μ μ
20 2 Theory

Multiplying from the left by a specific basis function φ∗ν and integrating over
the electronic coordinates results in the Roothaan-Hall equations, being the
Hartree-Fock equations in the atomic orbital basis:


K 
K
Fνμ Cμa = a Sνμ Cμa . (2.45)
μ μ

Equation 2.45 reduces the problem of calculating the HF MOs to the problem
of calculating a set of expansion coefficients Cμa , which can be derived by
diagonalizing8 the K × K Hermitian Fock- and overlap-matrix, Fνμ and Sνμ ,
in the set of basis functions.

Fνμ = φν (r1 )|fˆ(r1 )|φμ (r1 ) (2.46)

Sνμ = φν (r1 )|φμ (r1 ) (2.47)

Subsequently, this leads to the K × K square matrix of the expansion

coefficients Cμa and the K-dimensional diagonal Matrix a containing the
orbital energies. Each column of the coefficient matrix represents a MO
ϕa (ri ) consisting of different contributions of the K basis functions φμ (ri ) in
each row. Thus, the number of basis functions defines the number of MOs.
The Fock matrix can be expressed in form of a one-electron and a two-
electron contribution according to Equation 2.40:

Fνμ = Hνμ + Gνμ

= φν (r1 )|ĥ(r1 )|φμ (r1 )

N/2
+ Pλσ [2 φν (r1 )φλ (r2 )|ĝ(r1 , r2 )|φμ (r1 )φσ (r2 )
b λ,σ

− φν (r1 )φλ (r2 )|ĝ(r1 , r2 )|φσ (r1 )φμ (r2 )] (2.48)

with the coefficients presented as a density matrix: [89,90,101]

Pλσ = Cλb Cσb . (2.49)

2.4.6 The SCF Procedure

To solve the above introduced Roothaan-Hall equation (Equation 2.45), an
iterative algorithm is used - the SCF approach. The procedure is as follows:
8 This is addressed in detail in the next section.
 2.5 Electron Correlation Methods 21

1. Specify molecule and its properties,

K like atomic numbers Zα , number of
electrons N and basis set μ φμ (ri )
2. Calculate all required integrals that do not rely on the expansion coeffi-
cient: Sνμ , Hνμ and the two-electron integrals φν φλ ||φμ φσ .
3. Diagonalize the overlap matrix S to obtain a transformation matrix X.

4. Obtain a guess for the density matrix P , e.g. through extended Hückel
method. [102]
5. Calculate the two-electron matrix G from the density matrix and the two-
electron integrals and the entire Fock matrix by adding the one-electron
term (Equation 2.48).
6. Calculate the transformed Fock matrix F  = X † F X.

7. Diagonalize F  to obtain C  and  .

8. Calculate C = XC  .
9. Form a new density matrix P from the obtained coefficient matrix C.

10. Determine if the convergence criteria are fulfilled. If not, start at step (5)
with the new density matrix P .
Solving the SCF algorithm (for a closed shell system) produces a total of
K MOs with N/2 occupied MOs and K − (N/2) unoccupied or virtual
MOs. [89,90]

2.5 Electron Correlation Methods

The Hartree-Fock method generates solutions (with an appropriate basis set)
that are able to account for around 99% of the total energy. Unfortunately,
the remaining 1% are very important for describing chemical phenomena
with sufficient accuracy. The missing (non-relativistic) energy contribution
is called electron correlation (EC) energy

EC = Eexact − EHF . (2.50)

In general, the determination of this additional energy for a better description

of the system is achieved by incorporating excited Slater determinants
generated out of the ground state configuration.
22 2 Theory

Commonly, two kinds of EC are differentiated: i) static and ii) dynamic

electron correlation.
i) The static part is associated with electrons avoiding each other on
a permanent basis, i.e. occupying different spatial orbitals. It has to be
taken into account for describing near degeneracy structures and mesomeric
effects. Thus, it is especially relevant for dissociations, excited states and
transition metals. Static correlation is described by adding few excited SD
with substantial weight, representing different possible configurations.
ii) The dynamic contribution describes coupled motions of the electrons.
Based on electrostatic repulsion they try to avoid each other. Thus, con-
trary to the static correlation, more excited SDs each with less weight are
incorporated.
In this section, selected methods to describe EC will be introduced. [89,90,103]

2.5.1 Configuration Interaction

Since the HF method provides the best one determinant solution to the
electronic ground state, its wave function |ΨHF  is used as reference for
the variational Configuration Interaction (CI) method, where the trial wave
function is written as a linear combination of determinants.9
  
|ΨCI  = c0 |ΨHF  + cra |Ψra  + ab |Ψab  +
crs rs
abc |Ψabc  + · · ·
crst rst

ar a<b a<b<c
r<s r<s<t
  
= c0 |ΨHF  + cS |S + cD |D + cT |T  + · · · (2.51)
S D T

The wave function |Ψra , or |S in short form, describes a SD where one
electron in an occupied orbital a has been excited to a virtual orbital r with
respect to the |ΨHF  reference. The other terms in Equation 2.51 describe
doubly and triply excited determinants.
Similar to the HF derivation, the energy of the CI wave function is mini-
mized under the constraint that the total wave function |ΨCI  is normalized:

L = ΨCI |Ĥ|ΨCI  − λ (ΨCI |ΨCI ) . (2.52)

This results in i secular equations, one for each involved determinant, of the
form 
ci (Ei − λ) + cj Ψi |Ĥ|Ψj  = 0, (2.53)
j=0

9 Every determinant consists of molecular orbitals expanded in a set of basis functions.

2.5 Electron Correlation Methods 23

which can be written as matrix equation:

(Hij − EIij ) ci = 0
Hij ci = Eci (2.54)
⎛ ⎞⎛ ⎞ ⎛ ⎞
H00 − E H01 ··· H0j ··· c0 0
⎜ H11 − E ··· · · ·⎟ ⎟ ⎜0⎟
⎜ H10 H1j ⎟⎜ c ⎟ ⎜ ⎟
⎟⎜
1
⎜ .. .. .. .. ⎜ .. ⎟ ⎜ .. ⎟
⎜ · · ·⎟
⎜ . . . . ⎟⎜ ⎟
⎜ . =⎜
⎜.⎟
⎟ (2.55)
⎜ .. ⎟⎜ ⎟ ⎟ ⎜ ⎟
⎜ · · · Hjj − E · · ·⎟ c 0
Hj0 . ⎠ ⎝ . ⎠ ⎝.⎠
j
⎝
.. .. .. ..
··· ··· ··· . .

Solving the secular equations is equivalent to diagonalizing the CI matrix

Hij . The energy obtained as the lowest eigenvalue of the CI matrix and
the corresponding eigenvector ci describe the ground state and its compo-
sition based on the involved determinants. The second lowest eigenvalue
corresponds to the first excited state, etc. The diagonalization itself cannot
be done by standard methods because of the extensive matrix size. Thus,
special iterative methods are utilized for extracting one, or a few, eigenvalues
and their eigenvectors. The most commonly used versions are based on the
Davidson algorithm. [104]
Before diagonalizing the CI matrix, several simplifications resulting in
elements being zero can be made. First of all, matrix elements between
states of different spatial or spin symmetry are zero by definition.10 To take
these contributions into account, proper spin eigenfunctions have to be built
using the given determinants. These linear combinations of determinants
resulting in configurations that are eigenfunctions of the Ŝ 2 operator are
called Configurational State Functions (CSF). For example, two singlet
excited SD yield two CSFs - one singlet CSF (the negative linear combination)
and one triplet CSF (the positive linear combination). In addition, the
Brillouin theorem states that matrix elements between the HF ground state
and any singlet excited SD are zero: [105]

ΨHF |Ĥ|S = 0 . (2.56)

If two SDs differ by more than two (spatial) MOs, the matrix elements
are zero, as well; this connection is known as Slater Condon rule. [106,107]
Therefore, the main contribution for evaluating the ground state with CI are

10 Because the Hamiltonian is total symmetric.

24 2 Theory

doubly excited matrix elements. Equation 2.57 shows a possible CI matrix

with respect to the presented rules and its Hermitian character.11

⎛ ⎞
ΨHF |Ĥ|ΨHF  0 ΨHF |Ĥ|D 0 0 ···
⎜ S|Ĥ|S S|Ĥ|D S|Ĥ|T  · · ·⎟
⎜ 0 ⎟
⎜ ⎟
⎜ D|Ĥ|D D|Ĥ|T  D|Ĥ|Q · · ·⎟
⎜ ⎟
⎜ T |Ĥ|T  T |Ĥ|Q · · ·⎟
⎜ ⎟
⎜ Q|Ĥ|Q · · ·⎟
⎝ ⎠
..
.
(2.57)
If all possible excited SDs are considered, the wave function ΨCI is called
full CI wave function. Though, the factorial growth of the number of
determinants with the size of the basis set and the size of the molecule makes
the full CI method unfeasible for all but the very smallest systems.
Therefore, truncated CI methods are preferable. These methods stop the
linear combination of excited SDs after a specific excitation level. Whereas
CIS (truncating after singlet excited SDs) does not introduce any improve-
ment to the HF wave function due to the Brillouin theorem, CISD recovers
80-90% of the correlation energy. Incorporating more excitation levels, re-
sults in new methods, like CISDT, CISDTQ, etc. The more levels are used,
the higher the computational costs are. The main problem of all truncated
methods is the size inconsistency, which is rather disadvantageous especially
for the description of dissociation or reaction pathways. [89,90,103,108]

2.5.2 Multi-Configuration Self-Consistent Field

The Multi-Configuration Self-Consistent Field (MCSCF) method can be
considered a CI method, which is not only optimizing the coefficients for the
determinant contributions ci but also the MOs ϕa (ri ) used for constructing
the SDs (by optimizing the basis set expansion coefficients bμa ).

11 The Hermitian character is expressed in form of only one depicted half of the symmetric
CI matrix.
2.5 Electron Correlation Methods 25


|ΨMCSCF  = ci |Ψi  (2.58)
i
|Ψi  = |χ1 χ2 · · · χN −1 χN  = |ϕ1ϕ1 · · · ϕaϕa · · · ϕN/2 ϕN/2  (2.59)

K
|ϕa  = bμa φμ (2.60)
μ

This results in a much harder convergence to keep the method variational

with respect to both coefficient sets. For a normalized wave function the
energy is obtained as the expectation value of the Hamiltonian like seen for
the CI methods:
 
E = ΨMCSCF |Ĥ|ΨMCSCF  = hij Dij + gijkl Pijkl , (2.61)
i,j i,j,k,l

with a Hamiltonian in second quantification expression and Êij being exci-

tation operators:

 1 
Ĥ = hij Êij + gijkl Êij Êkl − δjk Êil . (2.62)
ij
2
i,j,k,l

The MCSCF matrix consists of one-electron hij and two-electron gijkl ma-
trices containing all information about the MOs and their coefficients. The
density matrices Dij and Pijkl contain the CI coefficients and, thus, are
coupling the individual SDs.12
MCSCF methods are - contrary to the CI methods discussed above - not
designed for obtaining a large fraction of dynamic correlation. They are
rather used for systems with a large contribution of static electron correlation,
which in general includes every system in which more than one non-equivalent
resonance structure is important.13 The different structures/configurations
|Ψi , represented by optimized SDs14 , generate the multiconfigurational wave
function |ΨMCSCF . Examples for systems with such a multiconfigurational
12 This describes a solution for a more-determinant system like seen in the section of
Configuration Interaction, but in this case through a second quantification approach.
13 This means that even the ground state wave function cannot be described qualitatively

correct by Hartree-Fock. Therefore, already the reference single determinant wave

function is poor and, thus, multi configuration (determinant) corrections to the ground
state are essential.
14 Since the MOs of each SD contributing to the total multiconfigurational wave function

are optimized individually, it is referred to as optimized SDs.

26 2 Theory

character are biradicals and conjugated π-systems. In conclusion, MCSCF

methods are mainly used for generating a qualitatively correct wave function
through recovering the static electron correlation and not to introduce
a dynamical correction to the energy. The latter one has to be carried
out with additional methods, like perturbation theory or multireference
calculations. [89,103,109]

Complete Active Space SCF

The major challenge of MCSCF methods is selecting the correct configura-
tions |Ψi . The most common approach for doing so is the Complete Active
Space Self-Consistent Field (CASSCF) method [110] , where the selection is
achieved by partitioning the MOs into active, inactive and frozen spaces.
Within the active space a full CI is done with optimizing all MOs; in the
inactive space only the MOs are optimized, while in the frozen space no
further optimizations are done.
The active space typically incorporates some of the highest occupied and
some of the lowest unoccupied MOs, whereas the inactive and frozen space
have either two (occupied) or zero (virtual) electrons. A common notation
is [n,m]-CASSCF indicating that n electrons are distributed in m orbitals
within the active space.
By selecting the active space, the contributing determinants are chosen
through the full CI permutations within it and, thus, static correlation
is introduced. When the active space is increased, more configurations
are taken into account and consequently a higher amount of dynamical
correlation is achieved.
The most challenging part is selecting the important MOs; thus, a good
chemical understanding of the system and its orbitals is indispensable to
investigate the processes of interest. For example, the right antibonding
orbitals for a dissociation, or possible excited triplet states for charge transfer
observations should be considered. Some general hints are the use of the
Highest Occupied Molecular Orbital (HOMO) and the Lowest Unoccupied
Molecular Orbital (LUMO), valence orbitals and corresponding pairs of
occupied and virtual orbitals. [89,103]

Restricted Active Space Self-Consistent Field

In general, a [n,m]-CASSCF with more than 16 electrons and/or orbitals
is not feasible due to an exponential increase in the CSFs. For example, a
[n,n]-CASSCF with n = 14 results in 2.76 million CSFs.
2.5 Electron Correlation Methods 27

An extension of CASSCF that aims to compensates the size problem,

is the so-called Restricted Active Space Self-Consistent Field (RASSCF)
method. [111,112] Here the active MOs are further divided into three categories
RAS1, RAS2 and RAS3 with restrictions to their occupation number and
excitation levels. The RAS1 space consists of selected doubly occupied MOs
(in the HF reference), and the RAS3 of virtual ones. The RAS2 is identical
to the active space in CASSCF and subjected to a full CI treatment. RAS1
and RAS3 are restricted in terms that only a specific number of electrons
are allowed to leave or respectively enter the subspaces. Usually, the number
ranges from zero to two, representing a truncated CI treatment. This results
in less CSFs; however, the computational setup and the selection of a suitable
AS is more complicated compared to CASSCF. Furthermore, the convergence
is worse than in CASSCF. [89,111,112]

State Average MCSCF

One of the most common applications of MCSCF methods is calculating
excited states for photophysical and photochemical processes. However,
especially excited state calculations with RASSCF tend to have convergence
problems, root flipping, and other difficulties. An approximation to overcome
these issues is to use a state average scheme. This means that the orbitals
are optimized for a number of N electronic states, which results in an energy
written as an average of the individual state energies:


N
Eav = N −1 Ei . (2.63)
i

Consequently, the energy is obtained by introducing average density matrices

(compare to Equation 2.61):
 
av av
Eav = hij Dij + gijkl Pijkl . (2.64)
i,j i,j,k,l

State average calculations are useful to describe near degeneracy states (of
the same symmetry) or even conical intersections. [103]

CAS Perturbation Theory through Second-Order

As discussed before, CASSCF and RASSCF include mainly static corre-
lation to the HF reference. As a result, dynamic correlation effects are
missing. A possible method to include such effects is perturbation theory.
28 2 Theory

Analogous to Møller-Plesset Perturbation Theory (MP) [113] , which adds

dynamic correlation to a HF reference, this can be achieved for a MC-
SCF reference with Complete Active Space Perturbation Theory through
Second-Order (CASPT2). [114] The combination of CASSCF (RASSCF) and
CASPT2 (RASPT2) is a very successful combination for treating both kinds
of EC and for solving a variety of problems.

Rayleigh-Schrödinger perturbation theory Both methods, MP and

CASPT2, are based on Rayleigh-Schrödinger perturbation theory, which
will be introduced briefly in the following.
Consider a TISE (Equation 2.2), to which the solutions are unknown.
First of all, the Hamiltonian is divided up into two parts
Ĥ = Ĥ0 + λĤ  (2.65)
with Ĥ0 being the unperturbed Hamiltonian (reference) with known eigen-
(0)
functions |ψi :
(0) (0) (0)
Ĥ0 |ψi  = Ei |ψi  , (2.66)
and Ĥ  the Hamiltonian for the unknown part (perturbation). In the
next step, the wave function and energy are expanded in form of a Taylor
expansion:
(0) (1) (2)
 (n)
Ei (λ) = Ei + λEi + λ2 Ei + ... = λn E i (2.67)
n=0
(0) (1) (2)
 (n)
2
|Ψi  (λ) = |ψi  + λ |ψi  +λ |ψi  + ... = λn |ψi  , (2.68)
n=0

under the condition of intermediate normalization:

(0) (n)
ψi |ψi  = 0 . (2.69)
The TISE is expanded by Equation 2.67 and Equation 2.68, and ordered
after the level of perturbation λn .
(0) (0) (0)
λ0 : Ĥ0 |ψi  = Ei |ψi  (2.70)
(1) (0) (0) (1) (1) (0)
λ1 : Ĥ0 |ψi  + Ĥ  |ψi  = Ei |ψi  + Ei |ψi  (2.71)
2 (2) (1) (0) (2) (1) (1) (2) (0)
λ : Ĥ0 |ψi  + Ĥ  |ψi  = Ei |ψi  + Ei |ψi  + Ei |ψi 
(2.72)
(n) (n−1)

n
(j) (n−j)
λn : Ĥ0 |ψi  + Ĥ  |ψi = Ei |ψi  (2.73)
j=0
2.5 Electron Correlation Methods 29

(0)
By multiplication from the left with a specific wave function ψi |, the
energy and wave function corrections of specific order n are obtained. A
general solution for the energy is:
(n) (0) (n−1)
Ei = ψi |Ĥ  |ψi . (2.74)

By expanding the first-order correction to the wave function in the known

complete set of unperturbed functions
(1)
 (1) (0)
|ψi  = cj |ψj  (2.75)
j

the explicit first, second- and third-order energy correction, as well as the
first-order wave function coefficients can be obtained:
(1) (0) (0)
Ei = ψi |Ĥ  |ψi  (2.76)
(0) (0)
(1) ψj |Ĥ  |ψi 
cj = (0) (0)
(2.77)
Ei − Ej

(2)
 (1) (0) (0)
 ψi(0) |Ĥ  |ψj(0)  ψj(0) |Ĥ  |ψi(0) 
Ei = cj ψi |Ĥ  |ψj  = (0) (0)
. (2.78)
j j Ei − Ej

In general, knowledge of the nth-order wave function allows a calculation of

the (2n + 1)th-order energy. [89,90,115,116]
In the MP perturbation theory, the unperturbed reference is HF and
the perturbation Hamiltonian the exact (not averaged) electron-electron
repulsion minus the averaged electron-electron potential from HF.
(0)
|ψi  = |ψHF  (2.79)
 1
Ĥ  = Ĥ − Ĥ0 = − 2 Vee  (2.80)
r
i<j ij

Applying this on Equation 2.78 results in MP2, which mixes excited de-
terminants with the HF reference. Because of the Slater Condon rule and
Brillouin theorem, only doubly excited SDs have to be taken into account.
This yields a dynamic energy correlation correction. [89,90,113]
occ. virt.
  ψHF |Ĥ  |ψ rs  ψ rs |Ĥ  |ψHF 
E (2) = ab ab
(2.81)
EHF − Eab
rs
a<b r<s
30 2 Theory

CASPT2 formalism For CASPT2 the same principle applies as for MP2.
The energy correction in second-order relies on the first-order wave function
and is constructed using doubly excited determinants generated from the
CASSCF reference wave function. But only doubly excited SDs that are not
already in the CASSCF reference are used.15
Therefore, the first-order wave function is written as:
 (K) (K)
|Ψ(1)  = tP X̂P |Ψ(0)  (2.82)
KP
(K)  
X̂P ∈ Êvjtu , Êvjti , Êatvx , Êaivx , Êvjai , Êbvat , Êbjat , Êbjai , (2.83)

where K is an index denoting the excitation type and P is the index of an

individual excitation of that K type. The different two-electron excitations
contributing are shown in Table 2.1. The notation is the following: The
inactive space is labeled i, j, k, l, the active t, u, v, x and the virtual/secondary
(K)
has the indices a, b, c, d. The parameters tP are called amplitudes and are
the coefficients that weight the excitations.
Looking at the contribution of each excitation can give an insight into the
quality of the chosen reference CASSCF wave function. Too high contribu-
tions indicate missing orbitals in the active space of the reference. [103,114]

Reference weight and level shift The reference weight is an important

value to estimate the quality of the CASPT2 wave function. If - apart from
the ground state - excited states are calculated, all states should have the
same reference weight. Moreover, so-called intruder states often result in a
small reference weight. These intruder states are states that are energetically
close to the reference with large amplitudes, which are, thus, close to the
eigenvalue of the zeroth-order wave function, and are a clear indication
that the perturbational treatment failed or is not feasible using the applied
computational setup. [103,114]
A way to overcome this problem of small or inconsistent reference weight
is a level shift. This shift separates the excited SDs from the reference state

15 In general, it is requested that the n-th order wave function lies in the n-th order
interacting space. This space is defined by the residual wave function

ρ = Ĥ − E0 |Ψ(0)  ,

which measures the error of the reference to the exact wave function Ψ and is spanned
by all SDs that are not in the reference, but differ by at most two-electron excitation
from at least on reference SD.
2.5 Electron Correlation Methods 31

in the unperturbed Hamiltonian matrix energetically. The concept is best

explained by a two-level system.
   
α 0 0 δ
H0 = H = (2.84)
0 β+γ δ −γ

Table 2.1: Labeling and contributions for the configurations in CASPT2 from
the different excitation operators Epqrs . P and M describes the
positive and negative linear combination.

Config. Excitation 1
VJTU Inactive (J) → Active (V)
VJTIP Inactive (J) → Active (V)
VJTIM Inactive (J) → Active (V)
ATVX Active (T) → Secondary (A)
AIVX Inactive (I) → Secondary (A)
or: Active (X) → Secondary (A)
VJAIP Inactive (J) → Active (V)
VJAIM Inactive (J) → Active (V)
BVATP Active (V) → Secondary (B)
BVATM Active (V) → Secondary (B)
BJATP Inactive (J) → Secondary (B)
BJATM Inactive (J) → Secondary (B)
BJAIP Inactive (J) → Secondary (B)
BJAIM Inactive (J) → Secondary (B)
Config. Excitation 2
VJTU Active (U) → Active (T)
VJTIP Inactive (I) → Active (T)
VJTIM Inactive (I) → Active (T)
ATVX Active (X) → Active (V)
AIVX Active (X) → Active (V)
or: Inactive (I) → Active (V)
VJAIP Inactive (I) → Secondary (A)
VJAIM Inactive (I) → Secondary (A)
BVATP Active (T) → Secondary (A)
BVATM Active (T) → Secondary (A)
BJATP Active (T) → Secondary (A)
BJATM Active (T) → Secondary (A)
BJAIP Inactive (I) → Secondary (A)
BJAIM Inactive (I) → Secondary (A)

The two states α and β are coupled through the perturbation δ and are
separated by a level shift γ. In the whole, exact Hamiltonian H = H0 + H 
the shift γ is gone and, therefore, does not influence the energy. However,
benchmark calculations have shown that the level shift cannot be removed
32 2 Theory

totally and a residue of up to 1% is possible. [103,117,118] Applied to second-

order perturbation theory for ψ0 as reference, the energetic level shift γ
appears in the equations as follows:

(1) ψj |Ĥ  |ψ0 


cj = (2.85)
E0 + γ − E j
 | ψj |Ĥ  |ψ0  |2
 (2) =
E (2.86)
j
E0 + γ − E j

The influence of the level shift is indicated by a tilde. By introducing

the level shift, intruder states will be removed, but the second-order energy
exhibits a dependency on γ, which has to be removed by an approximate back
transformation of the second-order energy to the unshifted value without
intruder states. Equation 2.87 shows a transformation of Equation 2.86 that
can be further simplified by the assumption Ej − E0 >> γ:
 (1)  γ

 (2)
E =E +γ (2)
|
cj | 1 + (2.87)
j
Ej − E0
 
 (2) − γ 1 − 1 ≡ E (2) ,
E (2) ≈ E (2.88)
LS
ω
where ω is the weight of the reference wave function in the level shifted
(2)
CASPT2 calculation and ELS the level shift corrected second-order energy,
which will only differ from its non-shifted equivalent when the assumption
above is violated.

2.6 Density Functional Theory

Until now, the central element of the discussed methods has been the wave
function Ψ containing all information about a system. For a system of N
electrons, this means a 4N dependency (three spatial coordinates and one
spin coordinate for each electron). A different approach is the use of the
electron density ρ(r1 ) instead.

