Matrix Isolation and Quantum-chemical Study of Molecules containing Transition Metals in High Oxidation States

Inauguraldissertation

zur Erlangung des akademischen Grades eines

Doctor rerum naturalium (Dr. rer. nat.)

der Fakult¨at fur¨ Chemie und Pharmazie der Albert-Ludwigs-Universit¨at Freiburg i. Br. vorgelegt von

Tobias Schl¨oder

aus Bonn 2013 Dekan: Prof. Dr. Bernhard Breit

Vorsitzender des Promotionsausschusses: Prof. Dr. Thorsten Koslowski

Betreuer der Arbeit: Dr. Sebastian Riedel

Referent: Dr. Sebastian Riedel

Korreferent: Prof. Dr. Ingo Krossing

Datum der mundlichen¨ Prufung:¨ 30. April 2013 Die vorliegende Arbeit wurde im Zeitraum von Januar 2010 bis April 2013 am Institut fur¨ Anorganische Chemie der Albert-Ludwigs-Universit¨at Freiburg unter Anleitung von Dr. Sebastian Riedel angefertigt. Bei ihm m¨ochte ich mich fur¨ die interessante Themenstellung, das Interes- se am Fortschritt der Arbeit, seine große Hilfsbereitschaft bei quantenchemischen und experimentellen Fragen, sowie die nette Betreuung im Allgemeinen bedanken. Prof. Ingo Krossing danke ich fur¨ die Ubernahme¨ des Korreferates, die ebenfalls gute Betreuung sowie die Aufnahme in seinen Arbeitskreis. Außerdem bedanke ich mich bei allen weiteren Personen, die zum Gelingen dieser Arbeit beigetragen haben: Heike Haller, Felix Brosi, Robin Bruckner¨ und Thomas Vent-Schmidt sowie allen Mit- gliedern der Arbeitskreise Krossing, Knapp, Kurz, Hillebrecht und R¨ohr fur¨ die angenehme Arbeitsatmosph¨are und den fachlichen Austausch, ”meinen” Mitarbeiterpraktikanten und Bachelorstudenten Andreas Sch¨oppach, Patrick Muller,¨ Thomas Vent-Schmidt, Marc J¨ager und Benjamin Freyh fur¨ die Mitarbeit an verschiedenen Projekten, Prof. Lester Andrews fur¨ die interessanten Kooperationsprojekte sowie seine große Hilfe beim L¨osen von experimentellen Problemen und schließlich Vera Bruksch und Brigitte J¨orger fur¨ das unkomplizierte Erledigen von organisatorischen und burokratischen¨ Aufgaben. Ganz besonderer Dank gilt weiterhin den Mitarbeitern der Feinmechanischen Werkstatt des chemischen Laboratoriums, insbesondere Reinhard Tomm, Christian Roll und Markus Melder, welche am Aufbau der neuen Matrixanlage wesentlich beteiligt waren und ohne deren Beitrag der experimentelle Teil dieser Arbeit nicht m¨oglich gewesen w¨are. Prof. Harald Hillebrecht und der Deutschen Forschungsgemeinschaft danke ich fur¨ finan- zielle Unterstutzung.¨

Table of Contents

List of Tables XI

List of Figures XIII

List of Abbreviations XV

1 Introduction 1

2 Background 3 2.1 Stabilisation of high oxidation states ...... 3 2.2 Internal reduction reactions ...... 4 2.3 Assignment of oxidation states ...... 5 2.4 The highest oxidation states of the transition metals ...... 6 2.5 Transition metal hexafluorides ...... 7

3 Methods 9 3.1 The matrix isolation technique ...... 9 3.1.1 Generation of the matrix sample ...... 9 3.1.2 Matrix IR spectroscopy ...... 10 3.1.3 Matrix host materials ...... 10 3.1.4 Reactions in the matrix ...... 11

V VI TABLE OF CONTENTS

3.2 Quantum-chemical description of molecules containing high-valent transition metals ...... 11 3.2.1 Electron correlation ...... 12 3.2.2 Relativistic effects ...... 16 3.2.3 Assignment of oxidation states ...... 19

4 Setup of the new matrix isolation apparatus 21 4.1 Cold head and cold window ...... 21 4.2 Thematrixchamber ...... 22 4.3 The high vacuum system ...... 23 4.4 Laser ablation of metals ...... 23 4.5 The spectrometer ...... 24

5 Fluorides of the 3d transition metals 25 5.1 Chromiumfluorides...... 26 5.1.1 Structures...... 27 5.1.2 Thermochemistry ...... 30 5.1.3 Vibrational frequencies ...... 31 5.1.4 Matrix isolation experiments ...... 33 5.1.5 Summary ...... 38 5.2 Manganese fluorides ...... 39 5.2.1 Structures...... 39 5.2.2 Thermochemistry ...... 42 5.2.3 Vibrational frequencies ...... 43 5.2.4 Summary ...... 44 5.3 Ironfluorides ...... 44 5.3.1 Structures...... 45 5.3.2 Thermochemistry ...... 47 TABLE OF CONTENTS VII

5.3.3 Matrix isolation experiments ...... 48 5.3.4 Vibrational frequencies ...... 51 5.3.5 Summary ...... 53 5.4 Cobalt fluorides ...... 53 5.4.1 Structures...... 54 5.4.2 Thermochemistry ...... 55 5.4.3 Vibrational frequencies ...... 56 5.4.4 Summary ...... 57 5.5 Conclusion and outlook ...... 58 5.6 Experimental and computational details ...... 59 5.6.1 Matrix isolation experiments ...... 59 5.6.2 Quantum-chemical calculations ...... 59

6 Oxides of Rh, Ir and Au 61 6.1 Rhodiumtetroxide ...... 61 6.1.1 Matrix isolation experiments ...... 62 6.1.2 Structures...... 62 6.1.3 Thermochemistry ...... 65 6.1.4 Vibrational frequencies ...... 65 6.1.5 Summary ...... 68 6.2 The iridium tetroxide cation ...... 68 6.2.1 Structures...... 68 6.2.2 Thermochemistry ...... 70 6.2.3 Vibrational frequencies ...... 71 6.2.4 Summary ...... 71 6.3 Thegolddioxidecation...... 73 6.3.1 Structures...... 73 VIII TABLE OF CONTENTS

6.3.2 Thermochemistry ...... 76

6.3.3 Vibrational frequencies ...... 77

6.3.4 Summary ...... 79

6.4 Conclusion and outlook ...... 79

6.5 Computational details ...... 80

7 Oxide fluorides of Hg, Au, U and Th 81

7.1 Mercuryoxidefluorides...... 82

7.1.1 Matrix isolation experiments ...... 82

7.1.2 Structures...... 83

7.1.3 Thermochemistry ...... 85

7.1.4 Vibrational frequencies ...... 85

7.1.5 Summary ...... 85

7.2 Goldoxidefluorides...... 87

7.2.1 Structures...... 87

7.2.2 Thermochemistry ...... 90

7.2.3 Vibrational frequencies ...... 91

7.2.4 Summary ...... 92

7.3 Thorium and uranium oxide difluoride ...... 92

7.3.1 Matrix isolation experiments ...... 93

7.3.2 Structures and bonding ...... 93

7.3.3 Vibrational frequencies ...... 97

7.3.4 Summary ...... 97

7.4 Conclusion and outlook ...... 97

7.5 Computational details ...... 98

8 Conclusion 101 TABLE OF CONTENTS IX

Appendix 103 A Technicaldrawings ...... 103 X TABLE OF CONTENTS List of Tables

5.1 Structural parameters of molecular chromium fluorides ...... 28

5.2 Calculated thermochemistry of CrF6 and CrF5 ...... 30 5.3 Wavenumbers of the Cr–F stretching modes of molecular chromium fluorides 31

5.4 Isotopic shifts of selected Cr–F stretching modes of CrF4, CrF5 and CrF6 . 32

5.5 Observed bands after the reaction of chromium atoms with F2 ...... 34 5.6 Structural parameters of molecular manganese fluorides ...... 41 5.7 Calculated thermochemistry of molecular manganese fluorides ...... 42 5.8 Wavenumbers of the Mn–F stretching modes of molecular manganese fluorides 43 5.9 Structural parameters of molecular iron fluorides ...... 46

5.10 CV correlation effects on the calculated bond lengths of FeF2, FeF3 and FeF4 46 5.11 Calculated thermochemistry of molecular iron fluorides ...... 47 5.12 Wavenumbers of the Fe–F stretching modes of molecular iron fluorides . . 51

5.13 CV correlation effects on the calculated wavenumbers of FeF2, FeF3 and FeF4 52 5.14 Isotopic shifts of selected Fe–F stretching modes of molecular iron fluorides 52 5.15 Structural parameters of molecular cobalt fluorides ...... 55 5.16 Calculated thermochemistry of molecular cobalt fluorides ...... 56 5.17 Wavenumbers of the Co–F stretching modes of molecular cobalt fluorides . 57

6.1 Structural parameters of molecular rhodium oxides ...... 64

6.2 Calculated thermochemistry of RhO4 and RhO2 ...... 65

XI XII LIST OF TABLES

6.3 Selected wavenumbers of molecular rhodium oxides ...... 66 6.4 Isotopic shifts of selected modes of molecular rhodium oxides ...... 67

+ 6.5 Structural parameters of different [IrO4] isomers and their argon complexes 69

+ 6.6 Calculated thermochemistry of [IrO4] and its argon complexes ...... 70

+ 6.7 Calculated argon complexation energies of the different [IrO4] isomers . . 71

+ 6.8 Selected wavenumbers of different [IrO4] isomers ...... 72

+ 6.9 Selected wavenumbers of the argon complexes of the different [IrO4] isomers 72

+ 6.10 Structural parameters of different [AuO2] isomers and their argon complexes 74 6.11 Calculated energy differences between different electronic states of [OAuO]+ 75

1 + 6.12 Calculated thermochemistry of [Au(η -O2)] and its argon complexes . . . 76

+ 6.13 Calculated argon complexation energies of the different [AuO2] isomers . 77

+ 6.14 Selected wavenumbers of different [AuO2] isomers ...... 78

+ 6.15 Selected wavenumbers of the argon complexes of the different [AuO2] isomers 78

7.1 Structural parameters of molecular mercury oxide fluorides ...... 84 7.2 Calculated thermochemistry of molecular mercury oxide fluorides . . . . . 85 7.3 Wavenumbers of the stretching modes of molecular mercury oxide fluorides 86 7.4 Isotopic shifts of selected stretching modes of FHgOF and OHgF ...... 86 7.5 Structural parameters of molecular gold oxide fluorides ...... 89 7.6 Calculated thermochemistry of molecular gold oxide fluorides ...... 90 7.7 Calculated argon complexation energies of AuF, AuO and AuOF ...... 90 7.8 Wavenumbers of the stretching modes of molecular gold oxide fluorides . . 91

7.9 Structural parameters of ThOF2 and UOF2 ...... 94

7.10 NBO analysis of the U–O and Th–O bonds in UOF2 and ThOF2 ...... 96

7.11 Wavenumbers of the stretching modes of ThOF2 and UOF2 ...... 97 List of Figures

2.1 The highest oxidation states of the transition metals in homoleptic fluoride andoxidecomplexes ...... 6

3.1 Schematic illustration of the Pauli repulsion between ligand orbitals and the outermost core shell in a transition metal complex ...... 16

4.1 Picture of the copper pieces for mounting the cold window to the cold head 22

4.2 Pictures of the upper and lower part of the new matrix chamber ...... 22

4.3 Pictures of the motor holder with the magnetic coupling ...... 24

5.1 The highest known neutral binary fluorides of the 3d elements ...... 25

5.2 Optimised structures of molecular chromium fluorides ...... 28

5.3 Scan of the ”mexican hat“ potential energy surface of CrF5 ...... 29

5.4 IR spectra obtained after co-deposition of chromium atoms with F2 diluted inneonorargon...... 35

5.5 IR spectra obtained after co-deposition of chromium atoms with F2 diluted in neon (UV irradiation experiments and different F2 concentrations) . . . 36

5.6 IR spectra obtained after co-deposition of chromium atoms with 5% F2 in neon and in neat fluorine ...... 37

5.7 Optimised structures of molecular manganese fluorides ...... 40

5.8 Optimised structures of molecular iron fluorides ...... 45

5.9 Born-Fajans-Haber cycle for the formation of solid iron tetrafluoride . . . . 48

XIII XIV LIST OF FIGURES

5.10 IR spectra obtained after co-deposition of iron atoms with F2 diluted in neonorargon ...... 49

5.11 IR spectra obtained after co-deposition of iron atoms with F2 diluted in neon (UV irradiation experiments and different F2 concentrations) . . . . . 50 5.12 Optimised structures of molecular cobalt fluorides ...... 54 5.13 Born-Fajans-Haber cycle for the formation of solid cobalt tetrafluoride . . . 56

6.1 Optimised structures of molecular rhodium oxides ...... 63

+ 6.2 Optimised structures of different [IrO4] isomers ...... 69

+ 6.3 Optimised structures of different [AuO2] isomers and their argon complexes 73

7.1 Optimised structures of molecular mercury oxide fluorides ...... 83 7.2 Calculated spin density of OHgF ...... 84 7.3 Optimised structures of molecular gold oxide fluorides and their argon complexes ...... 88

7.4 Calculated spin densities of AuOF2 and OAuF ...... 90

7.5 Optimised structures of ThOF2 and UOF2 ...... 93

7.6 CASPT2 orbitals for the U–O bond in UOF2 ...... 95

7.7 CASPT2 orbitals for the Th–O bond in ThOF2 ...... 96

A1 Technical drawing of the window holder ...... 103 A2 Technical drawing of the holding plate ...... 104 A3 Technical drawing of the upper part of the matrix chamber ...... 104 A4 Technical drawing of the lower part of the matrix chamber (part 1) . . . . 105 A5 Technical drawing of the lower part of the matrix chamber (part 2) . . . . 105 A6 Technical drawing of the motor holder ...... 106 A7 Technical drawing of the first magnet holder ...... 106 A8 Technical drawing of the second magnet holder ...... 107 List of Abbreviations

%TAE Percentage of triples contribution to atomisation energies

CASPT2 Complete active space 2nd order perturbation theory

CASSCF Complete active space SCF

CBS Complete basis set

CC Coupled cluster

CCSD Coupled cluster with single and double excitations

CCSD(T) Coupled cluster with single, double and perturbative triple excitations

CI Configuration interaction

CSF Configuration state function

CV Core-valence

DFT Density functional theory

ECP Effective core potential

EPR Electron paramagnetic resonance

FCI Full CI

FIR Far infrared

FT-IR Fourier transform infrared

GED Gas phase electron diffraction

HF Hartree-Fock

IR Infrared

XV XVI LIST OF ABBREVIATIONS

MBPT Multi-body perturbation theory

MCQDPT2 Multi-configurational quasi-degenerate 2nd order perturbation theory

MCSCF Multi-configurational SCF

MIR Mid-infrared

MPnn th order Møller-Plesset perturbation theory

NBO Natural bond orbital

NPA Natural population analysis

SCF Self-consistent field

SO Spin-orbit

UV Ultraviolet

UV/Vis Ultraviolet/visible

VSEPR Valence shell electron pair repulsion

ZPE Zero point energy 1 Introduction

The formal concept of oxidation states is one of the fundamental principles in chemistry and as such usually already taught in first year’s chemistry lectures. The reason for this is that redox reactions are ubiquitous and of fundamental importance in different fields of chemistry. They are for example crucial for all kinds of catalytic reactions, in artificial processes as well as in enzymatic systems where transition metals frequently play a decisive role. In the catalysis of oxidation reactions, highly oxidised intermediates are often involved in the key steps, and the investigation of high oxidation states can contribute to a better understanding of these processes. Compounds containing elements in high oxidation states are usually potent oxidisers which can also be used in preparative inorganic chemistry. The hexafluorides of the second and third row transition metals have for example high oxidiser strengths as shown by the [1] [2] pioneering work of Bartlett who used PtF6 for the oxidation of dioxygen and xenon, disproving by the latter reaction the inertness of xenon[3] and opening thereby the now vast field of noble gas chemistry.[4, 5] Transition metal hexafluorides were further used as oxidising agents in the syntheses of new and interesting species, as for example in the + [6] preparation of the [Cl4] cation obtained after the reaction of dichlorine with IrF6. Since highly oxidised compounds often tend to be very reactive, their stabilisation requires special experimental setups. Some compounds are only stable as isolated molecules at low temperatures. These conditions can be reached using the matrix isolation technique which was successfully applied to the stabilisation of highly oxidised transition metal compounds, [7] [8] as for example the IrO4 and HgF4 molecules representing the unprecedented respective VIII and IV oxidation states of iridium and mercury. As in many other fields, quantum chemical calculations have gained an important role in high oxidation state chemistry, too. During the last years the availability of more and more powerful computational (hardware) resources together with the development of increasingly sophisticated theoretical methods and implementations thereof has made computational chemistry an almost standard tool for today’s chemists. Quantum-chemical calculations on small model systems can help to build up knowledge about oxidation or reduction reactions and identify possible target molecules for synthetic chemists. A famous example for the predictive power[9] of these calculations in the successful experimental preparation of the above-mentioned mercury tetrafluoride molecule, 30 years after its prediction[10] and almost 20 years after the first quantum-chemical calculations of its thermochemical stability.[11, 12] Today, these calculations can not only be extended to larger systems but

1 2 CHAPTER 1: INTRODUCTION are also capable of very accurately predicting structural parameters and spectroscopic properties. Scientists have always been fascinated by exploring the edges of the possible, in reaching and going beyond limits and thereby pushing them further. The highest known oxidation state of any element today is VIII, as for example in the long-known tetroxides OsO4, RuO4 and XeO4 as well as their derivatives. Recently, this oxidation state could also be stabilised for iridium (in the IrO4 molecule, vide supra) and it was already suggested by Jørgensen that this element should have the highest probability of being further oxidised.[13] + Its IX oxidation state might indeed by stabilised in the [IrO4] cation, which was predicted to be stable by means of quantum-chemical calculations.[14] The cited examples show that high oxidation state chemistry is an active field of research and that there is still room for new discoveries. The goal of the present thesis was to characterise new molecules containing transition metals in high oxidation states, both on a computational level and experimentally. Before proceeding to the experiments, the thermochemical stability of the investigated compounds was evaluated by state-of-the-art quantum-chemical calculations. If their formation was predicted to be favourable, the corresponding experiments were conducted using the matrix isolation technique which was chosen because of its ability to stabilise the supposedly highly reactive product molecules. In order to make these experiments, a new matrix chamber equipped with a cold head and connected to a high vacuum system had to be installed together with a laser ablation setup for the evaporation of metals to be used as reactants in subsequent oxidation reactions. The so-formed products were analysed by infrared spectroscopy and the assignment of the new bands was supported by a comparison with the calculated vibrational spectra of the target molecules. This thesis is organised as follows: In chapter 2 some general information about high oxidation state transition metal chemistry and a survey of the highest known oxidation states of these elements is given. Next, the experimental and theoretical methods used in this work will be briefly described in chapter 3. The building up of the new matrix isolation apparatus with the connected laser ablation setup is subject of chapter 4, followed by the obtained computational and experimental results for binary fluorides in chapter 5, binary oxides in chapter 6 and mixed oxide fluorides in chapter 7. 2 Background

The content of this chapter is also part of a review on high oxidation state quantum chemistry by both the author and the supervisor of this thesis which is already accepted for publication (T. Schl¨oder, S. Riedel in Comprehensive Inorganic Chemistry II, J. Reedijk, K. Poeppelmeier (Eds.), Elsevier).

This chapter gives a short survey of high oxidation state transition metal chemistry and the highest oxidation states of the d-block elements will be presented along with the ligand systems necessary for their stabilisation. However, not every oxidation state can be stabilised and when approaching extreme oxidation numbers it is important to study the molecules more detailedly in order to avoid an erroneous assignment of the oxidation states. Further, the current developments in this field of chemistry will be discussed in order to set the background for the results presented in the later chapters of this thesis.

2.1 Stabilisation of high oxidation states

The highest possible oxidation state of a given transition metal (the same holds for any other element) is attained just before the oxidation power of the metal atom is high enough to oxidise its ligand sphere. Highly electronegative ligands (or counterions in the case of ionic compounds) are therefore necessary for the stabilisation of extreme oxidation states. Hence, fluorine is the ligand of choice and many elements reach their highest oxidation state in their fluorides. In other cases the highest oxidation numbers are achieved in the oxides, and this is because the lower formal –II oxidation state of the oxo ligand allows a smaller number of ligands (for the stabilisation of the same oxidation state) when compared with the fluorides where steric crowding, i. e. the repulsion between the electron clouds of the ligands, can easily become an important factor. It is also possible to stabilise high oxidation states in complex ions. This leads to an additional kinetic stabilisation because two ions bearing charges of the same sign repel each other electrostatically and bimolecular decomposition reactions hence have very high activation barriers. Together with an appropriate counterion, these complex ions can also form salts which are stabilised by higher lattice energies when compared with the corresponding binary compounds. In charged species, the number of ligands necessary for the stabilisation of a given oxidation state furthermore differs from that in the corresponding neutral compounds and this leads, on the one hand, to a reduction of steric crowding in cationic complexes. On the

3 4 CHAPTER 2: BACKGROUND other hand, the charge also leads to a variation of the bond lengths which increase when electrons are added. This can be explained by the slightly antibonding character of the lowest unoccupied orbitals of these complexes with respect to the metal-ligand bonds. The additional electrons thus lead to a reduction of the steric crowding in the anionic complexes. As these two effects are in opposite directions they partially cancel but in many cases the dominating effect is the stabilising one. Some examples for the stabilisation of + + + [15–21] high oxidation states in complex ions are the [CrF5] , [MnF4] and [CoF4] cations 2− [22, 23] as well as the [CuF6] anion for all of which the corresponding neutral molecules could not be characterised so far (vide infra for neutral chromium fluorides).

2.2 Internal reduction reactions

When the metal atom is the oxidised beyond the limit set by the redox stability of the ligands the latter are oxidised by the former. This process is referred to as internal reduction as the metal atom is reduced intramolecularly by its own ligand sphere. In the 2 1 case of oxygen ligands this can lead to the formation of peroxo- (η -O2), superoxo- (η -O2) and oxyl complexes. For example, the tetroxides of the group 8 elements osmium and ruthenium are stable molecules with the metals in their VIII oxidation states whereas VIII [24, 25] for their lighter homologue iron, reports about the Fe O4 molecule are probably incorrect.[26] In this hypothetical molecule, FeVIII would oxidise its ligand sphere, yielding VI 2 the [Fe O2(η -O2)] complex which was indeed isolated in cryogenic argon matrices and characterised by infrared and M¨ossbauer spectroscopy supported by quantum-chemical calculations.[27–29] By contrast, fluorine ligands are more resistant towards oxidation and thus stable in their –I oxidation state even in the vicinity of highly oxidised metal atoms and also in cases where oxide ligands are oxidised to higher oxidation numbers than –II. This is well illustrated by a comparison between the stability of the fluoro and oxo complexes of the late 5d transition metals (Pt–Hg). The highest possible oxidation state of mercury is IV and can only be stabilised in its tetrafluoride form whereas oxo complexes of HgIV are yet unknown. The situation is similar for platinum and gold, mercury’s left side neighbours V in the periodic table of elements: The binary fluoride Au F5 is a well-known molecule whereas compounds of AuV containing oxygen ligands could not yet be unambiguously + [30] characterised. The [AuO2] cation was observed in the gas phase, but it is most probably V VI not a Au species (see chapter 6). In the case of platinum, the hexafluoride Pt F6 is a stable molecule whereas only scarce information is available for the corresponding trioxide, VI [31] Pt O3. Although the latter is sometimes mentioned in textbooks as a brown-red substance, it might actually only be stable as an isolated molecule and a band was indeed assigned to molecular PtO3 in matrix IR spectra obtained after the reaction of laser-ablated [32] Pt atoms with O2. 2.3. Assignment of oxidation states 5

However, even if fluoride ligands are much less prone to internal reduction processes than oxide ligands, they can also be oxidised if the oxidation number of the metal atoms exceeds its highest possible value, as exemplified by the case of the gold heptafluoride molecule: Its preparation was published based on a characterisation by IR spectroscopy and elemental analysis.[33] Not long ago, it could be shown by a quantum-chemical reinvestigation that if the stoichiometry was indeed AuF7, the formulation as gold heptafluoride was misleading [34] as this species is better described as a AuF5 · F2 complex. This compound is indeed one of the rare examples in which F2 is coordinated to another molecule. Another example is given by the “tetrafluorides” of the group 13 elements which are better described as [35] (F2)MF2 complexes (M = B, Ga, Al, In, Tl).