ρ(
r1 ) = N ··· |Ψ (
r1 , ω1 ,  rN , ωN ) |2 dω1 , d
r2 , ω2 , ...,  r2 , dω2 , ..., d
rN , dωN

(2.89)
Equation 2.89 gives the probability of finding an electron with arbitrary spin
in the volume element r1 . The main advantage is the dependency on only
three coordinates regardless of the size of the system.
2.6 Density Functional Theory 33

The main goal of Density Functional Theory (DFT) is to design functionals

connecting the electron density with chemical properties. e.g. the energy. [89]

2.6.1 Hohenberg-Kohn Theorems

The foundation of DFT was provided by Hohenberg and Kohn and their
theorems. [119]

Theorem I
The first theorem proofs that the ground state electronic energy is determined
completely by the electron density ρ0 :

E0 = E0 [ρ0 ] = Ψ[ρ0 ]|Ĥ|Ψ[ρ0 ] . (2.90)

In the definition of E.B. Wilson [120] this means that the integral of the
density defines the number of electrons, the cusps in the density the position
of the nuclei and the height of the cusps the corresponding nuclear charges.

Theorem II
The expectation value of the Hamiltonian (Equation 2.90) follows the varia-
tional principle.
E[ρ0 ] ≤ E[ρ ] (2.91)
An arbitrary test density ρ can never result in a smaller energy than the
real electron density of the ground state ρ0 .

2.6.2 Kohn-Sham Theory

The energy functional can be divided into three parts (and the nuclear-nuclear
repulsion as a constant within the BOA), kinetic energy T [ρ], attraction
between nuclei and electrons Ene [ρ], and electron-electron repulsion Eee [ρ].
With reference to HF theory, the latter one can be separated into a Coulomb
and an exchange part.

E[ρ] = T [ρ] + Ene [ρ] + Eee [ρ] (2.92)

= T [ρ] + Ene [ρ] + J[ρ] + K[ρ] (2.93)
34 2 Theory

The Ene [ρ] and J[ρ] functionals can be expressed easily in terms of the
electron density.


K  α )ρ(r)
Zα (R
Ene [ρ] = − dr (2.94)
 α − r|
|R
α
1 ρ(r)ρ(r  )
J[ρ] = drdr  (2.95)
2 |r − r  |

However, the kinetic functional with a second derivative of the wave function
and the exchange functional cannot be expressed in dependence on the
electron density. [121]

Kohn-Sham Ansatz
A solution to this is given by Kohn and Sham, who suggest a division of the
electron kinetic energy into two parts. The first part TS [ρ] calculates an exact
expression for the energy with the assumption of non-interacting electrons,
whereas the second part is a small correction term TC [ρ] = T [ρ] − TS [ρ]. The
exact expression is gained through the reintroduction of orbitals in DFT.
Therefore, a Slater determinant composed of molecular orbitals ϕa (ri ) is
used to give the exact kinetic energy functional:


N
1
TS = ϕa (r1 )| − ∇2 (r1 )|ϕa (r1 ) , (2.96)
a
2

and a ground state density based on one-electron wave functions (MOs) as

one-electron densities:

(KS)

N
ρ0 (r) = |ϕa (ri )|2 . (2.97)
a

The remaining kinetic energy TC [ρ] is combined with the exchange functional
and the missing electron correlation to an exchange-correlation term Exc [ρ]:

EDFT [ρ] = TS [ρ] + Ene [ρ] + J[ρ] + Exc [ρ] (2.98)

Exc [ρ] = (T [ρ] − TS [ρ]) + (Eee [ρ] − J[ρ]) (2.99)

Unfortunately, the exact term of the exchange-correlation functional is

unknown. Deriving such an expression is considered the holy grail of DFT;
since it would make DFT exact within the BOA and the basis set limit. Thus,
2.6 Density Functional Theory 35

the various DFT methods (see. subsection 2.6.3) aim to find an approximate
solution.
Even though, the reintroduction of orbitals increases the coordinates from
3 to 3N , the computational costs are similar to HF theory and the results
are much better. And once the exchange-correlation functional has been
selected, the computational problem is very similar to HF. The energy is
optimized with orbitals ϕa (ri ), constrained to orthogonality by the Lagrange
method. Using the variational principle on the Lagrange function, like in
HF, gives an effective one-electron operator:
1 2
ĥKS (r1 ) = ∇ (r1 ) + V̂eff (r1 ) (2.100)
2
V̂eff (r1 ) = V̂ne (r1 ) + V̂J (r1 ) + V̂xc (r1 ) (2.101)
δExc [ρ] δxc (r2 )
V̂xc (r1 ) = = xc [ρ] + ρ(r2 ) dρ(r2 ), (2.102)
δρ(r1 ) δρ(r1 )

and a pseudo-eigenvalue equation, known as Kohn-Sham equation:

ĥKS (r1 )ϕa (r1 ) = a ϕa (r1 ), (2.103)

with canonical,16 , non-interacting Kohn-Sham orbitals. These orbitals are

expanded in an atomic basis set to give a matrix equation, similar to
Roothaan-Hall (Equation 2.45). The success of modern DFT methods is
based on Kohn and Sham and their suggestion to incorporate an auxiliary
set of orbitals to construct the electron density. All attempts to construct
orbital-free functionals depending directly on the electron density lack of
accuracy worse than in HF. Nevertheless, if such functionals could be derived,
the full potential of DFT in having only three size-independent coordinates
could be realized. [89,121,122]

2.6.3 Exchange-Correlation Functionals

It can be proven that the exchange-correlation energy is a unique functional
for all systems, but (so far) an explicit functional form could not be found.
The approximated functionals have in general a mathematical form contain-
ing parameters, which are obtained to fulfill certain criteria or by fitting to
experimental data (or a combination of both). The potential itself cannot
be evaluated analytically but must be generated by numerical integration.
16 Likein HF, a unitary transformation makes the Lagrange multiplier diagonal to yield
the pseudo-eigenvalue equations.
36 2 Theory

Thus, defining a grid with a specific number of grid-points is essential to DFT

calculations.17 Also, it is important to point out that - contrary to HF - there
is no systematic, theoretically founded way to improve a given functional,
among others due to the fact that the exchange-correlation functional are to
a large extent empirically fitted. [123,124]
Generally, there are three different kind of exchange-correlation functional
groups, namely Local Density Approximation (LDA), Generalized Gradient
Approximation (GGA) and hybrid methods.

Local Density Approximation

In the LDA it is assumed that the density is a slowly varying function. Thus,
the exchange-correlation functional and its energy xc are only functions of the
electron density itself. If the α and β spin functions are not in equal spatial
orbitals (unrestricted case), Local Spin Density Approximation (LSDA) is
used. For the exchange part, a Dirac model is used, whereas the correlation
part is parameterized.
LDA functionals are exact for the special case of a uniform electron gas,
good for metals and poor for ions. For molecular systems the exchange energy
is underestimated by around 10% and the correlation energy is overestimated
by a factor of two. Therefore, bond strengths are often overestimated by up
to 100 kJ/mol.
Examples for LDA functionals are VWN, PZ81 and PW92. [125–128]

Generalized Gradient Approximation

In the GGA method ρ and its first derivative, ∇ρ, are included as variable.
This method represents a strong improvement and is widely used nowa-
days, often in combination with hybrid functionals (see PBE0 composition,
Equation 2.104). There are different GGA functionals for exchange and
correlation part. [129–131]
Examples for GGA functionals are:
• exchange part, Ex : B(88), B97, OPTX
• correlation part, Ec : LYP
• both, Exc : PW86, PW91, PBE

17 Every quantum chemical program has a default grid setting, which can be adjusted.
2.6 Density Functional Theory 37

Hybrid Methods
Hybrid methods use an exact expression for the exchange functional via HF
exchange potential. This is done by using the already present Kohn-Sham
orbitals, which results in very accurate functionals, applicable to a lot of
chemical systems.
In general, the exact exchange is added to expressions from other func-
tionals. E.g. the PBE0 functional consists of the GGA PBE functional and
exact HF exchange: [132]

PBE0 1 HF 3 PBE
Exc = E + Ex + EcPBE . (2.104)
4 x 4
The exact amount of HF exchange potential is different for each functional.
Examples for hybrid functionals are, besides PBE0 (25% HF exchange),
B3LYP (20% HF exchange) and TPSSh (10% HF exchange). [132–136]

2.6.4 Time-Dependent Density Functional Theory

An extension of the ground state method DFT is Time-Dependent Density
Functional Theory (TDDFT), which uses a time-dependent, external, electric
potential Vext to allow the description of excited states.
A Hohenberg-Kohn like theorem for TDDFT is the Runge-Gross theorem,
which states that the unique one-to-one correspondence between an external
potential and the electron density holds both in time-dependent and time-
independent cases. Furthermore, the potential is also dependent on the wave
function |Ψ0  of the system at time t0 . This dependency vanishes if it is the
ground state wave function, |Ψ0  = |Ψgs  [137,138] :
|Ψ0 =|Ψgs 
V (r, t) = V [ρ, Ψ0 ](r, t) = V [ρ](r, t) (2.105)

The time-dependent Hamiltonian consists of a (time-dependent) Fock

operator fˆ(r, t) and the external potential V̂ext (r, t). These two can also be
written as a kinetic energy T̂ (r) and an effective potential operator V̂eff (r, t).
The latter one can be decomposed into a nuclear-electron term, Coulomb and
exchange potentials from the electron-electron interaction and the external
potential.

Ĥ(r, t) = fˆ(r, t) + V̂ext (r, t) (2.106)

= T̂ (r) + V̂eff (r, t) (2.107)
ˆ r, t) + K̂(r, t) + V̂ext (r, t)
V̂eff (r, t) = V̂ne (r, t) + J( (2.108)
38 2 Theory

Under the same assumptions as in DFT’s Kohn-Sham theory (section 2.6.2),

the time-dependent density is obtained from a set of Kohn-Sham orbitals
corresponding to non-interacting electrons in a one SD Ansatz:


N
ρ(KS) (r, t) = |ϕa (r, t)|2 . (2.109)
a

Considering the TDSE (Equation 2.1) this leads to a set of time-dependent

Hartree-Fock equations, which must be solved self-consistently.
∂ 1
i |ϕa (r, t) = T̂ (r) + V̂eff (r, t) |ϕa (r, t) = − ∇2 + V̂eff (r, t) |ϕa (r, t)
∂t 2
(2.110)
The main but significant difference to HF or DFT is the fact that the energy
is not a conserved quantity in the time-dependent case; thus, the variational
principle is not applicable. Instead a stationary condition for the action is
defined by:  
t
∂
S[Ψ] = Ψ| i − Ĥ |Ψ dt . (2.111)
t0 ∂t
Due to a Kohn-Sham like Ansatz, the Hamiltonian can be expressed in
the form of an exchange-correlation functional containing the remaining
kinetic correction as well as the exchange and correlation part from the
electron-electron-repulsion. This results in an effective potential of:
ˆ r, t) + V̂xc (r, t) + V̂ext (r, t)
V̂eff (r, t) = V̂ne (r, t) + J( (2.112)

Apart from the exchange-correlation functional, the terms in the equa-

tion above can be treated in the same way as in the time-independent
case (with the exception of using time-dependent wave functions |ϕa (r, t)).
However, the exchange-correlation functional is more complicated. In the
time-independent case it is given by the functional derivative of the exchange-
correlation energy (Equation 2.102), but in the time-dependent case, it is
the functional derivative of the exchange-correlation action Sxc [ρ].

δSxc [ρ]
Vxc [ρ, Ψ0 , ΨKS ](r, t) = (2.113)
δρ(r, t)

Thus, the exchange-correlation functional depends on the initial density at

time t0 , as well as on the time-interval t − t0 .18 Furthermore, Vxc depends
18 Because the action includes the integral over time for each wave function (Equa-
tion 2.111).
2.6 Density Functional Theory 39

on Ψ0 of the exact interacting system and the initial state of Kohn-Sham

system, ΨKS . Subsequently, the solutions for the time-dependent Hartree-
Fock equations (Equation 2.110) must also be solved self-consistently in the
time domain; it is referred to as the time-memory effect. This computational
highly demanding effect is essentially neglected by all common TDDFT
methods by making an adiabatic approximation. It assumes that the density
is a slowly varying function of time, which reduces the exchange-correlation
functional to its time-independent case in Equation 2.102. [89,122,139]

Linear response theory

For implementing TDDFT - like presented above - a code must deal with two
computational heavy tasks. Apart from a representation in space (numerical
or with a basis set expansion) a propagation in time using time-depending
wave functions and time evolution operators has to be applied. This is done
in real-time TDDFT and will not be of interest in this section.
An approach for small perturbations representing a small external potential
is the linear response theory. It is widely used for spectroscopy, where the
response to a weak probe (external potential) is used to determine spectral
properties. The goal of linear response theory is to directly calculate the
change of a certain observable subjected to a perturbation by first-order,
without calculating the change of the wave function.
A system of interacting particles in its ground state is considered, and
at t0 = 0 a perturbation is switched on. The response of any observable to
the perturbation (in this case an external potential) can be expressed in a
Taylor series for the density
ρ(r, t) = ρ0 (r) + ρ1 (r, t) + ρ2 (r, t) + ... (2.114)
Here, ρ0 (r) is the ground state density and ρ1 (r, t) the linear density response
to the small perturbation. Higher, non-linear terms are neglected in linear
response theory. The linear density response can be written as:
∞
ρ1 (r, t) = dt d3 r χ(r, t, r  , t )Vext (r  , t ), (2.115)
0

with χ(r, t, r  , t ) being the density-density response function, which is de-

fined as:
χ(r, t, r  , t ) = −iθ(t − t ) Ψgs | [ρ̂H0 (r, t), ρ̂H0 (r  , t )] |Ψgs  . (2.116)
The step function θ(t − t ) ensures that the response comes after the per-
turbation (casual function) and the commutator of density operators allows
40 2 Theory

mixing of states, which is seen better by inserting the identity in form of

the completeness of all interacting states and Fourier-transforming of the
time domain to the frequency domain. The result is a sum-over-states form,
known as the Lehmann representation:
∞

 Ψgs |ρ̂(r)|Ψn  Ψn |ρ̂(r  )|Ψgs 

χ(r, r , ω) =
n=1
ω − Ωn

Ψgs |ρ̂(r  )|Ψn  Ψn |ρ̂(r)|Ψgs 
− , (2.117)
ω + Ωn

with Ωn = En − Egs . The equation above shows that the resonant part of the
response function has poles at the exact excitation energy of state n of the
system. The remaining numerator at the poles represents transition moments
between the ground and excited state n. Simply put, if a perturbation
potential is applied and its frequency matches one of the excitation energies,
the response of the system in the frequency domain is very large, seen as a
peak in the spectrum.
In TDDFT, the linear density response is calculated as the response of the
(KS)
time-independent, non-interacting Kohn-Sham system ρ0 (Equation 2.97)
to an effective, time-dependent, perturbation potential (see Equation 2.112)
with a monochromatic dipole field along z direction as external potential:

V̂ext (r, t) = Aez sin(ωt) . (2.118)

Using Kohn-Sham orbitals yields:

 ϕj (r )ϕ∗k (r)ϕ∗j (r  )ϕk (r  )
χKS (r, r  , ω) = (fk − fj ) , (2.119)
ω − (j − k )
k,j

where fk and fj are the occupation numbers referring to the configuration

of the Kohn-Sham ground state. This Kohn-Sham density-density response
function is used together with the effective potential to give the linear
response density (Equation 2.115). Its poles yield the excitation energy of
different states, as described above.19 [89,122,139–142]
19 It is important to point out that this is only the Kohn-Sham density-density response
and Kohn-Sham linear response density. Therefore, it gives only Kohn-Sham (non-
interacting system) excitation energies. In computational programs, the real linear
response density is calculated by introducing further changes on the external potential
(for more see Ref. [139] - xc kernel) The exact solutions are eigenmodes of the so-called
Casida equations.
2.7 Spin-Orbit Coupling 41

2.6.5 (TD)DFT Problems

Despite the wide success of DFT and TDDFT, there are some issues where
these methods are known to fail.
First of all, (TD)DFT is based on a single determinant approach and,
therefore, performs poorly regarding static correlation, i.e. for multiconfigu-
rational problems. Further known problems arise from weak interactions due
to dispersion forces and long-range behavior in general. This results in bad
descriptions of loosely bound electrons, radicals, Rydberg states, dissociative
and charge transfer systems. General problems arising with boron20 contain-
ing systems have also been reported and will be addressed in this work. This
includes especially the calculations of excited state properties. [68,143–150]

2.7 Spin-Orbit Coupling

In spectroscopy and photophysics populating triplet states and other states
of higher multiplicity through a crossing with a closed shell system, is an
important effect. Since these higher spin states often exhibit long lifetimes
and charge separated properties, they are essential to describe the overall
photo-induced behavior of a system. An important step to treat this kind of
interaction between states of different multiplicity of a given molecule (or
atom) is Spin-Orbit Coupling (SOC), which offers a quantitative description
of ISC as relativistic treatment upon the Born-Oppenheimer Hamiltonian.
The Hamiltonian does not treat relativistic effects; thus, the spin is
added ad hoc to the wave function (recall subsection 2.4.1). To incorporate
relativistic effects, a distinction between two different effects has to be made:
i) One is the so-called (spin-independent) scalar-relativistic effect, which
is important for the chemistry of heavy atoms. Their core electrons move
with a velocity near the speed of light, which results in a contraction of the
low lying s and p orbitals and, therefore, a lowering of their energy. The
opposite effect can be seen for d and f orbitals. Because of the contraction
of the core orbitals, they "see" a smaller nuclear charge; known as shielding,
which leads to an energetic destabilization of the outer orbitals (d and f ).
ii) The second kind of relativistic effect is SOC; it describes mixing of
states of different multiplicities (ΔS = 0, ±1) and energetic splitting of the
different MS states of a given multiplicity S. This correction applies for heavy

20 The Boron atom has the electron configuration 1s2 2s2 2p1 with a single p-electron
that can be distributed in three different, but energetic degenerated p-orbitals. This
multiconfigurational property leads to problems in (TD)DFT.
42 2 Theory

and also for light elements and impacts photophysical and photochemical
properties. [151–153]
The exact, but also most expensive way, to implement relativistic ef-
fects would be the four-component Dirac equation [154,155] instead of the
Schrödinger equation. A common kind of simplification, including the preser-
vation of the Schrödinger equation as basis of all calculations, is given by
the Douglas-Kroll-Hess (DKH) transformation. [156] It transforms the Dirac
equation in a spin-independent scalar-Hamiltonian and a spin-dependent
SOC-Hamiltonian.21 [151,152,157]
The spin-independent scalar-relativistic effect is usually implemented
in form of a correction to the one-electron integrals and/or as Effective
Core Potential (ECP). For the spin-dependent SOC-Hamiltonian different
approaches are available. The most common one is the Breit-Pauli (BP)
Ansatz with the corresponding BP operator defined as follows: [151,152,158,159]
SOC (1) (2)
ĤBP = ĤBP + ĤBP (2.120)
(1)
 α 
ĤBP = ĥ1el−SOC
i = −3 ˆ
Zα riα liα ŝi (2.121)
i
2 i α
(2) (2) (2)
ĤBP = ĤSSO + ĤSOO (2.122)
 α  
2el−SOC −3 ˆ 2 −3 ˆ
= gij =− Zα rij lij ŝi − α rij lij ŝj .
2 i
i j=i j=i i j=i

The BP-Hamiltonian is separated in a one- and two-electron term. The

one-electron term describes the interaction of electron i at nucleus α with
the magnetic moment induced by the motion of nucleus α. The two-electron
Spin-Same Orbit (SSO) term represents the interaction of electron i with the
magnetic momentum induced by the movement of electron i in the field of
electron j. The counterpart Spin-Other Orbit (SOO) is the coupling of the
spin of electron i with the orbital angular momentum of electron j and vice
versa. [151,152,158,159] However, this results in a lot of two-electron four-center
integrals. Therefore, further approximations, especially for the two-electron
part, are implemented in different methods.
The most commonly used approximation is the so-called mean-field Ansatz.
It is implemented in programs like ORCA [160] and MOLCAS [161] and ap-
proximates the two-electron part as a mean field like the electron-electron
21 This is the result of a sequence of unitary transformations that removes the mathemat-
ical demanding coupling of the large and small components of the Dirac one-electron
through external potentials of nth-order (where n = ∞ would represent an exact
splitting).
2.7 Spin-Orbit Coupling 43

repulsion in HF. The SOC between two states is calculated by expressing the
two states as two orthonormal SDs with biorthonormal MOs, which differ
only by an excitation χa → χr .22 This offers an expression of the SOC energy
correction in terms of one-electron and two-electron integrals. [151,152,163–165]
Ψ1 |ĤSOC |Ψ2  (2.123)
= χa |ĥ1el−SOC |χr  (2.124)

2el−SOC 2el−SOC
+ (χa χj |g12 |χr χj  + χj χa |g12 |χj χr  (2.125)
j
2el−SOC 2el−SOC
− χa χj |g12 |χj χr  − χj χa |g12 |χr χj ) (2.126)
The two electron integrals can be divided into a Coulomb (SOO) part
(Equation 2.125) and an exchange (SOO) part (Equation 2.126).
The aim is now to generate an effective one-electron SOC operator, com-
parable to the Fock operator in HF. This means a separation of the spin
and orbital angular momentum part:

ĤSOC = ˆli ŝi . (2.127)
i

The one-electron part (Equation 2.124) is already in the desired form. The
two-electron parts can be separated by the following mechanism:
2el−SOC
χa χb |ĝ12 |χr χs  = (ϕa ϕb |ĝ SOC |ϕr ϕs ){δsb ss sa |ŝ|sr +2δsa sr sb |ŝ|ss } .
(2.128)
Here, s(ω) represents the spin function with spin α or β, whereas the orbital
angular momentum two-electron operator is given as:
α2 ˆ −3
ĝ SOC (i, j) = −
lij rij . (2.129)
2
The outcome of these conversions allows a treatment of SOC as effective one-
electron operator. [163,164] Qualitatively speaking, this results in a notation
of:
Ψ1 |ĤSOC |Ψ2  (2.130)
= ϕa sa (ω)|ĤSOC |ϕr sr (ω)

= ϕa |ˆli |ϕr sa |ŝi |sr  .
i
22 Thisis the Restricted Active Space State Interaction (RASSI) implemented in e.g.
MOLCAS. [162] There is a variety of operator that can be chosen for the interaction
between the states.
44 2 Theory

Equation 2.130 shows that depending on the symmetry of the system, SOC
does not occur between all states of a given spatial symmetry.23 [166]
Another implementation with two additional but reasonable approxi-
mations is Schimmelpfennig’s Atomic-Mean-Field Integral (AMFI) pro-
gram. [165,167] The approximations are that: i) only one-center terms are
retained, and ii) the "mean-field-orbitals" are replaced by atomic SCF or-
bitals obtained from spherically averaged atomic SCF calculations with pre-
determined valence shell occupations, instead of the complete non-spherical
molecular density.

23 The orbital angular momentum transforms as rotation. In C2v no rotation operator is

total symmetric (A1 ), which means that if two states are of the same spatial symmetry,
they do not interact through SOC.
3 Computational Details
In this chapter, the computational setups used for the quantum chemical
calculations in this thesis are introduced. The active space will be discussed
thoroughly. All methods and calculations were applied on the BODIPY dye
introduced in chapter 1 and shown in Figure 1.5.

3.1 Used Methods

Most of the computational details are based on the previous work by
Dr. Stephan Kupfer, [86] as well as the basis set and functional studies
by M. Momeni et al. [68]

Geometry optimization and Model system

The starting geometry for the BODIBY dye was taken from similar, slightly
modified structures investigated in Reference [86] and re-optimized with DFT
and the hybrid-functional PBE0. [135] The correlation-consistent triple-ζ-
basis set cc-pvtz [168] with the ECP MWB 46 [169] for Iodine was used. The
calculations were done in Gaussian 09. [170]
A vibrational analysis in the re-optimized structure yielded four imaginary
frequencies, which involve the torsion at the four methyl groups and the
phenyl ring (see Figure 3.1). But taking these geometry distortion into
account and re-optimizing the structure, means a loss of symmetry and,
thus, the system would not be applicable to RASPT2 due to the exceeding
computational costs. Furthermore, the excited state character in TDDFT
gets more complicated and diffuse due to a high amount of mixed transitions
contributing to each excited state, with weights of the desired excitation of
less than 50%. This would make a qualitative interpretation very challenging.
Furthermore, the crystal structure for this component obtained by Li
Fu et al. [171] shows no distortion of the methyl groups. Another indicator
that this might be a gas phase phenomenon is an optimization upon the
optimized gas phase geometry with a Polarizable Continuum Model (PCM)
for tetrahydrofurane ( = 7.4257, n = 1.4070) as solvent. Here, the system

© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2019

K. M. Ziems, Photophysics and Photochemistry of a BODIPY-Based
Photosensitizer, BestMasters, https://doi.org/10.1007/978-3-658-26188-7_3
46 3 Computational Details

ν = −81 cm−1 ν = −57 cm−1

ν = −51 cm−1 ν = −42 cm−1

Figure 3.1: Imaginary vibrations at ground state (GS) (A1 ) geometry

gets stabilized by 0.22 eV and loses two imaginary frequencies. The remaining
two vibrations are shown in Figure 3.2.
In conclusion, the negative frequencies were tolerated due to the stated
reasons above and a C2v model system introduced to lower the computational
costs and make the system applicable for state-of-the-art multiconfigurational
methods. In later sections, this phenomenon is revisited and will be discussed
in detail.
For the solvent calculation, the PCM was applied, which utilizes the
integral equation formalism (IEFPCM) and considers a solute cavity out of

ν = −84 cm−1 ν = −60 cm−1

Figure 3.2: Imaginary vibrations at GS (A1 ) geometry in THF (PCM).
3.1 Used Methods 47

a set of overlapping spheres centered at the atoms of the molecule. On this

surface the different contributions of the solvent-molecule interaction are
calculated; this includes electrostatic, dispersive and cavity forces. [51,172,173]
The singly reduced doublet of the investigated BODIPY dye was optimized
in the gas phase with the same computational setup and under the same
considerations regarding the imaginary frequency analysis. The displace-
ment vectors of the respective vibrational normal modes are illustrated in
Figure A.2 of the appendix.

DFT and TDDFT

All DFT- and TDDFT-calculations were performed in Gaussian 09. [170]
The hybrid-functional PBE0 [135] was used and the correlation-consistent
double-ζ-basis set cc-pvdz [168] . To reduce the computational costs the core
potential MWB 46 [169] was utilized for the Iodine atoms.
For excited state optimizations no symmetry restrictions were applied,
to allow a full relaxation of the respective state. For selected excited state
optimizations the initial geometry was slightly distorted to prevent relaxation
into a stationary point of the PES. No differences to an optimization starting
from a non-distorted geometry was found.
For each calculation, the parameters, e.g. maximum step-size, number of
states, SCF-convergence method and others, were adjusted to the problem.
The default starting values were: i) maximum size for an optimization step
of 0.01N Bohr with N = 30, ii) 20 excited states of each multiplicity, and
iii) a combination of EDIIS and CDIIS [174] in the SCF procedure.

MCSCF
All MCSCF calculations, namely (MS-)RASPT2 and (SA-)RASSCF, were
calculated with Molcas 8.0 service pack 1 [161] and the ano-rcc-vdzp [175–180]
basis set.
Relativistic ano-type basis sets include a priori scalar-relativistic effects;
therefore, the integrals are calculated with a DKH transformation of one-
electron integrals and correlation of semi-core orbitals. Furthermore, this
basis set was studied and optimized for SOC and the use in the AMFI ap-
proximation. Thus, ano-rcc-type basis sets are ideal for incorporating scalar-
and SO-relativistic effects. All these concepts were thoroughly discussed in
section 2.7. [175–180]
For speed-up, all calculations were subjected to the Cholesky decompo-
sition of the two-electron integrals with the default threshold of 10−4 a.u. [181]
48 3 Computational Details

For (SA-)RASSCF calculations the convergence criteria were set to 10−6

a.u. for the energy, 10−3 a.u. for the orbital rotation-matrix and 10−3 a.u.
for the energy gradient. To allow a straight forward convergence in the
RASSCF routine, pure super-CI iterations or a mix of super-CI and Quasi-
Newton update procedures were used. [182,183] In general, four roots for both
BODIPY systems, each AS, each spatial symmetry and each multiplicity
were included; if necessary, more roots were incorporated, which led to
state-average calculations with six roots.
The (MS-)RASPT2 calculations on the RASSCF reference function were
performed with a level shift of mostly 0.3 a.u., which prevented the intro-
duction of intruder states and was thoughtfully analyzed by means of the
reference weight of the electronic states. This will be discussed further in
subsection 3.2.2 and the result part.
For SOC and transition dipole moments (to obtain oscillator strength)
the RASSI module was used. [162]
All calculations were done within the above-introduced C2v model system.

MEP
The Minimum Energy Path (MEP) was obtained by MOLCAS 8.2 [161] on
the (single state) RASSCF level of theory with the ano-rcc-vdzp basis set, the
Cholesky-decomposition, analytical gradients and applying C2v symmetry.
A MEP is the steepest descendant minimum energy reaction path on the
PES starting from a given geometry in an initial state. The path is built
through a series of geometry optimizations, each requiring the minimization
of the potential energy on a hyperspherical cross-section of the PES and a
predefined hyperspherical radius from the previous optimization step (or the
initial structure for the first step). [184,185]

3.2 Active Spaces for RASSCF/RASPT2

calculations
For the multiconfigurational and RASSI calculations, a suitable active space
is needed. Due to the large size of the dye, the space had to be chosen wisely
to limit the number of CSFs to a computationally manageable amount; thus,
the RASSCF approach was chosen over CASSCF. In the following sections,
the different active spaces are introduced, followed by some considerations
regarding the level shift for the (MS-)RASPT2 calculations.
3.2 Active Spaces for RASSCF/RASPT2 calculations 49

3.2.1 Active spaces for dissociation and charge transfer

First, the orbitals of the active space regardless of the RAS distribution
have to be considered in the framework of elucidating the photochemistry of
the present BODIPY dye. This includes especially orbitals contributing to
charge transfer and dissociative transitions.
Hence, the complete π-system of the main (heteroaromatic) ring and
the π-system of the side (phenyl) ring without the complete bonding and
complete anti-bonding orbital were incorporated to describe the highest
occupied and lowest unoccupied orbitals, which fit for excitation and charge
transfer. The complete bonding and complete anti-bonding orbital of the side
ring were excluded because they are energetically unfeasible for excitations.
Furthermore, the p-orbitals of the Iodines in the π-main-system plane are
included. To describe dissociative processes, the bonding and anti-bonding
σ-orbitals of the C-I bond are also part of the active space.
In orbital and electron notation, this means a (12,11) space for the main
ring, a (4,4) space for the side phenyl ring, for each Iodine a p-orbital (2,1)
space and for each C-I bond a (2,2) space describing the σ-bond. This
resembles overall 24 electrons in 21 orbitals, a (24,21) active space. The
orbitals are all displayed in Figure 3.3.
Second, following a general active space, the RAS distribution of its 24
orbitals has to be determined, since a complete active space approach would
exceed the available resources. In total, four active spaces, which differ in
their RAS distribution, were investigated on the RASPT2 level of theory
to establish a RAS for each research interest; one to describe CT - one to
describe dissociation. Since the number and nature of orbitals in the total
active space (RAS1-RAS3) is the same for all of the four investigated RAS
distributions, only the orbitals taken in the full-CI RAS2 space can be seen
in Table 3.1. The remaining orbitals are in RAS1 and RAS3 corresponding
to their occupation number. A complete overview of all RAS distributions
for the different ASs can be found in the Appendix in Table A.1.

Table 3.1: Different RAS2 distributions and their CSFs for different Active
Spaces (AS)

AS RAS 2 CSFs
1 σ(a1 ), σ(b1 ), σ ∗ (a1 ), σ ∗ (b1 ), π3 (a2 ), π4∗ (b2 ) 222869
2 σ ∗ (a1 ), σ ∗ (b1 ), π2 (a2 ), π3 (a2 ), π4∗ (b2 ) 71140
3 σ ∗ (a1 ), σ ∗ (b1 ), π2 (a2 ), π3 (a2 ), π4∗ (b2 ), π5∗ (b2 ) 122948
4 π2 (a2 ), π3 (a2 ), π4∗ (b2 ), πph,2 (b1 ), πph,3 (a2 ), πph,4
∗ ∗
(a2 ), πph,5 (b1 ) 534332
50 3 Computational Details

HF

Figure 3.3: All orbitals included in the RAS. The RAS distribution differs
in AS1-AS4, recall Table 3.1. Dashed line (HF) labels the MOs
occupied in the electronic singlet ground state. Orbital symmetry
is labeled by a1 , b1 , a2 , b2 .
3.2 Active Spaces for RASSCF/RASPT2 calculations 51

AS1 includes the bonding and anti-bonding σ-orbitals σ(a1 ), σ(b1 ), σ ∗ (a1 ),
σ (b1 ), as well as the HOMO and LUMO (π3 (a2 ) and π4∗ (b2 )) in RAS2. Since
∗

the bonding σ-orbitals are very low in energy with a occupation number of
2.0, they are attributed to RAS1 in AS2. Instead, the HOMO-1 (π2 (a2 ))
is moved to RAS2 for a better treatment of the excitations starting from
the occupied HOMO (and HOMO-1) orbital(s). In the third considered
active space (AS3) the LUMO+1 (π5∗ (b2 )) is taken into RAS2, additionally.
However, this does not result in any improvement in comparison with AS2.
The final active space, AS4, consists of HOMO (π3 (a2 )), HOMO-1 (π2 (a2 )),
LUMO (π4∗ (b2 )) and the four π-orbitals of the side ring, πph,2 (b1 ), πph,3 (a2 ),
∗ ∗
πph,4 (a2 ), πph,5 (b1 ). In summary, AS1-AS3 are suitable for dissociative
calculations, and AS4 for CT.
After a comparison of the excited state at GS (A1 ) geometry in all four
ASs, AS2 and AS4 were chosen to be utilized in further investigations on
the RASPT2 level of theory. AS1 and AS3 were neglected in favor of AS2,
because they have more CSFs, but do not present any improvement in
excitation energies. From now on, AS2 is also referred to as the dissociative
AS , since this AS involves excitations into the antibonding C-I orbitals,
σ ∗ (a1 ) and σ ∗ (b1 ), within RAS2 and, thus, is used to describe dissociative
processes. AS4 is the charge transfer AS because it includes the important
side-ring π/π ∗ orbitals in RAS2, as well as the HOMO and LUMO of the
boron moiety.
All calculated roots with each AS (and different level shifts) are collected
in section A.2.

3.2.2 Level shift considerations

Initially, all RASPT2 calculations have been performed without a level shift,
since an equal reference weight was obtained for all roots. Unfortunately,
after expanding the state average calculations to six roots for selected spatial
symmetries in the non-reduced system1 and after starting to investigate the
reduced system, deviations from the reference weight of former calculations
occurred (see all reference weights in Appendix B). Therefore, all calcu-
lations were redone with a level shift for the sake of comparability of all
results. Based on experiences from the literature [117,118] and own benchmark
calculations, a level shift of 0.3 a.u. was applied. The energies of previously
converged RASPT2 simulations (equal reference weight for all roots) were
not significantly influenced (< 0.1 eV) upon introduction of a level shift
1 For example, in triplet spin state and A1 symmetry more roots were needed to address
all CT states with AS4.
52 3 Computational Details

of 0.3 a.u. More detail on the impact of the level shift is summarized in
Appendix A. However, SOCs were influenced by up to a factor of two (see
Appendix C). Thus, for calculations whose goal it was to obtain SOCs, the
smallest possible level shift with respect to a fitting reference weight was
used.
From now on, unless noted otherwise, all RASPT2 calculations include a
level shift of 0.3 a.u.
4 Results for non-reduced singlet
system
The preliminary step to asses the photochemistry of the present dye is to
obtain insight into its initial photo-excitation in the Franck-Condon region,
followed by possible relaxation mechanisms. This will be done for the
non-reduced singlet and the singly reduced doublet system of the BODIPY.
The first point of interest is the non-reduced singlet system. Here, the
postulated photoactivation mechanism is an excitation of the BODIPY in its
singlet ground state, followed by a relaxation in a dissociative or CT state
upon ISC into the triplet manifold, recall Figure 1.4.
In this chapter, a uniform color-code is used to label different excited
states of interest. Black represents the S0 ground state, blue excited singlet
states in general, red dissociative singlet states, green triplet states in general,
orange dissociative triplet states and cyan triplet CT states. Dissociative
state implicates an excitation into an anti-bonding σ ∗ C-I orbital, while CT
state features excitation into the phenyl ring from the main π-system or vice
versa.

4.1 TDDFT and RASPT2 benchmark

Like discussed in subsection 2.6.5, DFT lacks accuracy for describing multi-
configurational systems and boron-species in general. [143–147] Therefore, the
first step was to compare the excited states from TDDFT and RASPT2 at the
GS (A1 ) geometry. In Table 4.1 and Figure 4.1 all important excited states,
calculated with both methods, are compared. For the dissociative states
AS2 is used, for the CT states AS4 (recall subsection 3.2.1). An overview
of all calculated states with contributions, weights, oscillator strengths and
wavelengths is given in section A.2.

© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2019

K. M. Ziems, Photophysics and Photochemistry of a BODIPY-Based
Photosensitizer, BestMasters, https://doi.org/10.1007/978-3-658-26188-7_4
 54

Table 4.1: Comparison of TDDFT and RASPT2 results, namely excitation energies, oscillator strengths and excited
state characters obtained in the fully optimized GS structure

TDDFT RASPT2
Transition name
state E/eV f state E/eV f AS
π3 (a2 ) → π4∗ (b2 ) S1 (B1 ) 2.88 0.518 S1 (B1 ) 2.71 0.938 2 HOMO → LUMO

π2 (a2 ) → π4∗ (b2 ) S3 (B1 ) 3.45 0.311 S2 (B1 ) 3.65 0.094 2 HOMO-1 → LUMO

π3 (a2 ) → σ ∗ (a1 ) S8 (A2 ) 4.41 0.000 S3 (A2 ) 5.06 0.000 2 Diss.

π3 (a2 ) → σ ∗ (b1 ) S11 (B2 ) 4.55 0.000 S5 (B2 ) 5.42 0.001 2 Diss.

π3 (a2 ) → π4∗ (b2 ) T1 (B1 ) 1.55 - T1 (B1 ) 1.88 - 2 HOMO → LUMO

π2 (a2 ) → π4∗ (b2 ) T2 (B1 ) 2.58 - T2 (B1 ) 3.04 - 2 HOMO-1 → LUMO

πph,2 (b1 ) → π4∗ (b2 ) T6 (A2 ) 3.90 - T7 (A2 ) 4.93 - 4 CT

π3 (a2 ) → σ ∗ (a1 ) T7 (A2 ) 4.02 - T5 (A2 ) 4.83 - 2 Diss.
π3 (a2 ) → σ ∗ (b1 ) T8 (B2 ) 4.11 - T6 (B2 ) 5.09 - 2 Diss.
πph,3 (a2 ) → π4∗ (b2 ) T11 (B1 ) 4.29 - T8 (B1 ) 5.09 - 4 CT
∗
π3 (a2 ) → πph,5 (b1 ) T12 (B2 ) 4.39 - T10 (B2 ) 5.51 - 4 CT
∗
π3 (a2 ) → πph,4 (a2 ) T13 (A1 ) 4.46 - T11 (A1 ) 5.52 - 4 CT
4 Results for non-reduced singlet system
4.1 TDDFT and RASPT2 benchmark 55

S5(B2) T11(A1)
5.5 T10(B2)
T6(B2)
T8(B1)
5.0 S3(A2)
T13(A1) T7(A2)
S11(B2) T5(A2)
4.5 S8(A2) T12(B2)
T8(B2) T11(B1)
4.0 T7(A2) T6(A2)
S2(B1) S2(B1)
3.5 S3(B1) S3(B1)
Energy in eV

T2(B1) T2(B1)
3.0 S1(B1) f = 0.518 S1(B1) f = 0.518
S1(B1) S1(B1)
T2(B1) f = 0.938 T2(B1) f = 0.906
2.5

2.0 T1(B1) T1(B1)

1.5 T1(B1) T1(B1)

1.0

0.5

0.0 S0(A1) S0(A1) S0(A1) S0(A1)

TDDFT RASPT2-AS2 TDDFT RASPT2-AS4

Figure 4.1: Comparison of TDDFT and RASPT2 at GS (A1 ) geometry. Left:

comparison for dissociative states. Right: comparison for CT
states. Colorcode: ground state, singlet state, dissociative singlet
state, triplet state, dissociative triplet state, triplet CT state

The excitation energy of the bright S1 (B1 ) state is stabilized from 2.88 to
2.71 eV going from TDDFT to RASPT2, whereas an opposite trend is
observed for dissociative and CT states. TDDFT underestimates these
states by up to 1.1 eV. This shows clearly the problem of (TD)DFT to
describe this kind of systems.
Moreover, a population of the desired CT and dissociative states in the
GS geometry within the visible light spectrum can be excluded.
56 4 Results for non-reduced singlet system

4.2 Spin-Orbit coupling at ground state geometry

To determine a probability for an ISC between the bright S1 (B1 ) and
a dissociative triplet or CT state, SOCs were calculated at the GS (A1 )
geometry. In Table 4.2 and 4.3 the SOCs to different triplet states, namely of
dissociative and CT nature, are shown.1 The small couplings reveal a small
probability for an ISC due to SOC. These results support the hypothesis of
the previous section that at the GS (A1 ) geometry neither dissociative nor
CT states are populated.

Table 4.2: Results for SOCs from the bright S1 (B1 ) in AS2

Transition E / eV name SOC with

S1 in cm−1

T1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 1.88 HOMO → LUMO -

T2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 3.02 HOMO-1 → LUMO -
T5 (A2 ) π3 (a2 ) → σ ∗ (a1 ) 4.78 Diss. 0.6
T6 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 5.06 Diss. 39.2

Table 4.3: Results for SOCs from the bright S1 (B1 ) in AS4 with level shift of
0.3

Transition E / eV name SOC with

S1 in cm−1

T1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 1.90 HOMO → LUMO -

T2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 3.04 HOMO-1 → LUMO -
T7 (A2 ) πph,2 (b1 ) → π4∗ (b2 ) 4.93 CT 5.3
T8 (B1 ) πph,3 (a2 ) → π4∗ (b2 ) 5.09 CT -
T10 (B2 ) ∗
π3 (a2 ) → πph,5 (b1 ) 5.51 CT 0.0
T11 (A1 ) ∗
π3 (a2 ) → πph,4 (a2 ) 5.52 CT 0.5

Obviously, the calculation of SOCs exclusively at GS (A1 ) geometry is

a major simplification, since they might change along a dissociative or CT
coordinate. Thus, this issue will be addressed in section 4.4.2.
Further SOCs between S0 , S1 and calculated excited triplet states at GS
(A1 ) geometry with AS2 and AS4 can be found in Appendix C.

1 Like discussed in subsection 3.2.2 the level shift influences the SOCs; thus, the minimal
possible level shift regarding a reasonable reference weight has been chosen. This is
0.0 a.u. for AS2 and 0.3 a.u. for AS4.
4.3 TDDFT geometry-states correlation diagram 57

4.3 TDDFT geometry-states correlation diagram

In the first two sections, the calculations were carried out within the GS
(A1 ) geometry. This is a huge approximation regarding results for a possible
population of CT or dissociative states since, in real systems, wave packets
relax along specific coordinates on the PES and, therefore, the PESs cross
or come energetically close to each other at geometries, which might be very
different to the GS (A1 ) geometry.
To reveal if the energy levels of excited states undergo significant alter-
nations upon structural variations, so-called geometry-states correlation
diagrams were created. The different excited states, namely S1 (B1 ), S2 (A1 ),
S3 (B1 ), T1 (B1 ), T6 (A2 ), T11 (B1 ), T12 (B2 ), T13 (A1 ), S8 (A2 ), S11 (B2 ), T7 (A2 )
and T8 (B2 ), at GS (A1 ) were optimized at the TDDFT level of theory. Within
these geometries excited states were calculated and compared diabatically
to the states at GS (A1 ), which is displayed in Figure 4.2.
Two major observations can be drawn. Firstly, at no given geometry are
the CT states significantly lower in energy. Secondly, the dissociative states
undergo stabilization by up to 1.5 eV and are found even lower than the
bright S1 (B1 ) state, when going to the optimized dissociative geometries,
S8 (A2 ), S11 (B2 ), T7 (A2 ) and T8 (B2 ). This means that theoretically these
states can be populated by illumination under the restriction of having a
suitable coordinate leading to these optimized dissociative structures (in
more detail in subsection 4.4.2).
The optimized dissociative geometries are shown in Figure 4.3 and show
an elongation of one or two C-I bonds up to 2.7 Å for the antisymmetric,
S8 (A2 ) and T7 (A2 ), and 2.3 Å for the symmetric, S11 (B2 ) and T8 (B2 ),
dissociation. Thus, excitation into the symmetric orbital σ ∗ (a1 ) results in
a mono-dissociation, whereas the excitation into the antisymmetric σ ∗ (b1 )
yields a di-dissociation.
No pronounced structural changes were observed within the optimized
CT states. In addition, to exclude convergence into a wrong geometry, the
initial geometry for the CT optimization was distorted, i.e. by changing
the main/side ring dihedral from 90◦ to 85◦ , but no significant change upon
optimization was observed compared to the optimization starting from the
non-distorted GS (A1 ) geometry.
58 4 Results for non-reduced singlet system

Figure 4.2: TDDFT geometry-states correlation diagram. Asterisks indicate

non-converged geometries, where the last step of the optimization
procedure has been taken. Colorcode: ground state, singlet state,
dissociative singlet state, triplet state, dissociative triplet state,
triplet CT state
4.4 Dissociation 59

S8 (A2 ) S11 (B2 )

T7 (A2 ) T8 (B2 )

Figure 4.3: Optimized TDDFT structures for dissociative states with R(C-I)
bond length. At GS (A1 ) geometry R(C-I) = 2.082 Å

4.4 Dissociation
The next step was a more thorough investigation of the dissociation starting
from the GS (A1 ) equilibrium, since the TDDFT geometry-states correlation
diagram (Figure 4.2) showed already some promising results concerning the
population of the dissociative states.
In the following section, several methods and approaches have been used to
describe the dissociation of the non-reduced BODIPY species. A MEP with
RASSCF, diabatic coordinates for dissociation with TDDFT and RASPT2
and spin-orbit kinetics were applied.
60 4 Results for non-reduced singlet system

4.4.1 Minimum energy pathway

A MEP is a computational demanding procedure and, thus, an analytical
approach is necessary for larger systems with a high number of CSFs and
degrees of freedom, like the system investigated in this thesis. Unfortunately,
analytical gradients at the SA-RASSCF level are not implemented in Molcas
8.2 [161] for MEP calculations. Therefore, it can only be applied to the
ground state of a given spatial symmetry along the MEP (single state
RASSCF). This means that for the BODIPY dye system only the symmetric
di-dissociation, arising from excitation to state S5 (B2 )2 (π3 (a2 ) → σ ∗ (b1 )),
can be investigated. The mono-dissociation would require Cs instead of C2v
symmetry, and in the symmetry-reduced system the dissociative states are
no longer ground states of a spatial symmetry.
The MEP stops after nine steps due to an increase in energy. This means
that at this point either a minimum or a crossing point was found. The
energy pathway is shown in Figure 4.4, together with the final geometry.
The two C-I bonds dissociate slightly from 2.08 Å to 2.39 Å. This is in
consistency with the optimized TDDFT geometry of state S11 (B2 ), which is
characterized by the same transition and exposes also a symmetric elongation
from 2.08 Å to 2.33 Å (see Figure 4.3).

0.00
Energy
rel. Energy in Ha

−0.01

−0.02

−0.03

−0.04

1 2 3 4 5 6 7 8 9
MEP Step

Figure 4.4: Pathway of the energy (left) and the final structure at MEP step 9
(right).

2 The state S5 (B2 ) in RASSCF/RASPT2 notation is the same like state S11 (B2 ) in
TDDFT notation, defined through their excitation pattern. Therefore, the results ob-
tained in the section before regarding S11 (B2 ) and its relaxation into a di-dissociation
also hold for S5 (B2 ). Recall Table 4.1 for all important states.
4.4 Dissociation 61

4.4.2 Dissociation coordinate

To treat the dissociation in all detail, coordinates along the PES describing
the dissociation are needed. In the following, different approaches are
discussed and, consequently, different diabatic potentials are investigated.