2.3 Assignment of oxidation states

Usually the formal oxidation state of a central atom in a coordination sphere is defined as the charge remaining on the central atom after removing all coordinating ligands in their respective most stable forms. The bonding electron pairs between the metal atom and the ligands are therefore exclusively assigned to the more electronegative fragment. Due to the high electronegativities of both oxygen and fluorine, this assignment is usually unproblematic for oxides and fluorides. It was however already mentioned in the previous section that internal reduction reactions occur when the oxidation power of the metal is high enough to oxidise its ligand sphere. Care must therefore be taken when assigning oxidation states in compounds which possibly contain atoms in extreme oxidation states since the usual assignment of the oxidation states based on the stoichiometry only (relying on the standard –I and –II oxidation numbers of fluoro and oxo ligands respectively) is not longer possible a priori. A closer look at the structure generally gives better information about the oxidation state of the metal, no matter if the structure was observed by experimental methods like electron diffraction, deferred from vibrational spectra or calculated using quantum-chemical methods. The reason for this is that internal reduction processes often lead to significant structural changes which allow to correctly assign the oxidation state, as for example in the case of the dioxides and corresponding peroxides. Of course, other experimental techniques for the determination of oxidation states exist which directly depend on the electronic structure of the compounds, for example electron paramagnetic resonance (EPR) spectroscopy (if unpaired electrons are involved), X-ray absorption spectroscopic techniques (XAS, XANES) and also, at least for some nuclei like iron, tin or gold, M¨ossbauer spectroscopy. If however, as it is often the case in mass spectrometric investigations, no information is available about the structure of a molecule, the only reliable way to determine its molecular and electronic structure is the use of computational methods. The assignment of oxidation states on the basis of quantum-chemical calculations will be discussed in the next chapter. 6 CHAPTER 2: BACKGROUND 2.4 The highest oxidation states of the transition met- als

It was already mentioned above that the two main factors which influence the highest possible oxidation state of an element are its oxidation power (in its highly oxidised form) and its size which determines the highest possible number of ligands in its coordination sphere. These two influences can serve to explain the experimentally observed trends for the highest oxidation states of the transition metals, each row of which can be divided into three regions. At first, the highest possible oxidation state increases linearly and coincides with the group number which corresponds to the number of valence electrons. In this region both fluorides and oxides are in general stable compounds. The second part begins when the steric repulsion between the increasing number of fluoride ligands necessary to reach the high oxidation number becomes too important and where therefore oxide ligands are necessary to stabilise higher oxidation states. In principle, the linear increase of the oxidation numbers could proceed up to a hypothetical dodecavalent group 12 metal with its valence shell completely oxidised. For the late transition metals however, the ionisation potentials increase sharply with increasing group number and thus destabilise the high oxidation states. In this third region finally, the more electronegative fluorine is therefore again needed as a ligand for the stabilisation of the late transition metals’s extreme oxidation states if they are at all accessible.

5 4 3 4 2 3 3 3 4 4 5 5 6 7 6 6 4 4 2 2 Sc Ti V Cr (6) Mn(5) Fe (4) Co (5) Ni (4) Cu (4) Zn

Ox. Fl. 3 3 4 4 5 5 6 6 7 6 8 6 6 6 4 4 3 3 2 2 M Y Zr Nb Mo Tc Ru Rh Pd Ag Cd

6 3 3 4 4 5 5 6 6 7 7 8 8 6 6 6 3 5 2 2 La Hf Ta W Re Os (7) Ir Pt Au Zn

Figure 2.1 : The highest oxidation states of the transition metals in neutral (in parentheses: ionic) homoleptic fluoride and oxide complexes; values taken from ref. [36] and updated (see text).

The highest known oxidation states for neutral binary transition metal oxides and fluorides are displayed in figure 2.1 in which the trends outlined above can be easily recognised. The onset of the second region where the binary fluorides are unstable because of steric crowding shifts from group 6 for the 3d elements to group 8 for the 5d elements which have larger atomic radii. As the sizes of the atoms within one row further decrease with increasing Z the only stable heptafluoride known for all d-block elements is ReF7. However, [37] although not yet experimentally verified, TcF7 was predicted to be a stable molecule, too. Osmium heptafluoride was also calculated to be thermochemically stable,[38] but its 2.5. Transition metal hexafluorides 7 unambiguous experimental characterisation is still outstanding as its original synthesis[39] was shown not be reproducible.[40] Even higher oxidation numbers can be reached for the midtransition elements Ru, Os and Ir which form stable tetroxides (vide supra) with the metals in their respective VIII oxidation states. No example of any element in the next higher oxidation number (IX) was yet observed but quantum-chemical calculations predict + [14] that this oxidation state might be stabilised in the iridium tetroxide cation [IrO4] . Finally, for the late transition metals the highest possible oxidation numbers decrease again and are only reached for the third row elements in PtF6, AuF5 and HgF4 where they are stabilised due to relativistic effects.

2.5 Transition metal hexafluorides

One class of high-valent transition metal fluorides that has attracted special attention is the hexafluorides, and molecules of MF6 stoichiometry were experimentally characterised without doubt for nine elements (M = Mo, Tc, Ru, Rh, W, Re, Os, Ir and Pt).[41] As mentioned in the introduction, their high electron affinities make them interesting oxidation agents and as, depending on the metal atom, their electron affinities vary over a wide range, they can be used in many reactions for which different oxidiser strengths are needed. Three more hexafluorides are discussed in the literature: CrF6, PdF6 and AuF6. The case of chromium hexafluoride was subject of a vivid dispute. Its synthesis was first claimed by Roesky and Glemser[42] and the molecule was then further characterised by Odgen et al. using the matrix isolation technique.[43–45] The same infrared spectra were also obtained by Willner et al. but curiously, they came to the different interpretation [16, 46] that the observed bands were due to CrF5 instead of CrF6, and the question of the existence of CrF6 still remains unsolved. A report of the synthesis of PdF6 was published but its formation under the indicated conditions seems unlikely.[47] The molecule was also studied theoretically and DFT calculations suggest it to be stable,[48] but unfortunately no higher-level calculations are available to corroborate these results. By contrast, gold hexafluoride was never observed in any experiment and its instability was confirmed in a computational investigation at the CCSD(T) level of theory.[49] The intriguing properties of these molecules were also investigated in further quantum- chemical studies concerning both their structures and properties. The latter, especially their electron affinities, are subject of two recently published detailed theoretical investigations and the predicted values are in fairly good agreement with the experimental data.[50, 51] The question of the structures of these molecules was also raised. For chromium hexafluoride a trigonal prismatic form was suggested[52] but is now commonly accepted that it would have an octahedral structure.[53–56] The nine experimentally known transition metal hexafluorides all crystallise isostructurally and no significant deviations from ideal octahedral symmetry could be detected in the crystal structures[57] although a Jahn-Teller distortion is to be expected for the molecules having a d1, d2 or d4 electron configuration. In the gas phase 8 CHAPTER 2: BACKGROUND too, perfectly octahedral structures were detected by GED for the third row transition metal hexafluorides.[58] There are two possible reasons for this behaviour: On the one hand, the barriers of interconversion between two distorted minimum structures were found to be very small. The molecules might therefore fluctuate between different equivalent minima [58, 59] and this would lead to an averaged experimental structure of O h symmetry. On the other hand, the minimum structures of the third row transition metal hexafluorides are also influenced by SO coupling. Whereas at the scalar relativistic limit D 4h-symmetrical minimum structures were calculated for the triplet ground states of the hexafluorides of all group 10 metals,[60, 61] the consideration of SO coupling led to a perfectly octahedral structure for PtF6 and also confirmed the experimentally established closed shell electron configuration of this molecule.[61] 3 Methods

3.1 The matrix isolation technique

The matrix isolation technique was pionieered in the 1950s[62, 63] and is an experimental method which allows to study isolated molecules at low temperatures. It consists in trapping these molecules in a solid matrix of an unreactive material where they can be characterised by different analytical techniques. By this method, highly reactive species can be studied which are only stable at the conditions of the matrix isolation experiments. Spectacular new molecules were prepared using the matrix isolation technique, as for example the HArF molecule, the first and so far only compound containing an argon atom forming covalent bonds,[64] or the long-elusive mercury tetrafluoride.[8] One advantage of the isolation of the molecules is that bimolecular decomposition reactions resulting from the collision of two molecules are inhibited by the solid environment. Also, the low temperatures prevent activation barriers to be overcome and lead to a further kinetic stabilisation of the isolated species.

3.1.1 Generation of the matrix sample

Matrix samples are generated by condensing a diluted mixture of the active material in a large excess of the matrix host onto a cold surface, usually a KBr or CsI window if the matrix is to be analysed by IR spectroscopy. The large excess of the matrix host is necessary in order to ensure the isolation of the molecules and the formation of a homogenous matrix sample whose solid state order is not too perturbed by the guest molecules. If the active material is a gas at room temperature or if its vapour pressure is high enough, diluted gas mixtures can be prepared and directly deposited onto the cold surface. Another possibility is to condense the vapour pressure above a heated solid or liquid sample together with a flow of the matrix host where care must be taken that the sample evaporates without decomposition. It is also possible to generate the active material by a gas phase reaction directly before deposition, and metal-bearing molecules can for example be synthesised by co-depositing metal atoms with a reactive gas. Two complementary methods for the generation of gaseous metal atoms are possible: First, the metal can be evaporated thermally whereby the metal atoms are obtained in their ground states. Second, the evaporation of the metal can be done by a pulsed laser beam focussed onto a solid metal target. This is the laser ablation technique which usually yields atoms in more reactive excited states and which can also be applied to refractory metals for

9 10 CHAPTER 3: METHODS which thermal evaporation is not possible. However, as hard UV light is emitted when the laser beam hits the metal target additional photolysis reactions may occur.

3.1.2 Matrix IR spectroscopy

The matrix samples can be analysed by a great variety of spectroscopical techniques, ranging from vibrational and UV/Vis over EPR to M¨ossbauer spectroscopy. The most commonly used analytical method is however infrared spectroscopy which was also applied during the present thesis. As the molecules are isolated from each other, the conditions of the matrix isolation experiments resemble those of the gas phase with the difference that the molecules cannot rotate freely because they are trapped in cavities of the solid structure of the matrix host. Together with the low temperatures of the experiments, this usually leads to well resolved IR spectra with small bands widths. However, although the matrix material is usually chosen to be as inert as possible it still exerts an influence on the molecules which becomes manifest in a bathochromic shift of the IR absorptions; the stronger the interactions of the molecule with the matrix material are, the larger this so-called matrix shift is. Another characteristic of matrix IR spectra is that a splitting of bands can sometimes be observed and this is due to two main reasons: First, the molecules can be trapped in slightly different cavities and this results in a site-splitting of the bands which often disappears when annealing the matrix samples to slightly higher temperatures. Second, the shape of the cavity may distort the molecule from a regular high-symmetrical structure which can also lead to the appearance of new bands.

3.1.3 Matrix host materials

In order to characterise an unperturbed molecule it is desirable to reduce the interactions between the guest molecules and the matrix material to a minimum, and therefore the most commonly employed matrix hosts are the rare gases argon and neon. The use of argon as the matrix material has the advantage that the matrix samples can be made at higher temperatures and are thereby easier to be generated. The higher sublimation temperature of argon further allows the matrix samples to be annealed to higher temperatures whereas neon matrix samples are not stable above 10 K. One advantage of neon as the matrix host is that it is even less reactive than argon and several examples are known where the argon surrounding significantly influences the isolated compounds: For example, the argon matrix environment induces a change of the ground state of the uranium dioxide molecule when compared to neon matrices or the gas phase.[65] Argon atoms were furthermore shown to coordinate to the highly reactive coinage metal monohalides.[66, 67] In both cases, these interactions lead to significant and unusual changes in the observed infrared spectra when changing the matrix host from neon to argon. Besides these two standard matrix materials other alternatives exit as well, including for example N2, CO and CO2. These 3.2. Quantum-chemical description of molecules containing high-valent transition metals 11 host molecules usually interact more strongly with the guest compounds and larger matrix shifts result. In extreme cases, the matrix host can even take part in reactions, and dinitrogen can for example react with chromium atoms under formation the [Cr(N2)6] complex.[68] This concept of reactive matrices might also be applied to high oxidation state chemistry. For the stabilisation of extreme oxidation states the use of an extremely oxidising matrix material seems beneficial and first experiments were already done in which [69] F2 was used as a reactive matrix host.

3.1.4 Reactions in the matrix

Although the molecules are isolated in the solid matrix environment reactions can take place inside the matrix samples. This can be the case when two reaction partners are by chance trapped in the same cavity but in most cases the low temperatures effectively prevent reactions to proceed. There are two possible ways of providing the necessary activation energies, either by irradiating the sample with UV/Vis irradiation or by annealing it to slightly higher temperatures (up to 35 K for argon matrices). Annealing further allows the diffusion of atoms and small molecules through the matrix and reaction partners can meet this way. Besides bimolecular reactions, unimolecular reaction can occur too, as for example isomerisations or conformational changes which also need to be activated by some energy input. All bands corresponding to different vibrational modes of the same species show the same behaviour in each of these experiments which therefore provide valuable information for the identification of groups of absorptions belonging to the same molecule.

3.2 Quantum-chemical description of molecules con- taining high-valent transition metals

The content of this section is also part of a review on high oxidation state quantum chemistry by both the author and the supervisor of this thesis which is already accepted for publication (T. Schl¨oder, S. Riedel in Comprehensive Inorganic Chemistry II, J. Reedijk, K. Poeppelmeier (Eds.), Elsevier).

Quantum-chemical calculation have become an important complement to experimental studies and a plethora of different theoretical methods is available for calculating the structure and properties of a molecule.[70] The most widely used of these methods is density functional theory (DFT).[71] It often gives good approximations of structural parameters and reliably predicts the spectroscopic and thermochemical properties of molecules having a well-defined electronic structure. However, compounds containing transition metals are different and this is mainly due to low-lying excited electronic states. In order to accurately 12 CHAPTER 3: METHODS calculate their properties, an appropriate ab initio treatment of electron correlation is usually necessary.[72, 73] Furthermore, the consideration of relativistic effects is important for the third and to a lesser extent also for the second row transition metals elements.[74, 75]

3.2.1 Electron correlation

3.2.1.1 Fermi and Coulomb correlation

The correct description of the effects of electron correlation is one of the major challenges in non-relativistic electronic structure theory. The term ”electron correlation“ describes the fact that the motions of the electrons in a quantum-chemical system are not independent of each other.[76] Within Hartree-Fock (HF) theory,[77–80] the total wavefunction Φ of a N electron system is expressed as a Slater determinant of N one electron wavefunctions φi. This total wavefunction changes its sign with respect to the interchange of two electrons while the total electron density remains unchanged, thus meeting the requirements set by the indistinguishability of the electrons and their fermionic character. Beginning with an inital guess of orbitals (usually formed by a linear combination of atomic orbitals, each of which is described by a basis sets) the total energy of the system is optimised variationally with the energy of each electron being obtained from its interaction with the molecular field created by the nuclei and the other electrons. This leads to a new set of improved orbitals and the procedure is then repeated until the molecular field is self-consistent. The so-obtained energy is the Hartree-Fock energy EHF and accounts for Fermi correlation. The second source of electron correlation is the Coulomb repulsion between the electrons. Although the Hartree-Fock model already includes their electrostatic repulsion, is does so only in a mean field fashion which does not account for the dynamics of the interaction. As the Coulomb interaction between two electrons is always repulsive it leads to a reduced probability of finding two electrons close to each other, and hence the total energy of the system is always lowered (in the ground state). This behaviour is different from Fermi correlation which can lead to either an increase or a decrease of the total energy.[76] The difference between the energy obtained within the Hartee-Fock approximation and the exact energy of the system is defined[81] as the (Coloumb) correlation energy:

Ecorr = Eexact − EHF (3.1) where Eexact is the exact energy at the non-relativistic limit obtained for example by a FCI calculation with an infinite basis set. Due to its importance, Ecorr is generally referred to as simply ”the correlation energy“ although Fermi correlation is already included in EHF . One way to systematically improve the wavefunction and converge to the exact solution (within the Born-Oppenheimer approximation and the non-relativistic limit) is to allow virtual excitations of electrons to empty orbitals and thus to include other electronic configurations than the ground state. The wavefunction can then be written as a linear 3.2. Quantum-chemical description of molecules containing high-valent transition metals 13

combination of Slater determinants ΦK belonging to different orbital occupations: X ΨCI = cK Φk (3.2) K where the coefficients cK can be determined variationally. This approach is called configu- ration interaction[82, 83] (CI) and the inclusion of all possible configurations leads to the Full CI (FCI) method. Because of the huge number of possible configurations this method can however only be applied to very small systems; for ”real“ molecules it is thus necessary to use approximate methods like Møller-Plesset perturbation theory[84] (also: multi-body perturbation theory, MBPT), coupled cluster theory (CC)[85–87] or multi-configurational SCF (MCSCF)[88, 89] which will be presented below.

3.2.1.2 Static and Dynamical Correlation

It was suggested by Sinano˘glu[90] to split the description of Coulomb correlation into two different extremes, short-ranged dynamical and long-ranged static or non-dynamical correlation. Short-ranged correlation is associated with a tight pair of electrons and can be explained by a modified probability of finding two electron electrons in the same region of space. This results in a cusp in the wavefunction if the electron-electron interaction is repulsive, as depicted by the Coloumb hole. The cusp is moving with the electron and the correlation can therefore be called dynamical. The second extreme, namely the non-dynamical correlation arises for example when two electrons occupy two degenerate or near-degenerate electronic states. It hence describes the influence of one ore more low-lying electron configurations which cannot be neglected.[91] Here, the movement of the electrons is such that both electrons are never to be found in the same orbital. This effect is often small in closed shell systems near their equilibrium structures, but becomes important when molecules are distorted, as for example in the case of the homolytical breaking of a bond. Upon lengthening of the bond, the energy gap between the corresponding bonding and antibonding orbitals decreases and thus the weight of the formerly unoccupied antibonding orbital in the CI description becomes more important; at a certain point it can no longer be neglected and the antibonding orbital must be explicitly included in the description of the electronic structure. As this part of the electron correlation is also based on the dynamics of the electrons (in order to avoid a ionic situation the electrons which are delocalised over both nuclei have to move simultaneously) the term ”static“ correlation is somewhat arbitrary and mostly used for historical reasons. The two parts of the correlation energy can however not be strictly separated and are both described within the FCI treatment. The difference between static and dynamical correlation can also be expressed in terms of the CI coefficients cK : In the extreme case of all coefficients being either one or zero, one speaks of static correlation whereas the correlation is said to be dynamical if one large coefficient is accompanied by many small coefficients. For an accurate quantum-chemical description dynamical correlation is always necessary whereas static correlation is only required in the case of low-lying excited states. 14 CHAPTER 3: METHODS

3.2.1.3 Available methods

Many different theoretical approaches, ranging from DFT with various different functionals available to wavefunction-based ab initio methods, can be chosen for the calculations of the structures and properties of a molecule. Not all methods perform well when non-dynamical correlation is important, and it is therefore indispensable to assess the multi-reference character of the wavefunction in order to choose a suitable level of theory (the term ”multi-reference character“ refers to the fact that more than one orbital configuration, expressed as a Slater determinant and also called reference function, is necessary to correctly describe the electronic structure of the molecule). If static correlation is important the quantum-chemical description is usually done by multi-configurational HF theory whereas in cases of only moderate non-dynamical correlation reliable results can be obtained at the CCSD(T) level of theory. Assessment of the multi-reference character of the wavefunction. Several diag- nostics were developed in order to evaluate the ”amount“ of multi-reference character in the wavefunction. They usually analyse the sufficiency of a given level to describe the electronic situation in a molecule based on the calculation itself. The values examined 2 are the squared CI coefficent of the leading configuration (c0) in MCSCF calculations, the T1 and T2 amplitudes in CC calculations or the norms of the wavefunction in MBPT calculations. Further, the percentage of triples contribution to the atomisation energies [92] (%TAE) can also be used as a diagnostics value. For coupled cluster theory the T1 diagnostic was developed[93–95] and the single-reference treatment was found to yield reliable [94, 96] results for values of T1 ≤ 0.02. As an alternative, the highest T2 amplitude can serve as a diagnostic, too.[97] In order to provide a more accurate estimation, the modified [98–101] D1 and D2 diagnostics were proposed for both MP2 and CCSD calculations and good performances of these methods were observed for values of D1 ≤ 0.015 (MP2) and D1 ≤ 0.020 (CCSD) as well as for D2 ≤ 0.15 in both cases. Recently, a detailed analysis 2 of the c0 values, the T1 and D1 diagnostics, as well as the %TAE was published for 124 transition metal-bearing molecules.[102] Ab initio calculations. If the size of the molecules (for electronic structures the number of orbitals N is the important factor) allows the use of ab initio methods, these are usually to be preferred to semi-empirical methods like DFT (vide infra) because of the systematic treatment of electron correlation. Very reliable results can be obtained by the use of coupled cluster theory including single, double and perturbative triple excitations, CCSD(T). This method has been called ”the gold standard of quantum-chemistry“ because it often yields very good results and is a good compromise between computational cost (the perturbational 3 4 [103–105] calculation of the triple excitations scales as NoccNvirt) and achieved accuracy. Although it is a single-reference method, it can compensate for moderate multi-reference character and also in cases where other methods like MP2 or CCSD cannot be relied on. When low-lying excited states must explicitly be considered, multi-configurational SCF calculations are mandatory; the most common method here is CAS(n,m)SCF (complete 3.2. Quantum-chemical description of molecules containing high-valent transition metals 15 active space SCF)[106–108] where the number of included orbital configurations is reduced by restricting the occupancy variations to an active space of n electrons in m orbitals. CASSCF a priori only accounts for static correlation and further dynamical correlation can be included by an additional perturbation theory calculation as in the CASPT2 (complete active space 2nd order perturbation theory) method.[109, 110] Density functional theory calculations. If large molecules are investigated, the use of computational expensive ab initio methods is however no longer feasible. In these cases, DFT calculations can be used, at least for structure optimisations because they often provide good approximation of structural parameters at low computational cost. For closed shell species with a well-defined electronic structure in which the atoms are linked by covalent bonds, the predictions of thermochemical and spectroscopic data are often good and reliable. In many studies, DFT was also applied to compute structures and properties of a great variety of transition metal complexes.[111] However, the major drawback of DFT is that it accounts for electron correlation in an empirical manner and thus only poorly describes cases in which non-dynamical correlation or dispersive interactions are important;[112, 113] empirical corrected DFT methods are currently being developed in order to amend these deficiencies, see for example ref. [113–115]. Furthermore, the large number of different functionals can not be systematically ordered in terms of the treatment of correlation effects, and a validation of these functionals against experimental results or higher level quantum-chemical methods like CCSD(T) is therefore important.[116, 117]

3.2.1.4 Electron correlation in high-valent transition metal compounds

In contrast to the major part of the molecules containing only main group elements, the electronic structure of transition metal compounds is inherently more complicated due to the low-lying excited states. In the case of the high-valent fluorides, the crowded ligand spheres furthermore lead to stretched bonds because the ligands are pushed away from the metal atoms by inter-ligand repulsion. Another reason exists which prevents an ideal relaxation of the metal-ligand bonds (described by an ideal overlap of the (n – 1) d orbitals of the metal and the 2p orbitals of the ligand) as exemplarily shown by an analysis of − [118] the bonding situation in the [MnO4] anion: The outermost 3s and 3p core orbitals have a similar radial probability density as the 3d valence orbitals. This leads to stretched bond as an ideal relaxation is prohibited by Pauli repulsion between the core orbitals of the metal and the 2p orbitals of the ligands (see figure 3.1). This effect exists for all transition metals but is less pronounced for the 4d and 5d elements.[118, 119] As mentioned in section 3.2.1, non-dynamical correlation effects become more important in the case of strechted bonds. Together with the low-lying excited states due to the d orbitals this leads to fact that reliable results for high-valent transition metals can only be obtained when accounting for non-dynamical correlation, which for example plays an important role in [11, 12, 36] the stabilisation of HgF4. The reaction energy for its decomposition to HgF2 and F2 was calculated at different ab initio levels of theory and it was shown that, compared 16 CHAPTER 3: METHODS

Figure 3.1 : Schematic illustration of the Pauli repulsion between ligand orbitals and the outermost core shell in a transition metal complex. with benchmark CCSD(T) calculations, its stability is overestimated by both MP2 and MP4 whereas it is underestimated by MP3 and CCSD.[36] Several density functionals were validated againt these results, too, and it was found that the hybrid functionals B1LYP, B3LYP and MPW1PW91 performed best for this model reaction.[117]

3.2.2 Relativistic effects

3.2.2.1 Special relativity theory and quantum mechanics

For the 5d transition metals and also, albeit to a smaller extent, for the 4d transition metals, a correct quantum-chemical description needs to account for special relativity theory. This leads to the appearance of so-called relativistic effects for different properties which are defined as the deviations of the results obtained using the full relativistic Dirac equation from those of the non-relativistic Schr¨odinger equation.[74, 75, 120–123] The differences arising form the use of the relativistic Hamiltonian are the following: First, the kinematic effects, also called scalar relativistic effects, result from the proximity of the core electrons and the nucleus. The high velocity of the electrons leads to a relativistic increase of their kinematic mass and thus to a contraction of the corresponding orbitals. The associated energetic lowering of the orbitals is called the direct (kinematic) relativistic effect and affects all electrons. This direct relativistic stabilisation of an electron depends on its quantum numbers n and j and is larger if these are small.[121] Hence, it is most pronounced for the s and p1/2 orbitals but much smaller for the orbitals with higher angular momenta like the p3/2, d and f orbitals which have almost no density at the core and are therefore only marginally stabilised by the kinematic effect (the quantum number j and the splitting of the p orbitals into two subsets will be explained below). As the direct relativistic effect depends on the velocity of the electrons, it is more pronounced for the core-like orbitals and for the heavier elements. The contraction due to the direct 3.2. Quantum-chemical description of molecules containing high-valent transition metals 17 relativistic effect also leads to a more effective shielding of the Coulomb potential of the nuclei and hence to an expansion and destabilisation of the outer orbitals. This indirect (kinematic) relativistic effect outweighs the direct effect for the d and f as well as for the outer p3/2 orbitals which are thus destabilised. In turn, this destabilisation of the valence d orbitals leads to a small additional indirect stabilisation of the s and p orbitals of the next shell. This effect increases with the number of d electrons and partially explains why the relativistic stabilisation of the 6s and 7s orbitals reaches maximal values for Au [124, 125] th and 112Cn. The shift of this so-called ”gold-maximum“ in the 6 period (17.3% contraction) to the ”group-12 maximum“ in the 7th period (31% contraction) is related to the ground state electron configurations of these elements. Both 111Rg and Cn have the a dqs2 electron configuration whereas in the 6th row the ground states of Au and Hg are d10s and d10s2, respectively.[75, 126] The overall scalar relativistic effect consists thus in a contraction and stabilisation of the s, p1/2 and inner p3/2 orbitals whereas all other orbitals are destabilised. The second relativistic effect is the increased interaction between the electron spin and the orbital magnetic momentum. In contrast to the non-relativistic Schr¨odinger equation the use of the full relativistic Dirac Hamiltonian leads to a four-component wavefunction (spinor) including positronic states (charge degrees of freedom). Further, the odd and the even components of the four-spinor represent the spin-up (α) and spin-down (β) particles (spin degrees of freedom).[121] The Dirac equation thus explicitly introduces the spin quantum number of the electron and can describe its interaction with the orbital angular momenta, called spin-orbit (SO) coupling. This effect is important for higher angular momentum orbitals of p, d and f type because the corresponding magnetic moments can interact with the magnetic moment generated by the spin and this causes a splitting of the energy levels. The total angular momentum of an electron is described by the quantum number j, the vector sum of its spin and orbit angular momenta; the two possible values 1 j = l ± 2 correspond to the parallel and antiparallel orientation of the two momenta. In the cases of the p orbitals with l = 1, this leads to the aforementioned splitting into the p1/2 and p3/2 subsets. As for the scalar relativistic effect, the magnitude of the SO coupling also increases with increasing Z. For the superheavy elements the splitting of the levels reaches the values of typical bonding energies, making its consideration mandatory for the description of the 6d elements. By contrast, it is of minor importance for the 5d elements but can however not always be neglected for their correct description, as shown by the case of PtF6 for which the wrong ground state multiplicity is predicted at the scalar-relativistic limit[61] (compare chapter 2). It does also play an important role for the 6p and 7p orbitals which are however either fully occupied or empty in the case transition metal compounds.[127] 18 CHAPTER 3: METHODS