Different approaches for dissociation coordinates

Three different approaches for dissociation coordinates are feasible in theory:
i) A vibrational mode at GS (A1 ) describing the dissociation, ii) a Linear-
Interpolated Cartesian Coordinate (LICC) between two optimized states and
iii) a coordinate describing the C-I bond elongation, i.e. a scan of increasing
C-I bond length. Moreover, a differentiation between a symmetric mono- or
an antisymmetric di-dissociation has to be made.
i) The vibrational analysis reveals that two normal modes are associated
with the symmetric and anti-symmetric C-I dissociation; see Figure 4.5 and
4.6. The symmetric one displays a clear cleavage of the two C-I bonds,
whereas the antisymmetric one is pronouncedly mixed with vibration vectors
not associated with the cleavage.
ii) The LICCs were obtained, firstly, by transforming the coordinate
system of the two structures in such a way that both molecules are centered
identically within the coordinate system. Secondly, the Boron atom was
chosen as point of origin to avoid rotational and translational artifacts.
Subsequently, the linear interpolated vector connecting the two states was
calculated. The two LICCs are displayed in Figure 4.7 and 4.8. Both show
cleavage of the C-I bond(s), but the antisymmetric one is, like seen for the
vibrations, more complicated with mixed vectors.
As a result of i) and ii) only symmetric di-dissociations seem to be feasible
for further studies. Although, the vectors of the symmetric mode and LICC
seem to influence exclusively the C-I bond, going to larger displacements
reveals that also the small variations to the rest of the molecule have a great
impact on the structure. This is illustrated in Figure 4.9 for an elongation
to 7 Å for both symmetric dissociation coordinates (mode and LICC), which
leads to a chemical unreasonable structure. Thus, both coordinates are not
applicable for large elongations describing a full dissociation.
iii) In conclusion, the third option - a C-I bond elongation coordinate - has
to be considered. In this case, a simple elongation of one or both C-I bonds
serves as such a coordinate for the symmetric or antisymmetric dissociation.
These elongations can be done as a frozen or relaxed scan, i.e. without or
with geometry relaxation at each step of the scan.
62 4 Results for non-reduced singlet system

Figure 4.5: Symmetric vibration at Figure 4.6: Antisymm. vibration at

GS with νs = 147 cm−1 GS with νas = 207 cm−1

Figure 4.7: LICC between S0 (A1 ) Figure 4.8: LICC between S0 (A1 )
and S11 (B2 ) geometry and S8 (A2 ) geometry

Figure 4.9: Elongation of the C-I bond to 7 Ångström for the symmetric vi-
bration at GS (left) and the LICC between S0 (A1 ) and S11 (B2 )
geometry (right)
4.4 Dissociation 63

Symmetric di-dissociation scan coordinate

Both C-I bonds were elongated simultaneously and excited states were
monitored diabatically with TDDFT. First, this was done as a frozen scan
along the C-I bonds. The results can be seen in Figure 4.10 and show that

S0(A1) S1(B1) S8(A2) S11(B2) T7(A2) T8(B2) T6(A2) T11(B1) T12(B2) T13(A1)
13
12
11
10
9
Energy in eV

8
7
6
5
4
3
2
1
0
2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0
R(C-I) in Ångström

Figure 4.10: TDDFT potentials along the frozen symmetric C-I elongation.
Colorcode: ground state, singlet state, dissociative singlet state,
triplet state, dissociative triplet state, triplet CT state. A smaller
detail of this coordinate until an elongation of 3.0 Å can be seen
in Figure D.1

the potentials of the dissociative states are not dissociative. They rather
feature a minimum at approximately 2.32 Å after crossing the S1 (B1 ) state.
Furthermore, the degeneracy of the (now pseudo-)dissociative state at long
elongation can be seen. This is a reason why bigger elongations are not
accessible with (TD)DFT. It ends in a degenerated ground state that cannot
be described with this single determinant method. The potential energy
landscape of CT states is hardly influenced by this coordinate and follows
the S0 ground state potential.
The idea that this might be due to the poor description by TDDFT was
excluded after performing RASPT2 simulations along the frozen coordinate
led to a qualitatively similar picture; see PESs in Figure 4.11. Several
observations have to be mentioned in regard to the RASPT2 PES. First
64 4 Results for non-reduced singlet system

of all, like seen in Figure 4.1, the dissociative3 excited states are higher
in energy and the S1 (B1 ) state is lower in energy compared to TDDFT,
which, in consequence, leads to a crossing of the dissociative states with S1
at longer distances of approximately 2.6 Å. Secondly, the minimum of the
dissociative states is found at around 2.3 Å, which is in accordance with the
final MEP structure (see Figure 4.4). Thirdly, the SOCs were calculated
along the coordinate.4 As visualized in Figure 4.12 for S1 (B1 )/T5 (A2 ) and
S1 (B1 )/T6 (B2 ) the obtained SOCs show a small dependency with respect to
the C-I distance. Between the GS structure and the crossing at approximately
2.6 Å the coupling slightly rises and decreases again until values of 0.6 cm−1
and 37.0 cm−1 are obtained at the crossing. Not surprisingly, the SOCs
decrease for longer distances. But overall no significant high value can be
observed, which indicates a low probability for ISC.
A comparison of the frozen potentials with the states at the optimized,
dissociative S11 (B2 ) geometry on TDDFT and RASPT2 level of theory,
revealed a lowering of the dissociative (and CT) states by preservation of the
GS and S1 (B1 ) state energy, accompanied by an increase of the SOCs. For
the graphical display of these results, see section D.1 and also Figure 4.13.
This comparison indicates that a relaxation of the system results in an
energetic lowering of dissociative and CT states. Thus, the potentials were
calculated as relaxed scan along both C-I bond elongations, which is shown
for TDDFT in Figure 4.13 and reveals no significant improvement regarding
the conceptual question for dissociative states.
To summarize, a symmetric di-dissociation as result of an ISC (SOC)
between S1 (B1 ) and a dissociative state is impossible. This result is in
agreement with the fact that it was possible to find a minimum during the
optimization of these very dissociative states with TDDFT. The position
of the excited states of the optimized geometry S11 (B2 ) is in accordance
with the minima found in the diabatic potentials at around 2.32 Å, which
can be seen for the potentials along the symmetric elongation in Figure 4.13
and section D.1. Another interesting observation is that the minimum is
described equally, independent of the method or degree of relaxation. Merely
the excitation energy levels change and, thus, the position of the crossing
points.

3 Even though these states are not really dissociative, they will be called so for notation
purposes.
4 It is important to mention that following the results of subsection 3.2.2 the smallest

possible level shift was chosen to get only marginally influenced SOCs. This was at a
shift of 0.1 for the RASPT2 calculations.
4.4 Dissociation 65

S0(A1) S1(B1) S3(A2) S5(B2) T5(A2) T6(B2)

8
7
6
Energy in eV

5
4
3
2
1
0
2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0
R(C-I) in Å

Figure 4.11: RASPT2 potentials along the frozen symmetric C-I elongation.
Colorcode: ground state, singlet state, dissociative singlet state,
dissociative triplet state

1.25
SOC in cm−1

S1(B1)-T5(A2)
1.00
0.75
0.50
0.25
60
SOC in cm−1

S1(B1)-T6(B2)
40

20

0
2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0
R(C-I) in Å

Figure 4.12: SOCs for the diabatic potentials along the frozen symmetric C-I
elongation between the S1 (B1 ) state and the two dissociative
triplet states, T5 (A2 ) (above) and T6 (B2 ) (below); calculated
with RASPT2.
66 4 Results for non-reduced singlet system

S0(A1) S1(B1) S8(A2) S11(B2) T7(A2) T8(B2)

12
at optimized S11(B2) geometry
11
relaxed coordinate
10
9
8
Energy in eV

7
6
5
4
3
2
1
0
2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4.0
R(C-I) in Å

Figure 4.13: TDDFT PESs of S0 , S1 , S8 , S11 as well as of T7 and T8 along the

frozen (filled dots) and relaxed scan (stars). And the states at
the optimized, dissociative S11 (B2 ) geometry. Colorcode: ground
state, singlet state, dissociative singlet state, dissociative triplet
state

Antisymmetric mono-dissociation scan coordinate

The next step is the investigation of the mono-dissociation. Since stretching
of one C-I bond reduces the symmetry from C2v to Cs , while increasing the
computational costs, the calculations have only been done with TDDFT and
not with RASPT2.
Again, a frozen and a relaxed scan was applied. The resulting PESs can
be seen in Figure 4.14 and 4.15, respectively.
The frozen coordinate reveals the expected behavior. All states except for
the mono-dissociative ones, S8 (A2 ) and T7 (A2 ), follow the ground state, while
the mono-dissociative states hold a near-dissociative behavior. Until 2.7 Å
the dissociative potentials have a dissociative behavior, after that a small
increase in energy until the crossing point with the ground state can be noted
(≈ 0.18 eV). After that (TD)DFT enters a highly non-equilibrium region,
where the results are commonly poor (see below) and an interpretation is
difficult.
4.4 Dissociation 67

S0(A1) S1(B1) S8(A2) S11(B2) T7(A2) T8(B2) T6(A2) T11(B1) T12(B2) T13(A1)
9
8
7
6
Energy in eV

5
4
3
2
1
0
2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4.0
R(C-I) in Å

Figure 4.14: TDDFT PESs along the frozen mono C-I elongation. Colorcode:
ground state, singlet state, dissociative singlet state, triplet state,
dissociative triplet state, triplet CT state

S0(A1) S1(B1) S8(A2) S11(B2) T7(A2) T8(B2) T6(A2) T11(B1) T12(B2) T13(A1)
9
8
7
6
Energy in eV

5
4
3
2
1
0
2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4.0
R(C-I) in Å

Figure 4.15: TDDFT PESs along the relaxed mono C-I elongation. Colorcode:
ground state, singlet state, dissociative singlet state, triplet state,
dissociative triplet state, triplet CT state
68 4 Results for non-reduced singlet system

In general, this is a good example of one of the main problems of (TD)DFT.

At the crossing of the ground state S0 (A1 ) and the mono-dissociative triplet
state T7 (A2 ), the BOA breaks down and the wave function displays a
highly multiconfigurational character. Both issues cannot be handled with
(TD)DFT. Also, for bond elongations beyond this point, S0 (A1 ) is no longer
the ground state. This represents a system far away from its equilibrium
structure and (TD)DFT is also known for a non-sufficient description of
such systems. Therefore, an investigation with a method like RASPT2 is
essential for quantitatively more reliable results. Nevertheless, the qualitative
conclusion of a possible dissociation due to a crossing of the bright S1 (B1 )
state with the mono-dissociative S8 (A2 ) and T7 (A2 ) states can be drawn.
Prerequisite for this is an excess energy to overcome the slight energetic
barrier of the S1 (B1 ) pathway to the crossing point. Unfortunately, no SOCs
are available at the TDDFT level of theory in Gaussian. As a result, no
quantitative prediction of a population transfer among these states. i.e. by
means of a rate constant can be obtained.
The relaxed system in Figure 4.15 exhibits almost the same behavior like
the frozen scan, with the difference of a higher increase in energy between
the minimum of the dissociative state at 2.7 Å and the crossing point with
the GS (0.8 eV instead of 0.18 eV). This might be a result of the stabilization
of the ground state.
In conclusion, a shallow minimum for the mono-dissociative states was
found. To overcome this barrier of 0.18 eV along the frozen and 0.8 eV along
the relaxed coordinate enough energy is needed. Additionally, an excess
energy is necessary to reach the crossing between the bright S1 (B1 ) state
and the dissociative ones (0.40 eV for the frozen and 0.32 eV for the relaxed
scan). Furthermore, it can be observed that the shallow minimum is at
the same C-I distance as the optimized mono-dissociative TDDFT states in
Figure 4.3, explaining the minimum found for this optimization.

4.4.3 Spin-Orbit Kinetics

The golden-rule expression for radiationless transitions is adapted for ISC
rates and SOCs as interacting perturbation-potential [166,186]

2π 1 3 2 1 (ΔE + λ)2
if
kISC = |  Ψi |ĤSOC | Ψf  | √ exp − . (4.1)
h̄ 4πλRT 4λRT
4.4 Dissociation 69

This can be interpreted according to Marcus theory [187,188] with ΔE as

energy difference between the optimized states and λ as reorganization
energy.5
The Marcus theory formalism was applied to an ISC from S1 (B1 ) state at
GS (A1 ) geometry to the optimized mono- and di-dissociative triplet states,
T7 (A2 ) and T8 (B2 ), at their respective minimized geometry. All calculations
were obtained at the TDDFT level of theory (see Figure 4.2). The potentials
can be seen in Figure 4.16 and the results of the Marcus theory in Table 4.4.
SOCs were taken from Table 4.2.

5.00
6.0 S1(B1) S1(B1)
Potential S1(B1) 4.75 Potential S1(B1)
5.5
T7(A2) 4.50 T8(B2)
5.0 Potential T7(A2) Potential T8(B2)
4.25
Energy in eV

Energy in eV

4.5 λi 4.00 λf
4.0 3.75
λf
3.5 3.50 λi
3.25
3.0 ΔE
ΔE 3.00
2.5
2.75
S0(A1) T7(A2) S0(A1) T8(B2)
Geometry Geometry

Figure 4.16: Potentials "fitted" as quadratic equation according to Marcus

theory. Left: For an ISC from S1 (B1 ) at GS (A1 ) geometry to
T7 (A2 ); mono-dissociation. Right: For an ISC from S1 (B1 ) at GS
(A1 ) geometry to T8 (B2 ); di-dissociation

For both cases, the results show a very slow ISC of 1.69 × 104 s−1 for
S1 (B1 ) to T7 (A2 ) and 1.51 × 103 s−1 for S1 (B1 ) to T8 (B2 ), but still within
the range of literature values. [189] This means that the population of the
dissociative states occurs very slowly, which might be an explanation for the
slow cleavage after several hours of irradiation observed in experiments. [86]
The final and initial reorganization energies (λf and λi ) should ideally
be the same or at least vary little, which is not the case, especially for the
mono-dissociation. According to Marcus theory, both potentials for a given
radiationless transition should have the same curvature, which would result
5 This is the energy, that is required to get from one optimized geometry to another on
the diabatic surface of one state without crossing.
70 4 Results for non-reduced singlet system

Table 4.4: Results for ISC based on Marcus theory and Equation 4.1. f = final
and i = initial, representing the geometry from which the reorgani-
zation energy was taken

S1 (B1 ) to T7 (A2 ) S1 (B1 ) to T8 (B2 )

ΔE in eV -0.37 0.48
λ(f ) in eV 1.75 0.67
λ(i) in eV 1.51 0.75
SOC in cm−1 0.58 39.22
tISC (f ) in s 5.63 × 10−4 4.19 × 10−4
tISC (i) in s 5.92 × 10−5 6.62 × 10−4
kISC (f ) in s−1 1.78 × 103 2.39 × 103
kISC (i) in s−1 1.69 × 104 1.51 × 103

in consistent reorganization energies. Since only two points on the PESs were
used here, a fit to obtain potentials, which describe the system sufficiently,
was not possible, resulting in varying reorganization energies.
The crude approximation in the calculations is the use of the SOCs at
the GS (A1 ) geometry. The justification for this are the results of the SOCs
seen for the scan coordinates (Figure 4.12), which showed no pronounced
dependency on the C-I distance between ground state and crossing point.
Nevertheless, the SOCs were calculated with RASPT2 exemplarily at the
geometry derived from the crossing point of the two quadratic potentials of
S1 (B1 ) and T8 (B2 ) (recall Figure 4.16); the result is shown in Table 4.5. Two
conclusions can be derived from it. First, the big difference of the methods
TDDFT and RASPT2, which shifts the states energetically, consequences in
not (near) degeneracy for RASPT2 at the TDDFT crossing point. Secondly,
even though the energy gap between the bright S1 (B1 ) and the dissociative
T5 (A2 ) state gets halved compared to the GS (A1 ) geometry (Table 4.1),
the SOCs do not change significantly.
To obtain the SOC at the RASPT2 crossing point a scan along this
S1 (B1 )/T8 (B2 ) LICC would have to be performed with RASPT2; the same
holds for S1 (B1 )/T7 (A2 ). Moreover, a complete scan along these LICCs
would be needed to verify the use of Marcus theory.
4.5 Charge-Transfer 71

Table 4.5: RASPT2 results with SOCs at the crossing point of the quadratic
Marcus theory potentials S1 (B1 ) and T8 (B2 ), being S1 (B1 )/T5 (A2 )
in RASPT2 state notation (see Table 4.1)

Transition E / eV name SOC with

S1 in cm−1

S1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 2.72 HOMO → LUMO -

T5 (A2 ) π3 (a2 ) → σ ∗ (a1 ) 3.72 Diss. 1.7
T6 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 3.91 Diss. 53.0

4.5 Charge-Transfer
A further investigation of the photoactivation via charge transfer states is the
next step. Figure 4.2 already showed that CT states do not get energetically
stabilized at any optimized geometry. Furthermore, the optimized CT
geometries themselves do not expose any major change in their structure
compared to the GS (A1 ) geometry. Thus, an easily traceable coordinate
does not exist.
A reason for the high lying CT states might be the fact that the two
rings are orthogonal to each other and, thus, their orbitals do not overlap.
Therefore, a torsion around the dihedral of the main- and side-ring, illustrated
in Figure 4.17, might cause an energetic stabilization of the CT states.

Figure 4.17: Dihedral torsion around the two aromatic ring systems

For a first look, a frozen scan in the interval [90◦ , 10◦ ] and with a step size
of 5◦ around the dihedral was performed using TDDFT. The eclipsed, as
well as the staggered frozen scan, showed a small stabilization of the T6 (A2 )
72 4 Results for non-reduced singlet system

excitation energy of 0.55 eV and 0.65 eV, respectively (see Figure D.5).6
Consequently, a relaxed scan was conducted. The calculated PESs are
displayed in Figure 4.18, which reveal a significant maximum energetic
stabilization of 0.98 eV at 45◦ for the T6 (A2 ) CT state.

T13(A1)
5.5
T12(B2)
5.0
4.5 T11(B1)
4.0
Energy in eV

S1(B1)
3.5
T6(A2)
3.0
2.5
2.0
1.5
1.0 S0(A1)
0.5
0.0
15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90
Dihedral (◦)

Figure 4.18: TDDFT PESs along the relaxed scan around the dihedral be-
tween the two aromatic ring systems. Colorcode: ground state (o)
singlet state (o), triplet CT state (x)

The behavior of T6 (A2 ) at higher dihedral degrees is due to a mixed

excitation pattern for this state in TDDFT. Excitations out of the p orbitals
of the Iodines into the LUMO mix into this state.
Like discussed thoroughly in the computational details (section 3.1), em-
ploying a torsion on the main/side ring dihedral results in a relaxation
into a geometry with slightly distorted methyl groups, but a smaller total
energy than the GS (A1 ) geometry of our C2v model system. Thus, for
considering a possible torsion at room temperature, the geometry at 85◦ was
taken as reference. This results in a possible torsion at room temperature
(RT is 298 K and corresponds to an energy of 0.026 eV) of up to 69◦ , which
means that the CT state T6 (A2 ) gets lowered from 3.9 eV to around 3.0 eV.
Moreover, the crossing of the bright S1 (B1 ) and the CT state is at an energy
6 Eclipsed and staggered refer to the orientation of the two methyl groups of the main
ring to the side phenyl ring. Eclipsed means that a single hydrogen of the methyl
group points towards the side phenyl ring and is in-plane with the boron fragment.
Staggered is its 180◦ equivalent. Two hydrogens of the methyl groups point towards
the side phenyl ring with the main phenyl ring in the plain between them.
4.5 Charge-Transfer 73

of below 3 eV. As the experiment is conducted with an excitation energy of

around this energy, a population of the CT state is possible.
The major problem at this point is the level of theory. Like seen in
Figure 4.1 the S1 (B1 ) gets lowered in energy and the CT state energy is
increased, if TDDFT is compared to RASPT2. Consequently, the lowering
of the CT state in energy is overestimated with respect to RASPT2, as well
as the crossing point, which might be shifted to smaller dihedral degrees
and higher energy. It has been seen in section 4.4.2 for the symmetric
dissociation that the position of the crossing point changes going from
TDDFT to RASPT2. In conclusion, the population of the CT state T6 (A2 )
might be less likely to occur with respect to RASPT2 results - the more
accurate, but computational costly state-of-the-art theory for these systems.
To quantify a possible ISC from the S1 (B1 ) to the CT state T6 (A2 ),
SOCs are necessary. These were already computed at GS geometry with
RASPT27 , recall Table 4.3, and revealed no significant coupling. A reason
for that might be the perpendicular arrangement of the two (hetero)aromatic
rings. Thus, SOCs along the dihedral coordinate discussed above have
to be considered for a better insight into the SOCs for a CT crossing.
Unfortunately, the geometries of the relaxed scan lack of any symmetry,
which makes a calculation with RASPT2 to gain SOCs computationally
expensive. As a result, the couplings were simulated for a frozen torsion
around the main/side ring dihedral to maintain Cs symmetry. A dihedral of
55◦ was chosen, for which the result are displayed in Table 4.6.

Table 4.6: Results for SOC with RASPT2 AS4 with a level shift of 0.3 in the
fully optimized non-reduced geometry for a torsion of 55◦ around
the main/side ring dihedral.

Transition E / eV name SOC with

S1 in cm−1

S1 (B) π3 (a2 ) → π4∗ (b2 ) 2.68 HOMO → LUMO -

T1 (B) π3 (a2 ) → π4∗ (b2 ) 0.90 HOMO → LUMO 0.1
T2 (B) π2 (a2 ) → π4∗ (b2 ) 2.03 HOMO-1 → LUMO 0.4
T3 (A) πph,2 (b1 ) → π4∗ (b2 ) 3.62 CT 4.0

The results exhibit the same problems as were seen with the Marcus
theory (Table 4.5). Due to the overestimation of the S1 (B1 ) state and the
underestimation of the CT states with TDDFT, 55◦ does not represent a

7 The CT state T6 (A2 ) with TDDFT represents the T7 (A2 ) state with RASPT2-AS2.
74 4 Results for non-reduced singlet system

near degeneracy between those two states in RASPT2. Nevertheless, it can

be concluded that a decrease of the energy gap between S1 (B) and T3 (A)
from 2.2 eV at the GS (A1 ) geometry to 0.94 eV at a torsion of 55◦ does not
yield larger SOCs, contrary, the SOCs even decrease from 5.3 to 4.0 cm−1 .
In conclusion, a quantitatively efficient crossing to populate a CT state is
unlikely.
Finally, the nature of the relaxation of the geometries along the relaxed
scan has to be taken into consideration. A twist of the whole molecule
around the C-C bond between the two ring systems and an adjustment
in the methyl groups can be observed along the relaxed scan. This is in
accordance with the corresponding vibrational modes at GS (A1 ) geometry,
mimicking the behavior of the relaxation.

ν = 45 cm−1 ν = 71 cm−1 ν = 102 cm−1 ν = 170 cm−1

Figure 4.19: Vibrational modes at GS (A1 ) geometry corresponding to a tor-
sion around the main/side ring dihedral.
5 Results for reduced doublet
system
A different possible approach to the entire two-component catalytic system
is to start with the chemical reduction of the BODIPY dye by triethylamine.
This results in a reduced BODIPY structure with a doublet (B2 ) ground state.
Subsequent excitation can populate excited dissociative (photodegradation)
and charge transfer (photoactivation) states (recall mechanism in Figure 1.4).
Thus, all investigations done in chapter 4 are repeated for the singly
reduced system. Based on the results for the non-reduced system, the
research for the reduced system can be focused on, e.g., mono-dissociation
instead of di-dissociation.
The color-notation is adjusted as follows. Black represents the D0 (B2 )
ground state, blue general excited doublet states, red doublet dissociative
states, green doublet CT states, orange quartet dissociative states and cyan
quartet CT states. Dissociative state implicates an excitation into an anti-
bonding σ ∗ C-I orbital, whereas CT state represents an excitation into the
side phenyl ring from the main π-system or vice versa.

5.1 Comparison of TDDFT and RASPT2 results

Initially, the reduced doublet and quartet geometries were optimized using
DFT. The doublet optimization assessed the LUMO of the non-reduced sys-
tem, π4∗ (b2 ), to be the orbital containing the unpaired electron. The quartet
optimization yielded different ground states depending on the step size of
the optimization, displayed in the form of spin densities in Figure A.1. Thus,
a further study of the system with TDDFT and RASPT2 was indispensable.
The comparison of the excited states in doublet and quartet multiplicity
with TDDFT and RASPT2 is shown in Table 5.1 and graphically displayed
in Figure 5.1. Like before, AS2 represents dissociative states, AS4 CT states.

© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2019

K. M. Ziems, Photophysics and Photochemistry of a BODIPY-Based
Photosensitizer, BestMasters, https://doi.org/10.1007/978-3-658-26188-7_5
 76

Table 5.1: Comparison of TDDFT and RASPT2 results, namely excitation energies, oscillator strengths, and excited
state characters obtained in the fully optimized singly reduced doublet GS structure.

TDDFT RASPT2
Transition name
state E/eV f state E/eV f AS
∗
π4∗ (b2 ) → πph,5 (b1 ) D1 (B1 ) 2.04 0.000 D1 (B1 ) 2.11 0.000 4 CT
∗
π4∗ (b2 ) → πph,4 (a2 ) D2 (A2 ) 2.11 0.000 D5 (A2 ) 3.54 0.150 4 CT
π4∗ (b2 ) → σ ∗ (a1 ) D3 (A1 ) 2.12 0.000 D1 (A1 ) 2.58 0.000 2 Diss.
π4∗ (b2 ) → σ ∗ (b1 ) D4 (B1 ) 2.26 0.000 D3 (B1 ) 2.83 0.000 2 Diss.
π3 (a2 ) → π4∗ (b2 ) D5 (A2 ) 2.69 0.194 D2 (A2 ) 2.49 0.183 4 HOMO → LUMO
π2 (a2 ) → π4∗ (b2 ) D6 (A2 ) 3.38 0.003 D4 (A2 ) 3.35 0.010 4 HOMO-1 → LUMO
π3 (a2 ) → σ ∗ (a1 ) Q1 (B1 ) 3.55 - Q1 (B1 ) 4.26 - 2 Diss.
π3 (a2 ) → σ ∗ (b1 ) Q3 (A1 ) 3.64 - Q2 (A1 ) 4.43 - 2 Diss.
∗
π3 (a2 ) → πph,5 (b1 ) - - - Q2 (A1 ) 4.72 - 4 CT
5 Results for reduced doublet system
5.1 Comparison of TDDFT and RASPT2 results 77

5.0
Q2(A1)
Q2(A1)
4.5
Q1(B1)

4.0
Q3(A1)
Q1(B1) D5(A2)
3.5 D4(A2)
D6(A2) D4(A2)

3.0 D3(B1)
Energy in eV

f = 0.2630 f = 0.1942
D2(A2) D5(A2) f = 0.1831
D2(A2)
2.5 D1(A1) D4(B1)
D3(A1) D1(B1)
2.0 D2(A2)
D1(B1)
1.5

1.0

0.5
D0(B2) D0(B2) D0(B2)
0.0
RASPT2-AS2 TDDFT RASPT2-AS4

Figure 5.1: Comparison of TDDFT and RASPT2 (AS2 and AS4) states at
singly reduced GS doublet (B2 ) geometry. Colorcode: ground
state, doublet state, dissociative doublet state, CT doublet state,
dissociative quartet state, CT quartet state

A remarkable difference to the non-reduced system is that the states of

interest, dissociative and CT states, are already low lying at the GS geometry.
They are energetically below the bright D5 (A2 )/D2 (A2 )1 state. Thus, a
population of the interesting states is possible due to internal conversion
(via non-adiabatic coupling).
The comparison of TDDFT and RASPT2 shows the same trend as in the
case of the non-reduced system. The bright low-lying ππ ∗ state is stabilized
in energy and the dissociative and CT states are higher in energy, going from
TDDFT to RASPT2. Contrary to the non-reduced system, also a significant
reordering of the states occurs. Moreover, the high lying quartet states are
described very poorly with TDDFT.

1 Be aware of the TDDFT and RASPT2 state notation, seen in Table 5.1
78 5 Results for reduced doublet system

Beside the already mentioned differences, the RASPT2 state D5 (A2 ) has
to be discussed in detail. It can be seen in Table 5.1 that this state has an
oscillator strength of 0.15 a.u. Furthermore, the difference of the excitation
energy compared to the corresponding TDDFT state D2 (A2 ) is with 1.4 eV
very large. The problem, and presumably the underlying cause, is that the
weight of the important CT character is only 29 %. The main contribution
of this state is the transition π4∗ (b2 ) to π4∗ (a2 ) with 38 %, which might be
the transition contributing to the high transition dipole momentum. A
separation of these two transitions into separated electronic states could not
be achieved, e.g. by changing the level shift, input orbitals, convergence
criteria etc. (see Table A.13). To summarize, the conclusion of a population
of the RASPT2 state D5 (A2 ) at GS geometry via irradiation should not be
made.

5.2 TDDFT geometry-states correlation diagram

The excited states at the GS geometry were optimized at the TDDFT level of
theory and in analogy to the non-reduced system a TDDFT geometry-states
correlation diagram is presented in Figure 5.2.
Apart from small vibrations within the ring structures, the CT geometries
showed no significant structural change after optimization, whereas the
dissociative states exposed a mono- and di-dissociation (D3 (A1 ) to 2.91 Å
and D4 (B1 ) to 2.36 Å, respectively). Their geometries are presented in
Figure 5.3.
The mono-dissociation, i.e. the optimized D3 (A1 ) geometry, yields a
significant decrease in energy of the mono-dissociative D3 (A1 ) state, even
below the former ground state D0 (B2 ). However, at the di-dissociative
optimized geometry, D4 (B1 ), the respective di-dissociative state does not get
significantly lower in energy compared to D3 (A1 ) in its optimized geometry.
Combined with the fact that the di-dissociation has already been classified
non-dissociative for the non-reduced system, the di-dissociation can be
neglected, whereas the coordinate following the mono-dissociation is of high
interest.
Moreover, the CT states show a lowering in energy. Since their geometry
change involves only small variations within the ring structures, a population
of them, already at the doublet GS geometry, is likely.
Figure 5.2 exposes another issue of TDDFT. Double excitations cannot
be described with a single determinant method like (TD)DFT. Therefore,
the bright D5 (A2 ) state cannot be monitored when the ground state has
5.2 TDDFT geometry-states correlation diagram 79

changed to the dissociative state D3 (A1 ). For the diabatic follow up of the
bright state within this new ground state D3 (A1 ) a double excitation into
the π4∗ (b2 ) from the HOMO π3 (a2 ), as well as from the Singly Occupied
Molecular Orbital (SOMO) σ ∗ (a1 ) would be needed.

D0 (B2 ) D1 (B1 ) D2 (A2 ) D3 (A1 ) D4 (B1 ) D5 (A2 )

3.5

3.0

2.5
Energy (eV)

2.0

1.5

1.0

0.5

0.0
D0(B2) D1(B1) D2(A2) D3(A1) D4(B1)
Geometry

Figure 5.2: TDDFT geometry-states correlation diagram for the first five
doublet states. Colorcode: ground state, bright doublet state,
dissociative doublet state, CT doublet state

D3 (A1 ) D4 (B1 )

Figure 5.3: Optimized TDDFT structures for the dissociative states with
their R(C-I) bond length. At doublet GS (B2 ) geometry R(C-I)
= 2.095 Å
80 5 Results for reduced doublet system

5.3 Dissociation
Based on the previous results for the non-reduced system and the TDDFT
geometry-states correlation diagram (recall Figure 5.2), only the antisym-
metric mono-dissociation into the symmetric σ ∗ (a1 ) orbital was taken into
account. However, this means a loss in symmetry compared to the C2v
symmetry at the ground state. Hence, all calculations were done using
TDDFT exclusively.
Again, a scan coordinate (frozen and relaxed) describing the simple elonga-
tion of one C-I bond was chosen over the LICC between D0 (B2 ) and D3 (A1 )
and an antisymmetric vibrational mode at the GS, both displayed in Figure
Figure 5.4, for the sake of simplicity. The mono-dissociative vibrational
mode exhibits predominantly displacement vectors not associated with the
cleavage. The LICC differs only marginally from the used elongation scan
coordinate.

Figure 5.4: Left: vibrational mode at ν = 201 cm−1 . Right: LICC between
D0 (B2 ) and D3 (A1 ). Both coordinates of the reduced dye were not
investigated, in favor of an elongation scan.

The aim of the calculations was to clarify whether or not the population
of the mono-dissociative state D3 (A1 ) is leading to a C-I cleavage. The
result for the relaxed scan can be seen in Figure 5.5,2 for the frozen scan see
Figure D.6.
The state D3 (A1 ) is clearly dissociative until the crossing with the ground
state. In the region of the crossing, the results are poorly described due to
2 This coordinate shows the same problem of TDDFT regarding double excitations, as
seen in section 5.2. The diabatic follow up of the bright state would be a double
excitation into the π4∗ (b2 ) from the HOMO π3 (a2 ), as well as from the SOMO σ ∗ (a1 ),
and this is not possible with TDDFT.
5.3 Dissociation 81

4.0

3.5

3.0 D5(A2)
Energy (eV)

D4(B1)
2.5
D3(A1)
2.0 D2(A2)
D1(B1)
1.5

1.0

0.5

0.0 D0(B2)
2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0
R(C-I) in Å

Figure 5.5: TDDFT PESs along the relaxed mono C-I elongation. The arrows
indicate a possible relaxation mechanism. Colorcode: ground state,
bright doublet state, dissociative doublet state, CT doublet state

a break-down of the BOA and pronounced multiconfigurational character.

Afterwards, the system is in a state far away from the equilibrium and,
therefore, TDDFT results have to be handled with caution. Thus, the
energetic increase could be attributed to TDDFT artifacts in this non-
equilibrium region. Moreover, the increase in energy after the crossing with
the GS is small (0.42 eV) compared to the 3 eV of initial excitation energy.
Therefore, this small barrier can easily be overcome by the excess energy in
the system.
Furthermore, the population of D3 (A1 ) leads to a crossing with the GS
D0 (B2 ), from where a return to the initial GS geometry or a dissociation
is possible, offering an explanation for the experimental observation of a
cleavage foremost after several hours of irradiation. [86]
82 5 Results for reduced doublet system

3.5

3.0
D5(A2)
2.5 D4(B1)
D3(A1)
Energy (eV)

2.0 D2(A2)
D1(B1)
1.5

1.0

0.5

0.0 D0(B2)

10 20 30 40 50 60 70 80 90
Dihedral (◦)

Figure 5.6: TDDFT PESs along the relaxed scan around the two aromatic
ring systems. The arrows indicate a possible relaxation mechanism
starting from the bright D5 (A2 ) state. Colorcode: ground state,
bright doublet state, dissociative doublet state, CT doublet state

5.4 Charge Transfer

The optimization of the CT states showed only a small change within the
aromatic ring structure (see section 5.2). Thus, the population of the low
lying CT state(s) due to small vibrations in the ring structure seems likely.
The main relaxation channel can be approximated by a torsion around the
main/side ring dihedral. This concept has been introduced in section 4.5.
The relaxed scan around this dihedral is presented in Figure 5.6 (for the
frozen scan see Figure D.7) and exhibits the same stabilization due to the
imaginary frequencies along the dihedral in the ground state of the C2v
model system which was described extensively for the non-reduced system
(see section 5.2).
The first prominent result is that there is no further stabilization of the
CT states from the relaxed GS minimum, in contrast to the non-reduced
system with a stabilization of almost 1 eV. Secondly, taking the minimum
in the GS as reference, a torsion at room temperature of 55◦ is possible.
5.4 Charge Transfer 83

Thirdly, an interesting relaxation mechanism can be determined, as follows:

Upon initial excitation, the bright D5 (A2 ) state relaxes towards its minimum
at around 50◦ . On its relaxation pathway, it crosses the CT states, which
can get populated. The CT states can relax into their minima at around 80◦
while crossing the dissociative states along their pathway. The population of
these states might result in a dissociation or relaxation back to the ground
state (see section 5.3).
The results obtained with TDDFT regarding the lowest CT state are quite
trustworthy due to the fact that this state and the bright D5 (A2 ) do not
change significantly in energy with RASPT2 (see Figure 5.1).
Finally, there are several GS vibrational modes, which are displayed in
Figure 5.7, corresponding to the discussed torsion. They also coincide with
the geometry relaxation for the optimized CT states along the relaxed scan.

ν = 39 cm−1 ν = 71 cm−1 ν = 97 cm−1 ν = 168 cm−1

Figure 5.7: Vibrational modes at doublet GS B2 geometry corresponding to a
torsion around the main/side ring dihedral.
6 Conclusion
As stated in the introduction the goals of this thesis were i) the method
benchmark, namely for TDDFT vs. RASPT2, ii) to unravel the nature
of the photo-induced catalytic mechanism and iii) to obtain insight into
the photochemistry of the dye, as well as iv) the identification of unwanted
photodegradation channels and how to quench them.
Regarding the comparison of the two methods, TDDFT and RASPT2,
it can be concluded on the basis of both, non-reduced and singly reduced,
systems that TDDFT exhibits problems describing these systems. This is
shown best in Figure 4.1 and 5.1. Using RASPT2 instead of TDDFT results
in a decrease in energy of the bright state (HOMO → LUMO transition)
and an increase of the dissociative and CT states of up to 1.1 eV, which
is also observed in crossing points shifted towards higher energies and
bigger elongations/smaller dihedral degrees. Nevertheless, the qualitative
description of TDDFT is comparable to RASPT2 (see section 4.4.2).
The thorough investigation of the non-reduced singlet system indicates
that this mechanism is not contributing predominantly to the overall two-
component system. The analysis regarding a feasible dissociation rules out
a symmetric di-dissociation, whereas the mono-dissociation is likely to be
possible with enough excess energy given to overcome the barrier towards
the crossing point and the shallow minimum. However, the calculated ISC
rates do not favor a fast population of these excited triplet states, though, it
has to be mentioned that the SOCs for calculating the ISC rates were only
obtained at the GS geometry. On the other hand there are several features
that support this approximation of almost invariant SOCs, like seen for the
di-dissociation coordinate (Figure 4.12), as well as for the LICC in Table 4.5
conducted to study the time constants associated to ISC; and also for the
CT SOCs no significant change was observed for the dihedral coordinate
(see Table 4.6). For the population of the charge transfer states a possible
mechanism was presented on the basis of the potentials in Figure 4.18. But
two points contradict a pathway from the bright S1 to the low lying CT state
via torsion around the ring systems. Firstly, the fact that the calculations
were done with TDDFT and the real position of the crossing point might be

© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2019

K. M. Ziems, Photophysics and Photochemistry of a BODIPY-Based
Photosensitizer, BestMasters, https://doi.org/10.1007/978-3-658-26188-7_6
86 6 Conclusion

3.5
4.0
3.0
D5(A2) 3.5
2.5 D4(B1)
D3(A1) 3.0 D5(A2)

Energy (eV)
Energy (eV)

2.0 D2(A2) D4(B1)

2.5
D1(B1) D3(A1)
1.5 2.0 D2(A2)
D1(B1)
1.5
1.0
1.0
0.5
0.5
0.0 D0(B2) D0(B2)
0.0
10 20 30 40 50 60 70 80 90 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0
Dihedral (◦) R(C-I) in Å

Figure 6.1: Two possible relaxation mechanisms for the reduced system. Left:
Relaxation starting from the bright D5 towards CT and dissocia-
tive states along the relaxed scan around the ring dihedral. Right:
Dissociation of the C-I bond with crossing of the GS along the
relaxed C-I elongation. Colorcode: ground state, bright doublet
state, dissociative doublet state, CT doublet state

energetically less favorable and, secondly, that the SOC showed very small
values for the population of these states (recall Table 4.3 and 4.6).
The reduced doublet system allows a possible population of both, dissocia-
tive and charge transfer states, even at the GS geometry (see Figure 5.1). The
symmetric di-dissociation is again not favorable, but the mono-dissociation
yields a cleavage of the C-I bond with a crossing of the GS along its mono-
dissociative coordinate. Thus, a population of the dissociative state does not
result in an immediate photodegradation, coinciding with the reported 20 h
of photostability (recall chapter 1). Two possible mechanisms for photoacti-
vation were found. First of all, only small vibrations in the ring systems are
needed to populate the low lying CT states. Secondly, a scan around the
main/side ring dihedral revealed promising results for populating CT, but,
unfortunately, also for dissociative states. The excitation into the bright
doublet state and its relaxation towards a minimum on its PES results in
a crossing with the CT states, which cross the dissociative potentials by
relaxing towards their minima. The mono-dissociative state can then relax
according to the potential discussed before, which is depicted in Figure 6.1.
A major advantage of the reduced system is that all states involved have
the same spin symmetry and, therefore, do not depend on high SOCs. On
the other hand, no indications for the need of the Iodines due to a relativistic
6 Conclusion 87

heavy atom effect, as described by T. Beweries et al., [86] could be found.

In conclusion, the transfer of the electron from the sacrificial donor, TEA,
to the BODIPY is the predominant mechanism of this two-component system
to yield long-lived doublet CT states, suitable for transferring an electron
to the catalyst. It could also be shown that in this reduced doublet system
the dissociative states can be populated easily along the population of the
CT states, which result in a cleavage of one C-I bond. The conditions for
the cleavage of the second bond needs an examination of the single Iodine-
substituted BODIPY. Unfortunately, no approach could be determined to
prevent a dissociation of the C-I bond due to the strong interaction of the
respective dissociative states with the mechanistic important CT states.
Furthermore, the postulated mechanism by T. Beweries et al. with a CT
triplet state as starting point has to be treated with caution, because these
states are too high in energy. This hypothesis was based among others on the
TDDFT assignments for the excited state absorptions in transient absorption
spectroscopy (see chapter 1). Their calculations showed a big amount of
CT excitations for these peaks. This assignment might be influenced by
the energetic underestimation of CT states with TDDFT describing this
boron-containing system, as seen in this work on several occasions. Instead,
the mechanism starting with a chemical reduction of the BODIPY, which
was reported for other, non-Iodine BODIPY systems from T. Beweries et
al., is most likely the photophysical background of this catalytic system.
7 Summary
In the presented work quantum chemical methods were applied to investi-
gate the mechanism of a two-component solar hydrogen-generating system
with a BODIPY dye as photosensitizer, [Pd(PPh3 )Cl2 ]2 as precursor for
catalytically-active palladium nanoparticles, and TEA as sacrificial electron
donor. The focus is hereby on the BODIPY dye described by T. Beweries
et al. [86] (displayed in Figure 1.5), which exhibits an increased hydrogen
productivity compared to previously studied BODIPY dyes. Two mod-
ifications on the PS are most likely associated to this observation: i) a
sterical demanding mesityl-group at the meso-position of the BODIPY,
which also enhances the dye’s stability, and ii) a Iodine substitution at the
2,6-position, which might cause increased ISC kinetics due to a heavy atom
effect. Unfortunately, this dye shows photodegradation via C-I cleavage
after longer irradiation. Founded on experimental and quantum chemical
data, two possible mechanisms were proposed, based on the population of
long-lived charge-separated states via CT on the side ring of the BODIPY
(see Figure 1.4). The first approach involves a CT excitation of the BODIPY,
followed by electron transfer to the catalyst and reduction of the dye (non-
reduced singlet system), whereas the second theory suggests a preliminary
reduction of the PS followed by photo-excitation (reduced doublet system).
The aim of this thesis was to investigate the described adversary processes
of photoactivation and photodegradation within both mechanisms. The
likelihood of both mechanisms was assessed with different levels of theory
(TDDFT and RASPT2). Regarding the photoactivation, CT states popu-
lations were investigated, whereas regarding the photodegradation it was
aimed to find an indication for the cleavage of the C-I bond(s), as well as to
elucidate possibilities to enhance the dye’s photostability.

First of all, calculations with different RASPT2 active spaces were con-
ducted - one to treat dissociative states and one for CT transitions - and
subsequently compared to TDDFT results. This was done for several excited
states at the Franck-Condon region in the fully optimized non-reduced singlet
geometry (see Table 4.1), as well as in the fully optimized reduced doublet
geometry (see Table 5.1). Generally, TDDFT overestimated the bright
© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2019
K. M. Ziems, Photophysics and Photochemistry of a BODIPY-Based
Photosensitizer, BestMasters, https://doi.org/10.1007/978-3-658-26188-7_7
90 7 Summary

HOMO-LUMO transition related to the initial photoactivation slightly and

underestimated the CT and dissociative states by up to 1.1 eV.
For the non-reduced system i) excited states and SOCs were computed
at the GS geometry, ii) a TDDFT geometry-states correlation diagram
was prepared to follow the excited state characters of interest diabatically
at different equilibrium structures, iii) different dissociation coordinates
were investigated with a MEP, elongation scans and ISC kinetics, and iv) a
possible CT population coordinate was introduced. The analysis revealed
that a symmetric dissociation is not possible upon S1 excitation, while an
antisymmetric one is unlikely due to high barriers of up to 0.8 eV and very
slow ISCs. For the population of the CT states the SOCs are very small
and, even though the suggested coordinate of a torsion around the main and
side ring dihedral exposed a possible, energetically feasible crossing of the
bright S1 state with a lowered CT state, there might be a high error, since
the calculations were only performed with TDDFT in this case. Therefore,
the exact point of degeneracy among the PESs of the respective states is
uncertain.
The reduced system was studied with i) excited states at GS doublet
geometry, ii) a TDDFT geometry-states correlation diagram, iii) mono-
dissociation scans, and iv) a CT coordinate describing the torsion around
the two aromatic rings. At the GS geometry, the reduced doublet system
showed energetically low lying CT and dissociative states, which can be
populated via small displacements along low-lying vibrational modes within
the ring system and antisymmetric elongation of one C-I bond, respectively.
Moreover, the coordinates for the mono-dissociation and the CT torsion
revealed a possible excited state relaxation cascade starting from the bright
HOMO-LUMO transition. The relaxation of this ππ ∗ state towards its
minimum on the PES along the ring torsion results in a crossing with the
CT PESs, which cross the dissociative PESs by relaxing in their respective
minima. Thus, the dissociative, as well as the CT states, get populated by a
relaxation of the bright state along the ring torsion. For the mono-dissociative
potential, a dissociative behavior with a crossing of the ground state along the
elongation could be confirmed. Both mechanisms are illustrated in Figure 6.1.
Furthermore, all transitions occur with the same doublet multiplicity and,
therefore, proceed in the fast IC time scale and do not depend on SOCs, like
the non-reduced system.
In conclusion, i) the transfer of an electron from the sacrificial donor,
TEA, to the BODIPY dye was identified as the predominant photoactivation
mechanism of the two-component system, ii) the cleavage of the C-I bond
could be explained as closely coupled to the population of the CT states,
7 Summary 91

which prevents any approach to increase the dye’s photostability, and iii)
the poor quantitative description of boron-containing systems with TDDFT
as level of theory was confirmed. iv) A significant heavy atom effect of the
Iodines could not be observed.

The CT and the mono-dissociation scans lack a sufficient level of theory,

and therefore, a next step would be a computation of these potentials with
RASPT2 instead of TDDFT, even though this would imply high computa-
tional costs. An approach towards accessible SOCs at the level of TDDFT
would be the use of, e.g., ADF [190] as quantum chemical program, which
would allow an easy estimation regarding ISC rates along the coordinates.
To get a detailed and quantitative insight into the processes and mechanisms
of the BODIPY dye, dynamical simulations have to be applied. This could
be done e.g. by quantum dynamics on existing PESs or with molecular
dynamic, e.g. by on the fly surface hopping programs like SHARC. [191] Also
the calculations of non-adiabatic couplings between excited states would
give an insight into quantitative aspects within the reduced system. The
reason for the increased hydrogen production due to the incorporation of
Iodines could not be found; however, a heavy atom effect seems to be unlikely
based on this work’s results. Further investigations should, therefore, focus
on a comparison with the non-substituted dye, as well as on BODIPYs
substituted with non-relativistic, sterical demanding and heavy groups, like
CF3 or C6 H11 instead of Iodine.
8 Zusammenfassung
In der vorliegenden Arbeit wurden quantenchemische Methoden verwen-
det, um den Mechanismus eines Zwei-Komponenten Katalysesystem zur
licht-induzierten Wasserspaltung, bestehend aus einem BODIPY Farbstoff
als Photosensibilisator, [Pd(PPh3 )Cl2 ]2 als Vorstufe zu katalytisch-aktiven
Pd-Nanopartikeln und Triethylamin als Elektronen-Opferdonor, zu unter-
suchen. Den Fokus bildete der BODIPY Farbstoff, welcher von T. Bew-
eries et al. beschrieben wurde (siehe Abbildung 1.5) und sich durch eine
erhöhte Wasserstoffproduktion im Zwei-Komponenten System von bisher
studierten BODIPYs absetzte. Zwei Modifikationen sind der vermutete
Grund hierfür: i) eine sterisch anspruchsvolle Mesityl-Gruppe an der meso-
Position des BODIPYs, welche auch die Photostabilität des Farbstoffs erhöht,
und ii) die Iod-Substitution an der 2,6-Position des BODIPY, die einen
Schweratomeffekt mit verbessertem ISC zu langlebigen Triplett Zuständen
ermöglicht. Allerdings zeigen diese C-I Bindungen einen Bindungsbruch nach
längerer Bestrahlung, womit es zur Photozersetzung des Farbstoffes kommt.
Basierend auf experimentellen und quantenchemischen Daten wurden zwei
mögliche Mechanismen der Photoaktivierung vorgeschlagen, die beide auf
der Bevölkerung von langlebigen ladungstransferierten Zuständen am Sei-
tenring aufbauen (siehe Abbildung 1.4). Der erste Mechanismus erfordert
eine CT Anregung des Farbstoffes, gefolgt von einem Elektronentransfer
auf den Pd-Katalysator und einer Reduktion des BODIPYs (non-reduced
system), wohingegen der zweite Mechanismus eine vorgelagerte chemische
Reduktion des PS postuliert (reduced doublet system). Das Ziel dieser Arbeit
war es die gegensätzlichen Prozesse der Photoaktivierung und -zersetzung zu
erörtern. Beide Mechanismen wurden mit verschiedenen quantenchemischen
Methoden (TDDFT und RASPT2) untersucht, um ihre Relevanz für den
Gesamtprozess abzuschätzen. Bezüglich der Photoaktivierung wurde die
Bevölkerung von CT Zuständen fokussiert, wohingegen die Photozersetzung
über dissoziative Zustände erforscht wurde.