3.2.2.2 Available Methods

As mentioned above, the use of the full Dirac Hamiltonian leads to a four-component wavefunction. For electronic solutions the two upper components of the four-spinor are large whereas the two lower components are small. As chemists are usually interested in the electronic solution, the use of quasirelativistic methods in which the small component is neglected or approximated is common practice. Besides the electron-core interaction which is commonly treated in a relativistic manner, the electron-electron interaction too can be described using relativistic Hamiltonians. These effects are small in comparison with those in the one electron Dirac operator and the use of simple electrostatic Coulomb Hamiltonian is often enough (Dirac-Coulomb Hamiltonian). The use of a relativistic Hamiltonian leads to very cost-intensive calculations which are not always feasible, especially for large molecules. It is thus important to apply methods which allow the treatment of these effects in a more cost efficient way. A very common way to do so is to use effective core potentials (ECPs) which account for relativistic effects in the core region of the nuclei.[128–133] The advantage of the use of ECPs is not only the implicit accounting for the different relativistic effects but also the reduced number of explicitly treated electrons as the core region of the wavefunction is replaced by a pseudo potential. This method allows the consideration of scalar (spin-free) relativistic effects in the calculations of structures, vibrational frequencies, thermochemical stabilities and activation barriers in a straightforward manner and with an excellent accuracy. Furthermore, the comparison between non-relativistic and relativistic effective core potentials directly yields the relativistic effects on a given property. Another possibility to account for relativistic effects is the use of quasirelativistic methods in which the lower spinors are eliminated from the calculations, like in the Douglas-Kroll-Hess (DKH) approach[134–138] or the Zeroth-Order Regular Approximation (ZORA) to the Dirac- Coulomb(-Breit) Hamiltonian (also called Chang-P´elissier-Durand Hamiltonian).[139–142]

3.2.2.3 Relativistic Effects in high oxidation state chemistry

Relativistic effects directly influence chemical properties, such as ionisation potentials, electron affinities, excitation energies and thermochemical stabilities. The increased stabilities of the high oxidation states of the late third row transition metals is caused by the indirect relativistic destabilisation of the 5d electrons which makes them available for oxidation. For example, gold is the only group 11 element for which the V oxidation state can be reached (in the AuF5 molecule) whereas the highest neutral fluorides of silver [143] and copper are AgF3 and CuF3. This effect also contributes to the stability of the IV oxidation state of mercury in HgF4 which cannot be reached for its lighter homologues Cd and Zn. However, in this case the more important factor is the relativistic stabilisation of the 6s orbitals which destabilises the II oxidation state of mercury, and the crucial point for the endothermic elimination reaction of F2 from HgF4 is the relativistic instability [11, 12] of HgF2. For mercury’s heavier homologue copernicium an even larger relativistic 3.2. Quantum-chemical description of molecules containing high-valent transition metals 19

[144] stabilisation of the IV oxidation state was calculated in CnF4. An example for the importance of SO coupling effects on thermochemical values are the relative stabilities of + + the [IrO4] and [MtO4] cations. It was shown by quantum-chemical calculations that the + hypothetical [IrO4] cation which would be the first example of a metal oxidised to the IX [14] oxidation state is indeed thermochemically stable with respect to O2 elimination. This stability is mainly due to the aforementioned destabilisation of the 5d orbitals: Whereas –1 the use of a relativistic ECP yielded an endothermic O2 elimination (+176.4 kJ mol ) the same reaction was calculated to be exothermic by 165.3 kJ mol–1 when a non-relativistic ECP was used. The effects of SO coupling on the thermochemistry were shown to be + + negligible for [IrO4] but led to a significantly decreased stability of the [MtO4] cation which can be explained by a more difficult oxidation of Mt due to the energetic lowering [14] of the d3/2 orbitals. As a conclusion, the consideration of scalar relativistic effects is very important in order to obtain a correct quantum-chemical description of transition metal compounds, especially for the heavier elements of the third and forth row transition metals. By contrast the effects of SO coupling must always be taken into account for the superheavy 6d elements only, whereas they can in many cases be neglected for the lighter transition metal elements.

3.2.3 Assignment of oxidation states

As mentioned in chapter 2, the assignment of oxidation states is not always a straight- forward task; while for some compounds structural parameters allow the determination of oxidation numbers, the correct assignment of the oxidation state is not that easy in many other cases because of, e.g. small electronegativity differences, delocalised bonds or non-innocent ligands. In cases where the oxidation states cannot be determined otherwise and also when there are doubts about the correct oxidation numbers, quantum-chemical calculations may provide the necessary information.

3.2.3.1 Spin density analysis

If unpaired electrons are involved, the assignment of oxidation states can be done by an analysis of the spin density which gives information about where the unpaired electrons are localised. It thus allows to assess whether the central atom or the ligands of a complex are oxidised, i. e. to distinguish between innocent and non-innocent ligands. A prominent example for the wrong assignment of an oxidation state was the believed stabilisation of HgIII in the [Hg(cyclam)]3+ complex (cyclam = 1,4,8,11-tetraazacyclotetradecane) prepared by electrochemical oxidation of the corresponding dication by Allred et al. in 1976.[145] More than 30 years later an electronic structure analysis done by quantum-chemical calculations proved that the presumption of a redox-inactive[13, 145] cyclam ring was wrong 20 CHAPTER 3: METHODS and that the ligand system was oxidised instead of the Hg2+ cation. An analysis of the spin density distribution in the trication showed the unpaired electron density to be mainly located at the cyclam ring; based on these findings, the [Hg(cyclam)]3+ complex must be described as a HgII cation coordinated by a radical cationic ligand.[146]

3.2.3.2 Atomic charges

By contrast, it was demonstrated that atomic charges obtained by e.g. natural population − analysis (NPA) are not adequate to assign oxidation states. In the [Cu(CF3)4] complex for example, the copper atom should be assigned the oxidation state III, as it was inferred from the square-planar structure typical for d8 complexes.[147] Later, the calculation of an NPA charge of q ≈ 1 for copper in this complex led to the conclusion that the compound was better described as an CuI complex.[148, 149] Kaupp and von Schnering however reaffirmed the assignment of the III oxidation state of copper, arguing that the formal concept of the oxidation states is fundamentally different from partial charges obtained from quantum- chemical calculations.[150] A more detailed investigation of these two concepts and their interdependencies was published recently by Aull´onand Alvarez:´ The atomic charges on 0 3− 2− − the metal atom in the isoelectronic d complexes [VO4] , [CrO4] and [MnO4] , which all represent the metals in their highest possible formal oxidation states, namely +V, +VI and +VII, were calculated to be 1.05, 1.09 and 0.96, respectively. In the case of the fluoro III − IV complexes [Sc F4] and Ti F4, the calculated charge of 1.97 is however significantly higher for both molecules although they formally represent lower oxidation states of the metals.[151] The atomic charge of the metal atoms does thus not reflect its formal oxidation state but depends mainly on the electronegativity of the ligand. 4 Setup of the new matrix isolation apparatus

In order to perform the experiments described in chapter 5, a new matrix isolation apparatus was installed. It consists of a cold head with a connected matrix window and placed inside a matrix chamber. The experimental work done in the present thesis was based on the reactions of laser-evaporated metal atoms with reactive gases, and therefore the new matrix chamber was equipped with a corresponding metal target onto which a laser beam can be focussed.

The custom-made pieces used for the construction of the new experimental setup were made in the technical workshop of the chemistry department of the university of Freiburg by Reinhard Tomm, Christian Roll and Markus Melder who also provided technical drawings for these pieces.

4.1 Cold head and cold window

The low temperatures necessary for freezing out the inert gases are generated by a RDK- 205D cold head purchased from Sumitomo Heavy Industries, Ltd. This cold head is based on a Gifford McMahon (GM) closed-cycle helium cryocooler with which temperatures as low as 4 K can be reached. A holder for the matrix window was fixed to the cold head via a copper block with an integrated 50 Ω heater and a DT-670A1-CU silicon diode (Janis Research, LLC). The IR transparent cold window (CsI or KBr, 30 mm × 3 mm, Korth Kristalle GmbH) was then mounted to the window holder by a holding plate. In order to ensure good heat transfer all components were made from oxygen/halogen-free copper (OHFC) and indium foil was used for the connections between the different pieces. Pictures of the copper parts for mounting the matrix window to the cold head are shown in figure 4.1 whereas the technical drawings of the window holder as well as the holding plate are given in appendix A. The temperature of the cold head is measured with a DT-670C-BO silicon diode sensor (Lake Shore Cryotronics, Inc.) and monitored by a temperature control unit (model 331, Lake Shore Cryotronics, Inc.). The temperature controller can furthermore be connected to the heater which allows to adjust the temperature of the cold window during the experiments.

21 22 CHAPTER 4: SETUP OF THE NEW MATRIX ISOLATION APPARATUS

Heater block

Cold window

Window holder Holding plate

Figure 4.1 : Picture of the copper pieces for mounting the cold window to the cold head.

4.2 The matrix chamber

The cold head was fitted into a two-piece custom-made aluminium matrix chamber (figure 4.2, see appendix A for technical drawings). Its upper part has two ports to which a flange for the connection to the high vacuum system as well as an inlet for the cables of the temperature sensor and heater can be attached. At the level of the cold window, the lower part of the matrix chamber has eight openings to which different windows or sample inlet system can be mounted. The top part of the chamber was fixed to the cold head whereas

Figure 4.2 : Pictures of the upper (left) and lower (right) part of the new matrix chamber. 4.3. The high vacuum system 23 the bottom part is moveable and the two pieces are held together by underpressure during the experiments. With this setup, the lower part of the matrix chamber can be freely rotated against the upper part, thereby allowing it to remain in the same position when the cold window is turned to one of the different orientations necessary for deposition, irradiation or recording of the spectra.

4.3 The high vacuum system

In order to perform the matrix isolation experiments, the matrix chamber needs to be evacuated so that only the matrix gases are frozen out on the cold window. Furthermore, the high vacuum also provides thermal isolation of the cold head against room temperature. The vacuum is generated by a water-cooled oil diffusion pump (DO-501, Leybold) with which pressures of < 5 · 10−7 torr can in principle be attained. A primary vacuum of ≈ 2 · 10−1 torr is necessary for the operation of the oil diffusion pump and is provided by rotary vane backing pump. The use of a turbomolecular secondary pump was also tested but the oil diffusion pump proved more robust against difluorine which was used as the reactive gas in the experiments. The pressure in the matrix chamber was measured by a TPG 261 SingleGauge (Pfeiffer Vacuum GmbH) connected to a PKR 251 double gauge head (consisting of a Pirani gauge head and a cold cathode, Pfeiffer Vacuum GmbH).

4.4 Laser ablation of metals

The setup for the laser ablation experiments consists of a rotating metal disc (the target) onto which the 1064 nm infrared fundamental of a pulsed neodymium doped yttrium aluminum garnet (Nd:YAG) laser (Minilite II, Continuum Inc.) is focussed using a plano convex lens (ThorLabs, N-BK7 C, focal length 125.0 mm). The repetition rate of the laser can be varied between 1 and 15 Hz at a constant pulse length of 5 ± 2 ns and the maximum pulse energy of 50 mJ and can be gradually attenuated. The rotary motion of the target which is glued onto an aluminum holder of adjustable length is driven by a DC-Micromotor (brush motor 1016M012G, gear box 101 256:1, Faulhaber GmbH & Co. KG) placed outside the matrix chamber. In order to assemble the different components of the experimental setup, a motor holder (figure 4.3, technical drawings in appendix A) was devised which can be mounted to one of the ports of the matrix chamber. It has two slots into which the motor as well as the target holder can be placed on the outer and inner side of the matrix chamber, respectively. Both motor and target holder are equipped with permanent magnets and the motion of the motor is thus transmitted to the target by magnetic coupling. The motor holder is furthermore equipped with a capillary through which the matrix gases can be admitted to the matrix chamber in proximity of the ablated metal atoms. In the setup chosen for the experiments, the metal target as well as the cold 24 CHAPTER 4: SETUP OF THE NEW MATRIX ISOLATION APPARATUS window are aligned perpendicularly to the laser beam laser which therefore has to enter the matrix chamber by the port opposite to that of the motor holder and pass through a hole in the cold window before hitting the target.

(a)

(b) (c) (e) (d) (d)

Figure 4.3 : Pictures of the motor holder with the magnetic coupling: (a) metal target, (b) target holder, (c) ball bearing, (d) magnets holders with magnets, (e) brush motor.

4.5 The spectrometer

The infrared spectra of the matrix samples were measured using a Vertex 70 FT-IR spectrometer (Bruker Optik GmbH) which was purged with dry air in order to reduce the concentrations of possibly interfering IR active atmospheric molecules (e.g. H2O). The cover of the sample compartment of the spectrometer was removed and the matrix chamber was fixed to the ceiling of the laboratory so that the cold window was positioned in the optical path of the spectrometer. Two opposite ports of the matrix chamber were equipped with CsI windows whereas the KBr windows between the sample compartment and the compartments of the interferometer and the detector were removed in order to improve the intensity of the IR beam and to widen the spectral range. The optical path between interferometer compartment, matrix chamber and detector compartment was protected by two custom made connecting flanges which are also purged with dry air. The source of the spectrometer is a tungsten halogen lamp which provides far infrared to ultraviolet irradiation. For the respective MIR and FIR spectral regions, a KBr (7500– 370 cm–1) and a multilayer beam splitter (680–30 cm–1) are available. Two types of detectors are possible for measuring the spectra: First, one of two room temperature deuterated l-alanine-doped triglycine sulfate (RT-DLaTGS) detectors can be used which are equipped with either a CsI (12000–160 cm–1) or a PE window (700–10 cm–1) for the recording of MIR and FIR spectra respectively. The second possibility is the use of a liquid-nitrogen-cooled mercury cadmium telluride (LN-MCT) detector (KRS-5 window, 12000–420 cm–1) which is more sensitive and thus allows a faster recording of the spectra. 5 Fluorides of the 3d transition metals

By contrast to their heavier homologues no hexafluoride was yet undoubtedly characterised for the first row transition metals. The highest possible oxidation state reached for any [36] 3d element in a neutral binary fluoride molecule is V as examplified by VF5 and CrF5. As mentioned in chapter 2, the case of chromium is however unclear and chromium hexafluoride might have actually been synthesised.[42–44] Its characterisation was based on the observation of one single IR absorption but the assignment of this band to the CrF6 [16, 46] molecule is doubtful as it might as well originate from CrF5. The thermochemical stability of the high-valent chromium fluorides as well as their vibrational spectra were therefore calculated quantum-chemically in order to solve this still open question, and the results are presented in section 5.1 together with new matrix isolation experiments. The highest-valent neutral fluorides of manganese, iron and cobalt characterised so far [36] are MnF4, FeF3 and CoF4 (figure 5.1). Crystalline manganese tetrafluoride is a well- [152] characterised compound whereas the only convincing evidence for molecular MnF4 comes from mass spectrometric experiments where the corresponding molecular ion could be detected. The next higher fluoride, MnF5, was never observed in any experiment but speculations about its stability were made.[19] The question whether or not manganese pentafluoride is a stable molecule which might be experimentally observed is discussed based

Figure 5.1 : The highest known neutral binary fluorides of the 3d elements.

25 26 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS on quantum-chemical results in section 5.2. In contrast to the case of manganese, solid iron tetrafluoride is unknown and no compelling evidence was yet published for molecular FeF4, making iron the only 3d midtransition metal whose tetrafluoride is not known. High-valent iron fluorides were therefore studied both theoretically and experimentally, and the results of this investigation are subject of section 5.3. Finally, solid cobalt tetrafluoride is unknown, too, whereas however molecular CoF4 could also be characterised by the mass spectrometric observation of its molecular ion.[20, 21] Section 5.4 deals with the quantum-chemical calculations of the structures and stabilities of high-valent cobalt fluorides.

5.1 Chromium fluorides

Is chromium hexafluoride a stable compound? Almost 50 years after the first report of its synthesis by Roesky and Glemser in 1963[42] this question could not yet been umambiguously answered. The VI oxidation state of chromium is however well known,[36] 2− 2− as for example in the anionic oxochromates(VI) [CrO4] and [Cr2O7] as well as in their anhydrous form, chromium trioxide. Further, the oxide fluorides CrO2F2 and CrOF4 are stable molecules, too and were studied in the solid state[153–155] as well as in the gas phase.[156, 157] The highest neutral binary fluoride of chromium characterised beyond doubt ◦ [153] + is CrF5, a deep red solid which melts at 30 C. Furthermore, the [CrF5] cation with chromium in the VI oxidation state was observed in mass spectrometric experiments,[15] but it is unclear whether it was formed by ionisation of neutral chromium pentafluoride or by fluoride abstraction from chromium hexafluoride.

The parent compound, CrF6, was claimed to be synthesised after high-pressure fluorination of either elemental chromium or CrO3. The first synthesis starting from chromium powder was conducted at 400 ◦C and a fluorine pressure of 350 atm and yielded a lemon yellow product together with deep red chromium pentafluoride; based on elemental analysis the [42] yellow product was identified as CrF6. In the second synthesis, the use of chromium trioxide as the starting material allowed somewhat milder reaction conditions (170 ◦C, [43, 44] 25 atm) and in this also led to the formation of both yellow CrF6 and red CrF5. Both chromium hexafluoride and chromium pentafluoride were subsequently evaporated and isolated in inert gas matrices. The recorded infrared spectra showed strong absorptions –1 –1 –1 at 767.7 cm (neon matrices), 763.2 cm (argon matrices) and 758.9 cm (N2 matrices) which all exhibited the typical isotopic pattern due to the different chromium isotopes in their natural abundances. In addition, the absorptions due to CrF4 were also observed in the spectra and were much stronger when CrF5 was evaporated. Odgen et al. concluded from these observations that the new band is caused by the octahedral CrF6 molecule, whose expectation IR spectrum contains one single IR active mode in the region of the [43–45] Cr–F stretching vibrations, and that CrF5 disproportionates upon evaporation. 5.1. Chromium fluorides 27

Later, the experiments were repeated by Willner et al. who offered a different explanation for the observed spectra which were very similar to those obtained before:[16, 46] They concluded that the new band should be assigned to CrF5 instead of CrF6 and based their argumentation upon the observation that gaseous CrF5 showed no tendency to disproportionate and did not decompose when being expanded into vacuum. However, the expectation spectrum of the C 2v-symmetrical chromium pentafluoride molecule consists of several IR active Cr–F stretching fundamentals. Willner et al. therefore suggested that the intensities of the three A1 modes would be too low to be observed and that the two remaining bands (corresponding to the B1 and B2 modes) would have almost the same wavenumber.[16] In their more detailed second investigation a second broad band could be assigned to CrF5 which overlaps with the bands of CrF4 and which might therefore have been overlooked previously.[46] Besides in these experimental studies, the chromium hexafluoride molecule was also investigated by quantum-chemical methods. However, the main question addressed in these studies was the structure of CrF6 and not its stability: Despite the initial proposition [52] of a trigonal prismatic structure (D 3h symmetry) there is no doubt today that, if it [53–56] existed, CrF6 would be octahedral. The thermochemistry of the higher chromium fluorides was investigated to a lesser extend and only DFT results are available which [56] indicate CrF6 to be stable against the loss of one fluorine atom. During the time of this thesis, a new DFT investigation of chromium fluoride molecules was published by Siddiqui in which the eliminations of a fluorine atom or molecule from CrF6 were predicted to be endothermic.[158] However, the findings presented in this study are unlikely to be reliable, as for example a triplet (!) ground state was predicted for chromium hexafluoride, which is impossible for the d0 electron configuration of CrVI and furthermore in disagreement with all previous investigations of this molecule.

5.1.1 Structures

The structures of the binary chromium fluoride molecules were optimised at CCSD(T) level and are shown in figure 5.2. All molecules were found to have high-spin ground states with a maximum number of unpaired electrons (table 5.1). For CrF2, a C 2v-symmetrical bent ◦ structure (d Cr–F = 179.8 pm, F-Cr-F = 143.8 ) was calculated and this result agrees well [163] with a previous computational investigation. Both chromium trifluoride and chromium tetrafluoride have highly symmetrical structures of D 3h and T d symmetry respectively, and the bond lengths were calculated to be 173.7 pm in CrF3 and 170.9 pm in CrF4. For CrF5, the single unpaired electron leads to a Jahn-Teller distorted trigonal bipyramidal structure 2 00 of C 2v symmetry. The vibronicly unstable E state of the regular D 3h-symmetrical 2 2 structure reduces to A2 + B1 in C 2v symmetry and both states were optimised at 2 CCSD(T) level. The A2 state was calculated to be a minimum lying marginally lower in –1 2 energy (0.6 kJ mol at CCSD(T)/CBS level) than the B1 state which was found to be a transition state. Due to the threefold rotation axis in the D 3h point group a ”mexican hat“ 28 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS

F(ax)

F(eq1) F(eq2)

CrF6 CrF6 CrF5 1 1 ’ 2 (Oh, A1g) (D3h, A1 ) (C2v, A2)

CrF4 CrF3 CrF2 3 4 ’ 5 (Td, A2) (D3h, A1 ) (C2v, B2)

Figure 5.2 : Optimised structures of molecular chromium fluorides; see table 5.1 for bond lengths and angles.

Table 5.1 : Calculated and experimental structural parameters of molecular chromium fluorides[a]

Molecule Parameter[b] B3LYP[c] CCSD(T)[c] Exp.[d] 5 CrF2 (C 2v, B2) d Cr–F 178.0 179.8 179.2(5)0 [159] F-Cr-F 136.7 143.8 4 0 CrF3 (D3h, A2) d Cr–F 173.7 173.7 173.2(2)0 [160] 3 CrF4 (T d, A2) d Cr–F 171.4 170.9 170.6(2)0 [161] 2  CrF5 (C 2v, A2) d Cr–F(ax) 174.7 173.8 174.2(10)  d 168.2 167.3  [162] Cr–F(eq1) 169.5(6)0 d Cr–F(eq2) 169.8 170.1  F(eq1)-Cr-F(ax) 091.4094.2 F(eq1)-Cr-F(eq2) 121.8 119.7 1 CrF6 (Oh, A1g) d Cr–F 172.9 172.4 1 0 CrF6 (D3h, A1) d Cr–F 173.9 173.4 [e] F(eq1)-Cr-X 050.4050.4 [a] Bond lengths in pm, angles in ◦. [b] See figure 5.2 for atom labelling. [c] aVTZ basis sets. [d] r g values from GED measurements; nozzle temperatures: 1520 K for CrF2, 1220 K for CrF3, ◦ ◦ 195–220 C for CrF4 and 80 C for CrF5. [e] Angle between the Cr–F bond and the C 3 axis. 5.1. Chromium fluorides 29

Figure 5.3 : Scan of the ”mexican hat“ potential energy surface of CrF5 at B3LYP/aVTZ level. The energy is plotted as a function of the displacement of the chromium atom from the centre of a trigonal bipyramidal structure (see text); values in kJ mol–1.

2 like potential energy surface results on which three A2 minima are connected by three 2 B1 transition states, and due to the very low barrier for the interconversion between two of the minima the molecule is most probably fluxional at already moderate temperatures. Figure 5.3 shows a scan of this potential energy surface which was calculated by optimising the structure of CrF5 in D 3h symmetry and subsequently displacing the chromium atom in the equatorial plane of the trigonal bipyramidal transition state. As the positions of the fluorine atoms were held constant in the calculations, only an approximate potential energy surface was obtained but nevertheless, its trigonal symmetry as well as the three minima 2 and transition states can be recognised. The calculated bond lengths in the A2 minimum structure are d Fax = 173.8 pm, d Feq(1) = 167.3 pm and d Feq(2) = 170.1 pm. Finally, an octahedral structure (O h symmetry) with a chromium-fluorine distance of 172.4 pm was calculated for chromium hexafluoride. The Cr–F bond lengths in the lower chromium fluorides decrease with increasing oxidation state of the metal until reaching a minimum for CrF4 and CrF5 which can explained by the higher charge on the central atom which leads to more polarised, stronger bonds. By contrast, the bond length increases again for CrF6, thus hinting at the steric crowding in the ligand sphere. The trigonal prismatic conformer of this molecule was also calculated. It lies 45.6 kJ mol–1 higher in energy at the CCSD(T)/CBS level of theory and furthermore shows an imaginary frequency of 95.2 cm–1 at CCSD(T)/aVTZ level, thereby corroborating the above-mentioned previous studies in which the O h-symmetrical conformer was also obtained as the minimum structure. For all of these molecules except the elusive CrF6, experimental gas phase structures 30 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS

[159–162] were obtained by electron diffraction measurements. With the exception of CrF5, the calculated bond lengths were slightly smaller than the experimental ones (table 5.1). These differences between the experimental and the computed values would increase if the effects of the elevated temperatures of the experiments were considered, especially for the lower fluorides for the evaporation of which higher temperatures were necessary. The agreement between the calculated and measured bond lengths thus significantly increases with increasing oxidation state of the metal which can be explained by the reduced importance of core-valence correlation in the higher-valent fluorides (see section 5.3 for more details).