Zunächst wurden mit verschiedenen active spaces - einer für dissoziative

und einer für CT Zustände - RASPT2 Berechnungen durchgeführt und mit
den TDDFT Ergebnissen verglichen. Dies wurde für mehrere angeregte
© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2019
K. M. Ziems, Photophysics and Photochemistry of a BODIPY-Based
Photosensitizer, BestMasters, https://doi.org/10.1007/978-3-658-26188-7_8
94 8 Zusammenfassung

Zustände in der Franck-Condon Region sowohl für die optimierte non-

reduced Singulett-Geometrie (siehe Abbildung 4.1), als auch für die optimierte
reduced Doublett-Geometrie (siehe Abbildung 5.1), durchgeführt. Es wurde
allgemein festgestellt, dass TDDFT den hellen HOMO-LUMO Übergang,
welcher die initiale Photoaktivierung widerspiegelt, leicht überschätzt und
die dissoziativen und CT Zustände um bis zu 1.1 eV überschätzt.
Für das non-reduced system wurden i) angeregte Zustände und SOCs in
der Grundzustandsgeometrie berechnet, ii) ein TDDFT Geometrie-Zustands
Korrelationsdiagramm erstellt, um den angeregten Zuständen bei verschie-
denen Gleichgewichtsgeometrien diabatisch zu folgen, iii) verschiedene Dis-
soziationskoordinaten mit einem MEP, Scans entlang der C-I Bindungen
und ISC Kinetiken untersucht, und iv) eine mögliche CT Reaktionskoordi-
nate zur Population derselbigen Zustände vorgestellt. Die Untersuchungen
ergaben, dass eine symmetrische Dissoziation als Folge der Anregung in
den S1 Zustand nicht möglich ist, während ihr antisymmetrisches Pendant
aufgrund von energetischen Barrieren bis zu 0.8 eV und sehr langsamen ISC
Raten unwahrscheinlich ist. Bezüglich der Bevölkerung von CT Zuständen
wurden ebenfalls nur kleine SOCs gefunden. Des Weiteren konnte zwar
festgestellt werden, dass die Torsion um die beiden Ringsysteme in einer
möglichen, energetisch niedrigen Kreuzung des hellen S1 und eines stabil-
isierten CT Zustandes führt, jedoch wurden diese Ergebnisse nur mit TDDFT
erhalten. Der exakte Kreuzungspunkt ist daher nicht bekannt und liegt
höchstwahrscheinlich energetisch höher.
Das reduced system wurde mit Hilfe von i) angeregten Zuständen in der
Doublett Grundzustandsgeometrie, ii) einem TDDFT Geometrie-Zustands
Korrelationsdiagramm, iii) mono-Dissoziationsscans, und iv) einer CT-
Torsions-Koordinate um die beiden Ringsysteme beschrieben. Bereits in
der Grundzustandsgeometrie zeigte das System tief liegende CT bzw. dis-
soziative Zustände, die durch geringe Vibrationen in den Ringsystemen bzw.
durch die Dehnung einer C-I Bindung bevölkert werden können. Darüber
hinaus offenbarten die Koordinaten für die mono-Dissoziation und die CT-
Torsion eine mögliche Relaxationskaskade, beginnend mit einer Anregung in
den hellen HOMO-LUMO Zustand: Dieser Zustand kreuzt bei der Relax-
ation in sein Minimum entlang der Ringtorsionskoordiante die PESs der CT
Zustände, wodurch diese bevölkert werden können. Diese wiederum kreuzen
die dissoziativen PESs bei Relaxation in ihre Minima. Der mono-dissoziative
Zustand kann im Folgenden entlang einer antisymmetrischen C-I Bindungs-
dehnung dissoziieren und schneidet dabei den GS. Diese beiden Mechanismen
sind in Abbildung 6.1 veranschaulicht. Ein intrinsisch vorhandener Vorteil
des reduced systems über das non-reduced system ist, dass alle relevanten
8 Zusammenfassung 95

Zustände in Doublett Spin-Symmetrie vorliegen und somit auf der schnellen

Zeitskala einer IC ablaufen und nicht auf hohe SOCs angewiesen sind.
Zusammenfassend lässt sich feststellen, dass i) die vorgelagerte chemische
Reduktion des BODIPY durch TEA der dominante Mechanismus der Pho-
toaktivierung im Zwei-Komponentensystems ist, ii) die Photozersetzung
durch den C-I Bindungsbruch als eng gekoppelt mit der Bevölkerung der CT
Zustände erklärt werden konnte und somit kein Ansatz zur Steigerung der
Photostabilität ermittelt werden konnte, und iii) die schlechte Beschreibung
von Bor-Systemen mit TDDFT bestätigt werden konnte. iv) Ein erheblicher
Schweratomeffekt der Iodatome konnte nicht beobachtet werden.
Die Scans für die CT und dissoziativen Zustände entbehren einer hin-
reichend guten methodischen Beschreibung, weswegen als nächster Schritt
- trotz der hohen Kosten - eine Berechnung dieser mit RASPT2, anstatt
von TDDFT, anstünde. Eine Möglichkeit zur Generierung von SOCs mit
TDDFT als Methode stellt das Quantenchemie-Programm ADF [190] dar und
würde eine einfache Abschätzung von ISC Raten entlang der Koordinaten
erlauben. Des Weiteren sind für einen detaillierteren und quantitativen
Einblick in die ablaufenden Prozesse quantendynamische Simulationen von
Nöten. Beispiele hierfür wären Quantendynamik auf den errechneten PESs
oder Molekulardynamik, wie das on the fly surface hopping Programm
SHARC. [191] Als weitere Möglichkeit für eine vertiefte quantitative Betrach-
tung des reduced systems würde die Berechnung von non-adiabatic couplings
zwischen angeregten Zuständen dienen. Um den genauen Grund für die
erhöhte Wasserstoffproduktion bei Iod-Substitution zu erörtern, müssten die
Ergebnisse mit dem nicht-substituierten Farbstoff verglichen werden. Auch
die Verwendung von nicht-relativistischen, sterisch anspruchsvollen und
schweren Gruppen als Substituenten, z.B. CF3 oder C6 H11 statt Iodatomen,
wäre eine Betrachtung wert.
Bibliography
[1] Lewis, N. S.; Nocera, D. G. Proceedings of the National Academy of
Sciences 2006, 103, 15729–15735.

[2] Kim, D.; Sakimoto, K. K.; Hong, D.; Yang, P. Angewandte Chemie
International Edition 2015, 54, 3259–3266.

[3] Esswein, A. J.; Nocera, D. G. Chemical reviews 2007, 107, 4022–4047.

[4] Chu, S.; Majumdar, A. nature 2012, 488, 294.

[5] Yuan, Y.-J.; Yu, Z.-T.; Chen, D.-Q.; Zou, Z.-G. Chemical Society
Reviews 2017, 46, 603–631.

[6] Liao, W.-M.; Zhang, J.-H.; Hou, Y.-J.; Wang, H.-P.; Pan, M. Inorganic
Chemistry Communications 2016, 73, 80–89.

[7] Barber, J. Chemical Society Reviews 2009, 38, 185–196.

[8] Tollefson, J. Nature News 2010, 464, 1262–1264.

[9] Pagliaro, M.; Konstandopoulos, A. G.; Ciriminna, R.; Palmisano, G.

Energy & Environmental Science 2010, 3, 279–287.

[10] Armaroli, N.; Balzani, V. ChemSusChem 2011, 4, 21–36.

[11] Xu, Y.; Kraft, M.; Xu, R. Chemical Society Reviews 2016, 45, 3039–
3052.

[12] Oshima, T.; Lu, D.; Ishitani, O.; Maeda, K. Angewandte Chemie
International Edition 2015, 54, 2698–2702.
[13] Joya, K. S.; Joya, Y. F.; Ocakoglu, K.; van de Krol, R. Angewandte
Chemie International Edition 2013, 52, 10426–10437.

[14] Wang, X.; Maeda, K.; Chen, X.; Takanabe, K.; Domen, K.; Hou, Y.;
Fu, X.; Antonietti, M. Journal of the American Chemical Society 2009,
131, 1680–1681.

© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2019

K. M. Ziems, Photophysics and Photochemistry of a BODIPY-Based
Photosensitizer, BestMasters, https://doi.org/10.1007/978-3-658-26188-7
98 Bibliography

[15] Zhang, P.; Zhang, J.; Gong, J. Chemical Society Reviews 2014, 43,
4395–4422.
[16] Yang, J.; Wang, D.; Han, H.; Li, C. Accounts of chemical research
2013, 46, 1900–1909.
[17] Chang, K.; Mei, Z.; Wang, T.; Kang, Q.; Ouyang, S.; Ye, J. ACS nano
2014, 8, 7078–7087.
[18] Wu, B.; Liu, D.; Mubeen, S.; Chuong, T. T.; Moskovits, M.;
Stucky, G. D. J. Am. Chem. Soc 2016, 138, 1114–1117.
[19] Zhang, X.; Jin, Z.; Li, Y.; Li, S.; Lu, G. The Journal of Physical
Chemistry C 2009, 113, 2630–2635.
[20] Gong, L.; Wang, J.; Li, H.; Wang, L.; Zhao, J.; Zhu, Z. Catalysis
Communications 2011, 12, 1099–1103.
[21] Zhang, W.; Hong, J.; Zheng, J.; Huang, Z.; Zhou, J.; Xu, R. Journal
of the American Chemical Society 2011, 133, 20680–20683.
[22] Lazarides, T.; McCormick, T.; Du, P.; Luo, G.; Lindley, B.; Eisen-
berg, R. Journal of the American Chemical Society 2009, 131, 9192–
9194.
[23] McCormick, T. M.; Calitree, B. D.; Orchard, A.; Kraut, N. D.;
Bright, F. V.; Detty, M. R.; Eisenberg, R. Journal of the American
Chemical Society 2010, 132, 15480–15483.

[24] Lehn, J. M.; Sauvage, J.-P. Nouveau Journal De Chimie-New Journal

of Chemistry 1977, 1, 449–451.
[25] Kirch, M.; Lehn, J.-M.; Sauvage, J.-P. Helvetica Chimica Acta 1979,
62, 1345–1384.
[26] Brown, G. M.; Chan, S.; Creutz, C.; Schwarz, H. A.; Sutin, N. Journal
of the American Chemical Society 1979, 101, 7638–7640.

[27] Yuan, Y.-J.; Yu, Z.-T.; Chen, D.-Q.; Zou, Z.-G. Chemical Society
Reviews 2017, 46, 603–631.

[28] Chan, S.-F.; Chou, M.; Creutz, C.; Matsubara, T.; Sutin, N. Journal
of the American Chemical Society 1981, 103, 369–379.
Bibliography 99

[29] Kalyanasundaram, K.; Kiwi, J.; Grätzel, M. Helvetica Chimica Acta

1978, 61, 2720–2730.
[30] DeLaive, P. J.; Sullivan, B.; Meyer, T.; Whitten, D. Journal of the
American Chemical Society 1979, 101, 4007–4008.
[31] Krishnan, C.; Sutin, N. Journal of the American Chemical Society
1981, 103, 2141–2142.
[32] Krishnan, C.; Brunschwig, B. S.; Creutz, C.; Sutin, N. Journal of the
American Chemical Society 1985, 107, 2005–2015.
[33] Brown, G. M.; Brunschwig, B. S.; Creutz, C.; Endicott, J. F.; Sutin, N.
Journal of the American Chemical Society 1979, 101, 1298–1300.
[34] Tschierlei, S.; Karnahl, M.; Presselt, M.; Dietzek, B.; Guthmuller, J.;
González, L.; Schmitt, M.; Rau, S.; Popp, J. Angewandte Chemie
2010, 122, 4073–4076.
[35] Balzani, V.; Credi, A.; Venturi, M. ChemSusChem 2008, 1, 26–58.
[36] Amouyal, E. Solar Energy Materials and Solar Cells 1995, 38, 249–
276.
[37] Esswein, A. J.; Nocera, D. G. Chemical reviews 2007, 107, 4022–4047.
[38] Meyer, T. J. Nature 2008, 451, 778–779.
[39] Balzani, V.; Juris, A.; Venturi, M.; Campagna, S.; Serroni, S. Chemical
Reviews 1996, 96, 759–834.
[40] Rau, S.; Schäfer, B.; Gleich, D.; Anders, E.; Rudolph, M.; Friedrich, M.;
Görls, H.; Henry, W.; Vos, J. G. Angewandte Chemie 2006, 118, 6361–
6364.
[41] Tschierlei, S.; Presselt, M.; Kuhnt, C.; Yartsev, A.; Pascher, T.; Sund-
ström, V.; Karnahl, M.; Schwalbe, M.; Schäfer, B.; Rau, S. Chemistry-A
European Journal 2009, 15, 7678–7688.
[42] Chen, N.-Y.; Xia, L.-M.; Lennox, A. J.; Sun, Y.-Y.; Chen, H.; Jin, H.-
M.; Junge, H.; Wu, Q.-A.; Jia, J.-H.; Beller, M. Chemistry-A European
Journal 2017, 23, 3631–3636.
[43] Kirch, M.; Lehn, J.-M.; Sauvage, J.-P. Helvetica Chimica Acta 1979,
62, 1345–1384.
100 Bibliography

[44] Du, P.; Knowles, K.; Eisenberg, R. Journal of the American Chemical
Society 2008, 130, 12576–12577.
[45] Metz, S.; Bernhard, S. Chemical Communications 2010, 46, 7551–
7553.
[46] DiSalle, B. F.; Bernhard, S. Journal of the American Chemical Society
2011, 133, 11819–11821.
[47] Whang, D. R.; Sakai, K.; Park, S. Y. Angewandte Chemie International
Edition 2013, 52, 11612–11615.
[48] Mak, C. S.; Wong, H. L.; Leung, Q. Y.; Tam, W. Y.; Chan, W. K.;
Djurišić, A. B. Journal of Organometallic Chemistry 2009, 694, 2770–
2776.
[49] Esswein, A. J.; Nocera, D. G. Chemical reviews 2007, 107, 4022–4047.
[50] Eckenhoff, W. T.; Eisenberg, R. Dalton Transactions 2012, 41, 13004–
13021.
[51] Cossi, M.; Rega, N.; Scalmani, G.; Barone, V. Journal of computational
chemistry 2003, 24, 669–681.
[52] Lakadamyali, F.; Kato, M.; Muresan, N. M.; Reisner, E. Angewandte
Chemie International Edition 2012, 51, 9381–9384.
[53] Sakai, T.; Mersch, D.; Reisner, E. Angewandte Chemie International
Edition 2013, 52, 12313–12316.
[54] Cheng, M.; Yang, X.; Li, J.; Zhang, F.; Sun, L. ChemSusChem 2013,
6, 70–77.
[55] Zhang, P.; Wang, M.; Li, C.; Li, X.; Dong, J.; Sun, L. Chemical
Communications 2010, 46, 8806–8808.
[56] Han, Z.; McNamara, W. R.; Eum, M.-S.; Holland, P. L.; Eisenberg, R.
Angewandte Chemie International Edition 2012, 51, 1667–1670.
[57] Lazarides, T.; McCormick, T.; Du, P.; Luo, G.; Lindley, B.; Eisen-
berg, R. Journal of the American Chemical Society 2009, 131, 9192–
9194.
[58] Zhang, W.; Hong, J.; Zheng, J.; Huang, Z.; Zhou, J.; Xu, R. Journal
of the American Chemical Society 2011, 133, 20680–20683.
Bibliography 101

[59] Hartley, C. L.; DiRisio, R. J.; Screen, M. E.; Mayer, K. J.; McNa-
mara, W. R. Inorganic chemistry 2016, 55, 8865–8870.
[60] Sabatini, R. P.; McCormick, T. M.; Lazarides, T.; Wilson, K. C.;
Eisenberg, R.; McCamant, D. W. The Journal of Physical Chemistry
Letters 2011, 2, 223–227.
[61] Bartelmess, J.; Francis, A. J.; El Roz, K. A.; Castellano, F. N.;
Weare, W. W.; Sommer, R. D. Inorganic chemistry 2014, 53, 4527–
4534.
[62] Manton, J. C.; Long, C.; Vos, J. G.; Pryce, M. T. Physical Chemistry
Chemical Physics 2014, 16, 5229–5236.
[63] Luo, G.-G.; Fang, K.; Wu, J.-H.; Dai, J.-C.; Zhao, Q.-H. Physical
Chemistry Chemical Physics 2014, 16, 23884–23894.
[64] Luo, G.-G.; Fang, K.; Wu, J.-H.; Mo, J. Chemical Communications
2015, 51, 12361–12364.
[65] Luo, G.-G.; Lu, H.; Zhang, X.-L.; Dai, J.-C.; Wu, J.-H.; Wu, J.-J.
Physical Chemistry Chemical Physics 2015, 17, 9716–9729.
[66] Loudet, A.; Burgess, K. Chemical reviews 2007, 107, 4891–4932.
[67] Ulrich, G.; Ziessel, R.; Harriman, A. Angewandte Chemie International
Edition 2008, 47, 1184–1201.
[68] Momeni, M. R.; Brown, A. Journal of chemical theory and computation
2015, 11, 2619–2632.

[69] Ni, Y.; Zeng, W.; Huang, K.-W.; Wu, J. Chemical Communications
2013, 49, 1217–1219.
[70] Gresser, R.; Hartmann, H.; Wrackmeyer, M.; Leo, K.; Riede, M.
Tetrahedron 2011, 67, 7148–7155.
[71] Hayashi, Y.; Obata, N.; Tamaru, M.; Yamaguchi, S.; Matsuo, Y.;
Saeki, A.; Seki, S.; Kureishi, Y.; Saito, S.; Yamaguchi, S. Organic
letters 2012, 14, 866–869.

[72] Jiang, X.-D.; Fu, Y.; Zhang, T.; Zhao, W. Tetrahedron Letters 2012,
53, 5703–5706.
102 Bibliography

[73] Yu, C.; Xu, Y.; Jiao, L.; Zhou, J.; Wang, Z.; Hao, E. Chemistry-A
European Journal 2012, 18, 6437–6442.
[74] Berhe, S. A.; Rodriguez, M. T.; Park, E.; Nesterov, V. N.; Pan, H.;
Youngblood, W. J. Inorganic chemistry 2014, 53, 2346–2348.
[75] Sarkar, S. K.; Mukherjee, S.; Thilagar, P. Inorganic chemistry 2014,
53, 2343–2345.
[76] Kamkaew, A.; Lim, S. H.; Lee, H. B.; Kiew, L. V.; Chung, L. Y.;
Burgess, K. Chemical Society Reviews 2013, 42, 77–88.
[77] Awuah, S. G.; You, Y. Rsc Advances 2012, 2, 11169–11183.
[78] Nepomnyashchii, A. B.; Bard, A. J. Accounts of chemical research
2012, 45, 1844–1853.
[79] López Arbeloa, F.; Banuelos, J.; Martínez, V.; Arbeloa, T.; López Arbe-
loa, I. International Reviews in Physical Chemistry 2005, 24, 339–374.
[80] Duran-Sampedro, G.; Agarrabeitia, A. R.; Garcia-Moreno, I.;
Costela, A.; Bañuelos, J.; Arbeloa, T.; López Arbeloa, I.; Chiara, J. L.;
Ortiz, M. J. European Journal of Organic Chemistry 2012, 2012,
6335–6350.
[81] Boens, N.; Leen, V.; Dehaen, W. Chemical Society Reviews 2012, 41,
1130–1172.
[82] Gonçalves, M. S. T. Chemical reviews 2008, 109, 190–212.
[83] Yuan, L.; Lin, W.; Zheng, K.; He, L.; Huang, W. Chemical Society
Reviews 2013, 42, 622–661.
[84] Vendrell, M.; Zhai, D.; Er, J. C.; Chang, Y.-T. Chemical reviews 2012,
112, 4391–4420.
[85] Batat, P.; Cantuel, M.; Jonusauskas, G.; Scarpantonio, L.; Palma, A.;
O’Shea, D. F.; McClenaghan, N. D. The Journal of Physical Chemistry
A 2011, 115, 14034–14039.
[86] Dura, L.; Wächtler, M.; Kupfer, S.; Kübel, J.; Ahrens, J.; Höfler, S.;
Bröring, M.; Dietzek, B.; Beweries, T. Inorganics 2017, 5, 21.
[87] Dura, L.; Ahrens, J.; Pohl, M.-M.; Höfler, S.; Bröring, M.; Beweries, T.
Chemistry-A European Journal 2015, 21, 13549–13552.
Bibliography 103

[88] Ziems, K.; Gräfe, S.; Kupfer, S. Catalysts 2018, 8, 520.

[89] Jensen, F. Introduction to computational chemistry; John wiley & sons,
2017.
[90] Szabo, A.; Ostlund, N. S. Modern quantum chemistry: introduction to
advanced electronic structure theory; Courier Corporation, 2012.
[91] Handy, N. C.; Lee, A. M. Chemical physics letters 1996, 252, 425–430.

[92] Born, M.; Oppenheimer, R. Annalen der Physik 1927, 389, 457–484.
[93] Sutcliffe, B. Advances in Chemical Physics 2000, 114, 1–122.
[94] Bochevarov, A. D.; Valeev, E. F.; David SheRrill, C. Molecular Physics
2004, 102, 111–123.
[95] Hartree, D. R. The wave mechanics of an atom with a non-Coulomb
central field. Part I. Theory and methods. Mathematical Proceedings
of the Cambridge Philosophical Society. 1928; pp 89–110.
[96] Pople, J.; Nesbet, R. The Journal of Chemical Physics 1954, 22,
571–572.
[97] Kobus, J. Advances in Quantum Chemistry 1997, 28, 1–14.

[98] Hariharan, P. C.; Pople, J. A. Theoretical Chemistry Accounts: Theory,

Computation, and Modeling (Theoretica Chimica Acta) 1973, 28, 213–
222.
[99] McLean, A.; Chandler, G. The Journal of Chemical Physics 1980, 72,
5639–5648.

[100] Weigend, F.; Furche, F.; Ahlrichs, R. The Journal of chemical physics
2003, 119, 12753–12762.
[101] Roothaan, C. Hall GG (1951) Proc R Soc (London) A 1951, 205, 541.
[102] Hoffmann, R. The Journal of Chemical Physics 1963, 39, 1397–1412.
[103] Roos, B. O. Multiconfigurational quantum chemistry. 2005.

[104] Davidson, E. R. Journal of Computational Physics 1975, 17, 87–94.

[105] Brillouin, L. Nos 1934, 71, 1933–1934.
104 Bibliography

[106] Slater, J. C. Physical Review 1929, 34, 1293.

[107] Condon, E. Physical Review 1930, 36, 1121.
[108] Sherrill, C. D.; Schaefer, H. F. Advances in quantum chemistry 1999,
34, 143–269.
[109] Roos, B. O. Lecture notes in quantum chemistry; Springer, 1992; pp
177–254.
[110] Roos, B. O.; Taylor, P. R.; Si, P. E. Chemical Physics 1980, 48,
157–173.
[111] Malmqvist, P. A.; Rendell, A.; Roos, B. O. Journal of Physical Chem-
istry 1990, 94, 5477–5482.
[112] Olsen, J.; Roos, B. O.; Jørgensen, P.; Jensen, H. J. A. The Journal of
chemical physics 1988, 89, 2185–2192.
[113] Møller, C.; Plesset, M. S. Physical Review 1934, 46, 618.

[114] Andersson, K.; Malmqvist, P. A.; Roos, B. O.; Sadlej, A. J.; Wolin-
ski, K. Journal of Physical Chemistry 1990, 94, 5483–5488.
[115] Schrödinger, E. Annalen der physik 1926, 385, 437–490.

[116] baron Raleigh, J. W. S. The Theory of Sound; Macmillan & Company,

Limited, 1937.
[117] Roos, B. O.; Andersson, K.; Fülscher, M. P.; Serrano-Andrés, L.;
Pierloot, K.; Merchán, M.; Molina, V. Journal of Molecular Structure:
THEOCHEM 1996, 388, 257–276.

[118] Roos, B. O.; Andersson, K. Chemical physics letters 1995, 245, 215–
223.
[119] Hohenberg, P.; Kohn, W. Physical review 1964, 136, B864.
[120] Löwdin, P.-O. International Journal of Quantum Chemistry 1985, 28,
19–37.
[121] Kohn, W.; Sham, L. J. Physical review 1965, 140, A1133.

[122] Ullrich, C. A.; Yang, Z.-h. Brazilian Journal of Physics 2014, 44,
154–188.
Bibliography 105

[123] Pérez-Jordá, J. M.; Becke, A. D.; San-Fabián, E. The Journal of

chemical physics 1994, 100, 6520–6534.
[124] Gill, P. M.; Johnson, B. G.; Pople, J. A. Chemical Physics Letters
1993, 209, 506–512.
[125] Giuliani, G.; Vignale, G. Quantum theory of the electron liquid; Cam-
bridge university press, 2005.
[126] Vosko, S. H.; Wilk, L.; Nusair, M. Canadian Journal of physics 1980,
58, 1200–1211.
[127] Ceperley, D.; Alder, B.; Perdew, J. Phys. Rev. B 1981, 23, 5048.

[128] Perdew, J. P.; Wang, Y. Physical Review B 1992, 45, 13244.

[129] Becke, A. D. Physical review A 1988, 38, 3098.
[130] Lee, C.; Yang, W.; Parr, R. G. Physical review B 1988, 37, 785.
[131] Perdew, J. P.; Burke, K.; Ernzerhof, M. Physical review letters 1996,
77, 3865.
[132] Perdew, J. P.; Ernzerhof, M.; Burke, K. The Journal of Chemical
Physics 1996, 105, 9982–9985.
[133] Becke, A. D. The Journal of chemical physics 1992, 96, 2155–2160.

[134] Curtiss, L. A.; Raghavachari, K.; Trucks, G. W.; Pople, J. A. The

Journal of chemical physics 1991, 94, 7221–7230.
[135] Adamo, C.; Barone, V. The Journal of chemical physics 1999, 110,
6158–6170.
[136] Staroverov, V. N.; Scuseria, G. E.; Tao, J.; Perdew, J. P. The Journal
of chemical physics 2003, 119, 12129–12137.

[137] Runge, E.; Gross, E. K. Physical Review Letters 1984, 52, 997.
[138] van Leeuwen, R. Time-Dependent Density Functional Theory 2006,
17–31.

[139] Marques, M. A.; Maitra, N. T.; Nogueira, F. M.; Gross, E. K.; Rubio, A.
Fundamentals of time-dependent density functional theory; Springer
Science & Business Media, 2012; Vol. 837.
106 Bibliography

[140] Giuliani, G.; Vignale, G. Quantum theory of the electron liquid; Cam-
bridge university press, 2005.
[141] Ullrich, C. A. Time-dependent density-functional theory: concepts and
applications; OUP Oxford, 2011.
[142] Tsuneda, T. Density functional theory in quantum chemistry; Springer
Science & Business Media, 2014.
[143] Meijer, E. J.; Sprik, M. The Journal of chemical physics 1996, 105,
8684–8689.
[144] Cybulski, S. M.; Seversen, C. E. The Journal of chemical physics 2005,
122, 014117.
[145] Dreuw, A.; Weisman, J. L.; Head-Gordon, M. The Journal of chemical
physics 2003, 119, 2943–2946.
[146] Gritsenko, O.; Baerends, E. J. The Journal of chemical physics 2004,
121, 655–660.
[147] Becke, A. D. The Journal of chemical physics 2002, 117, 6935–6938.
[148] Spiegel, J. D.; Kleinschmidt, M.; Larbig, A.; Tatchen, J.; Marian, C. M.
Journal of chemical theory and computation 2015, 11, 4316–4327.
[149] Chibani, S.; Laurent, A. D.; Le Guennic, B.; Jacquemin, D. Journal
of chemical theory and computation 2014, 10, 4574–4582.
[150] Ji, S.; Ge, J.; Escudero, D.; Wang, Z.; Zhao, J.; Jacquemin, D. The
Journal of organic chemistry 2015, 80, 5958–5963.
[151] Neese, F.; Petrenko, T.; Ganyushin, D.; Olbrich, G. Coord. Chem. Rev.
2007, 251, 288.

[152] Malmqvist, P. A. Department of Theoretical Chemistry - Chemical

Center Lund University 2009, Workshop 5 .
[153] Strange, P. Relativistic Quantum Mechanics; Cambridge University
Press, 1998.
[154] Dirac, P. A. The quantum theory of the electron. Proceedings of the
Royal Society of London A: Mathematical, Physical and Engineering
Sciences. 1928; pp 610–624.
Bibliography 107

[155] Dirac, P. A. A theory of electrons and protons. Proceedings of the

Royal Society of London A: Mathematical, Physical and Engineering
Sciences. 1930; pp 360–365.
[156] Jansen, G.; Heß, B. A. Physical Review A 1989, 39, 6016.
[157] Hess, B. A. Phys. Rev. 1985, A 32, 756.
[158] Breit, G. Phys. Rev. 1932, 39, 616.
[159] Bethe, H.; Salpeter, E. Quantum Mechanics of One- and Two-Electron
Atoms; Springer, 1957.
[160] Neese, F. Wiley Interdisciplinary Reviews: Computational Molecular
Science 2012, 2, 73–78.
[161] Aquilante, F.; Autschbach, J.; Carlson, R. K.; Chibotaru, L. F.; Del-
cey, M. G.; De Vico, L.; Ferré, N.; Frutos, L. M.; Gagliardi, L.;
Garavelli, M. Journal of computational chemistry 2016, 37, 506–541.
[162] Malmqvist, P.-Å.; Roos, B. O. Chemical physics letters 1989, 155,
189–194.
[163] Neese, F. J. Chem. Phys. 2005, 122, 034107.
[164] Malmqvist, P. Å.; Roos, B. O.; Schimmelpfennig, B. Chemical physics
letters 2002, 357, 230–240.
[165] Heß, B. A.; Marian, C. M.; Wahlgren, U.; Gropen, O. Chemical Physics
Letters 1996, 251, 365–371.
[166] Beljonne, D.; Shuai, Z.; Pourtois, G.; Bredas, J. The Journal of
Physical Chemistry A 2001, 105, 3899–3907.
[167] Schimmelpfennig, B. University of Stockholm, Stockholm, Sweden
1996,
[168] Godbout, N.; Salahub, D. R.; Andzelm, J.; Wimmer, E. Canadian
Journal of Chemistry 1992, 70, 560–571.
[169] Figgen, D.; Rauhut, G.; Dolg, M.; Stoll, H. Chemical Physics 2005,
311, 227–244.
[170] Frisch, M.; Trucks, G.; Schlegel, H. B.; Scuseria, G.; Robb, M.; Cheese-
man, J.; Scalmani, G.; Barone, V.; Mennucci, B.; Petersson, G. Inc.,
Wallingford, CT 2009, 200 .
108 Bibliography

[171] Fu, L.; Jiang, F.-L.; Fortin, D.; Harvey, P. D.; Liu, Y. Chemical
Communications 2011, 47, 5503–5505.
[172] Tomasi, J.; Mennucci, B.; Cammi, R. Chemical reviews 2005, 105,
2999–3094.
[173] Mennucci, B.; Tomasi, J.; Cammi, R.; Cheeseman, J.; Frisch, M.;
Devlin, F.; Gabriel, S.; Stephens, P. The Journal of Physical Chemistry
A 2002, 106, 6102–6113.
[174] Kudin, K. N.; Scuseria, G. E.; Cancès, E. The Journal of chemical
physics 2002, 116, 8255–8261.
[175] Roos, B. O.; Veryazov, V.; Widmark, P.-O. Theoretical Chemistry
Accounts 2004, 111, 345–351.
[176] Roos, B. O.; Lindh, R.; Malmqvist, P.-Å.; Veryazov, V.; Widmark, P.-
O. The Journal of Physical Chemistry A 2004, 108, 2851–2858.
[177] Roos, B. O.; Lindh, R.; Malmqvist, P.-Å.; Veryazov, V.; Widmark, P.-
O. The Journal of Physical Chemistry A 2005, 109, 6575–6579.
[178] Roos, B. O.; Lindh, R.; Malmqvist, P.-Å.; Veryazov, V.; Widmark, P.-
O. Chemical physics letters 2005, 409, 295–299.
[179] Roos, B. O.; Lindh, R.; Malmqvist, P.-Å.; Veryazov, V.; Widmark, P.-
O.; Borin, A. C. The Journal of Physical Chemistry A 2008, 112,
11431–11435.
[180] Roos, B.; Malmqvist, P.-Å.; Gagliardi, L. Fundamental World of
Quantum Chemistry 2003,
[181] Aquilante, F.; Lindh, R.; Bondo Pedersen, T. The Journal of chemical
physics 2007, 127, 114107.
[182] Malmqvist, P. A.; Rendell, A.; Roos, B. O. Journal of Physical Chem-
istry 1990, 94, 5477–5482.
[183] Roos, B. O. International Journal of Quantum Chemistry 1980, 18,
175–189.
[184] Fukui, K. Accounts of chemical research 1981, 14, 363–368.
[185] Quapp, W.; Heidrich, D. Theoretical Chemistry Accounts: Theory,
Computation, and Modeling (Theoretica Chimica Acta) 1984, 66, 245–
260.
Bibliography 109

[186] Marian, C. M. Wiley Interdisciplinary Reviews: Computational Molec-

ular Science 2012, 2, 187–203.
[187] Marcus, R. A. The Journal of Chemical Physics 1956, 24, 966–978.
[188] Marcus, R. A. The Journal of Chemical Physics 1956, 24, 979–989.

[189] Kasha, M. Discussions of the Faraday society 1950, 9, 14–19.

[190] te Velde, G.; Bickelhaupt, F. M.; Baerends, E. J.; Fonseca Guerra, C.;
van Gisbergen, S. J. A.; Snijders, J. G.; Ziegler, T. J. Comput. Chem.
2001, 22, 931–967.
[191] Richter, M.; Marquetand, P.; González-Vázquez, J.; Sola, I.;
González, L. Journal of chemical theory and computation 2011, 7,
1253–1258.
A TDDFT and RASPT2 states
A.1 Different ASs and their RAS distributions

Table A.1: Different RAS distributions and their CSFs for different ASs

AS RAS 1
1 p(a2 ), p(b2 ), π1 (a2 ), π1 (b2 ), πph,2 (b1 ), πph,3 (a2 ), π2 (a2 ), π2 (b2 ), π3 (b2 )
2 σ(a1 ), σ(b1 ), p(a2 ), p(b2 ), π1 (a2 ), π1 (b2 ), πph,2 (b1 ), πph,3 (a2 ), π2 (b2 ), π3 (b2 )
3 σ(a1 ), σ(b1 ), p(a2 ), p(b2 ), π1 (a2 ), π1 (b2 ), πph,2 (b1 ), πph,3 (a2 ), π2 (b2 ), π3 (b2
4 σ(a1 ), σ(b1 ), p(a2 ), p(b2 ), π1 (a2 ), π1 (b2 ), π2 (b2 ), π3 (b2
AS RAS 2
1 σ(a1 ), σ(b1 ), σ ∗ (a ∗ ∗
1 ), σ (b1 ), π3 (a2 ), π4 (b2 )
2 σ ∗ (a1 ), σ ∗ (b1 ), π2 (a2 ), π3 (a2 ), π4∗ (b2 )
3 σ ∗ (a1 ), σ ∗ (b1 ), π2 (a2 ), π3 (a2 ), π4∗ (b2 ), π5∗ (b2 )
4 π2 (a2 ), π3 (a2 ), π4∗ (b2 ), πph,2 (b1 ), πph,3 (a2 ), πph,4
∗ ∗
(a2 ), πph,5 (b1 )
AS RAS 3
1 π4∗ (a2 ), πph,4
∗ ∗
(a2 ), πph,5 (b1 ), π5∗ (a2 ), π5∗ (b2 ), π6∗ (b2 )
2 π4∗ (a2 ), πph,4
∗ ∗
(a2 ), πph,5 (b1 ), π5∗ (a2 ), π5∗ (b2 ), π6∗ (b2 )
3 ∗ ∗
π4 (a2 ), πph,4 (a2 ), πph,5∗ (b1 ), π5∗ (a2 ), π6∗ (b2 )
4 π4∗ (a2 ), σ ∗ (a1 ), σ ∗ (b1 ), π5∗ (a2 ), π5∗ (b2 ), π6∗ (b2 )
AS CSFs
1 222869
2 71140
3 122948
4 534332

© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2019

K. M. Ziems, Photophysics and Photochemistry of a BODIPY-Based
Photosensitizer, BestMasters, https://doi.org/10.1007/978-3-658-26188-7
112 A TDDFT and RASPT2 states

A.2 Non-reduced singlet system

In the appendix the abbreviation HOMO → LUMO (H-L) and
HOMO-1 → LUMO (H-1-L) are used.

Table A.2: TDDFT results for the fully optimized non-reduced GS structure

weight E λ
Transition f name
% eV nm
S0 (A1 ) - - - - - HF
S1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 89 2.88 430 0.518 H-L
π2 (a2 ) → π4∗ (b2 ) 11 H-1-L
S2 (A1 ) π2 (b2 ) → π4∗ (b2 ) 98 3.44 360 0.045
S3 (B1 ) π2 (a2 ) → π4∗ (b2 ) 88 3.45 359 0.311 H-1-L
π3 (a2 ) → π4∗ (b2 ) 12 H-L
S8 (A2 ) π3 (a2 ) → σ ∗ (a1 ) 93 4.41 281 0.000 Diss.
S11 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 87 4.55 272 0.000 Diss.

T1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 96 1.55 800 - H-L

T2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 89 2.58 480 - H-1-L
T3 (A1 ) π2 (b2 ) → π4∗ (b2 ) 90 2.75 451 -
T4 (A1 ) π3 (b2 ) → π4∗ (b2 ) 80 3.36 369 -
p(b2 ) → π4∗ (b2 ) 12
T6 (A2 ) πph,2 (b1 ) → π4∗ (b2 ) 82 3.90 318 - CT
T7 (A2 ) π3 (a2 ) → σ ∗ (a
1) 70 4.02 308 - Diss.
π2 (b2 ) → σ ∗ (b1 ) 17
π2 (a2 ) → σ ∗ (a1 ) 10
T8 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 58 4.11 302 - Diss.
π2 (b2 ) → σ ∗ (a1 ) 26
π2 (a2 ) → σ ∗ (b1 ) 11
T11 (B1 ) πph,3 (a2 ) → π4∗ (b2 ) 100 4.29 289 - CT
T12 (B2 ) ∗
π3 (a2 ) → πph,5 (b1 ) 94 4.39 282 - CT
T13 (A1 ) π3 (a2 ) → ∗
πph,4 (a2 ) 85 4.46 278 - CT
A TDDFT and RASPT2 states 113

Table A.3: RASPT2 results for the fully optimized non-reduced GS structure
with AS1 and no level shift

weight E λ
Transition f name
% eV nm
S0 (A1 ) HF 81 - - - HF
S1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 80 2.75 451 1.059 H-L
S2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 70 4.50 276 0.002 H-1-L
DE 10
S3 (A2 ) πph,2 (b1 ) → π4∗ (b2 ) 84 4.77 260 0.000 CT
S4 (A1 ) DE: π3 (a2 ) → π4∗ (b2 ) 48 4.93 251 0.001 H-L
π2 (b2 ) → π4∗ (b2 ) 25
S5 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 78 5.39 230 0.001 Diss.
S6 (B2 ) ∗
π3 (a2 ) → πph,5 (b1 ) 81 6.29 197 0.000 CT
S7 (A2 ) π3 (a2 ) → π4∗ (b2 ) 8.04 154 0.000 H-L
∗
πph,2 (b1 ) → πph,4 (a2 ) 62 intra Ph
DE 21

T1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 82 1.97 630 - H-L

T2 (A1 ) π2 (b2 ) → π4∗ (b2 ) 76 3.66 339 -
T3 (B1 ) π2 (a2 ) → π4∗ (b2 ) 77 4.00 310 - H-1-L
T4 (A1 ) π3 (b2 ) → π4∗ (b2 ) 77 4.42 281 -
T5 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 74 5.20 239 - Diss.
T6 (A2 ) πph,2 (b1 ) → π4∗ (b2 ) 82 5.66 219 - CT
T7 (B2 ) ∗
πph,3 (a2 ) → πph,5 (b1 ) 33 6.09 204 - intra Ph
∗
πph,2 (b1 ) → πph,4 (a2 ) 48 intra Ph
T8 (A2 ) DE - CT
114 A TDDFT and RASPT2 states

Table A.4: RASPT2 results for the fully optimized non-reduced GS structure
with AS2 and no level shift

weight E λ
Transition f name
% eV nm
S0 (A1 ) HF 83 - - - HF
S1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 78 2.66 467 0.954 H-L
S2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 75 3.62 342 0.058 H-1-L
S3 (A2 ) π3 (a2 ) → σ ∗ (a 1) 74 5.01 247 0.000 Diss.
S4 (A1 ) DE: π3 (a2 ) → π4∗ (b2 ) 59 5.05 245 0.003 H-L
π3 (b2 ) → π4∗ (b2 ) 13
S5 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 77 5.37 231 0.001 Diss.
S6 (A2 ) πph,2 (b1 ) → π4∗ (b2 ) 86 5.80 214 0.000 CT
S7 (A2 ) π2 (a2 ) → σ ∗ (a 1) 65 5.84 212 0.000 Diss.
S8 (B2 ) ∗
π3 (a2 ) → πph,5 (b1 ) 84 6.29 197 0.000 CT
S9 (A2 ) DE 51 7.14 174 0.000 Diss.
DE 11

T1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 82 1.88 660 - H-L

T2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 78 3.02 411 - H-1-L
T3 (A1 ) π3 (b2 ) → π4∗ (b2 ) 24 4.04 307 -
π2 (b2 ) → π4∗ (b2 ) 47
T4 (A1 ) π3 (b2 ) → π4∗ (b2 ) 49 4.74 262 -
π2 (b2 ) → π4∗ (b2 ) 24
T5 (A2 ) π3 (a2 ) → σ ∗ (a1 ) 73 4.78 260 - Diss.
T6 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 61 5.06 245 - Diss.
π2 (a2 ) → σ ∗ (b1 ) 13
T7 (A2 ) π2 (a2 ) → σ ∗ (a1 ) 63 5.75 216 - Diss.
T8 (A2 ) πph,2 (b1 ) → π4∗ (b2 ) 86 5.85 212 - CT
T9 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 15 5.93 209 - Diss.
π2 (a2 ) → σ ∗ (b1 ) 50
T10 (A2 ) DE 41 7.19 172 - Diss.
DE 11
DE 12
A TDDFT and RASPT2 states 115

Table A.5: RASPT2 results for the fully optimized non-reduced GS structure
with AS2 and a level shift of 0.3 a.u.

weight E λ
Transition f name
% eV nm
S0 (A1 ) HF 83 - - - HF
S1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 76 2.71 457 0.938 H-L
S2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 75 3.65 340 0.094 H-1-L
S3 (A2 ) π3 (a2 ) → σ ∗ (a 1) 74 5.06 245 0.000 Diss.
S4 (A1 ) DE: π3 (a2 ) → π4∗ (b2 ) 59 5.06 245 0.003 H-L
π3 (b2 ) → π4∗ (b2 ) 13
S5 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 77 5.42 229 0.001 Diss.
S6 (A2 ) πph,2 (b1 ) → π4∗ (b2 ) 86 5.91 210 0.000 CT
S7 (A2 ) π2 (a2 ) → σ ∗ (a 1) 64 5.93 209 0.000 Diss.
S8 (B2 ) ∗
π3 (a2 ) → πph,5 (b1 ) 84 6.33 196 0.000 CT
S9 (A2 ) DE 51 7.22 172 0.000 Diss.
DE 11

T1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 82 1.88 658 - H-L

T2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 78 3.04 408 - H-1-L
T3 (A1 ) π3 (b2 ) → π4∗ (b2 ) 24 4.09 303 -
π2 (b2 ) → π4∗ (b2 ) 47
T4 (A1 ) π3 (b2 ) → π4∗ (b2 ) 49 4.78 260 -
π2 (b2 ) → π4∗ (b2 ) 24
T5 (A2 ) π3 (a2 ) → σ ∗ (a1 ) 72 4.83 257 - Diss.
T6 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 60 5.09 244 - Diss.
π2 (a2 ) → σ ∗ (b1 ) 14
T7 (A2 ) π2 (a2 ) → σ ∗ (a1 ) 62 5.83 213 - Diss.
T8 (A2 ) πph,2 (b1 ) → π4∗ (b2 ) 86 5.95 208 - CT
T9 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 16 5.99 207 - Diss.
π2 (a2 ) → σ ∗ (b1 ) 49
T10 (A2 ) DE 41 7.23 172 - Diss.
DE 11
DE 12
116 A TDDFT and RASPT2 states

Table A.6: RASPT2 results for the fully optimized non-reduced GS structure
with AS3 and no level shift

weight E λ
Transition f name
% eV nm
S0 (A1 ) HF 82 HF
S1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 77 2.63 472 0.921 H-L
S2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 74 3.62 343 0.060 H-1-L
S3 (A2 ) π3 (a2 ) → σ ∗ (a 1) 73 5.00 248 0.000 Diss.
S4 (A1 ) DE: π3 (a2 ) → π4∗ (b2 ) 57 5.03 246 0.000 H-L
π3 (b2 ) → π4∗ (b2 ) 13
π3 (a2 ) → π4∗ (a2 ) 11
S5 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 77 5.34 232 0.001 Diss.
S6 (A2 ) πph,2 (b1 ) → π4∗ (b2 ) 85 5.76 215 0.000 CT
S7 (A2 ) π2 (a2 ) → σ ∗ (a 1) 64 5.84 212 0.000 Diss.
S8 (B2 ) ∗
π3 (a2 ) → πph,5 (b1 ) 83 6.02 206 0.000 CT
S9 (A2 ) DE 51 Diss.
DE 11

T1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 82 1.88 660 - H-L

T2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 77 3.01 412 - H-1-L
T3 (A1 ) π3 (b2 ) → π4∗ (b2 ) 13 4.05 306 -
π2 (b2 ) → π4∗ (b2 ) 57
T4 (A1 ) π3 (b2 ) → π4∗ (b2 ) 52 4.73 262 -
π2 (b2 ) → π4∗ (b2 ) 14
π3 (a2 ) → π4∗ (a2 ) 14
T5 (A2 ) π3 (a2 ) → σ ∗ (a1 ) 73 4.78 259 - Diss.
T6 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 61 5.05 245 - Diss.
π2 (a2 ) → σ ∗ (b1 ) 13
T7 (A2 ) π2 (a2 ) → σ ∗ (a1 ) 62 5.75 216 - Diss.
T8 (A2 ) πph,2 (b1 ) → π4∗ (b2 ) 85 5.83 213 - CT
T9 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 15 5.93 209 - Diss.
π2 (a2 ) → σ ∗ (b1 ) 50
T10 (A2 ) DE 41 7.19 172 - Diss.
DE 11
DE 12
A TDDFT and RASPT2 states 117

Table A.7: RASPT2 results for the fully optimized non-reduced GS structure
with AS4 and no level shift

weight E λ
Transition f name
% eV nm
S0 (A1 ) HF 79 - - - HF
S1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 75 2.64 471 0.915 H-L
S2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 70 3.60 345 0.059 H-1-L
S3 (A1 ) DE: π3 (a2 ) → π4∗ (b2 ) 17 4.49 276 0.056 H-L
π2 (b2 ) → π4∗ (b2 ) 53

T1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 72 1.88 661 - H-L

T2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 75 2.99 414 - H-1-L
T3 (A1 ) π2 (b2 ) → π4∗ (b2 ) 70 4.10 303 -
T4 (A1 ) ∗
πph,2 (b1 ) → πph,5 (b1 ) 33 4.31 288 - intra Ph
∗
πph,3 (a2 ) → πph,4 (a2 ) 49
T5 (B2 ) ∗
πph,3 (a2 ) → πph,5 (b1 ) 40 4.73 262 - intra Ph
∗
πph,2 (b1 ) → πph,4 (a2 ) 34
T6 (A1 ) π3 (b2 ) → π4∗ (b2 ) 76 4.82 257 -
T7 (A2 ) πph,2 (b1 ) → π4∗ (b2 ) 81 4.93 252 - CT
T8 (B1 ) πph,3 (a2 ) → π4∗ (b2 ) 72 4.93 252 - CT
T9 (A1 ) ∗
πph,2 (b1 ) → πph,5 (b1 ) 45 5.06 245 - intra Ph
∗
πph,3 (a2 ) → πph,4 (a2 ) 29
T10 (B2 ) ∗
π3 (a2 ) → πph,5 (b1 ) 68 5.39 230 - CT
T11 (A2 ) π3 (a2 ) → σ ∗ (a
1) 47 5.49 226 - Diss.
π2 (a2 ) → σ ∗ (a1 ) 18
π2 (b2 ) → σ ∗ (b1 ) 13
T12 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 55 5.76 215 - Diss.
π3 (b2 ) → σ ∗ (a1 )
T13 (B2 ) ∗
πph,3 (a2 ) → πph,5 (b1 ) 37 5.86 212 - intra Ph
∗
πph,2 (b1 ) → πph,4 (a2 ) 43
T14 (A2 ) π3 (a2 ) → σ ∗ (a1 ) 23 6.45 192 - Diss.
π2 (a2 ) → σ ∗ (a1 ) 22
π2 (b2 ) → σ ∗ (b1 ) 17
T15 (B1 ) DE 6.62 187 - CT
T16 (A2 ) DE 6.96 178 - CT
118 A TDDFT and RASPT2 states

Table A.8: RASPT2 results for the fully optimized non-reduced GS structure
with AS4 and a level shift of 0.3 a.u.

weight E λ
Transition f name
% eV nm
S0 (A1 ) HF 79 - - - HF
S1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 73 2.69 460 0.906 H-L
S2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 68 3.62 342 0.099 H-1-L
S3 (A1 ) DE: π3 (a2 ) → π4∗ (b2 ) 17 4.52 275 0.056 H-L
π2 (b2 ) → π4∗ (b2 ) 53

T1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 72 1.90 654 - H-L

T2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 75 3.04 408 - H-1-L
T3 (A1 ) π2 (b2 ) → π4∗ (b2 ) 64 4.16 298 -
T4 (A1 ) ∗
πph,2 (b1 ) → πph,5 (b1 ) 31 4.45 279 - intra Ph
∗
πph,3 (a2 ) → πph,4 (a2 ) 40
T5 (B2 ) ∗
πph,3 (a2 ) → πph,5 (b1 ) 39 4.84 256 - intra Ph
∗
πph,2 (b1 ) → πph,4 (a2 ) 36
T6 (A1 ) π3 (b2 ) → π4∗ (b2 ) 65 4.85 256 -
T7 (A2 ) πph,2 (b1 ) → π4∗ (b2 ) 81 4.93 252 - CT
T8 (B1 ) πph,3 (a2 ) → π4∗ (b2 ) 72 5.09 243 - CT
T9 (A1 ) ∗
πph,2 (b1 ) → πph,5 (b1 ) 43 5.23 237 - intra Ph
∗
πph,3 (a2 ) → πph,4 (a2 ) 21
T10 (B2 ) ∗
π3 (a2 ) → πph,5 (b1 ) 68 5.51 225 - CT
T11 (A1 ) π3 (a2 ) → ∗
πph,4 (a2 ) 68 5.52 225 - CT
T12 (A2 ) π3 (a2 ) → σ ∗ (a
1) 42 5.54 224 - Diss.
π2 (a2 ) → σ ∗ (a1 ) 20
π2 (b2 ) → σ ∗ (b1 ) 15
T13 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 56 5.82 213 - Diss.
π3 (b2 ) → σ ∗ (a1 ) 11
A TDDFT and RASPT2 states 119

A.3 Reduced doublet system

a) b) c)

Figure A.1: Spin Densities for optimized singly reduced ground state geometry
with DFT for different multiplicities (doublet and quartet) and
optimization step size. a) doublet optimization with 0.3 Bohr step
size, b) quartet optimization with 0.3 Bohr step size, c) quartet
optimization with 0.1 Bohr step size

ν = −136 cm−1 ν = −135 cm−1 ν = −70 cm−1 ν = −40 cm−1

Figure A.2: Imaginary vibrations at GS doublet (B2 ) geometry.
 120

Table A.9: TDDFT results for the fully optimized singly reduced GS structure

Transition weight E / eV λ / nm f 2S+1 name

D0 (B2 ) - - - - - - -
D1 (B1 ) ∗
π4∗ (b2 ) → πph,5 (b1 ) 99 2.04 607 0.000 2.032 CT
D2 (A2 ) ∗
π4∗ (b2 ) → πph,4 (a2 ) 98 2.11 587 0.000 2.034 CT
D3 (A1 ) π4∗ (b2 ) → σ ∗ (a1 ) 95 2.12 584 0.000 2.108 Diss.
D4 (B1 ) π4∗ (b2 ) → σ ∗ (b1 ) 94 2.26 549 0.000 2.124 Diss.
D5 (A2 ) π3 (a2 ) → π4∗ (b2 ) 87 2.69 460 0.194 2.052 H-L
D6 (A2 ) π2 (a2 ) → π4∗ (b2 ) 93 3.38 367 0.003 2.079 H-1-L

Q1 (B1 ) π3 (a2 ) → σ ∗ (a1 ) 60 3.55 350 - 3.246 Diss.