5.1.2 Thermochemistry

As to answer the question of the thermochemical stability of chromium hexafluoride, three different possible decomposition reactions were calculated (table 5.2). Whereas at DFT level only the bimolecular F2 elimination reaction is found to be endothermic for this molecule, the unimolecular elimination of difluorine (yielding CrF4) too is thermochemically favoured at CCSD(T) level. For comparison, the same decomposition reactions were all calculated to be exothermic for the experimentally verified CrF5 molecule. The claimed synthesis of chromium hexafluoride was conducted at high temperatures and high fluorine pressures, and according to the principle of Le Chˆatelier, these conditions favour the formation of CrF6. Furthermore, there are good arguments for a kinetic stability of this molecule: First, the unimolecular F2 elimination leading to CrF4 is spin-forbidden as this latter molecule has a triplet ground state. The lowest-lying singlet state of CrF4 was therefore also calculated and found to lie 156.1 kJ mol–1 above the triplet ground state (CCSD(T)/CBS level) so that the corresponding spin-allowed F2 elimination reaction would be endothermic. Second, the bimolecular elimination of F2 probably has a high activation barrier as for it to proceed, the negatively charged ligand spheres of two CrF6 molecules need to come in

[a] Table 5.2 : Calculated thermochemistry of CrF6 and CrF5

Reaction B3LYP[b] CCSD(T)[c]

CrF6 → CrF4 + F2 023.60(14.5) –22.0 (–32.3) CrF6 → CrF5 + F 055.30(47.7)030.30(23.2) 1 CrF6 → CrF5 + 2 F2 –22.4 (–26.8) –46.1 (–50.4) CrF5 → CrF3 + F2 339.7 (329.6) 303.1 (294.6) CrF5 → CrF4 + F 123.6 (115.8) 100.40(91.7) 1 CrF5 → CrF4 + 2 F2 046.00(41.3)024.10(18.1) 1 1 CrF5 → 2 CrF6 + 2 CrF4 034.20(34.0)052.20(55.5) [a] Energies in kJ mol–1; values in parentheses are ZPE corrected. [b] aVTZ basis sets. [c] Ex- trapolated to the CBS limit and corrected for CV effects, ZPE correction using the aVTZ basis sets. 5.1. Chromium fluorides 31 close contact. If matrix-isolated molecules are considered, this reaction is further inhibited by the solid inert environment. It can therefore not be excluded that CrF6 was actually synthesised and survived the transfer to the inert matrices as a metastable species. As the experimental matrix IR spectra of the vapour phases above samples of CrF5 and the supposed CrF6 were identical, it was suggested that CrF5 disproportionates to yield CrF4 and CrF6. This disproportionation was therefore also calculated and the energy of reaction of +56.4 kJ mol–1 indicates this reaction to be unlikely to proceed, thus confirming the experimental observation that gaseous CrF5 is stable against disproportionation. However, lattice effects were not considered in this calculation and the difference between the lattice enthalpies of CrF4 and CrF5 might actually lead to an overall exothermicity of the disproportionation reaction due the additional stabilisation of solid CrF4 as the product.

5.1.3 Vibrational frequencies

The vibrational spectra of the matrix-isolated lower chromium fluorides CrF2, CrF3 and [44, 46, 164, 165] CrF4 are known and there is no doubt about the assignment of their bands which all show the typical splitting due to the different isotopes of chromium in their natural abundances (50Cr: 4.3%, 52Cr: 83.8%, 53Cr: 9.5% and 54Cr: 2.4%). By contrast,

Table 5.3 : Calculated and observed wavenumbers of the Cr–F stretching modes of molecular chromium fluorides[a] Calc. Exp. (matrix) Molecule Mode B3LYP[b] CCSD(T)[b] Ne Ar Ref. 5 CrF2 (C 2v, B2) A1 617.80(56) B2 721.4 (235) 679.5 654.3 tw 4 0 0 CrF3 (D3h, A2) A1 669.5 00(–) 676.8 00(–) E0 759.0 (469) 761.9 (537) 762.8 749.2 tw 3 CrF4 (T d, A2) A1 713.9 00(–) 722.1 00(–) T2 783.5 (602) 799.7 (669) 790.3 784.4 tw 2 CrF5 (C 2v, A2) A1 608.9 00(0) 623.9 00(0) A1 720.6 00(0) 720.1 00(0)  B1 727.7 (310) 790.1 (337)  [c] B2 763.5 (324) 809.5 (222) 767.7 763.9 [16] A1 828.3 (209) 844.2 (211)  1 CrF6 (Oh, A1g) Eg 587.2 00(–) 589.9 00(–) A1g 708.2 00(–) 716.8 00(–) T1u 761.0 (939) 785.3 (1019) [a] Values in cm–1 for 52Cr isotopomers; numbers in parentheses are infrared intensities in –1 [43, 44] km mol . [b] aVTZ basis sets. [c] This band was alternatively assigned to CrF6. tw = this work. 32 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS the origin of the band observed at 763.0 cm–1 in argon and 767.7 cm–1 in neon matrices is as yet unclear.[16, 43–46] This band also showed the typical isotopic pattern of chromium and was observed in the matrix IR spectra obtained after the thermal evaporation of samples of either CrF5 or the putative CrF6. It was subsequently attributed to both of these molecules and the correct assignment is still under debate. In order to solve this long-standing problem, the harmonic frequencies of the chromium fluoride molecules were calculated at CCSD(T)/aVTZ level and the results are given in table 5.3. Unfortunately, the coupled cluster frequency calculations for CrF2 only yielded unreasonable results but the computed wavenumbers of CrF3 and CrF4 agree nevertheless well with the experimental values. The difference between the calculated and experimental wavenumber is only 0.9 cm–1 for chromium trifluoride whereas it amounts to 5.9 cm–1 for chromium tetrafluoride. The larger deviation in the case of CrF4 can be explained by the reduced importance of core-valence correlation for this molecule (see section 5.3 for details). Despite the small deviations, these frequency calculations are nevertheless very accurate and can

Table 5.4 : Calculated and experimental isotopic shifts of selected Cr–F stretching modes of [a] CrF4, CrF5 and CrF6

Calc. Exp. (matrix) [b] Molecule Mode B3LYP[c] CCSD(T)[c] Ne Ar Ref. 3 50  CrF4 (T d, A2) T2 ( Cr) 788.80(5.3) 805.105.1 795.40(5.2) 789.50(5.2) 52  T2 ( Cr) 783.50(0.0) 799.70(0.0) 790.20(0.0) 784.30(0.0)  53 [44, 46] T2 ( Cr) 781.0 (–2.5) 797.1 (–2.6) 787.7 (–2.5) 781.8 (–2.5)  54  T2 ( Cr) 778.5 (–5.0) 794.7 (–5.0) 785.3 (–4.9) 779.5 (–4.8)  2 50  CrF5 (C 2v, A2) B1 ( Cr) 732.00(4.3) 796.90(6.8) 774.20(6.5) 769.50(6.5) 52  B1 ( Cr) 727.70(0.0) 790.10(0.0) 767.70(0.0) 763.00(0.0)  53 [16] B1 ( Cr) 725.7 (–2.0) 786.1 (–4.0) 764.6 (–3.1) 760.2 (–2.8)  54  B1 ( Cr) 723.7 (–4.0) 783.7 (–6.4) 761.6 (–6.1) 757.0 (–6.0)  50 B2 ( Cr) 770.10(6.1) 815.60(6.1) 52 B2 ( Cr) 763.50(0.0) 809.50(0.0) 53 B2 ( Cr) 760.3 (–3.2) 806.6 (–2.9) 54 B2 ( Cr) 757.3 (–5.2) 803.8 (–5.7) 50 A1 ( Cr) 834.50(6.2) 850.40(6.2) 52 A1 ( Cr) 828.30(0.0) 844.20(0.0) 53 A1 ( Cr) 825.3 (–3.0) 841.2 (–3.0) 54 A1 ( Cr) 822.5 (–5.8) 738.3 (–5.9) 1 50 CrF6 (Oh, A1g) T1u ( Cr) 767.10(6.1) 791.70(6.4) 52 T1u ( Cr) 761.00(0.0) 785.30(0.0) 53 T1u ( Cr) 758.2 (–2.8) 782.2 (–3.1) 54 T1u ( Cr) 755.4 (–5.6) 779.3 (–6.0) [a] Values in cm–1; in parentheses: isotopic shifts compared to the 52Cr isotopomers. [b] See text for the assignment of the CrF5 band. [b] aVTZ basis sets. 5.1. Chromium fluorides 33

be expected to give reliable results for CrF5 and CrF6, too. One single IR active Cr–F stretching fundamental is expected for octahedral CrF6 and the calculated harmonic wavenumber of this mode is 785.3 cm–1 at CCSD(T)/aVTZ level. This value agrees well with the ambiguously assigned experimental band, especially as the consideration of anhamonic corrections would further decrease the difference between the experimental and the calculated wavenumber. Moreover, the computed isotopic shifts are also in good agreement with the experimental data (table 5.4). All Cr–F stretching vibrations of the C 2v-symmetrical CrF5 molecule are in principle IR active but not all of them are expected to have high IR intensities. Willner et al. assigned the band in question to chromium pentafluoride, arguing that only two of its Cr–F stretching modes would show a significant IR intensity and that furthermore these two absorptions would overlap as to form a single [16] band. Hence, CrF5 and CrF6 would be indistinguishable by IR spectroscopy. However, the CCSD(T) frequency calculations for CrF5 yielded three Cr–F stretching vibrations with considerable IR intensities and three significantly different wavenumbers, and thereby support the assignment of the doubtful band to chromium hexafluoride. Thus, on the one hand the spectroscopic arguments favouring the CrF6 hypothesis are supported by the CCSD(T) calculations. On the other hand, the general observations made by Willner et al. about the stability of CrF5, which did not show any tendency to disproportionate, are also confirmed by these calculations which predict the disproportionation of CrF5 to be endothermic, at least in the gas phase. The question of the correct assignment of the band thus remains open. In their second investigation Willner et al. were able to identify a second broad band which they could also attribute to CrF5. If the corresponding third band were even broader it might not be observed in the spectra at all, thus giving an alternative explanation for the concordance of the experimental spectrum of CrF5 with the expectation spectrum of CrF6. This interpretation is supported by the calculated wavenumber and isotopic shifts for the B1 Cr–F stretching mode of CrF5 which are very similar to those of the T1u mode of CrF6 and therefore also agree well with the experimental values. This B1 mode corresponds to the antisymmetric F(ax)-Cr-F(ax) stretching vibration of CrF5 whereas the remaining two high intensity A1 and B2 modes correspond to Cr–F stretching vibrations in the equatorial plane. One might thus speculate that the flat potential energy surface in the equatorial plane is responsable for the extreme broadness of the two latter bands and that the experimentally observed absorption corresponds to the B1 vibration of CrF5.

5.1.4 Matrix isolation experiments

Chromium fluoride molecules were already investigated in matrix isolation experiments and two different methods were used in these studies: First, pieces of chromium were reacted with elemental fluorine at elevated temperatures and the highest chromium fluoride obtained [166] this way was CrF4. The second approach was to thermally evaporate pre-synthesised [16, 43–46, 164, 165] chromium fluoride samples from CrF2 to CrF5 or even CrF6. In search 34 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS of conclusive experimental evidence for the assignment of the debatable IR absorption, new matrix isolation experiments were done using a third method for the generation of molecular chromium fluorides, namely the gas phase reaction of excited chromium atoms with elemental fluorine. Laser-ablated chromium atoms were thus condensed together with F2 diluted in neon or argon onto a CsI window cooled to 5.0 or 10.0 K respectively for deposition. Using a F2 concentration of 0.5% in neon, three major groups of absorptions were observed in the spectra measured directly after deposition (figure 5.4Ne) which could be assigned to the lower fluorides CrF2, CrF3 and CrF4 by comparison with published values[46, 165] (table 5.5). When the matrix samples were annealed to 9.0 K, the bands of CrF4 slightly increased at the expense of the CrF2 absorptions and further UV irradiation of the matrices using the full spectrum of a mercury arc lamp led to a growth of all observed chromium fluoride bands. The photochemistry of the matrix samples was investigated in more detail by irradiating the sample using different high-pass filter of successive lower wavenumber edges (figure 5.5UV). No changes were observed in the spectra when using λ > 515 nm irradiation whereas the use of lower wavelengths led to an increase of all chromium fluoride bands. In the corresponding experiments with argon as the matrix host, only [164] the bands of CrF2 could be observed directly after deposition whereas the bands of [165] [44] CrF3 and CrF4 only appeared in the spectra after annealing to 25 K and broadband irradiation (figure 5.4Ar). However, in contrast to the published spectra, a group of three absorptions was observed for CrF4 instead of one band in the region of the Cr–F stretching modes. Different fluorine concentrations ranging from 0.25% to 20% F2 in neon were tested in the experiments and, as expected, the intensities of the CrF4 band increased

Table 5.5 : Observed bands in the matrix IR spectra obtained after co-deposition of laser-ablated chromium atoms with elemental fluorine in different matrix hosts[a]

[b] Molecule Ne Ar F2 50 CrF2 685.0 52 CrF2 679.6 654.3 53 CrF2 676.9 652.2 54 CrF2 674.3 50 CrF2 768.1 52 CrF3 762.8 749.2 53 CrF3 760.1 746.7 54 CrF3 757.7 50 CrF4 795.5 52 CrF4 790.3 782.6 / 784.4 / 787.5 780.7 / 783.7 / 789.6 53 CrF4 788.2 54 CrF4 785.5 52 CrF5 764.3 53 CrF5 761.1 [a] Values in cm–1. [b] Neat fluorine matrices. 5.1. Chromium fluorides 35

Ne Ar

CrF2 CrF4

(e) * (e)

(d) CO (d) 2

(c)

(c) (b)

CrF3

* * (b) * COF2 CO2 *

CrF3 * COF2 CrF2 CrF4 (a) (a) *

Figure 5.4 : IR spectra in the 820–640 cm–1 (Ne) and 810–630 cm–1 regions (Ar) obtained after co-deposition of laser-ablated chromium atoms with 0.5% F2 diluted in the respective noble gas. (a) After 30 min of sample deposition at 5.0 K (Ne) and 10.0 K (Ar). (b) After annealing to 9.0 K (Ne) and 25 K (Ar). (c) Difference spectrum (spectrum after annealing minus spectrum directly after deposition). (d) After broadband UV irradtion with λ > 200 nm. (e) Difference spectrum (Spectrum after irradiation minus spectrum after annealing). * denotes impurities of F2.

with increasing F2 content of the matrix (figure 5.5conc). Furthermore, neat elemental fluorine was also used as a reactive matrix host because the formation and stabilisation of high-valent chromium fluorides is expected to be favoured in this extremely oxidising environment. The general features of the spectra did not change until a F2 concentration of 5% in neon was used when a new absorption at 756 cm–1 (labelled A in figure 5.6Ne) appeared in the spectra as a shoulder of the CrF3 band. This new band grew upon annealing at the expense of the CrF3 and CrF4 absorptions and can therefore be expected to correspond to a higher chromium fluoride. Upon broadband UV irradiation, the new band disappeared whereas a broad new band (labelled B in figure 5.6Ne) grew at 784 cm–1. 36 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS

UV conc.

(g)

(f)

(e) (e) (d)

(c)

(d) (b)

CrF4 CrF2 (c) CrF3

* (b) COF2 * * CO2 CO2

CrF3 CrF2 CrF4 * (a)

* (a) *

Figure 5.5 : IR spectra in the 820–640 cm–1 region obtained after co-deposition of laser-ablated chromium atoms with F2 in excess neon. Left side (UV): (a) After 30 min of sample deposition with a F2 concentration of 0.5%. (b)–(g) Difference spectra (spectra after successive UV irradiation with different high pass filters minus spectrum with the respective previous filter); (b) 630 nm filter, (c) 515 nm filter, (d) 400 nm filter, (e) 320 nm filter (f) 280 nm filter and (g) without filter (λ > 220 nm). Right side (conc.): Spectra after 30 min of sample deposition at 5.0 K with fluorine concentrations of (a) 0.2%, (b) 0.5%, (c) 1.0%, (d) 2.5% and (e) 5%. * denotes impurities of F2.

Already at this fluorine concentration the quality of the matrices was significantly reduced, and at even higher fluorine concentrations no useful information could retrieved from the spectra which consisted only of broad and unresolved bands. By contrast, the use of pure F2 as the matrix material led again to spectra which were better resolved than those obtained using high fluorine concentrations in noble gases. In the spectra of the matrices measured directly after deposition of laser-ablated chromium atoms together with pure F2 at 10 K, two main groups of absorptions could be identified as being metal-dependant while all other bands could also be observed when neat F2 was deposited without chromium. The first group of absorptions consists of three bands at 789.5, 784.0 and 780.7 cm–1 which 5.1. Chromium fluorides 37

Ne F2

B B (e) (e)

(d)

(d)

A (c) (c)

(b)

CrF4 (b)

* A * CrF3 CrF4 (a) CrF * * * 2 (a) * CO2

–1 –1 Figure 5.6 : IR spectra in the 840–660 cm (Ne) and 850–670 cm (F2) regions region obtained after co-deposition of laser-ablated chromium atoms with 5% F2 in Ne and neat F2 respectively. (a) After sample deposition at 5.0 K (Ne) and 10.0 K (F2. (b) After annealing to 9.0 K (Ne) and 20 K (F2). (c) Difference spectrum (spectrum after annealing minus spectrum after deposition). (d) After irradiation with λ > 220 nm (Ne) and λ > 400 nm (F2). (e) Difference spectrum (spectrum after irradiation minus spectrum spectrum after annealing). * denotes impurities of F2.

were assigned to the CrF4 molecule by a comparison with the spectra of the argon matrices. The matrix shifts observed in fluorine and argon matrices were found to be similar in general, and the assignment of bands measured in F2 matrices can therefore be made based on the band positions detected in argon matrices. For example, the wavenumbers of –1 –1 the ν2 mode of matrix-isolated CO2 are 667.1 cm in neon and 663.5 cm in both argon and fluorine matrices. The second new absorption was a rather narrow band observed at –1 –1 764.3 cm with an isotopic counterpart at 761.2 cm (labelled A in figure 5.6F2). After annealing of the matrix to 20 K, this band grew at the expense of the CrF4 bands. In the irradiation experiments with visible light of λ > 400 nm the 764.3 cm–1 band vanished and 38 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS

some of the absorptions previously attributed CrF4 (labelled B in figure 5.6F2) grew. The behaviour of the new group A absorptions in the experiments with either 5% F2 in neon or neat F2 as the matrix material is similar, and both bands probably correspond to the same higher chromium fluoride species. The position of the band in fluorine matrices agrees very –1 well with the 763.0 cm previously observed in argon matrices, attributed to both CrF5 and CrF6. Furthermore, the behaviour in the irradiation experiments is the same as that published for the 767.7 cm–1 band in neon.[46] The calculated thermochemical properties of the high-valent chromium fluorides suggest CrF5 to be an intermediate species on the way to CrF6 if the latter can be formed at all. The absence of any other bands which can be attributed to a higher chromium fluoride thus indicates the group A absorptions to be due to CrF5. In new experiments further evidence was found for this assignment as the band [167] correlates with the newly assigned absorptions of the Cr2F10 dimer. The photolysis of CrF5 led to the formation of new bands which could not be clearly distinguished from those of CrF4 and there is evidence that CrF5 actually isomerises upon UV irradiation [167] instead of decomposing to CrF4. The same observations were made in the spectra of the fluorine matrices where the bands formed after UV/Vis irradiation cannot unambiguously be assigned to chromium tetrafluoride neither.

5.1.5 Summary

Based on the results of the quantum-chemical calculations chromium hexafluoride is an endothermic compound at 0 K in the gas phase. However, its claimed synthesis was done at very high F2 pressures and this might have led to the formation of CrF6 which could, as a metastable species, have survived the transfer to the inert matrices. The IR spectra of the chromium fluoride molecules were also calculated. The results indicate CrF5 to have three distinct IR active Cr–F stretching fundamentals, and thus the debatable experimental spectrum fits the CrF6 molecule much better. However, both CrF5 and the putative CrF6 were published to yield the same matrix IR spectra, and as the disproportionation of CrF5 was calculated to be endothermic there is no good argument for the observation of chromium hexafluoride after the evaporation of CrF5. If the absorption were instead due to chromium pentafluoride, an explanation for the absence of the two other bands might be the flat ”mexican hat“ potential energy surface for this molecule. In the matrix isolation experiments with fluorine diluted in argon or neon at low concentrations only the bands of the lower fluorides CrF2, CrF3 and CrF4 could be observed. At 5% fluorine concentration in neon a new band corresponding to a higher fluoride appeared in the spectra. When neat fluorine was used as the matrix material, a band ad 764.3 cm–1 could be detected and this band most probably corresponds to the previously published debatable bands at 763.0 cm–1 in argon and 767.7 cm–1 in neon matrices. Due to various reasons, the author of this thesis ventures to assign these three bands to chromium pentafluoride:

• The quantum chemical calculations and experimental observations agree in that CrF5 does not disproportionate. 5.2. Manganese fluorides 39

• The observation of only one band in the Cr–F stretching region of the IR spectrum of CrF5 can be explained by the flat hypersurface of this molecule.

• No band which could be attributed to another higher-valent chromium fluoride could be observed in the IR spectra, and according to the quantum-chemical calculations CrF5 can be expected to be an intermediate on the way to chromium hexafluoride.

However, even if the dispute about the assignment of the ambiguous band seems to be settled, the formation of chromium hexafluoride at high fluorine pressures cannot be excluded. The lemon-yellow product which was observed after the fluorination of either CrO3 or elemental chromium might indeed have been CrF6, and further experimental studies are currently being made in order to confirm or refute its existence.

5.2 Manganese fluorides

Manganese has seven valence electrons which can all be oxidised and accordingly its highest possible oxidation state is VII[36] as for example in the well-known permanganate anion − [168] [MnO4] or in the oxide fluoride MnO3F. By contrast, the binary fluoride MnF7 is unknown and contrary to the case of chromium, there is no doubt about the fact that manganese hexafluoride was never observed neither. Only indirect evidence was found for + MnF5 in mass spectrometric experiments where the [MnF4] cation with manganese in its formal V oxidation state could be detected but it is not clear whether this cation originated [17–19] from MnF4 or from MnF5. The highest known neutral manganese fluoride is MnF4, a well-known solid[152] which upon heating decomposes under release of difluorine, and this behaviour has led to applications in the purification and storage of F2. As MnF4 cannot be evaporated without decomposition, its molecular structure could not be determined in GED experiments and the only experimental data available for molecular MnF4 is a matrix isolation IR spectrum from which a tetrahedral structure was inferred, albeit based on a somewhat unclear argument and without the support of quantum-chemical calculations.[169]

5.2.1 Structures

The optimised structures of the high-spin ground states of the binary molecular manganese fluorides from MnF2 to MnF7 are shown in figure 5.7. MnF2 was calculated to be a linear molecule (D ∞h symmetry) with a bond length of 181.6 pm at CCSD(T)/aVTZ level. For manganese trifluoride a distorted trigonal planar, C 2v-symmetrical structure with one shorter (d Mn–F(1) = 173.0 pm) and two longer (d Mn–F(2) = 175.3 pm) Mn–F bonds was obtained. This result agrees well with previous investigations as the MnF3 molecule is indeed a prominent example for a Jahn-Teller distortion which has already attracted the 40 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS

F(2)

F(1)

F(3)

MnF7 MnF7 MnF6 1 ’ 1 2 (D5h, A1 ) (C2v, A1) (D3d, A1g)

MnF6 MnF5 2 3 ’ F(1) (D4h, B2g) (D3h, A1 ) F(1)

F(2) F(2)

MnF4 MnF3 MnF2 4 5 6 + (C2v, B2) (C2v, A1) (D∞h, Σg )

Figure 5.7 : Optimised structures of molecular manganese fluorides; see table 5.6 for bond lengths and angles.

[172, 173] [171, 174] attention of both theoreticians and experimentalists. For both MnF2 and [170, 171] MnF3 GED experiments are reported in the literature, and the computational results are in good agreement with the experimental data. As for the chromium fluorides, the calculated bond lengths are slightly too large and again the effect is less pronounced for the higher oxidation state of the metal (table 5.6). In agreement with a recent theoretical investigation of MnIV complexes,[173] the structure of manganese tetrafluoride was calculated to be distorted tetrahedral (C 2v symmetry) with two longer (d Mn–F(1) = 173.0 pm) and two shorter (d Mn–F(2) = 169.2 pm) Mn–F bonds. The only experimental data available for this molecule is a matrix IR spectrum from which a tetrahedral structure was deduced although the recorded spectrum fits a tetragonal planar structure much better (vide infra).[169] In addition, a regular tetrahedral structure is highly improbable IV 3 for Mn because its d electron configuration would lead to partially filled t2 set of orbitals from which a Jahn-Teller distortion would result. In a square planar arrangement of ligands an orbital occupation without partially filled degenerate orbitals is possible. However, a structure optimisation of MnF4 in D 4h symmetry only led to a saddle point –1 with two imaginary frequencies (111.3 kJ mol above the C 2v-symmetrical minimum at 5.2. Manganese fluorides 41

Table 5.6 : Calculated and experimental structural parameters of molecular manganese fluorides[a] Molecule Parameter[b] B3LYP[c] CCSD(T)[d] Exp.[e] Ref. 6 + MnF2 (D∞h, Σg ) d Mn–F 180.1 181.6 181.1(4)0 [170] MnF (C , 5A ) d 173.3 173.0 172.8(14) 3 2v 1 Mn–F(1) [171] d Mn–F(2) 175.6 175.3 175.4(8)0 F(1)-Mn-F(2) 106.7 106.3 4 MnF4 (C 2v, B2) d Mn–F(1) 173.6 173.0 d Mn–F(2) 169.9 169.2 F(1)-Mn-F(1) 136.7 138.9 F(2)-Mn-F(2) 107.0 105.6 3 0 MnF5 (D3h, A2) dMn–F(ax) 174.8 174.1 d Mn–F(eq) 167.9 167.4 2 MnF6 (D4h, B2g) d Mn–F(ax) 172.1 d Mn–F(eq) 173.4 2 MnF6 (D3d, A1g) d Mn–F 172.7 [e] F-Mn-X 054.9 1 MnF7 (C 2v, A1) d Mn–F(1) 193.0 d Mn–F(2) 175.4 d Mn–F(3) 175.3 F(1)-Mn-F(2) 074.8 F(3)-Mn-F(3) 077.4 1 0 MnF7 (D5h, A1) dMn–F(ax) 171.3 d Mn–F(eq) 180.4 [a] Bond lengths in pm, angles in ◦. [b] See figure 5.7 for atom labelling. [c] aVTZ basis sets. [d] r g values from GED measurements; nozzle temperatures: 1100 K for MnF2 and 1000 K for MnF3. [e] Angle between the Mn–F bond and the S 6 axis.