π2 (a2 ) → σ ∗ (a1 ) 10
Q2 (B2 ) ∗
πph,2 (b1 ) → πph,5 (b1 ) 52 3.62 342 - 3.46 intra Ph
∗
πph,3 (a2 ) → πph,4 (a2 ) 38
Q3 (A1 ) π3 (a2 ) → σ ∗ (b1 ) 52 3.64 341 - 3.274 Diss.
π2 (b2 ) → σ ∗ (a1 ) 15
π2 (a2 ) → σ ∗ (b1 ) 11
Q4 (A1 ) ∗
πph,2 (b1 ) → πph,4 (a2 ) 81 4.68 265 - 3.473 intra Ph
∗
πph,3 (a2 ) → πph,5 (b1 ) 17
Q5 (B2 ) ∗
πph,3 (a2 ) → πph,4 (a2 ) 60 4.73 262 - 3.461 intra Ph
∗
πph,2 (b1 ) → πph,5 (b1 ) 37
A TDDFT and RASPT2 states
A TDDFT and RASPT2 states 121

Table A.10: RASPT2 results for the fully optimized singly reduced GS struc-
ture with AS2 and no level shift

weight E λ
Transition f name
% eV nm
D0 (B2 ) - - - - - -
D1 (A1 ) π4∗ (b2 ) → σ ∗ (a1 ) 70 2.53 490 0.001 Diss.
D2 (A2 ) π3 (a2 ) → π4∗ (b2 ) 49 2.59 478 0.259 H-L
π2 (a2 ) → π4∗ (b2 ) 29 H-1-L
D3 (B1 ) π4∗ (b2 ) → σ ∗ (b1 ) 73 2.76 449 0.000 Diss.
D4 (A2 ) π2 (a2 ) → π4∗ (b2 ) 48 3.40 365 0.001 H-1-L
π3 (a2 ) → π4∗ (b2 ) 31 H-L
D5 (B1 ) π4∗ (b2 ) → πph,5
∗ (b1 ) 86 3.59 346 0.000 CT
D6 (A1 ) π3 (a2 ) → σ ∗ (b1 ) 53 4.56 272 0.001 Diss.
DE 21

Q1 (B1 ) π3 (a2 ) → σ ∗ (a1 ) 62 4.2 295 - Diss.

π2 (a2 ) → σ ∗ (a1 ) 14
Q2 (A1 ) π3 (a2 ) → σ ∗ (b1 ) 59 4.38 283 - Diss.
π2 (a2 ) → σ ∗ (b1 ) 15
Q3 (B1 ) π2 (b2 ) → σ ∗ (a1 ) 51 4.91 253 - Diss.
π3 (a2 ) → σ ∗ (a1 ) 17
π2 (b2 ) → σ ∗ (b1 ) 12
Q4 (A1 ) π2 (a2 ) → σ ∗ (b1 ) 48 5.09 244 - Diss.
π3 (a2 ) → σ ∗ (b1 ) 19
π2 (b2 ) → σ ∗ (a1 ) 13
122 A TDDFT and RASPT2 states

Table A.11: RASPT2 results for the fully optimized singly reduced GS struc-
ture with AS2 with a level shift of 0.3 a.u.

weight E λ
Transition f name
% eV nm
D0 (B2 ) - 83 - - - -
D1 (A1 ) π4∗ (b2 ) → σ ∗ (a1 ) 70 2.58 480 0.000 Diss.
D2 (A2 ) π3 (a2 ) → π4∗ (b2 ) 47 2.63 471 0.263 H-L
π2 (a2 ) → π4∗ (b2 ) 30 H-1-L
D3 (B1 ) π4∗ (b2 ) → σ ∗ (b1 ) 73 2.83 438 0.000 Diss.
D4 (A2 ) π2 (a2 ) → π4∗ (b2 ) 46 3.44 360 0.001 H-1-L
π3 (a2 ) → π4∗ (b2 ) 32 H-L
D5 (B1 ) π4∗ (b2 ) → πph,5
∗ (b1 ) 86 3.64 341 0.000 CT
D6 (A1 ) π3 (a2 ) → σ ∗ (b1 ) 54 4.61 269 0.001 Diss.
DE 21

Q1 (B1 ) π3 (a2 ) → σ ∗ (a1 ) 58 4.26 291 - Diss.

π2 (a2 ) → σ ∗ (a1 ) 17
Q2 (A1 ) π3 (a2 ) → σ ∗ (b1 ) 55 4.43 280 - Diss.
π2 (a2 ) → σ ∗ (b1 ) 18
Q3 (B1 ) π2 (b2 ) → σ ∗ (a1 ) 48 4.96 250 - Diss.
π3 (a2 ) → σ ∗ (a1 ) 21
π2 (b2 ) → σ ∗ (b1 ) 11
Q4 (A1 ) π2 (a2 ) → σ ∗ (b1 ) 45 5.13 242 - Diss.
π3 (a2 ) → σ ∗ (b1 ) 23
π2 (b2 ) → σ ∗ (a1 ) 11
A TDDFT and RASPT2 states 123

Table A.12: RASPT2 results for the fully optimized singly reduced GS struc-
ture with AS4 and no level shift

weight E λ
Transition f name
% eV nm
D0 (B2 ) - - - - - -
D1 (A2 ) π3 (a2 ) → π4∗ (b2 ) 40 2.48 500 0.125 H-L
π2 (a2 ) → π4∗ (b2 ) 26 H-1-L
π4∗ (b2 ) → π4∗ (A2 ) 12
D2 (B1 ) π4∗ (b2 ) → πph,5
∗ (b1 ) 83 2.58 480 0.000 CT
D3 (A2 ) π2 (a2 ) → π4∗ (b2 ) 50 3.33 372 0.002 H-1-L
π3 (a2 ) → π4∗ (b2 ) 23 H-L
D4 (B1 ) π4∗ (b2 ) → σ ∗ (b1 ) 67 3.59 346 0.000 Diss.
D5 (B2 ) DE 72 4.58 271 0.000 CT
DE 11

Q1 (B2 ) ∗
πph,3 (a2 ) → πph,4 (a2 ) 56 4.29 289 - intra Ph
∗
πph,2 (b1 ) → πph,5 (b1 ) 27
Q2 (A1 ) ∗
π3 (a2 ) → πph,5 (b1 ) 80 4.64 267 - CT
Q3 (B1 ) π3 (a2 ) → σ ∗ (a
1) 37 4.85 256 - Diss.
π2 (a2 ) → σ ∗ (a1 ) 22
π2 (b2 ) → σ ∗ (b1 ) 13
Q4 (A1 ) ∗
πph,3 (a2 ) → πph,5 (b1 ) 40 4.98 249 - intra Ph
∗
πph,2 (b1 ) → πph,4 (a2 ) 38
Q5 (B2 ) ∗
πph,2 (b1 ) → πph,5 (b1 ) 53 5.11 243 - intra Ph
∗
πph,3 (a2 ) → πph,4 (a2 ) 24
124 A TDDFT and RASPT2 states

Table A.13: RASPT2 results for the fully optimized singly reduced GS struc-
ture with AS4 with a level shift of 0.3 a.u.

weight E λ
Transition f name
% eV nm
D0 (B2 ) - 81 - - - -
D1 (B1 ) π4∗ (b2 ) → πph,5
∗ (b1 ) 83 2.11 587 0.000 CT
D2 (A2 ) π3 (a2 ) → π4∗ (b2 ) 72 2.49 497 0.183 H-L
D3 (B1 ) π4∗ (b2 ) → σ ∗ (b1 ) 67 3.03 409 0.000 Diss.
D4 (A2 ) π2 (a2 ) → π4∗ (b2 ) 66 3.35 370 0.010 H-1-L
D5 (A2 ) π4∗ (b2 ) → π4∗ (a2 ) 38 3.54 350 0.150
π4∗ (b2 ) → πph,4
∗ (a2 ) 29 CT
D6 (A2 ) π4∗ (b2 ) → πph,4
∗ (a2 ) 50 4.36 284 0.150 CT
π4∗ (b2 ) → π4∗ (a2 ) 23

Q1 (B2 ) ∗
πph,3 (a2 ) → πph,4 (a2 ) 60 4.38 283 - intra Ph
∗
πph,2 (b1 ) → πph,5 (b1 ) 23
Q2 (A1 ) ∗
π3 (a2 ) → πph,5 (b1 ) 82 4.72 262 - CT
Q3 (B1 ) π3 (a2 ) → σ ∗ (a
1) 34 4.89 254 - Diss.
π2 (a2 ) → σ ∗ (a1 ) 23
π2 (b2 ) → σ ∗ (b1 ) 13
Q4 (A1 ) ∗
πph,3 (a2 ) → πph,5 (b1 ) 41 5.06 245 - intra Ph
∗
πph,2 (b1 ) → πph,4 (a2 ) 39
Q5 (B2 ) ∗
πph,2 (b1 ) → πph,5 (b1 ) 57 5.23 237 - intra Ph
∗
πph,3 (a2 ) → πph,4 (a2 ) 20
B Reference weight

Table B.1: Reference weight for AS2 of the non-reduced singlet system. Big
deviations are shown in italic and red

state 0.0 0.3

S0 0.44447 0.49827
S1 0.43820 0.49664
S2 0.43699 0.49574
S3 0.43728 0.49379
S4 0.44163 0.49374
S5 0.43101 0.49257
S6 0.42352 0.48918
S7 0.41632 0.49267
S8 0.43567 0.49410
S9 0.41316 0.49371
T1 0.44371 0.49301
T2 0.44027 0.49280
T3 0.43572 0.49798
T4 0.42478 0.49581
T5 0.43854 0.49421
T6 0.43779 0.49308
T7 0.42201 0.48956
T8 0.42275 0.49320
T9 0.13975 0.49569
T10 0.43446 0.49406

© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2019

K. M. Ziems, Photophysics and Photochemistry of a BODIPY-Based
Photosensitizer, BestMasters, https://doi.org/10.1007/978-3-658-26188-7
126 B Reference weight

Table B.2: Reference weight AS4 of the non-reduced singlet system. Big devia-
tions are shown in italic and red

state 0.0 0.3

S0 0.44588 0.49941
S1 0.40215 0.49715
S2 0.44178 0.49721
S3 0.43864 0.49533
T1 0.43608 0.49688
T2 0.41794 0.49518
T3 0.43798 0.49681
T4 0.35945 0.49422
T5 0.27289 0.49425
T6 0.43820 0.49392
T7 0.43334 0.49864
T8 0.43956 0.49623
T9 0.39909 0.49488
T10 0.43699 0.49665
T11 0.35386 0.49299
T12 0.41622 0.49354
T13 0.43122 0.49425
T14 0.42723 0.49234
T15 0.42980 0.49196
T16 0.41214 0.49361
T17 - 0.49470
T18 - 0.49325
B Reference weight 127

Table B.3: Reference weight for AS2 of the reduced doublet system

state 0.0 0.3

D0 0.43594 0.49199
D1 0.43622 0.49219
D2 0.43395 0.49084
D3 0.43459 0.48990
D4 0.43860 0.49372
D5 0.43658 0.49204
D6 0.44274 0.49639
D7 0.43694 0.49372
Q1 0.43570 0.49257
Q2 0.43401 0.49142
Q3 0.43566 0.49218
Q4 0.43289 0.49109
Q5 0.43367 0.49004
Q6 0.42697 0.48846
Q7 0.43724 0.49277
Q8 0.43911 0.49393
128 B Reference weight

Table B.4: Reference weight for AS4 of the reduced doublet system. Big devia-
tions are shown in italic and red

state 0.0 0.3

D0 0.44393 0.49461
D1 0.44021 0.49259
D2 0.43725 0.49477
D3 0.42950 0.49201
D4 0.44017 0.48849
D5 0.43806 0.48910
D6 0.15086 0.49552
D7 0.32595 0.49328
D8 - 0.49754
D9 - 0.49618
Q1 0.43389 0.49171
Q2 0.42320 0.49089
Q3 0.43766 0.49242
Q4 0.39150 0.49094
Q5 0.36087 0.49310
Q6 0.42935 0.49039
Q7 0.43199 0.49467
Q8 0.42803 0.49399
C Spin-Orbit couplings of the
non-reduced singlet system

Table C.1: SOCs with RASPT2 AS2 within the optimized non-reduced GS C2v
model geometry with no level shift

E SOC with S0 SOC with S1

Transition name
eV in cm−1 in cm−1
T1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 1.88 H-L 0.1 -
T2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 3.02 H-1-L 1.0 -
T3 (A1 ) π3 (b2 ) → π4∗ (b2 ) 4.04 - 0.3
π2 (b2 ) → π4∗ (b2 )
T5 (A2 ) π3 (a2 ) → σ ∗ (a1 ) 4.78 Diss. 716.3 0.6
T6 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 5.06 Diss. 26.6 39.2
π2 (a2 ) → σ ∗ (b1 )
T7 (A2 ) π2 (a2 ) → σ ∗ (a1 ) 5.75 Diss. 971.1 8.2
T9 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 5.93 Diss. 33.8 158.1
π2 (a2 ) → σ ∗ (b1 )

© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2019

K. M. Ziems, Photophysics and Photochemistry of a BODIPY-Based
Photosensitizer, BestMasters, https://doi.org/10.1007/978-3-658-26188-7
130 C Spin-Orbit couplings of the non-reduced singlet system

Table C.2: SOCs with RASPT2 AS2 within the optimized non-reduced GS C2v
model geometry with a level shift of 0.3 a.u.

E SOC with S0 SOC with S1

Transition name
eV in cm−1 in cm−1
T1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 1.88 H-L 0.1 -
T2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 3.04 H-1-L 1.0 -
T3 (A1 ) π3 (b2 ) → π4∗ (b2 ) 4.09 - 0.3
π2 (b2 ) → π4∗ (b2 )
T5 (A2 ) π3 (a2 ) → σ ∗ (a1 ) 4.83 Diss. 778.0 0.2
T6 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 5.09 Diss. 27.4 37.3
π2 (a2 ) → σ ∗ (b1 )
T7 (A2 ) π2 (a2 ) → σ ∗ (a1 ) 5.83 Diss. 925.5 7.3
T9 (B2 ) π3 (a2 ) → σ ∗ (b1 ) 5.95 Diss. 33.3 153.7
π2 (a2 ) → σ ∗ (b1 )

Table C.3: SOCs with RASPT2 AS4 within the optimized non-reduced GS C2v
model geometry with no level shift

E SOC with S0 SOC with S1

Transition name
eV in cm−1 in cm−1
T1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 1.88 H-L 0.0 -
T2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 2.99 H-1-L 0.9 -
T3 (A1 ) π2 (b2 ) → π4∗ (b2 ) 4.10 - 0.4
T7 (A2 ) πph,2 (b1 ) → π4∗ (b2 ) 4.93 CT 4.6 5.0
T8 (B1 ) πph,3 (a2 ) → π4∗ (b2 ) 4.93 CT 0.4 -
T10 (B2 ) π3 (a2 ) → ∗
πph,5 (b1 ) 5.39 CT 6.4 1.0
C Spin-Orbit couplings of the non-reduced singlet system 131

Table C.4: SOCs with RASPT2 AS4 in the optimized non-reduced gGS C2v
model geometry with a level shift of 0.3 a.u.

E SOC with S0 SOC with S1

Transition name
eV in cm−1 in cm−1
T1 (B1 ) π3 (a2 ) → π4∗ (b2 ) 1.90 H-L 0.0 -
T2 (B1 ) π2 (a2 ) → π4∗ (b2 ) 3.04 H-1-L 0.9 -
T3 (A1 ) π2 (b2 ) → π4∗ (b2 ) 4.16 - 0.3
T7 (A2 ) πph,2 (b1 ) → π4∗ (b2 ) 4.93 CT 2.3 5.3
T8 (B1 ) πph,3 (a2 ) → π4∗ (b2 ) 5.09 CT 0.4 -
T10 (B2 ) π3 (a2 ) → ∗
πph,5 (b1 ) 5.51 CT 5.2 0.0
T11 (A1 ) ∗
π3 (a2 ) → πph,4 (a2 ) 5.52 CT - 0.5
D Further diabatic potentials
D.1 Symmetric di-dissociation scan coordinate of
the non-reduced singlet system

© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2019

K. M. Ziems, Photophysics and Photochemistry of a BODIPY-Based
Photosensitizer, BestMasters, https://doi.org/10.1007/978-3-658-26188-7
134 D Further diabatic potentials

S0(A1) S1(B1) S8(A2) S11(B2) T7(A2) T8(B2) T6(A2) T11(B1) T12(B2) T13(A1)
8
7
6
Energy in eV

5
4
3
2
1
0
2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0
R(C-I) in Å

Figure D.1: TDDFT potentials along the frozen symmetric C-I elongation for
the non-reduced system. Colorcode: ground state, singlet state,
dissociative singlet state, triplet state, dissociative triplet state,
triplet CT state

S0(A1) S1(B1) S8(A2) S11(B2) T7(A2) T8(B2) T6(A2) T11(B1) T12(B2) T13(A1)
8
at optimized S11(B2) geometry
7
6
Energy in eV

5
4
3
2
1
0
2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0
R(C-I) in Å

Figure D.2: TDDFT potentials along the frozen symmetric C-I elongation
for the non-reduced system including the states at the optimized,
dissociative S11 (B2 ) geometry. Colorcode: ground state, singlet
state, dissociative singlet state, triplet state, dissociative triplet
state, triplet CT state
D Further diabatic potentials 135

S0(A1) S1(B1) S3(A2) S5(B2) T5(A2) T6(B2)

8
at optimized S11(B2) geometry
7
6
Energy in eV

5
4
3
2
1
0
2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0
R(C-I) in Å

Figure D.3: RASPT2 potentials along the frozen symmetric C-I elongation
for the non-reduced system including the states at the optimized,
dissociative S11 (B2 ) geometry. Colorcode: ground state, singlet
state, dissociative singlet state, dissociative triplet state

2.00
1.75
SOC in cm−1

S1(B1)-T5(A2)
1.50
1.25
1.00
0.75
0.50
0.25
60
SOC in cm−1

S1(B1)-T6(B2)
40

20

0
2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0
R(C-I) in Å

Figure D.4: SOCs for the diabatic potentials along the frozen symmetric
C-I elongation and at optimized S11 (B2 ) geometry between
the S1 (B1 ) state and the two dissociative triplet states, T5 (A2 )
(above) and T6 (B2 ) (below). Calculated with RASPT2 for the
non-reduced system.
136 D Further diabatic potentials

D.2 Charge-transfer of the non-reduced singlet

system

30
28 T13(A1) T12(B2)
26 T11(B1)
24 T6(A2)
22
20 S0(A1)
Energy in eV

18
16
14
12
10
8
6
4
2
0
10 20 30 40 50 60 70 80 90
Dihedral (◦)

20
18 T (A ) T12(B2)
13 1
16
T11(B1)
14 T6(A2)
Energy in eV

12 S0(A1)
10
8
6
4
2
0
10 20 30 40 50 60 70 80 90
Dihedral (◦)

Figure D.5: TDDFT PESs along the frozen scan around the dihedral be-
tween the two aromatic ring systems for the non-reduced system.
Eclipsed above and staggered below. Colorcode: ground state (o),
triplet CT state (x)
D Further diabatic potentials 137

D.3 Antisymmetric mono-dissociation and CT

coordinate of the reduced doublet system

D0(B2) D1(B1) D2(A2) D3(A1) D4(B1) D5(A2)

4.5
4.0
3.5
3.0
Energy in eV

2.5
2.0
1.5
1.0
0.5
0.0
2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4.0
R(C-I) in Å

Figure D.6: TDDFT PESs along the frozen mono C-I elongation for the re-
duced system. Colorcode: ground state, bright doublet state,
dissociative doublet state, CT doublet state
D0(B2) D1(B1) D2(A2) D3(A1) D4(B1) D5(A2)

22
20
18
16
Energy in eV

14
12
10
8
6
4
2
0
10 20 30 40 50 60 70 80 90
Dihedral (◦)

Figure D.7: TDDFT PESs along the frozen scan around the two aromatic ring
systems for the reduced system. Colorcode: ground state, bright
doublet state, dissociative doublet state, CT doublet state
Danksagung
Zunächst möchte ich mich bei Frau Prof. Dr. Stefanie Gräfe für die heraus-
fordernde und interessante Aufgabenstellung, sowie die unterstützenden
Besprechungen bedanken.
Weiterhin gilt mein großer Dank Herrn Dr. Stephan Kupfer ohne dessen
sehr gute Betreuung diese Arbeit nicht Zustande gekommen wäre. Ein
weiterer Dank geht an die gesamte Arbeitsgruppe für die Unterstützung,
die hilfreichen Gespräch und die angenehme Arbeitsatmosphäre. Besonders
möchte ich Johannes Steinmetzer herausstellen, dessen Wissen im Bereich der
Quantenchemie im Allgemeinen und v.a. im Bereich von Python-gestützter
Auswertung und Verarbeitung von Daten meine Arbeit um ein vielfaches
erleichtert und auch abwechslungsreicher gemacht hat. Herrn Matthias
Paul danke ich für die anregenden Gespräche und (fast) ausgeglichenen
Tischtennispartien.
Für die Betreuung der Computer- und Server-Systeme, die für diese Arbeit
verwendet wurden, bedanke ich mich bei Dr. Dirk Bender und Frau Sabine
Irmer.
Prof. Dr. Michael Schmitt möchte ich für die Anfertigung des Zweitgutacht-
ens danken.
Mein herzlicher Dank gilt besonders meiner Familie, die mich bei all
meinen Vorhaben und meinem Studium im Allgemeinen stets unterstützt
und motiviert hat. Schließlich danke ich meinen Freunden und insbesondere
Nina Hagmeyer für die große Hilfe v.a. bezüglich des Korrekturlesens dieser
Arbeit.
Ein großer Dank geht an alle Stiftungen, die mich während meines Stu-
diums nicht nur finanziell, sondern auch ideell gefördert haben. Ich danke
daher der August-Wilhelm-von-Hofmann-Stiftung, der Konrad-Adenauer-
Stiftung und der Studienstiftung des deutschen Volkes.

© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2019

K. M. Ziems, Photophysics and Photochemistry of a BODIPY-Based
Photosensitizer, BestMasters, https://doi.org/10.1007/978-3-658-26188-7