CCSD(T)/aVTZ level). For MnF5, a D 3h-symmetrical trigonal bipyramidal minimum structure and bond lengths of d Mn–F(ax) = 174.1 pm and d Mn–F(eq) = 167.4 pm were calculated. In analogy to the chromium fluorides, the bond lengths in the manganese fluorides also decrease with increasing oxidation state of the metal until reaching a minimum for the pentafluoride after which steric crowding in the ligand sphere causes the bonds lengths to increase again. In the VI oxidation state, the d1 electron configuration of manganese induces a distortion from regular O h symmetry. Two different distortions are possible which lead to a D 4h- and a D 3d-symmetrical structure, respectively. Both conformers represent minima on the potential energy surface at the B3LYP/aVTZ level of theory with the tetragonal bipyramidal isomer being marginally favoured (0.5 kJ mol–1 at CCSD(T)/aVTZ//B3LYP/aVTZ level). The calculated bond lengths are d Mn–F(ax) = 172.1 pm and d Mn–F(eq) = 173.4 pm for the tetragonal bipyramidal and 172.7 pm for the trigonal antiprismatic structure. Finally, two minima on the PES could be found for MnF7, too: a regular pentagonal bipyramidal (D 5h symmetry) as well as a monocapped trigonal 42 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS

prismatic structure (C 2v symmetry). The energy gap between these two isomers was calculated to be 37.9 kJ mol–1 at CCSD(T)/aVTZ//B3LYP/aVTZ level with the latter isomer lying lower in energy. For the C 2v-symmetrical structure bond lengths of d Mn–F(2) = 175.4 pm (4×) and d Mn–F(3) = 175.3 pm (2×) were found in trigonal prismatic part with the remaining unique fluorine atom being more loosely bound (dMn-F(1) = 193.0 pm) at B3LYP level. The calculated bond lengths in the pentagonal bipyramidal isomer are 171.3 pm for the axial and 180.4 pm for the equatorial Mn–F bonds.

5.2.2 Thermochemistry

The thermochemical stability of the high-valent manganese fluorides molecules was evalu- ated by calculating several possible decomposition reactions and the results are summarised in table 5.7. As expected, the experimentally known MnF3 and MnF4 molecules only showed endothermic fluorine elimination reactions. Furthermore, the manganese penta- fluoride molecule was also calculated to be stable, and its least exothermic decomposition –1 reaction was found to be the bimolecular elimination of F2 requiring 12.9 kJ mol at 0 K. It should therefore be possible to prepare and characterise this molecule, for example using the matrix isolation technique. By contrast, all decomposition pathways for both MnF6 and MnF7 were calculated to be exothermic and these molecules are thus thermo- chemically unstable species. The decreased stability of the hexafluoride (compared with CrF6, section 5.1) can be explained by the smaller size of the central manganese atom as

Table 5.7 : Calculated thermochemistry of molecular manganese fluorides[a]

Reaction B3LYP[b] CCSD(T)[b,c]

MnF7 → MnF5 + F2 –394.5 (–399.9) –357.5 MnF7 → MnF6 + F –194.1 (–201.1) –119.3 1 MnF7 → MnF6 + 2 F2 –271.7 (–275.6) –195.3 MnF6 → MnF4 + F2 0–67.90(–76.5) –141.5 MnF6 → MnF5 + F 0–45.20(–49.8)0–86.6 1 MnF6 → MnF5 + 2 F2 –122.8 (–124.3) –162.2 MnF5 → MnF3 + F2 0217.40(206.5)0131.7 (119.9) MnF5 → MnF4 + F 0132.50(122.2) 0096.50(86.4) 1 MnF5 → MnF4 + 2 F2 0054.9 00(47.7) 0020.20(12.9) MnF4 → MnF2 + F2 0405.00(397.8)0304.1 (295.3) MnF4 → MnF3 + F 0240.10(233.2)0187.7 (180.5) 1 MnF4 → MnF3 + 2 F2 0162.50(158.7)0111.4 (107.0) MnF3 → MnF2 + F 0320.20(313.6)0268.9 (261.8) 1 MnF3 → MnF2 + 2 F2 0242.60(239.1)0192.6 (188.3) [a] Energies in kJ mol–1; values in parentheses are ZPE corrected. [b] aVTZ basis sets. [c] Single point energies at B3LYP/aVTZ structures for the MnF7 and MnF6 reactions. 5.2. Manganese fluorides 43 well as by the closed shell electron configuration of chromium hexafluoride. It is also in line with the stability trends of the hexafluorides of the 4d and 5d transition metals which become more reactive with increasing Z.[41, 50, 51] The stability of manganese pentafluoride in the condensed phase was also estimated. The calculation of the standard Gibbs energy 1 –1 for the MnF4 + 2 F2 → MnF5 gas phase reaction yielded a value of +12.2 kJ mol . Hence, the difference between the sublimation enthalpies of MnF4 and MnF5 would have to compensate for the exergonic gas phase reaction and as the sublimation enthalpies are usually larger for the lower fluorides, MnF5 is most probably not stable in the condensed phase.

5.2.3 Vibrational frequencies

Both MnF2 and MnF3 were unambiguously characterised by vibrational spectroscopy in matrix isolation experiments and the calculated frequencies are in good agreement with the experimentally obtained values. In addition, the observation of three Mn-F stretching fundamentals provides further evidence for the Jahn-Teller distortion discussed above. By contrast, the published spectrum of matrix-isolated MnF4 significantly differs from the calculated one. Three absorptions were observed and attributed to MnF4, one intense band at 794.5 cm–1 and two weak absorptions in the bending region of the spectrum.[169] The authors a priori considered two possible structures, a tetrahedral and a square planar one,

Table 5.8 : Calculated and observed wavenumbers of the Mn–F stretching modes of molecular manganese fluorides[a]

Calc. Exp. (matrix) Molecule Mode B3LYP[b] CCSD(T)[b] Ne Ar Ref. 6 + + MnF2 (D∞h, Σg ) Σg 575.3 00(–) 560.90(–) − Σu 723.5 (237) 721.3 (209) 722.1 700.1 [164] 5 MnF3 (C 2v, A1) A1 646.1 00(9) 656.70(15) 644.0 A1 720.9 (126) 729.9 (134) 728.0 712.0 [175] B2 764.2 (256) 779.3 (287) 774.0 759.0 MnF (C , 4B ) A 675.00(18) 681.50(13)  4 2v 2 1  A 749.30(85) 755.50(83)  1 794.5 [169] B1 752.3 (261) 777.9 (299)  B2 792.8 (123) 800.0 (105)  3 0 0 MnF5 (D3h, A2) A1 603.4 00(–) 617.50(–) 0 A1 704.1 00(–) 703.50(–) 00 A2 742.7 (299) 768.5 (331) E0 816.9 (272) 824.4 (235) [a] Values in cm–1, numbers in parentheses are infrared intensities in km mol–1. [b] aVTZ basis sets. 44 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS for which they stated correctly that two and three IR active fundamentals are expected respectively; the spectrum showing three bands, they concluded however that the molecule is tetrahedral (!). The calculated minimum structure of MnF4 molecule has however only C 2v-symmetry and its expectation IR spectrum consists of four active Mn–F stretching modes, three of which were predicted to show considerable IR intensities (table 5.8). In principle, the experimental spectrum would fit to a tetragonal planar molecule but no minimum structure could be found for a D 4h-symmetrical isomer of MnF4 (vide supra). A new experimental study of molecular manganese tetrafluoride seems therefore necessary in order to clarify these discrepancies. As the thermochemical analysis showed MnF5 to be a stable molecule, its vibrational spectrum was also calculated at CCSD(T)/aVTZ level. As expected, it shows two IR active Mn–F stretching modes which should in principle allow to characterise the molecule in a matrix isolation experiment.

5.2.4 Summary

It was shown that all molecular manganese fluorides from MnF2 up to MnF5 are thermo- chemically stable against elimination of either fluorine atoms or molecules. The two lower fluorides MnF2 and MnF3 were already thoroughly characterised by GED and matrix isolation experiments, and the quantum-chemical results are in good agreement with the experimental observations. By contrast, the experimental findings for MnF4 are not conclusive and MnF5 was never observed in any experiment. Thus, an experimental reinvestigation of molecular MnF4 could clarify the mismatch between the calculated and observed IR spectra of this molecule. In addition, the predicted thermochemical stability of manganese pentafluoride makes it a viable target molecule for future experiments. As this molecule was predicted to be stable as a molecular species at low temperatures only, the matrix isolation technique represents an adequate method for its characterisation.

5.3 Iron fluorides

The results presented in this section were already published in T. Schl¨oder, T. Vent- Schmidt, S. Riedel, Angew. Chem. Int. Ed. 2012, 51, 12063. All work presented hereafter was done by the author of this thesis except for the recording of the spectra shown in figure 5.11conc which was done by M.Sc. Thomas Vent-Schmidt.

Iron is the first 3d transition metal whose valence shell cannot be completely oxidised and its highest known oxidation state is VI[36] as for example in the tetraoxoferrate(VI) anion 2− [176] 2 [FeO4] or the neutral iron dioxide peroxide molecule [FeO2(η -O2)] which could be stabilised in rare gas matrices.[29] Furthermore, iron nitrido complexes containing iron in its V and VI oxidation states were successfully prepared in solution at low temperatures 5.3. Iron fluorides 45 and characterised by M¨ossbauer and X-ray absorption spectroscopy.[177, 178] However, the highest oxidation state of iron reached so far in its neutral fluorides is only III. Whereas the tetrafluorides of all other first row midtransition metals (Cr, Mn, Co) are known, FeF4 has so far eluded unambiguous characterisation. In a combined mass-spectrometric and matrix isolation investigation, its formation was postulated based on the relative + + abundance of the [FeF2] and [FeF3] fragments in the mass spectrum as well as a on new weak absorption at 758.5 cm–1 in the matrix IR spectra.[179] As no other related band could be observed, the molecule was assumed to have a highly symmetrical either square planar or tetrahedral structure but no quantum-chemical calculation were done in order to support this assignment.

5.3.1 Structures

The optimised structures of the iron fluoride molecules from FeF2 to FeF6 in their respective high-spin ground states are shown in figure 5.8. Both iron difluoride and iron trifluoride are well known molecules of D ∞h and D 3h symmetry respectively, whose structures were already studied in detailed computational investigations[181, 182] as well as in gas phase electron diffraction experiments.[170, 171] At CCSD(T)/aVTZ level, the calculated equilibrium bond lengths for these two molecules are 177.1 pm in FeF2 and 175.8 pm in FeF3 (table 5.9). In order to compare the experimental and calculated bond lengths, the effects of rotational and vibrational averaging at the temperatures of the experiment have to be accounted for in the calculations. This leads to larger atomic distances (table 5.10) and the calculated r g

F(ax)

F(ax)

F(eq) F(eq)

FeF6 FeF5 FeF4 3 4 5 (D4h, A2g) (C4v, B2) (D2d, B2)

FeF3 FeF2 6 ’ 5 (D3h, A1 ) (D ∞h, Δg)

Figure 5.8 : Optimised structures of molecular iron fluorides; see table 5.9 for bond lengths and angles. 46 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS

Table 5.9 : Calculated and experimental structural parameters of molecular iron fluorides[a]

Molecule Parameter B3LYP[b] MP2[b] CCSD(T)[b] Exp.[c] Ref. 5 FeF2 (D∞h, ∆g) d Fe–F 176.2 177.8 177.1 176.9(4) [170] 6 0 FeF3 (D3h, A1) d Fe–F 176.4 175.9 175.8 176.3(4) [180] 5 FeF4 (D2d, B2) d Fe–F 171.9 170.9 171.5 F-Cr-F 138.5 140.2 139.7 4 FeF5 (C 4v, B2) d Fe–F(ax) 168.4 169.6 169.3 d Fe–F(eq) 171.9 171.9 171.4 F(ax)-Cr-F(eq) 099.5 100.0098.8 3 FeF6 (D4h, A2g) dFe–F(ax) 172.7 164.5 d Fe–F(eq) 173.6 170.0 ◦ [a] Bond lengths in pm, angles in . [b] aVTZ basis sets. [c] r g values from GED measurements; nozzle temperatures: 1050 K for FeF2 and 1260 K for FeF3.

Table 5.10 : Core-valence correlation effects on the calculated bond lengths of FeF2, FeF3 and [a] FeF4 Molecule Parameter[b] CCSD(T)[c] CCSD(T)[d] CCSD(T)[e] Exp. Ref. 5 FeF2 (D∞h, ∆g) r e 177.1 176.7 176.5 r g 178.8 178.4 178.2 176.9(2) [170] 6 0 FeF3 (D3h, A1) r e 175.8 175.6 175.5 r g 177.3 177.1 176.3(4) [180] 5 FeF4 (D2d, B2) r e 171.5 171.3

[a] Values in pm. [b] r g values at 1050 K for FeF2 and 1260 K FeF3. [c] aVTZ basis sets. [d] aVTZ basis set for F and wCVTZ-NR basis set for Fe. [e] aVTZ basis set for F and awCVTZ-NR basis set for Fe. values are significantly too large when compared with the experiment. The deviation, which is larger for FeF2 than for FeF3, could be decreased for both molecules when core-valence correlation was considered in the CCSD(T) calculations by the use of the awCVTZ-NR basis set for iron and a corresponding smaller frozen core (∆d Fe–F = 0.6 pm for FeF2 and ∆d Fe–F = 0.3 pm for FeF3). It could be shown that these values can in principle be further improved by using larger basis sets and accounting for relativistic effects.[182] For the iron tetrafluoride molecule, the structure optimisation yielded a flattened tetrahedral structure of D 2d symmetry and a bond length of 171.5 pm at CCSD(T)/aVTZ level. As expected from the d4 electron configuration of FeIV no higher symmetrical structure was obtained and the initial presumption, derived from the observation of only one infrared absorption in the region of the Fe–F stretching modes, of a either D 4h- or T d-symmetrical structure is thus doubtful. The FeF5 molecule was calculated to have a square pyramidal structure (C 4v-symmetry). The calculated bond lengths in this molecule of d Fe–F(ax) = 169.3 pm and d Fe–F(eq) = 171.4 pm are in average the shortest of all iron fluoride molecules. Finally, 5.3. Iron fluorides 47

a Jahn-Teller distorted flattened octahedral structure of D 4h symmetry was calculated for iron hexafluoride. As for CrF6 and MnF6 the bond lengths in this molecule (d Fe–F(ax) = 172.7 pm and d Fe–F(ax) = 173.6 at B3LYP level) are again larger than in the corresponding pentafluorides and indicate an increased inter-ligand repulsion.

5.3.2 Thermochemistry

As to evaluate the thermochemical stability of the high-valent iron fluorides, three different possible decomposition reactions were calculated for each molecule. As expected, the experimentally known FeF3 molecule only shows endothermic decomposition pathways (table 5.11). Furthermore, iron tetrafluoride too was calculated to be a thermochemically stable species with the most probable decomposition reaction being the bimolecular –1 elimination of F2 (∆rE = + 42.5 kJ mol at 0 K). By contrast, this reaction was calculated to be exothermic for iron pentafluoride which is thus not a stable molecule in the gas phase. If FeF5 is present as an isolated molecule, only unimolecular decompositions need to be considered, and the homolytic breaking of one Fe–F bond was predicted to be exothermic. –1 For the unimolecular F2 elimination, a reaction energy of –0.5 kJ mol was calculated –1 and, the error of these calculations lying within a few kJ mol , FeF5 might therefore be marginally stable under matrix isolation conditions. As expected, the calculated thermochemistry of iron hexafluoride yielded only exothermic decomposition pathways and the molecule was found to be even less stable than MnF6. The stability of bulk iron tetrafluoride was estimated by constructing a Born-Fajans-Haber cycle starting from solid iron trifluoride and elemental fluorine (figure 5.9). The Gibbs standard sublimation enthalpy of FeF3 was taken from ref. [183] and the Gibbs reaction enthalpy for the gas phase

Table 5.11 : Calculated thermochemistry of molecular iron fluorides[a]

Reaction B3LYP[b] CCSD(T)[b,c]

FeF6 → FeF4 + F2 –143.0 (–153.2) –215.3 FeF6 → FeF5 + F 0–70.10(–78.7) –101.1 1 FeF6 → FeF5 + 2 F2 –147.7 (–153.2) –177.0 FeF5 → FeF3 + F2 0130.70(120.3) 0009.60(–0.5) FeF5 → FeF4 + F 0082.4 00(74.4) 0037.90(30.5) 1 FeF5 → FeF4 + 2 F2 0004.8 00(–0.1)0–38.4 (–43.0) FeF4 → FeF2 + F2 0377.30(369.5)0326.6 (317.8) FeF4 → FeF3 + F 0203.60(194.9)0124.2 (116.0) 1 FeF4 → FeF3 + 2 F2 0125.90(120.4) 0048.00(42.5) FeF3 → FeF2 + F 0329.00(323.6)0354.9 (348.8) 1 FeF3 → FeF2 + 2 F2 0251.40(249.1)0278.6 (275.3) [a] Energies in kJ mol–1; values in parentheses are ZPE corrected. [b] aVTZ basis sets. [c] Single point energies at B3LYP/aVTZ structures for the FeF6 reactions. 48 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS

ΔrG = –23.2 FeF3 (g) + ½ F2 (g) FeF4 (g)

ΔsubG = 159.6 –ΔsubG < –136.4

FeF3 (s) + ½ F2 (g) FeF4 (s) ΔrG < 0.0

Figure 5.9 : Born-Fajans-Haber cycle for the formation of solid iron tetrafluoride; values in kJ mol–1. reaction was calculated at CCSD(T)/aVTZ level. The result indicates that a minimum of –1 ∆subG(FeF4) = 136.4 kJ mol would be needed to be released upon condensation of gaseous FeF4 if it were to be stable as bulk material. The enthalpy of sublimation of a unknown tetrafluoride MF4 can be estimated by adding ∆subH(MnF4) − ∆subH(MnF3) to the [20] sublimation enthalpy of the corresponding trifluoride. Thereby, a value of ∆subH(FeF4) –1 ≈ 136 kJ mol was calculated for FeF4. However, the entropy of sublimation (which –1 amounts to 61.4 kJ mol for FeF3) needs still to be added and it is hence unlikely that iron tetrafluoride is a stable compound in the condensed phase.

5.3.3 Matrix isolation experiments

In order to synthesise and unambiguously characterise the iron tetrafluoride molecule, new matrix isolation experiments were done using the laser ablation technique for the generation of highly reactive excited iron atoms which were then reacted with elemental fluorine. The IR spectra measured directly after co-deposition of laser-evaporated iron atoms with F2 and excess argon or neon onto a KBr or CsI window cooled to 10.0 and 3.8 K respectively for deposition are shown in figure 5.10. In the spectra of the neon matrices three major groups of absorptions could be observed directly after deposition [164, 184] two of which were previously assigned to FeF2 and FeF3 whereas the third band at 778.6 cm–1 was so far unknown. This additional band is relatively strong directly after sample deposition and grows upon annealing to 9.0 K whereas the 752.5 and 743.6 cm–1 bands of FeF2 and FeF3 become slightly weaker. In contrast to the annealing experiments, irradiation of the neon matrices using the full spectrum of a mercury arc lamp led to a –1 decrease of the 778.6 cm band and an increase of the bands corresponding to FeF2 and FeF3. Based on the results of the quantum-chemical calculations (vide infra) this new band –1 at 778.6 cm was assigned to the FeF4 molecule, which upon annealing of the matrices is formed by the reaction between F2 molecules and adjacent iron atoms or lower fluorides, and upon λ > 220 nm irradiation decomposes to FeF2 and FeF3. The photochemistry of iron tetrafluoride was further investigated by irradiating the matrix samples using different UV/Vis high-pass filters with successive lower wavenumber edges, and these experiments 5.3. Iron fluorides 49

Ne Ar

(e)

(e)

(d)

(d)

(c) (c) 54 56 FeF3 FeF3 56 FeF4 56 FeF3 56 FeF4 (b) (b)

56 57 FeF2 FeF2 54 56 FeF2 FeF2 54FeF 2 (a) (a)

Figure 5.10 : IR spectra in the 730–790 cm–1 region (Ne) and the 710–770 cm–1 region (Ar) obtained after co-deposition of laser-ablated iron atoms with F2 in excess of the respective noble gas. a) After 1 h of sample deposition at 3.8 K (Ne) and 10.0 K (Ar). b) After annealing to 9 K (Ne) and 30 K (Ar). c) Difference spectrum (spectrum measured after annealing minus spectrum directly after deposition). d) After UV irradiation with λ > 220 nm. e) Difference spectrum (spectrum measured after irradiation minus spectrum after annealing).

showed that a wavelength of λ < 320 nm is necessary for the decomposition of FeF4 (figure 5.11conc). Several experiments were done in which the fluorine concentration was varied from 0.5% to 3.0%, and as expected the intensity of the FeF4 band increased with higher fluorine content of the matrix. Due to the Jahn-Teller distortion in FeF4 a second IR active Fe–F stretching vibration is expected for this molecule, and a very weak band could indeed be observed at 656.7 cm–1 when using high fluorine concentrations (figure 5.11conc). In the spectra of the argon matrices obtained directly after deposition, only the bands of [164] [166] FeF2 and FeF3 could be observed (figure 5.10Ar). Annealing of the matrix to 30 K –1 led to the growth of a new broad band centred at 757 cm whereas the bands of FeF2 50 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS

conc. UV

(g)

FeF2

(f) FeF3 FeF4 FeF4

(e)

(d) (d)

(c) (c)

(b)

FeF2 (b)

FeF4 FeF3

(a) (a)

Figure 5.11 : Infrared spectra obtained after co-deposition of laser-evaporated iron atoms with F2 in excess neon. Left side (conc.): (a) 1 h of sample deposition at 3.8 K. (b)–(g) Difference spectra (spectra after successive UV irradiation with different high pass filters minus spectrum with the respective previous filter); (b) 630 nm filter, (c) 515 nm filter, (d) 400 nm filter, (e) 320 nm filter (f) 280 nm filter and (g) without filter (λ > 220 nm). Right side (UV): After 1 h of sample deposition at 3.8 K using (a) 0.5% F2, (b) 1.0% F2, (c) 2.0% F2 and (d) 3.0% F2.

and FeF3 became weaker. As for the neon matrices, UV irradiation of the matrix samples led to a decrease of the new band accompanied by an increase of the bands of the lower fluorides. Unfortunately, the second absorption, which was identified in the spectra of the neon matrices, could be not observed when using argon as the matrix host. Nevertheless, the new band at 757 cm–1 showed the same behaviour as the 778.6 cm–1 band in the neon matrices and can therefore also be assigned to the FeF4 molecule. 5.3. Iron fluorides 51

5.3.4 Vibrational frequencies

The vibrational spectra of the iron fluoride molecules were calculated in order to support the experimental identification of FeF4. The calculated harmonic frequency at CCSD(T)/aVTZ level is in excellent agreement with the experimental neon matrix values for the FeF2 molecule (∆ν˜ = 0.7 cm–1). However, the differences between the experimental and the calculated values increase when going to the higher iron fluorides (table 5.3). It was shown above that the consideration of core-valence correlation is important for the calculation of the structures of these molecules and neither can it be neglected for the frequency calculations when very accurate results are desired. However, at first sight the agreement between the experimental and harmonic wavenumbers decreases when accounting for CV correlation (table 5.13). This is due to the harmonic approximation in the frequency calculations, and when anharmonic frequencies are calculated the difference between the experimental and computed values were very small for both FeF2 and FeF3. Thus, the core- valence error is actually compensated by the harmonic error for FeF2 at CCSD(T)/aVTZ level, and as the core-valence correlation becomes less important for the higher-valent fluorides, the two errors only partly cancel for FeF3. As CV correlation is less important for FeF3 and FeF4, very good results were obtained for these molecules using the wCVTZ-NR basis set for iron, and the predicted wavenumber for FeF4 confirmed the assignment of the –1 778.6 cm band to this molecule. For FeF4 the anharmonic corrections were calculated at the MP2 level of theory (using the same basis set combinations and frozen cores) which proved to reliably reproduce the anharmonic corrections calculated at CCSD(T) level for both FeF2 and FeF3. Finally, isotopic shifts were also calculated and the results are

Table 5.12 : Calculated and observed wavenumbers of the Fe–F stretching modes of molecular iron fluorides[a] Calc. Exp. (matrix) Molecule Mode B3LYP[b] CCSD(T)[b] Ne Ar Ref. 5 + FeF2 (D∞h, ∆g) Σg 601.1 00(–) 594.8 00(–) − Σu 751.8 (237) 751.8 (217) 752.5 730.5 tw 6 0 0 FeF3 (D3h, A1) A1 644.4 00(–) 663.6 00(–) E0 727.5 (334) 751.2 (389) 743.6 728.5 tw 5 FeF4 (D2d, B2) B2 653.10(54) 660.00(52) 651.9 tw A1 669.1 00(–) 661.8 00(–) E 771.2 (402) 791.4 (395) 778.6 757.0 tw 5 FeF5 (C 4v, B2) B2 578.8 00(–) 586.6 00(–) A1 667.1 00(1) 648.2 00(0) A1 728.80(55) 685.00(17) E 768.5 (403) 786.7 (359) [a] Values in cm–1 for 56Fe isotopomers; numbers in parentheses are infrared intensities in km mol–1. [b] aVTZ basis sets. tw = this work. 52 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS

Table 5.13 : Core-valence correlation effects on the calculated wavenumbers of FeF2, FeF3 and [a] FeF4 CCSD(T)[b] CCSD(T)[c] CCSD(T)[d] Exp. (matrix)

Molecule Mode ν˜harm. ν˜anh. ν˜harm. ν˜anh. ν˜harm. ν˜anh. Ne Ar Ref. − FeF2 Σu 751.8 742.7 756.2 747.4 759.4 750.4 752.5 730.5 tw 0 FeF3 E 751.2 742.8 753.4 744.4 755.1 743.6 728.5 tw FeF4 B2 660.0 650.9 665.6 656.9 651.9 tw E 791.4 779.0 794.4 782.1 778.6 757.0 tw Values in cm–1 for 56Fe isotopomers. [b] aVTZ basis sets. [c] aVTZ basis set for F and wCVTZ-NR basis set for Fe. [d] aVTZ basis set for F and awCVTZ-NR basis set for Fe. tw = this work.

Table 5.14 : Calculated and experimental isotopic shifts of selected Fe–F stretching modes of molecular iron fluorides[a] Calc. Exp. (matrix) Molecule Mode B3LYP[b] CCSD(T)[b] Ne Ar 5 − 54 FeF2 (D∞h, ∆g) Σu ( Fe) 757.40(5.6) 757.40(5.6) 758.00(5.5) 736.10(5.6) − 56 Σu ( Fe) 751.80(0.0) 751.80(0.0) 752.50(0.0) 730.50(0.0) − 57 Σu ( Fe) 749.1 (–2.7) 749.1 (–2.7) 750.0 (–2.5) 6 0 0 54 FeF3 (D3h, A1) E ( Fe) 732.10(4.6) 756.00(4.8) 748.30(4.7) E0 (56Fe) 727.50(0.0) 751.20(0.0) 743.60(0.0) 728.50(0.0) E0 (57Fe) 725.3 (–2.2) 748.9 (–2.3) 5 54 FeF4 (D2d, B2) B2 ( Fe) 654.80(1.7) 661.50(1.5) 56 B2 ( Fe) 653.10(0.0) 660.00(0.0) 651.90(0.0) 57 B2 ( Fe) 652.4 (–0.7) 659.2 (–0.8) E (54Fe) 776.90(5.7) 797.20(5.8) 784.40(5.8) 763.0 (6).0 E (56Fe) 771.20(0.0) 791.40(0.0) 778.60(0.0) 757.0 (0).0 E (57Fe) 768.5 (–2.7) 788.6 (–2.8) 5 54 FeF5 (C 4v, B2) A1 ( Fe) 732.80(4.0) 688.00(3.0) 56 A1 ( Fe) 728.80(0.0) 685.00(0.0) 57 A1 ( Fe) 726.9 –(1.9) 683.6 (–1.4) E (54Fe) 773.90(5.7) 792.50(5.8) E (56Fe) 768.20(0.0) 786.70(0.0) E (57Fe) 765.4 (–2.8) 783.9 (–2.8) [a] Values in cm–1; in parentheses: isotopic shifts compared to the 56Fe isotopomers. [b] aVTZ basis sets. 5.4. Cobalt fluorides 53

in good agreement with the experimentally found values (table 5.14). As FeF5 might also be stable, its IR spectrum was calculated at coupled cluster level, resulting in two Fe–F stretching modes of significant IR intensity which might support the experimental characterisation of this molecule.

5.3.5 Summary

The iron tetrafluoride molecule was synthesised by the reaction of laser-ablated excited iron atoms with F2 and subsequently trapped in rare gas matrices. It was characterised by its IR absorptions at 778.6 and 651.9 cm–1 in neon matrices and the assignment of the bands was supported by quantum-chemical calculations at CCSD(T) level. The observation of a second band in the Fe–F stretching region confirmed the quantum-chemical prediction of a distorted tetrahedral, D 2d-symmetrical structure for this molecule. When argon was –1 used as the matrix host, only one broad band at 757 cm could be assigned to FeF4. Due to the width of this band it cannot be excluded that the previously observed absorption at 758.5 cm–1 was also caused by iron tetrafluoride.[179] The preparation of matrix-isolated FeF4 is in line with its calculated thermochemical stability as a molecular species. However, it is unlikely to exists as bulk material at ambient conditions as its decomposition to solid FeF3 and F2 was predicted to be exothermic. The coupled cluster calculations further suggest that iron pentafluoride might be marginally stable but no evidence was found for this molecule in the matrix isolation experiments. A possible way for its future stabilisation might be the use of neat fluorine as the matrix host, which was also necessary for the formation of CrF5 in the laser ablation experiments (section 5.1) as no evidence was found for this molecule when noble gas matrices with small fluorine concentrations were used. Finally, iron hexafluoride was calculated to be unstable and will probably remain a hypothetical compound.

5.4 Cobalt fluorides

As for the previously discussed metals, the difluoride and the trifluoride of cobalt are well known compounds, too, both in the gas phase and in the solid state. CoF3 even found an industrial application as a fluorinating agent for organic substrates: In the Fowler process, gaseous hydrocarbons or fluorohydrocarbons are passed over solid CoF3 to yield perfluorocarbons and the cobalt fluoride can then be regenerated by fluorinationation with [185] F2. Furthermore, evidence for cobalt tetrafluoride was found in mass spectrometric + [20] experiments where the [CoF4] cation could be detected and an IR spectrum of matrix- [21] isolated CoF4 was also measured. Yet, no quantum-chemical calculations were made in order to support the assignment of the band to CoF4. Higher oxidation states of cobalt were not yet unambiguously characterised:[36] The cobalt dioxide peroxide molecule, 2 [CoO2(η -O2)], with cobalt in its VI oxidation state was detected in matrix isolation 54 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS experiments[186] but as quantum chemical calculations suggest the diperoxide complex 2 [187] [Co(η -O2)2] to be the most stable of the CoO4 isomers, a reinvestigation of this molecule might dispel the doubts.

5.4.1 Structures

All cobalt fluoride molecules from CoF2 to CoF6 were calculated to have high-spin ground states and their optimised structures are shown in figure 5.12. Cobalt difluoride is a linear 4 molecule (D ∞h symmetry) with a calculated bond length of 174.1 pm in its ∆g ground state. This result is confirmed by a previous theoretical investigation of cobalt dihalides, in which several low-lying excited states were also calculated.[188] The first excited state 4 − –1 ( Σg ) of this molecule was calculated to lie only 5.6 kJ mol above the ground state and for this electronic configuration a bond length of 173.0 pm was obtained. The computed structure of CoF3 has D 3h symmetry and the bond length in this molecule was calculated to be 172.2 pm. The gas phase structure of both molecules were determined by electron diffraction measurements,[189, 190] and as for the fluorides of chromium, manganese and iron the calculated bond lengths at CCSD(T)/aVTZ level are slightly too long when compared with the experimental values (table 5.15). For cobalt difluoride, the experimental bond 4 length compares better to the one calculated for its ∆g ground state than to that obtained 4 − for its Σg first excited state but due to the small energy gap between the two states they were both present in the high temperature experiment. For cobalt tetrafluoride a 5 regular tetrahedral structure (T d symmetry) was calculated as expected for the d electron

F(ax)

F(eq)

CoF6 CoF5 CoF4 4 5 ’ 6 (Oh, A2g) (D3h, A1 ) (Td, A1)

CoF3 CoF2 5 ’ 4 (D3h, A1 ) (D ∞h, Δg)

Figure 5.12 : Optimised structures of molecular cobalt fluorides; see table 5.15 for bond lengths and angles. 5.4. Cobalt fluorides 55 configuration of CoIV. With a Co–F distance of 171.8 pm, the bonds in this molecule were calculated to be the shortests of all cobalt fluorides. CoF5 was calculated to be a trigonal bipyramidal molecule with bond lengths of d Co–F(ax) = 172.2 pm and d Co–F(eq) = 173.9 pm. Finally, the structure of cobalt hexafluoride was also optimised and the molecule was found to be regular octahedral (d Co–F = 173.2 pm at B3LYP/aVTZ level). The longer bonds in CoF5 and CoF6 indicate that, in contrast to the lighter 3d elements, only four ligands can be placed around the smaller cobalt atom before inter-ligand repulsion becomes dominant.

Table 5.15 : Calculated and experimental structural parameters of molecular cobalt fluorides[a]

Molecule Parameter B3LYP[b] CCSD(T)[b] Exp.[c] Ref. 4  CoF2 (D∞h, ∆g) d Co–F 174.0 174.1 4 + 175.4(3) [189] CoF2 (D∞h, Σg ) d Co–F 172.3 173.0 5 0 CoF3 (D3h, A1) d Co–F 173.3 172.2 173.2(4) [190] 6 CoF4 (T d, A1) d Co–F 173.3 171.8 5 0 CoF5 (D3h, A1) d Co–F(ax) 170.5 172.2 d Co–F(eq) 174.6 173.9 4 CoF6 (Oh, A2g) d Co–F 173.2 ◦ [a] Bond lengths in pm, angles in . [b] aVTZ basis sets. [c] r g values from GED measurements; nozzle temperatures: 1373 K for CoF2 and 812 K for CoF3.

5.4.2 Thermochemistry

The thermochemistry of the cobalt fluorides was evaluated by calculating the energy of reaction for several different decomposition pathways, and the three experimentally verified molecules CoF2, CoF3 and CoF4 showed only endothermic decomposition reactions. By contrast, all decomposition reactions of cobalt penta- and hexafluoride were calculated to be endothermic at CCSD(T)/aVTZ level, and these molecules are therefore unlikely to exist. However, the trend of the decreasing stabilities of the hexafluorides with increasing Z of the metal is not followed for CoF6 which was calculated to be more stable than FeF6. The reason for this enhanced stability probably is the d3 electron configuration of CoVI which leads to a half-filled t2g set of orbitals. The thermochemistry of solid CoF4 was calculated by constructing a Born-Haber cycle starting from solid cobalt trifluoride and F2. The Gibbs enthalpy of the gas phase reaction was calculated at CCSD(T)/aVTZ level while the free enthalpy of sublimation of solid CoF3 was previously found to be –1 [191] ∆subHCoF3 = 216 kJ mol . The entropy of sublimation was calculated as difference [183] between the experimental entropy of CoF3(s) and the calculated entropy of CoF3(g) at –1 CCSD(T) level, leading to a value of ∆subG(CoF3) = 156.4 kJ mol for the free enthalpy –1 of sublimation of CoF3. Thus, a minimum energy of 117.7 kJ mol would need to be released upon condensation of gaseous CoF4 if it were a stable compound in the condensed –1 [20] phase. The lattice energy of CoF4 was estimated to be ∆subH(CoF4) ≈ 131 kJ mol 56 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS

Table 5.16 : Calculated thermochemistry of molecular cobalt fluorides[a]

Reaction B3LYP[b] CCSD(T)[b,c]

CoF6 → CoF4 + F2 –144.1 (–151.5) –197.7 CoF6 → CoF5 + F 0004.0 00(–3.7) 00–9.2 1 CoF6 → CoF5 + 2 F2 0–73.70(–78.2)0–85.1 CoF5 → CoF3 + F2 0031.6 00(24.7)0–48.9 CoF5 → CoF4 + F 0097.3 000(1.2)0–37.4 1 CoF5 → CoF4 + 2 F2 0–70.40(–73.3) –113.7 CoF4 → CoF2 + F2 0328.60(321.8)0280.1 (271.7) CoF4 → CoF3 + F 0179.60(172.3)0141.0 (133.5) 1 CoF4 → CoF3 + 2 F2 0102.0 00(97.8) 0064.80(60.0) CoF3 → CoF2 + F 0304.30(298.5)0291.6 (285.2) 1 CoF3 → CoF2 + 2 F2 0226.60(224.0)0215.3 (211.7) [a] Energies in kJ mol–1; values in parentheses are ZPE corrected. [b] aVTZ basis sets. [c] Single point energies at B3LYP/aVTZ structures for the CoF6 reactions.

ΔrG = –38.7 CoF3 (g) + ½ F2 (g) CoF4 (g)

ΔsubG = 156.4 –ΔsubG < –117.7

CoF3 (s) + ½ F2 (g) CoF4 (s) ΔrG < 0.0

Figure 5.13 : Born-Fajans-Haber cycle for the formation of solid cobalt tetrafluoride; values in kJ mol–1.

–1 but the entropy contribution, amounting to 59.6 kJ mol for CoF3, needs also to be considered. It is hence unlikely that cobalt tetrafluoride is stable as bulk material and this result is confirmed by a previous investigation in which solid CoF4 was also predicted to be unstable.[20]

5.4.3 Vibrational frequencies

The vibrational frequencies of the molecular cobalt fluorides were also calculated and the results are summarised in table 5.17. In contrast to the fluorides of chromium, manganese and iron, the agreement between the calculated and measured wavenumbers is good 4 only for CoF3. At CCSD(T)/aVTZ level, the calculated wavenumber of the ∆g ground –1 state of CoF2 differs by more than 40 cm from the experimental value. As the matrix environment induces a change of the ground state of the UO2 molecule (compare chapter 3) 5.4. Cobalt fluorides 57

Table 5.17 : Calculated and observed wavenumbers of the Co–F stretching modes of molecular cobalt fluorides[a] Calc. Exp. (matrix) Molecule Mode B3LYP[b] CCSD(T)[b] Ne Ar Ref. 4 + CoF2 (D∞h, ∆g) Σg 602.3 00(–) 598.3 00(–) − Σu 784.1 (184) 788.9 (183) 745.8 723.5 [164] 4 + + CoF2 (D∞h, Σg ) Σg 629.9 00(–) 662.4 00(–) − Σu 775.4 (237) 773.7 (228) 745.8 723.5 [164] 5 0 0 CoF3 (D3h, A1) A1 651.5 00(–) 673.7 00(–) E0 729.4 (300) 754.4 (368) 748.2 737.2 [192] 6 CoF4 (T d, A1) A1 654.3 00(–) 676.0 00(–) T2 717.6 (286) 751.4 (307) 767.8 [21] [a] Values in cm–1; numbers in parentheses are infrared intensities in km mol–1. [b] aVTZ basis sets.

4 − the vibrational frequencies were also calculated for the Σg excited state of CoF2. This led to a slightly better agreement of ∆ν˜ = 27.9 cm–1 between the calculated and experimental wavenumbers but nevertheless the difference is still large. One possible explanation of the deviation might be the multi-reference character of the wavefunction of CoF2. However, the calculated wavenumbers at the MCQDPT2 level gave even slightly worse values[188] and further theoretical investigations of this molecule are thus necessary. By contrast, the agreement of the calculated and experimental wavenumbers (neon matrices) is good –1 (∆ν˜ = 6.2 cm ) for CoF3, especially when taking into account that a consideration of anharmonic frequencies would lead to a lowering of the calculated wavenumber. In the case of cobalt tetrafluoride, the calculated harmonic wavenumbers at CCSD(T) level are 16.4 cm–1 too low when compared with the experimental value in argon matrices and this difference would even increase if anharmonic frequencies were calculated. Furthermore, it can be expected that the experimental wavenumber would be larger if neon were used as the matrix host. The assignment of this band to the CoF4 molecule is thus doubtful and a new study of matrix-isolated cobalt tetrafluoride could clarify whether the 767.8 cm–1 absorption was indeed due to this molecule.

5.4.4 Summary

The calculated thermochemistry for the cobalt fluorides showed only endothermic de- composition reactions for the experimentally known CoF2, CoF3 and CoF4 molecules. However, the experimental wavenumber of the IR absorption attributed to CoF4 in matrix isolation experiment significantly deviates from the calculated value, and a reinvestigation of matrix-isolated cobalt tetrafluoride might clarify this assignment. The stability of bulk CoF4 was also estimated, and it was found that this compound is probably not 58 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS stable in the condensed phase. Finally, the penta- and hexafluoride of cobalt were also studied. However, these two molecules were found to be thermochemically unstable, as all considered decomposition pathways were computed to be exothermic.

5.5 Conclusion and outlook

The structures and stabilities of the high-valent first row transition metal fluorides were investigated by quantum-chemical calculations at coupled cluster level. Further, matrix isolation experiments were done for iron and chromium fluorides using the laser ablation technique for the generation of excited gaseous metal atoms which were subsequently reacted with elemental fluorine.

Hexafluorides. The thermochemical analysis of the MF6 molecules (M = Cr, Mn, Fe, Co) at CCSD(T) level showed all of these molecules to be unstable against elimination of fluorine and thus corroborates the lack of conclusive experimental evidence for any of these compounds. With the exception of CoF6, the stabilities of the hexafluorides decrease with increasing Z of the metal, and CrF6 was calculated to be the most stable of all considered 3d transition metal hexafluorides. Its formation was debated in the literature and the existence of CrF6 cannot be excluded, as the published syntheses involved extreme fluorine pressures and high temperatures. However, it could be shown by new matrix isolation experiments that the IR band previously attributed to this molecule actually corresponds to CrF5 although the experimental spectrum fits the one expected for CrF6 much better. Pentafluorides. The highest chromium fluoride so far characterised beyond doubt is thus CrF5. Although manganese pentafluoride could never be observed in any experiment, its computed thermochemistry showed it to be a stable compound, at least if present as an isolated molecule at low temperatures. By contrast it is most probably not stable in the condensed phase. MnF5 thus represents a viable target molecule for future experimental efforts. As for the hexafluorides, the stabilities of the pentafluorides also decrease with increasing Z of the metal. For FeF5 the quantum-chemical calculations suggest that it might be marginally stable. Therefore, harsher conditions are probably needed for its experimental stabilisation which might be realised in neat fluorine matrices. Tetrafluorides. The reaction of laser-evaporated iron atoms with fluorine diluted in noble gases at low concentrations led to the formation of iron tetrafluoride which could be unambiguously characterised for the first time. As expected for its D 2d-symmetrical structure, two IR absorptions in the Fe–F stretching region could be assigned to FeF4 in the neon matrix IR spectra. The synthesis of this molecule bridged the gap between the previously known tetrafluorides of Cr, Mn and Co. However, although there is no doubt about the existence of both MnF4 and CoF4, the matrix IR spectra of these molecules do not correspond to predicted ones and a reinvestigation of these molecules could help to clarify the assignment of the bands. 5.6. Experimental and computational details 59 5.6 Experimental and computational details

5.6.1 Matrix isolation experiments

Matrix samples were prepared by co-deposition of laser-ablated excited metal atoms (Fe: 99.9%, Strem chemicals; Cr: 99.9%, Smart elements) with F2 (99.8 %, Solvay) diluted at different concentrations in neon (99.999%, Air Liquide) or argon (99.999%, Sauerstoffwerk Friedrichshafen) as well as with neat fluorine. The gases were mixed in a custom-made stainless mixing chamber equipped with a manometer to which the neon and argon bottles as well as a stainless steel F2 storage cylinder were connected. During the preparation of the gas mixtures the fluorine cylinder was cooled to 77 K in order to freeze out impurities. The mixing vessel was connected to the matrix chamber by a stainless steel capillary. Reactants were condensed onto KBr and CsI windows cooled to 3.8–5.0 K (neon) or 10.0 K (argon and F2) using a closed-cycle helium cryostat (Sumitomo Heavy Industries, RDK-205D) inside the vacuum chamber. The cold windows were coated with a protective argon layer before condensing neat fluorine matrices. For the laser ablation of metals, the 1064 nm fundamental of a Nd:YAG laser (Continuum, Minilite II; repetition rate: 10 Hz, pulse width: 5 ns) was focussed onto a rotating iron or chromium target through a hole in the cold window. The pulse energy of the laser was adjusted to the metal and the best spectra were obtained using laser energies of 32 and 18 mJ per pulse for chromium and iron, respectively. Irradiation of the matrix samples was done using a mercury arc street lamp (Osram HQL 250) with the outer globe removed and different high-pass filters (Schott; types: N-WG280, N-WG320, GG400, OG515 and RG630). Infrared spectra were recorded on a Bruker Vertex 70 FT-IR spectrometer purged with dry air at 0.5 cm–1 resolution in the region between 4000 and 450 cm–1 using a RT-DLaTGS or liquid-nitrogen-cooled MCT detector. The spectra were evaluated using the OPUS 6.5 software and plotted with OriginPro 8.1G.

5.6.2 Quantum-chemical calculations

The structures of all molecules were fully optimised (by relaxing all parameters) at density functional theory level using the B3LYP[193–196] hybrid functional which was shown to give reliable results for comparable molecules.[117] Dunning’s correlation-consistent triple-ζ aug-cc-pVTZ (denoted as aVTZ for brevity) basis sets were used for both fluorine and the different metals.[197, 198] Relativistic effects were not considered in the calculations as they are of minor importance for these light elements. The ground states of all molecules were determined by calculating the structures for all possible allowed spin multiplicities arising form the respective dn electron configurations of the metal atoms. Highly symmetrical stationary points on the potential energy surface were calculated within the restrictions of the respective point groups, especially when not corresponding to the (global) minimum structures. The subsequent ab initio structure optimisations at MP2 and CCSD(T) 60 CHAPTER 5: FLUORIDES OF THE 3D TRANSITION METALS level were done starting from the structures optimised at DFT level while retaining the molecular symmetries and using ROHF reference functions in the case of open shell electron configurations. In the ab initio calculations the frozen core approximation was used with the 1s orbitals of fluorine as well as the 1s, 2sp and 3sp orbitals of the metals not being considered in the evaluation of the correlation energies. Harmonic frequency calculations were carried out for the stationary points on the potential energy surface for different possible isotopomers. Furthermore, anharmonic frequencies were calculated at ab initio levels using second order vibrational perturbation theory (VPT2). The extrapolation of the thermochemical values to the complete basis set (CBS) limit[199] was done using different correlation-consistent basis sets (aug-cc-pVXZ, X = D, T, Q).[197, 198] Whereas the Hartree Fock energy was extrapolated by an exponential fit

EHF (X) = EHF (CBS) + Bexp(−cX) (5.1) the correlation energy at the CBS limit was obtained using an expression the of the form

−3 Ecorr(X) = Ecorr(CBS) + BX . (5.2)

As to judge the influence of core-valence correlation on structures and frequencies at CCSD(T) level, further optimisations were done using the (aug)-cc-pwCVTZ-NR basis sets [denoted as (a)wCVTZ-NR] [198, 200] and corresponding smaller frozen cores (F: none, Cr–Co: 1s2sp). For the evaluation of core-valence effects on the thermochemistry, two single point CCSD(T)/awCVTZ//CCSD(T)/aVTZ calculations were done with both the small (sFC) and the large frozen core (lFC) and the difference ∆ECV = EsF C − ElF C was used as an additive correction. The calculations at DFT and ab initio level were done using the Gaussian09[193] and CFOUR[201] program packages respectively. 6 Oxides of Rh, Ir and Au

As mentioned in chapter 2, the highest oxidation state reached for a transition metal in its binary fluorides is VII. Higher oxidation numbers are however possible for those midtransition metals having enough electrons which can formally be oxidised. For these elements the stabilisation of the VIII oxidation state is possible with oxygen ligands, as for example in the well-known tetroxides of ruthenium and osmium. Furthermore, IrO4 was prepared by the reaction of laser-ablated iridium atoms with dioxygen and subsequently trapped in inert gas matrices. The assignment of its IR bands was supported VIII by quantum-chemical calculations which also showed that IrO4 is a genuine Ir compound with a d1 electron configuration at the iridium atom.[7] However, this molecule is a low temperature species and was predicted not to be stable in the condensed phase. For 2 rhodium, the analogous reaction with O2 only led to the formation of the [RhO2(η -O2)] complex with rhodium in its VI oxidation state.[202] After the successful preparation of iridium tetroxide, the stability of its lighter homologue RhO4 was reinvestigated and the results of the combined matrix isolation and theoretical investigation are subject of section 6.1. Having nine valence electrons, iridium could in principle be further oxidised. + The [IrO4] cation was already predicted to be a stable molecule which would represent the unprecedented IX oxidation state of iridium.[14] In order to support the gas phase characterisation of this cation by photodissociation IR spectroscopy[203, 204] of its argon complexes the vibrational spectra of the free cations as well as of their argon adducts + were calculated (section 6.2). A cation of [AuO2] stoichiometry was already detected in a mass spectrometric investigation[30] but no further information is available for this species. It is therefore not clear whether this cation is an example for the V oxidation state of gold and this question is discussed in section 6.3 on the basis of quantum-chemical calculations.

6.1 Rhodium tetroxide

The results presented in this section are part of a collaboration with professor Mingfei Zhou (Fudan University, Shanghai, China) and were already published in Y. Gong, M. Zhou, L. Andrews, T. Schl¨oder, S. Riedel, Theor. Chem. Acc. 2011, 129, 667. All calculations shown below were done by the author of this thesis whereas the matrix isolation experiments were carried out in the Zhou group. Hence, the latter will only be summarised briefly and more detailed information can be found in the given reference.

61 62 CHAPTER 6: OXIDES OF RH, IR AND AU

The highest oxidation state so far known for rhodium is VI, as for example in its hexafluoride [36] RhF6. Rhodium oxide molecules were already studied in a matrix isolation investigation 2 [205] in which the [RhO2(η -O2)] complex could be identified by IR spectroscopy. For its heavier homologue iridium, an analogous dioxide peroxide complex was characterised in an [206] early matrix isolation study while in a more recently investigation evidence for the IrO4 molecule with iridium in its VIII oxidation state could be found.[7] It was formed by the 2 rearrangement of the [IrO2(η -O2)] complex under λ > 500 nm irradiation and rhodium tetroxide might by synthesised in a similar way. However, this irradiation experiment was 2 not yet done when [RhO2(η -O2)] was prepared.

6.1.1 Matrix isolation experiments

Laser-ablated rhodium atoms were co-deposited with O2 diluted at 0.5% in argon or neon onto a CsI window cooled to 6 K (argon) and 4 K (neon) for deposition. In the IR spectra of the matrix samples measured directly after deposition, only the known bands of the RhO2 − and [RhO2] molecules could be observed. Annealing of the argon matrices to 35 K led to –1 a decrease of the RhO2 absorption while three new bands at 1116.5, 890.6 and 837.7 cm appeared. When the matrix samples where exposed to infrared irradiation of the source of the spectrometer, these new absorptions decreased simultaneously while the known bands 2 of the [RhO2(η -O2)] complex grew, and this process could be reversed by near infrared irradiation of λ > 850 nm. When lower laser energies were used for the ablation of rhodium, no RhO2 was formed, and the major absorptions in the resulting spectra were instead observed at 1048.4, 959.5, 922.3 and 551.8 cm–1. The two former bands were previously 2 2 assigned to the [Rh(η -O2)2] and [Rh(η -O2)] complexes, respectively. By contrast, the two latter absorptions were so far unknown and converted to the 1048.4 cm–1 band of 2 the [Rh(η -O2)2] complex upon infrared irradiation from the source of the spectrometer. When the matrix was subsequently irradiated with visible light, the 922.3 and 551.8 cm–1 absorptions increased again at the expense of the 1048.4 cm–1 band. Experiments were 18 repeated using O2 in order to facilitate product identification, and the new bands at –1 1116.5, 890.6 and 837.7 cm could be assigned to the previously unknown end-on O2 1 –1 complex [RhO2(η -O2)] whereas the absorptions at 922.3 and 551.8 cm were attributed 2 to a second isomer of [Rh(η -O2)2].

6.1.2 Structures

The structures of the different RhO4 and RhO2 isomers were optimised at DFT level and are shown in figure 6.1. For rhodium dioxide a bent, C 2v-symmetrical structure 2 was obtained for its A1 ground state and the calculated bond length using the B3LYP − functional is 167.6 pm. The corresponding [RhO2] anion was also observed in the spectra and therefore its structure was calculated, too. In its singlet ground state, this molecule 6.1. Rhodium tetroxide 63

O(2) O(1) O(2) O(1)

2 1 RhO4 [RhO2(η -O2)] [RhO2(η -O2)] 2 2 2 ’ (D2d, A1) (C2v, A2) (Cs, A)

O(1)

O(2)

1 2 2 [RhO2(η -O2)] [Rh(η -O2)2] [Rh(η -O2)2] 4 ’ 2 4 (Cs, A) (D2d, B1) (D2h, B1u)

– 2 RhO2 [RhO2] [Rh(η -O2)] 2 1 2 (C2v, A1) (C2v, A1) (C2v, A2)

Figure 6.1 : Optimised structures of molecular rhodium oxides; see table 6.1 for bond lengths and angles.

was found to have a C 2v-symmetrical structure, too, and the calculated bond length of 171.0 pm is 3.4 pm larger than the one in neutral RhO2. For a second isomer of this 2 latter molecule, namely the side-on O2 complex [Rh(η -O2)], a symmetrical cyclic structure 2 (point group C 2v) and a A2 ground state were calculated. The computed bond lengths in this molecule are d Rh–O = 192.5 pm and d O–O = 133.4 pm. These results are in good agreement with previous computational studies in which similar structures were obtained for these three molecules.[205, 207] By contrast, the rhodium tetroxide molecule was not [7] yet investigated. In analogy to its heavier homologue IrO4, it was calculated to have a 2 Jahn-Teller distorted, flattened tetrahedral structure of D 2d symmetry. In its A1 ground state, the bond length was computed to be 170.1 pm and thus significantly larger than that in RhO2. Several other isomers of RhO4 stoichiometry with rhodium in lower formal oxidation states were also taken into consideration. Of these, both the side-on and end-on [205] O2 complexes of RhO2 were already investigated previously and the obtained results are in good agreement with the literature data: For the experimentally known dioxide 2 2 peroxide complex [RhO2(η -O2)], a C 2v-symmetrical structure and a A2 ground state were calculated. In this molecule, the bond lengths were computed to be d Rh–O(1) = 169.1 pm, d Rh–O(2) = 191.3 pm as well as d O–O = 139.7 pm. In the case of the corresponding end-on 64 CHAPTER 6: OXIDES OF RH, IR AND AU

Table 6.1 : Calculated structural parameters of molecular rhodium oxides[a]

Molecule Parameter B3LYP[b] BP86[b] 2 RhO2 (C 2v, A1) d Rh–O 167.6 169.0 O-Rh-O 158.4 159.4 − 1 [RhO2] (C 2v, A1) d Rh–O 171.0 172.1 O-Rh-O 163.9 167.3 2 2 [Rh(η -O2)] (C 2v, A2) d Rh–O 192.5 193.3 d O–O 133.4 134.5 2 RhO4 (D2d, A1) d Rh–O 170.7 172.6 O-Rh-O 112.1 112.3 2 2 [RhO2(η -O2)] (C 2v, A2) dRh–O(1) 169.1 171.0 O(1)-Rh-O(1) 118.1 114.8 d Rh–O(2) 191.3 191.7 d O–O 139.7 141.5 1 2 0 [RhO2(η -O2)] (C s, A ) d Rh–O(1) 170.1 170.4 d Rh–O(2) 200.6 189.5 d O–O 125.8 132.2 Rh-O-O 101.4099.7 O(1)-Rh-O(1) 130.7 131.0 O(1)-Rh-O(2) 112.0 112.3 1 4 0 [RhO2(η -O2] (C s, A ) dRh–O(1) 170.4 172.8 d Rh–O(2) 198.1 197.1 d O–O 125.4 127.0 Rh-O-O 126.1 125.4 O(1)-Rh-O(1) 125.5 124.0 O(1)-Rh-O(2) 115.9 116.6 2 4 [Rh(η -O2)2] (D2h, B1u) d Rh–O 201.9 201.2 d O–O 130.9 133.6 2 2 [Rh(η -O2)2] (D2d, B1) d Rh–O 190.8 191.8 d O–O 135.6 137.5 [a] Bond lengths in pm, angles in ◦. [b] aVTZ(-PP) basis sets.

2 0 O2 complex, C s-symmetrical structures were found for both the lowest-lying doublet ( A ) and quartet (4A0 ) states which were calculated to have very similar total energies. As for other molecules,[208, 209] the high-spin state is favoured when the B3LYP hybrid functional is used (∆E = 3.7 kJ mol–1), whereas the use of the BP86 functional leads to a lower energy for the low-spin state (∆E = 25.1 kJ mol–1). The ground state of this molecule can thus not be reliably predicted by these calculations, but the spectroscopic data suggests it to be the doublet state (vide infra). In both states, the coordinated O2 molecule lies in a plane perpendicular to that of the RhO2 unit and the bond lengths in the doublet ground state were computed to be d Rh–O(1) = 170.1 pm, d Rh–O(2) = 200.6 pm and d O–O = 125.8 pm. For 2 the diperoxide complex [Rh(η -O2)2] too, two different minimum structures could be found 6.1. Rhodium tetroxide 65

1 on the doublet and quartet energy surfaces respectively. As for the [Rh(η -O2)] complexes, the relative stability of the two spin states depends on the chosen functional, and whereas the high-spin state is favoured in the B3LYP calculations, the low-spin state is preferred when BP86 is used. For the quartet state, a planar, D 2h-symmetrical minimum structure 4 (electronic state B1u) was obtained and the calculated bond length in this complex are d Rh–O = 201.9 pm and d O–O = 130.9 pm. By contrast, the two O2 units are twisted by ◦ 2 90 in the doublet minimum structure (D 2d symmetry, electronic state B1). Compared to the planar isomer, the Rh–O bond in this molecule is shortened by about 11 pm while the O–O bond is 4.7 pm longer, both values indicating more strongly bound O2 ligands.

6.1.3 Thermochemistry

The thermochemical stability of RhO4 was evaluated by calculating several possible decomposition reactions (table 6.2). Whereas all decomposition and internal reduction reactions were computed to be exothermic at DFT level using the B3LYP functional, rhodium tetroxide was found to be a stable molecule in the BP86 calculations. At CCSD(T) 2 level, both the elimination of O2 and the rearrangement to the [RhO2(η -O2)] complex were calculated to be exothermic by 4.0 and 26.7 kJ mol–1 respectively. The latter complex is thus the thermochemically most stable isomer of RhO4 stoichiometry. For comparison, the thermochemistry of the RhO2 molecule was also calculated and as expected, it was found to be thermochemically stable.

[a] Table 6.2 : Calculated thermochemistry of RhO4 and RhO2

Reaction B3LYP[b] BP86[b] CCSD(T)[b,c] 2 RhO4 → [RhO2(η -O2)] –47.9 (–49.4)018.40(17.5) –26.7 1 [d] RhO4 → [RhO2(η -O2)] –17.6 (–21.5)056.80(54.1)069.6 2 [e] RhO4 → [Rh(η -O2)2] –15.5 (–17.8) 134.9 (132.6)050.8 RhO4 → RhO2 + O2 –35.2 (–42.6) 104.00(96.8)0–4.0 2 RhO2 → [Rh(η -O2)] 184.7 (182.7) 227.8 (226.6) 213.8 RhO2 → Rh + O2 297.2 (294.3) 428.1 (425.1) 347.3 [a] Values in kJ mol–1; numbers in parentheses are ZPE corrected. [b] aVTZ(-PP) basis sets. 2 0 [c] Single point energies at B3LYP/aVTZ(-PP) structures. [d] A state. [e] D2h isomer.

6.1.4 Vibrational frequencies

The calculated and experimental wavenumbers of the 16O and 18O isotopomers of the different RhO2 and RhO4 isomers are listed in tables 6.3 and 6.4. Compared to experimental values, the calculated wavenumbers are consistently shifted to higher wavenumbers but the new bands could nevertheless be assigned based on the 16O to 18O isotopic shifts. The spin 66 CHAPTER 6: OXIDES OF RH, IR AND AU

Table 6.3 : Selected calculated and experimental wavenumbers of molecular rhodium oxides[a]

Calc. Exp. (matrix) Molecule Mode B3LYP[b] BP86[b] Ne Ar Ref. 2 RhO2 (C 2v, A1) B2 0986.0 (281)0962.3 (215)0908.60899.9 [205] A1 0957.6 00(7)0917.4 00(5) − 1 [RhO2] (C 2v, A1) B2 0948.9 (356)0921.9 (282)0898.60893.6 [207] A1 0887.50(12)0859.7 00(6) 2 2 [Rh(η -O2)] (C 2v, A2) A1 1072.1 (186) 1064.70(81)0975.80959.5 [202] A1 0547.2 00(2)0553.8 00(0) 2 RhO4 (D2d, A1) E 0912.9 (189)0881.5 (134) A1 0896.1 00(–)0852.0 00(–) B2 0882.30(99)0843.10(76) 2 2  [RhO2(η -O2)] (C 2v, A2) A1 1019.10(79)0968.40(84)0930.40928.6  A1 0951.00(14)0908.50(23)0869.70865.0 [205] B1 0874.6 (132)0851.6 (130)0838.80831.1  A1 0525.3 00(3)0537.4 00(1) 1 2 0 0  [RhO2(η -O2)] (C s, A ) A 1296.8 (256) 1061.1 (154) 1108.0 1116.5  A00 0938.9 (215)0893.7 (146)0897.70890.6 tw A0 0903.30(31)0865.1 (101)0845.70837.7  A0 0425.70(10)0554.80(11) 1 4 0 0 [RhO2(η -O2)] (C s, A ) A 1227.8 (658) 1221.0 (212) A0 0910.1 00(5)0867.50(14) A00 0866.5 (137)0851.8 (115) A0 0426.1 00(9)0454.5 00(1) 2 4 [c] [Rh(η -O2)2]2 (D2h, B1u) Ag 1216.8 00(–) 1052.3 00(6) B1u 1172.5 (302) 1061.3 (281) 1047.4 1048.4 [202] B1u 0444.50(17)0549.10(12) Ag 0415.7 00(–)0585.1 00(0) 2 2 [Rh(η -O2)2]2 (D2d, B1) A1 1084.3 00(–) 1036.4 00(–) B 0993.0 (912) 1018.7 (305)0929.10922.3  2 tw B2 0606.70(33)0585.5 00(6)0551.8 A1 0493.7 00(–)0481.7 00(–) [a] Values in cm–1 for 16O isotopomers; numbers in parentheses are infrared intensities in km mol–1. 4 [b] aVTZ-PP basis sets. [c] C 2v symmetry with BP86 (electronic state B2). tw = this work.

1 state of the [RhO2(η -O2)] complex could not be determined based on the DFT energy calculations. However, the calculated IR spectra of the two spin states show a decisive difference, namely the relative position of the symmetric (A0 ) and antisymmetric (A00 ) O-Rh-O stretching modes which can be distinguished in the spectra by their relative intensities. For the doublet state, both employed functionals predicted the A00 mode to have the higher wavenumber whereas the order is reversed for the quartet state. Based on the relative intensities of the two experimentally observed bands at 890.6 and 837.7 cm–1 6.1. Rhodium tetroxide 67

Table 6.4 : Calculated and experimental isotopic shifts of selected modes of molecular rhodium oxides[a] Calc. Exp. (matrix) Molecule Mode B3LYP[b] B3LYP[b] Ne Ar 2 16 RhO2 (C 2v, A1) B2 ( O)0986.0 00(0.0)0962.3 00(0.0)0908.6 00(0.0)0899.9 00(0.0) 18 B2 ( O)0942.8 (–43.2)0920.2 (–42.1)0869.5 (–39.1)0861.1 (–38.8) − 1 16 [RhO2] (C 2v, A1) B2 ( O)0948.9 00(0.0)0921.9 00(0.0)0898.6 00(0.0)0893.6 00(0.0) 18 B2 ( O)0907.7 (–41.3)0881.1 (–40.2)0859.9 (–38.7)0855.3 (–38.3) 2 2 16 [Rh(η -O2)] (C 2v, A2) A1 ( O) 1072.1 00(0.0) 1064.7 00(0.0)0975.8 00(0.0)0959.5 00(0.0) 18 A1 ( O) 1011.2 (–60.9) 1004.0 (–60.7)0923.1 (–42.7)0907.9 (–51.6) 2 2 16 [RhO2(η -O2)] (C 2v, A2) A1 ( O) 1019.1 00(0.0)0968.4 00(0.0)0930.4 00(0.0)0928.6 00(0.0) 18 A1 ( O)0962.2 (–56.9)0914.3 (–54.1)0880.1 (–50.3)0878.7 (–49.9) 16 A1 ( O)0951.0 00(0.0)0908.5 00(0.0)0869.7 00(0.0)0865.0 00(0.0) 18 A1 ( O)0900.6 (–50.4)0861.2 (–47.3)0826.7 (–43.0)0820.3 (–44.7) 16 B1 ( O)0874.6 00(0.0)0851.6 00(0.0)0838.8 00(0.0)0831.1 00(0.0) 18 B1 ( O)0833.1 (–41.5)0811.0 (–40.6)0799.6 (–39.2)0792.1 (–39.0) 1 2 0 0 16 [RhO2(η -O2)] (C s, A ) A ( O) 1296.8 00(0.0) 1061.1 00(0.0) 1108.0 00(0.0) 1116.5 00(0.0) A0 (18O) 1222.5 (–74.3) 1000.0 (–61.1) 1045.4 (–62.6) 1053.2 (–63.3) A00 (16O)0938.9 00(0.0)0893.7 00(0.0)0897.7 00(0.0)0890.6 00(0.0) A00 (18O)0896.4 (–42.5)0852.8 (–40.9)0858.1 (–39.6)0851.3 (–39.3) A0 (16O)0903.3 00(0.0)0865.1 00(0.0)0845.7 00(0.0)0837.7 00(0.0) A0 (18O)0854.6 (–48.7)0818.3 (–46.8)0799.8 (–45.9)0792.1 (–45.6) 1 4 0 0 16 [RhO2(η -O2)] (C s, A ) A ( O) 1227.8 00(0.0) 1221.0 00(0.0) A0 (18O) 1157.6 (–70.2) 1151.0 (–70.0) A0 (16O)0910.7 00(0.0)0851.8 00(0.0) A0 (18O)0861.1 (–49.0)0812.5 (–39.3) A00 (16O)0866.5 00(0.0)0867.5 00(0.0) A00 (18O)0826.5 (–40.0)0821.2 (–46.4) 2 4 [c] 16 [Rh(η -O2)2]2 (D2h, B1u) B1u ( O) 1172.5 000.0 1061.3 00(0.0) 1047.4 00(0.0) 1048.4 00(0.0) 16 B1u ( O) 1105.5 (–67.0) 1001.1 (–60.2)0989.2 (–58.2)0990.0 (–58.1) 2 2 16 [Rh(η -O2)2]2 (D2d, B1) B2 ( O)0993.0 00(0.0) 1018.4 00(0.0)0929.1 00(0.0)0922.3 00(0.0) 16 B2 ( O)0938.6 (–54.4)0962.4 (–56.3)0879.6 (–49.5)0873.3 (–49.0) 16 B2 ( O)0606.7 00(0.0)0585.5 00(0.0)0551.8 00(0.0) 16 B2 ( O)0584.0 (–22.7)0563.9 (–21.6)0531.3 (–20.5) [a] Values in cm–1; in parentheses: isotopic shifts compared to the 16O isotopomers. [b] aVTZ(-PP) 4 basis sets. [c] C 2v symmetry with BP86 (electronic state B2).

1 in argon matrices, the [RhO2(η -O2)] molecule probably has a doublet ground state. The –1 bands at 922.3 and 551.8 cm in argon matrices were assigned to the D 2d-symmetrical 2 –1 isomers of [Rh(η -O2)2]2 for which two IR active bands above 500 cm were calculated whereas the known D 2h-symmetrical isomer only shows one band in this spectral region. Finally, no evidence could be found in the spectra for the RhO4 molecule which is in line 68 CHAPTER 6: OXIDES OF RH, IR AND AU with its predicted thermochemical instability but kinetic reasons preventing its formation can however not be excluded with certainty.

6.1.5 Summary

Unlike its heavier homologue iridium tetroxide, the RhO4 molecule with rhodium in its VIII oxidation state was calculated to be thermochemically unstable with respect to an internal 2 reduction reaction yielding the known rhodium dioxide peroxide complex [RhO2(η -O2)] with rhodium in the oxidation state VI. In line with this prediction, no evidence for RhO4 could be found in the matrix IR spectra. Instead, two other new product molecules could be identified, namely an end-on O2 complex of RhO2 as well as a second isomer of the 2 known [Rh(η -O2)2] complex, and the assignment of the bands to these new species was supported by quantum-chemical calculations. The former complex can be converted into the corresponding known side-on O2 complex by infrared irradiation from the source of the spectrometer. The second new molecule is a D 2d-symmetrical isomer of the diperoxide 2 complex [Rh(η -O2)2] which can be transformed to its known planar isomer under the same conditions. These two reactions can furthermore be reversed by λ > 850 nm and λ > 500 nm irradiation respectively.

6.2 The iridium tetroxide cation

Recently, the new VIII oxidation state of iridium could be stabilised in IrO4, the third transition metal tetroxide after the well-known RuO4 and OsO4 molecules. Contrary to ruthenium and osmium, iridium has nine valence electrons and could in principle be further oxidised. It was already predicted by Jørgensen, that this element would have the greatest chance of being oxidised beyond the VIII oxidation state.[13] The formal removal + of the remaining d electron from iridium tetroxide would lead to the [IrO4] cation with iridium in the IX oxidation state. This species was already predicted to be stable in the gas phase and, if combined with a weakly coordinating anion, in the solid state as [14] + well. As a gaseous cation, [IrO4] could be observed using mass spectrometry but the detection of a signal at the corresponding mass alone would not provide sufficient evidence for the formation of this species as it does not allow to distinguish between the different possible isomers. As to circumvent this problem, the IR spectrum of the ion beam could be measured using photodissociation IR spectroscopy of the corresponding argon complexes.

6.2.1 Structures

+ The optimised structures of the three possible [IrO4] isomers are shown in figure 6.2. As expected, and in analogy to the isoelectronic OsO4 molecule, the iridium tetroxide 6.2. The iridium tetroxide cation 69

O (2) O (1) O (1)

O (2)

+ 2 + 1 + [IrO4] [IrO2(η -O2)] [IrO2(η -O2)] 1 1 3 ” (Td, A1) (C2v, A1) (Cs, A )

+ Figure 6.2 : Optimised structures of different [IrO4] isomers; see table 6.5 for bond lengths and angles.

+ Table 6.5 : Calculated structural parameters of different [IrO4] isomers and their argon complexes[a]

CCSD(T)[b] B3LYP-D3[b] Molecule[c] Parameter[d] n = 0 n = 1 n = 0 n = 1 n = 2 n = 3 + 1 [IrO4 · Arn] (T d, A1) d Ir–O 170.8 170.8 168.6 168.6 168.7 168.6 d Ir–Ar 352.2 351.9 354.1 354.5 2 + 1 [IrO2(η -O2) · Arn] (C 2v, A1) d Ir–O(1) 168.0 166.5 166.5 166.5 166.5 O(1)-Ir-O(1) 119.4 119.9 120.1 120.9 121.2 d Ir–O(2) 188.8 188.4 188.3 188.3 188.1 d O–O 141.2 139.4 139.6 139.8 139.9 d Ir–Ar 352.1 350.0 358.6 1 + 3 00 [IrO2(η -O2) · Arn] (C s, A ) d Ir–O(1) 167.9 166.5 166.6 166.7 167.0 d Ir–O(2) 209.6 207.4 208.5 209.7 210.0 d O–O 122.1 121.2 121.2 121.3 121.4 O(1)-Ir-O(1) 136.2 137.4 138.7 140.9 146.4 O(1)-Ir-O(2) 111.9 111.2 110.6 109.5 106.8 O-O-Ir 124.0 129.4 127.5 125.4 123.6 d Ir–Ar 344.0 343.5 339.7 [a] Averaged bond lengths in pm, angles in ◦. [b] aVTZ(-PP) basis sets. [c] Point group and state for the free cations. [d] See figure 6.2 for atom labelling.

1 cation has a tetrahedral structure and a singlet A1 ground state. At CCSD(T) level, the iridium-oxygen distance was calculated to be 170.8 pm, a value which is about 2 pm larger than the one obtained at DFT level using the B3LYP functional. For the side-on O2 2 + complex [IrO2(η -O2)] , a C 2v-symmetrical minimum structure was calculated and the 1 molecule was found to have a singlet A1 ground state. The computed bond lengths in this isomer are d Ir–O(1) = 168.0 pm and d Ir–O(2) 188.8 pm for the Ir–O bonds as well as 141.2 pm for the peroxide bond. In the C s-symmetrical minimum structure of the end-on O2 complex 1 + [IrO2(η -O2)] , the coordinated O2 molecule lies in a plane perpendicular to that of the + 3 00 [IrO2] unit. This molecule was found to have a triplet A electron configuration and 70 CHAPTER 6: OXIDES OF RH, IR AND AU

Ir–O bond lengths of d Ir–O(1) = 167.9 pm and d Ir–O(2) = 209.6 pm. The O–O distance of d O–O = 122.1 pm was computed to be only slightly larger than in the free O2 molecule (d O–O = 121.3 pm at the same level of theory). The structures of the argon complexes of + the different [IrO4] isomers with up to three argon atoms attached to each of the three + [IrO4] isomers were optimised at dispersion corrected DFT level as well. No significant + changes occur in the [IrO4] substructures of the complexes when comparing the latter with the free cations (table 6.5). The validity of these DFT calculations is confirmed by a + structure optimisation at CCSD(T) level for [IrO4 · Ar] which also led to a structure in which the Ir–O bond length is the same as in the free cation. Further evidence for the + weak interaction between the argon atoms and the different [IrO4] isomers is provided by the large computed iridium-argon distances of d Ir–Ar > 335 pm.

6.2.2 Thermochemistry

The thermochemical stability of the iridium tetroxide cation was evaluated by calculating the possible internal reduction pathways as well as decomposition reactions involving the loss of oxygen (table 6.6). All possible decomposition reactions were found to be + endothermic and the [IrO4] cation with iridium in the IX oxidation state is thus a thermochemically stable species. As compared to the CCSD(T) results, the stability of + [IrO4] is underestimated at B3LYP level. Despite the considerable differences between the structures obtained at the two levels of theory, the computed thermochemistry at CCSD(T) level is nevertheless well reproduced by calculating single point CCSD(T) energies at the structures optimised at B3LYP level. Furthermore, the effect of argon complexation on the thermochemistry was also calculated. Since no significant changes in the energetic ordering + of the three isomers could be observed, the [IrO4] molecule should be an observable species as an argon complex, too. The computed complexation energies for the attachment + of the argon atoms to the [IrO4] isomers are listed in table 6.7. Each complexation step leads to a release of a about 10 kJ mol–1 for either isomer, underlining the weak interaction + between the [IrO4] cations and the argon atoms. It can furthermore be noted that the

+ [a] Table 6.6 : Calculated thermochemistry of [IrO4] and its argon complexes

B3LYP-D3[b] CCSD(T)[b,c] Reaction n = 0 n = 1 n = 2 n = 3 n = 0 n = 1 n = 2 + + [IrO4 · Arn] → [IrO2 · Arn] + O2 200.9 239.0 (241.6) + + [IrO4 · Arn] → [IrO3 · Arn] + O 282.9 296.9 (299.5) + 2 + [IrO4 · Arn] → [IrO2(η -O2) · Arn] 059.6 60.1 60.7 63.8086.30(89.2)086.7087.9 + 1 + [IrO4 · Arn] → [IrO2(η -O2) · Arn] 075.2 75.8 77.2 78.2 126.8 (129.0) 129.1 131.8 [a] Values in kJ mol–1. [b] aVTZ(-PP) basis sets. [c] Single point energies at B3LYP- D3/aVTZ(-PP) structures; in parentheses: CCSD(T) optimisations. 6.2. The iridium tetroxide cation 71

+ [a] Table 6.7 : Calculated argon complexation energies of the different [IrO4] isomers

Reaction B3LYP-D3[b] CCSD(T)[b,c] + + [IrO4] + Ar → [IrO4 · Ar] –14.4 (–13.5) –12.5 2 + 2 + [IrO2(η -O2)] + Ar → [IrO2(η -O2) · Ar] –13.8 (–12.8) –12.0 1 + 1 + [IrO2(η -O2)] + Ar → [IrO2(η -O2) · Ar] –13.8 (–13.2) –10.2 + + [IrO4 · Ar] + Ar → [IrO4 · Ar2] –13.9 (–13.2) –12.3 2 + 2 + [IrO2(η -O2) · Ar] + Ar → [IrO2(η -O2) · Ar2] –13.3 (–12.8) –11.2 2 + 2 + [IrO2(η -O2) · Ar] + Ar → [IrO2(η -O2) · Ar2] –12.5 (–12.3)0–9.6 + + [IrO4 · Ar2] + Ar → [IrO4 · Ar3] –13.7 (–12.8) 2 + 2 + [IrO2(η -O2) · Ar2] + Ar → [IrO2(η -O2) · Ar3] –10.70(–9.9) 2 + 2 + [IrO2(η -O2) · Ar2] + Ar → [IrO2(η -O2) · Ar3] –12.7 (–11.5) [a] Energies in kJ mol–1; values in parentheses are ZPE corrected. [b] aVTZ(-PP) basis set combination. [c] Single point energies at B3LYP-D3/aVTZ(-PP) structures. values obtained at B3LYP-D3 level are very similar to those calculated at CCSD(T) level. The interaction between the complex cations and the argon atoms is thus well described by dispersion-corrected density functional theory.

6.2.3 Vibrational frequencies

+ 2 + 2 + The calculated vibrational spectra of [IrO4] , [IrO2(η -O2)] and [IrO1(η -O2)] cations are given in table 6.8 for both 16O and 18O isotopomers. The computed wavenumbers at CCSD(T) level substantially differ from those obtained at B3LYP level and these deviations correlate with the significantly different minimum structures, particularly the different bond lengths obtained using the two different methods. However, despite these deviations the isotopic ratios are the same at both levels of theory. The vibrational spectra of the argon complexes were also computed and the results are summarised in table 6.9. As expected from the very small structural changes between the free cations and their argon complexes, the calculated wavenumbers of the stretching vibrations of the different + [IrO4] cations are only marginally affected by argon coordination. It can thus be deduced that the computed wavenumbers for the free cations are a good approximation for those of the corresponding argon complexes.

6.2.4 Summary

+ The structures and vibrational frequencies for three different isomers of [IrO4] stoichio- metry were calculated at the CCSD(T) level of theory. In agreement with ref. [14], the true iridium tetroxide cation with iridium in the oxidation state IX was found to be stable against both internal reductions and elimination of O2. Argon atoms were calculated to 72 CHAPTER 6: OXIDES OF RH, IR AND AU

+ [a] Table 6.8 : Selected calculated wavenumbers of different [IrO4] isomers

B3LYP[b] CCSD(T)[b] Molecule Mode 16O 18O 16O 18O + 1 [IrO4] (T d, A1) A1 1007.1 00(–)0949.3 916.40(–)0863.9 T2 1008.1 (118)0956.3 963.0 (35)0913.5 2 + 1 [IrO2(η -O2)] (C 2v, A1) A1 1031.60(60)0973.5 954.9 (55)0901.2 B1 1025.00(72)0972.8 996.5 (56)0945.8 A1 1060.40(12) 1001.8 1000.6 (11)0945.6 1 + 3 00 00 [IrO2(η -O2)] (C s, A ) A 1023.4 (133)0972.2 976.9 (86)0927.9 A0 1065.7 00(3) 1006.1 999.40(0)0943.6 A0 1486.30(25) 1401.1 1458.7 (25) 1375.1 [a] Values in cm–1 for 193Ir isotopomers; numbers in parentheses are infrared intensities in km mol–1. [b] aVTZ(-PP) basis sets.

+ Table 6.9 : Selected calculated wavenumbers of the argon complexes of the different [IrO4] isomers[a,b] Molecule[c] Mode n = 0 n = 1 n = 2 n = 3 + 1 [IrO4] · Arn (T d, A1) T2 1008.50(41) 1007.50(43) 1007.20(43) 1007.20(45) T2 1008.50(41) 1009.00(39) 1007.60(37) 1007.20(45) T2 1008.50(41) 1009.00(39) 1009.40(35) 1010.10(29) A1 1009.8 00(–) 1008.9 00(0) 1008.5 00(5) 1008.7 00(8) 2 + 1 [IrO2(η -O2) · Arn] (C 2v, A1) B2 1026.80(71) 1025.00(72) 1023.90(72) 1024.00(74) A1 1032.80(60) 1027.90(59) 1026.60(55) 1027.50(58) A1 1062.70(12) 1061.90(19) 1059.30(13) 1061.60(13) 1 + 3 00 00 [IrO2(η -O2) · Arn] (C s, A ) A 1026.3 (133) 1023.7 (137) 1024.10(136) 1023.0 (137) A0 1067.4 00(3) 1065.4 00(3) 1063.5 00(3) 1059.5 00(1) A0 1489.20(22) 1483.50(26) 1477.90(34) 1476.20(46) [a] Calculated at B3LYP-D3/aVTZ(-PP) level. [b] Values in cm–1; numbers in parentheses are infrared intensities in km mol–1. [c] Point group and symmetry labels for the free cations.

+ only weakly interact with each of the [IrO4] isomers. Accordingly, the effect of argon + complexation on both thermochemistry and the vibrational spectra of the different [IrO4] complexes is negligible. Argon thus represents an ideal messenger atom for the photodis- sociation IR spectroscopic characterisation of the iridium tetroxide cation which should in principle be observable as the most stable species of this composition. However, if the + 2 + formation of [IrO4] proceeds by a rearrangement of an initially formed [IrO2(η -O2)] complex, a kinetic barrier to the reorganisation needs to be overcome. As this barrier was [14] + calculated to be very high, the formation of [IrO4] might be hindered kinetically and the molecule might not be detected even though it is thermochemically the most stable isomer. 6.3. The gold dioxide cation 73 6.3 The gold dioxide cation

The highest known oxidation state of gold is V, as represented by the well-known neutral − [36] + AuF5 molecule and the corresponding [AuF6] anion. In 1987 a monocation of [AuO2] stoichiometry was observed in mass spectrometric experiments after sputtering of a gold [30] target in an Ar-O2 discharge. This cation was later mentioned as an example for the V oxidation state of gold stabilised in a binary oxide compound[36] although the structure of this species was not investigated. However, isomers containing gold atoms in lower oxidation states cannot be excluded a priori, especially when considering that the AuO I [210] + molecule is best described as a Au species. As for the [IrO4] isomers, the experimental distinction between the true gold dioxide cation and possible internal reduction products could be made by measuring the photodissociation IR spectra of the corresponding argon complexes.

6.3.1 Structures

+ Three different isomers are possible for a cation with [AuO2] stoichiometry: the end-on and 1 + 2 + side-on O2 complexes [Au(η -O2)] and [Au(η -O2)] as well as a molecule with an O-Au-O atom connectivity. The structures of all three molecules were optimised at CCSD(T) level + and the results are shown in figure 6.3. For the [OAuO] isomer a linear structure (D ∞h 1 + symmetry) and a singlet Σg ground state were calculated at coupled cluster level whereas

+ 2 + 1 + [AuO2] [Au(η -O2)] [Au(η -O2)] 1 + 1 3 ’’ (D∞h, Σg ) (C2v, A1) (Cs, A )

2 + 2 + 1 + [Au(η -O2) · Ar] [Au(η -O2) · Ar2] [Au(η -O2) · Ar] 1 1 3 ’’ (C2v, A1) (C2v, A1) (Cs, A )

+ Figure 6.3 : Optimised structures of different [AuO2] isomers and their argon complexes; see table 6.10 for bond lengths and angles.. 74 CHAPTER 6: OXIDES OF RH, IR AND AU

+ Table 6.10 : Calculated structural parameters of different [AuO2] isomers and their argon complexes[a]

CCSD(T)[b] CASPT2[b,c] B3LYP-D3[b] Molecule[d] Parameter n = 0 n = 0 n = 0 n = 1 n = 2 n = 3 0 + 1 + [AuO2 · Arn] (D∞h, Σg ) d Au–O 181.2 175.6 171.0 171.7 172.3 172.5 0 d Au–Ar 375.3 374.6 373.5 0 d O–Ar 266.2 268.6 274.8 + 3 + [AuO2] (D∞h, Σu ) d Au–O 190.2 179.9 182.6 + 3 [AuO2] (D∞h, ∆u) d Au–O 184.4 177.7 183.8 + 5 + [AuO2] (D∞h, Σg ) d Au–O 198.6 194.5 197.4 2 + 1 [Au(η -O2)] (C 2v, A1) d Au–O 226.0 211.8 d O–O 126.4 128.3 2 + 1 [ArAu(η -O2)] (C 2v, A1) d Au–O 208.5 208.0 d O–O 129.8 129.1 d Au–Ar 244.2 247.3 2 0 + 1 [Ar2Au(η -O2) · Arn] (C 2v, A1)d Au–O 203.9 205.6 206.0 d O–O 132.5 130.8 130.7 d Au–Ar 256.5 261.5 262.6 Ar-Au-Ar 086.6086.3086.3

0 d Au–Ar 330.2 1 + 3 00 [Au(η -O2)] (C s, A ) d Au–O 226.6 225.1 d O–O 121.2 120.4 Au-O-O 122.1 124.1 1 0 + 3 00 [ArAu(η -O2) · Arn] (C s, A ) d Au–O 216.2 215.8 217.8 220.3 d O–O 121.4 121.0 121.0 120.9 d Au–Ar 243.8 248.1 250.6 253.5 Au-O-O 121.2 124.0 123.4 123.0 Ar-Au-O 178.4 178.0 178.4 177.6 0 d Au–Ar 324.1 323.8 [a] Averaged bond lengths in pm, angles in ◦. [b] aVTZ(-PP) basis sets. [c] CAS(6,8). [d] Point group and state for the free cations; Ar0 denotes the weakly coordinated argon atoms.

5 at DFT level a bent structure and a quintet A1 ground state were obtained. In the CCSD(T) calculations the highest double amplitudes were |T2| = 0.248 for the excitations from the non-bonding πu to the O-Au-O antibonding πg orbitals, indicating multi-reference character of the wavefunction. Therefore, additional CAS(6,8)PT2 calculations were done in which the active space consisted of the bonding, non-bonding and antibonding O-Au-O π orbitals (πu, 2 × πg). The resulting CI coefficient of the leading configuration 2 (c0 = 0.610) confirmed the multi-reference character of the wavefunction. In the closed shell singlet configuration the πu orbitals and are fully occupied while the πg antibonding orbitals are empty, but the energy difference between these two sets of orbitals is very 6.3. The gold dioxide cation 75

small. Therefore other configurations arising from virtual πg ← πu excitations as accounted for in the CCSD(T) and even more so the CASPT2 calculations significantly contribute to the total singlet wavefunction. Because of the small energy gap between the frontier orbitals a high-spin quintet state in which 4 electrons are evenly distributed over the 5 + 3 3 + πu and πg orbitals ( Σg ) as well as two intermediate triplet states ( ∆u and Σu ) were also calculated. Whereas the two triplet wavefunctions also have significant multi-reference character, only one CSF contributes to the quintet wavefunction. At both CCSD(T) and CASPT2 level, linear structures were calculated for the three additional states which were all found to lie higher in energy than the singlet ground state (table 6.11). By contrast, 1 + the quintet state is preferred at DFT level which underestimates the stability of the Σg state. This can be explained by the accounting for non-leading configurations at CCSD(T) and CASPT2 level which leads to a stabilisation of the singlet state. Furthermore, the minimum structures obtained at DFT level for the triplet and quintet states were found to + have only C 2v symmetry. The calculated bond length in [AuO2] strongly depends on the level of theory and the electronic state (table 6.10). In general, it was found to increase with increasing population of the antibonding πg orbitals in the intermediate and high-spin states as well as in the CCSD(T) and CASPT2 calculations of the low-spin ground state. Due to the multi-reference character of the wavefunction, the best estimate for the bond length is probably given by the CASPT2 calculations which explicitly account for both static and dynamic electron correlation and which yielded a value of d Au–O = 175.6 pm. 2 + 1 + By contrast, the electronic structures of the [Au(η -O2)] and [Au(η -O2)] isomers are much simpler as both molecules have single-reference wavefunctions. The C 2v-symmetrical 1 side-on O2 complex has a singlet A1 ground state and the calculated bond lengths are d Au–O = 226.0 pm and d O–O = 126.4 pm at CCSD(T) level. For the end-on O2 complex, a 3 00 C s-symmetrical structure and a triplet A ground state were calculated. The computed bond lengths in this isomer are 226.6 pm for the gold-oxygen bond and 121.2 pm for the oxygen-oxygen bond, a value which is only very slightly smaller than that for the free O2 1 + molecule (d O–O = 121.3 pm at the same level of theory). The [Au(η -O2)] complex can + thus be described as a Au cation coordinated by an almost unperturbed O2 molecule.

Table 6.11 : Calculated energy differences between different electronic states of [OAuO]+ [a]

Point group State CCSD(T)[b] CASPT2[b,c] B3LYP-D3[b] 5 + [d] D∞h Σg 086.4 151.1 –46.8 5 [d] C 2v A1 –55.4 3 [d] D∞h ∆u 210.2 175.8072.1 3 [d] C 2v A2 019.5 3 + [d] D∞h Σu 101.5 102.9019.7 3 [d] C 2v B2 007.2 1 + [d] D∞h Σg 000.0 000.0 000.0 [a] Values in kJ mol–1. [b] aVTZ-PP basis sets. [c] CAS(6,8). [d] no minimum structure. 76 CHAPTER 6: OXIDES OF RH, IR AND AU

+ The argon complexes of the different [AuO2] isomers with up to three argon atoms were optimised at dispersion corrected DFT level, too (table 6.10). Only the singlet ground state, as determined by the higher level ab initio calculations, was considered for the [OAuO]+ isomer, and the in silico coordination of the argon atoms did not lead + to considerable structural changes of the [AuO2] substructure. In these complexes the argon atoms coordinate to the oxygen ligands rather than to the central gold atom and the large gold-argon as well as oxygen-argon distances indicate a weak interaction. By 1 + contrast, one argon atom strongly binds to the [Au(η -O2] complex, leading to the 1 + [ArAu(η -O2] cation with a gold-argon distance of 243.8 pm. This result is in agreement with the previously found strong interaction between monocoordinated AuI compounds and argon atoms.[66, 67, 211–214] Compared to the corresponding free cation, the Au–O bond is shortened by more than 10 pm in the argon complex while the effect is small for the 2 + O–O bond. In the case of [Au(η -O2] , small gold-argon distances were calculated for both the monoargon and the diargon complex. The argon complexation furthermore affects the computed O–O bond length which increases with the number of coordinated argon atoms. The attachment of more argon atoms did not lead to further changes in the structures 2 + 1 + of the [Ar2Au(η -O2)] and [ArAu(η -O2)] complexes and the Au–Ar distances were calculated to be much larger for the additional argon atoms.

6.3.2 Thermochemistry

+ The calculated thermochemistry of the different [AuO2] isomers at CCSD(T) level shows 1 + the end-on O2 complex [Au(η -O2)] with gold in the oxidation state I to be the only isomer which eliminates O2 endothermicly (table 6.12). However, the computed reaction energy + for the O2 elimination from [AuO2] is somewhat flawed as coupled cluster theory does not adequately account for the multi-reference character of the corresponding wavefunction. Nevertheless, the inclusion of a correct description of static correlation in the calculations would probably not change the overall exothermicity of the reaction. The effect of argon coordination on the thermochemistry was also calculated and contrary to the case of + + the [IrO4] cations, the decomposition energies of the [AuO2] isomers are considerably

1 + [a] Table 6.12 : Calculated thermochemistry of [Au(η -O2)] and its argon complexes

B3LYP[a] CCSD(T)[b] Reaction n = 0 n = 1 n = 2 n = 3 n = 0 n = 1 n = 2 1 + + [Au(η -O2) · Arn] → [Au · Arn] + O2 0059.6 0069.1 0026.8 0026.3 0044.300(44.4)0059.7 0014.7 2 + + [Au(η -O2) · Arn] → [Au · Arn] + O2 –126.6 –102.0 –125.2 –125.1 –114.8(–113.5)0–88.0 –110.4 + + [AuO2 · Arn] → [Au · Arn] + O2 –266.3 –296.2 –336.7 –338.4 –176.1(–156.0) [a] Values in kJ mol–1. [b] Single point energies at B3LYP-D3/aVTZ(-PP) structures; in paren- theses CCSD(T) optimisations. 6.3. The gold dioxide cation 77

+ + [a] Table 6.13 : Calculated argon complexation energies of Au and the different [AuO2] isomers

Reaction B3LYP-D3[b] CCSD(T)[b,c] CCSD(T)[b] Au+ + Ar → [AuAr]+ –50.1 (–49.2) –44.0 –44.1 (–43.2) + + [AuAr] + Ar → [AuAr2] –57.9 (–56.1) –57.1 –57.5 (–55.5) + + [AuAr2] + Ar → [AuAr3] –15.8 (–15.5) –11.8 –11.8 (–12.7) + + [AuO2] + Ar → [AuO2 · Ar] –20.3 (–19.1) + + [AuO2 · Ar] + Ar → [AuO2 · Ar2] –17.4 (–16.2) + + [AuO2 · Ar2] + Ar → [AuO2 · Ar3] –14.2 (–13.3) 2 + 2 + [Au(η -O2)] + Ar → [ArAu(η -O2)] –74.6 (–72.4) –70.8 –69.7 (–66.4) 2 + 2 + [ArAu(η -O2)] + Ar → [Ar2Au(η -O2)] –34.8 (–32.5) –34.6 –35.4 (–32.5) 2 + 2 + [Ar2Au(η -O2)] + Ar → [Ar2Au(η -O2) · Ar] –16.0 (–15.3) 1 + 1 + [Au(η -O2)] + Ar → [ArAu(η -O2)] –59.6 (–57.4) –59.4 –59.7 (–57.5) 1 + 1 + [ArAu(η -O2)] + Ar → [ArAu(η -O2) · Ar] –15.6 (–15.1) –12.0 1 + 1 + [ArAu(η -O2) · Ar] + Ar → [ArAu(η -O2) · Ar2] –15.4 (–14.7) [a] Energies in kJ mol–1; values in parentheses are ZPE corrected. [b] aVTZ(-PP) basis sets. [c] Single point energies at B3LYP-D3/aVTZ(-PP) structures.

1 + changed by argon complexation. Although all calculated argon adducts of [Au(η -O2)] were found to be stable against O2 elimination, the highest stability was computed for the 1 + monoargon complex [ArAu(η -O2)] . This can be explained by the different complexation 1 + + energies for [Au(η -O2)] and the free Au cation which are listed in table 6.13. As deferred from the significant structural changes, the first argon atom strongly binds to both the end-on and side-on O2 complex. By contrast, the second argon atom is only weakly 1 + 2 + bound to [ArAu(η -O2)] whereas the interaction between argon and [ArAu(η -O2)] is 1 + stronger. Compared to those of [Au(η -O2)] , the energies of argon complexation of the free Au+ cation are smaller for the first and larger for the second argon atom, explaining 1 + the increased relative stability of the [ArAu(η -O2)] complex. The respective third argon + atoms only weakly coordinate to the end-on and side-on O2 complexes as well as to Au and do therefore not influence the thermochemical stabilities. Finally, the interaction + between the [AuO2] isomer and the argon atoms was calculated to be only weak.

6.3.3 Vibrational frequencies

+ 16 The vibrational spectra of the different [AuO2] isomers were calculated for both O and 18O isotopomers in order to support their characterisation by IR spectroscopy (table 6.14). + The calculated wavenumbers of the [AuO2] isomer strongly depend on the employed level of theory (table 6.14) and as for the structures of this molecule the best estimate of its vibrational frequencies is probably given by the CASPT2 calculations. By contrast, the frequency calculations at both DFT and CCSD(T) level are much more reliable for 1 + 2 + [Au(η -O2)] and [Au(η -O2)] as well as for their argon complexes. The calculated 78 CHAPTER 6: OXIDES OF RH, IR AND AU

+ Table 6.14 : Selected calculated wavenumbers of different [AuO2] isomers and their strongly bound argon complexes[a]

B3LYP[b] CCSD(T)[b] CASPT2[b,c] Molecule Mode 16O 18O 16O 18O 16O 18O + 1 + + [AuO2] (D∞h, Σg ) Σg 0907.70(–)0855.70424.50(–)0400.2 1265.4 (–) 1192.9 − Σu 1031.3 (10)0980.70756.10(6)0719.00897.2 (1)0851.6 2 + 1 [Au(η -O2)] (C 2v, A1) A1 1266.50(8) 1194.0 1302.60(3) 1228.0 2 + 1 [ArAu(η -O2)] (C 2v, A1) A1 1251.1 (16) 1179.5 1192.7 (18) 1124.5 2 + 1 [Ar2Au(η -O2)] (C 2v, A1) A1 1198.2 (32) 1129.7 1120.6 (30) 1056.3 1 + 3 00 0 [Au(η -O2)] (C s, A ) A 1551.7 (28) 1462.7 1629.2 (24) 1535.9 1 + 3 00 0 [ArAu(η -O2)] (C s, A ) A 1541.6 (14) 1453.3 1599.5 (19) 1507.8 [a] Values in cm–1; numbers in parentheses are infrared intensities in km mol–1. [b] aVTZ(-PP) basis sets. [c] CAS(6,8).

+ Table 6.15 : Selected calculated wavenumbers of the argon complexes of the different [AuO2] isomers[a,b] Molecule[c] Mode n = 0 n = 1 n = 2 n = 3 0 + 1 + + [AuO2 · Arn] (D∞h, Σg ) Σg 0908.30(–)0882.6 (42)0858.8 (26)0849.4 (13) − Σu 1031.90(9) 1014.70(5)0999.40(5)0994.20(2) 2 + 1 [Au(η -O2)] (C 2v, A1) A1 1269.20(8) 2 + 1 [ArAu(η -O2)] (C 2v, A1) A1 1252.0 (16) 2 0 + 1 [Ar2Au(η -O2) · Arn] (C 2v, A1) A1 1198.2 (32) 1202.1 (33) 1 + 3 00 0 [Au(η -O2)] (C s, A ) A 1558.1 (26) 1 0 + 3 00 0 [ArAu(η -O2) · Arn] (C s, A ) A 1545.7 (13) 1546.6 (13) 1551.8 (14) [a] Calculated at B3LYP-D3/aVTZ(-PP) level [b] Values in cm–1; numbers in parentheses are infrared intensities in km mol–1. [c] Point group and symmetry labels for the free cations.

–1 wavenumber of the O–O stretching vibration of the side-on O2 complex is 1266.5 cm at –1 CCSD(T) level whereas a value of 1551.7 cm was computed for the end-on O2 complex, and the two isomers should thus be distinguishable by IR spectroscopy. Furthermore, the influence of argon coordination on the wavenumbers was calculated, too (table 6.15). Despite the low energies of argon complexation for the [OAuO]+ isomer, the wavenumbers − of its Σu mode significantly decreases with increasing number of coordinated argon atoms. The O–O bond of the side-on O2 complex was substantially lengthened upon coordination of the first two argon atoms and these structural changes are reflected by a decrease of 1 + the wavenumber of the corresponding O–O stretching vibration. In the [ArAu(η -O2)] complex the O–O bond length was computed to be marginally shorter than in the free 1 + [Au(η -O2)] cation and accordingly the effect on the O–O stretching mode is smaller, too. Finally, the additional weakly coordinated argon atoms only have a minor effect on the calculated wavenumbers, and the computed vibrational spectra of the free cations 6.4. Conclusion and outlook 79

(including the strongly bound argon atoms) give good approximate values for those of the corresponding weak argon complexes.

6.3.4 Summary

1 + According to the quantum-chemical calculations, the [Au(η -O2)] cation is the only + thermochemically stable isomer of [AuO2] stoichiometry. This complex ist best described + as an Au cation coordinated to a O2 molecule and the formal oxidation state of gold in this I 1 + cation is thus only I. In analogy to other Au species, [Au(η -O2)] is strongly coordinated 1 + by an argon atom and the resulting [ArAu(η -O2)] complex was calculated to be about 15 kJ mol–1 more stable than the free cation. The coordination of the second argon atom is only weak and the proposed structure might thus be experimentally verified by a 2 + photodissociation IR spectrum of the diargon complex. By contrast, both the [Au(η -O2)] and the [OAuO]+ isomer were calculated to be unstable with respect to elimination of dioxygen. Hence, the V oxidation state of gold is probably not stable in cationic gold oxides.

6.4 Conclusion and outlook

By the contrast to its heavier homologue IrO4, no evidence could be found in new matrix isolation experiments for the rhodium tetroxide molecule. This molecule with rhodium in its formal VIII oxidation state was furthermore predicted to be thermochemically unstable by means of CCSD(T) calculations. Nevertheless, two new molecules, namely the 1 end-on O2 complex [RhO2(η -O2)] and a new D 2d-symmetrical isomer of the known planar 2 [Rh(η -O2)2] molecule could be identified by their IR absorptions. Although the VIII oxidation state of rhodium can thus not be reached, it was already predicted previously that the IrO4 molecule can even be further oxidised. The resulting iridium tetroxide cation might be experimentally characterised by photodissociation IR spectroscopy of its + argon complexes. The vibrational spectra of the different [IrO4] isomers were therefore calculated at coupled cluster level in order to support the assignment of the IR bands. + Furthermore, the interaction of the different [IrO4] isomers with argon was calculated to be weak and accordingly, the vibrational spectra of these complexes are almost not + affected by argon coordination. In the case of [AuO2] , the quantum-chemical calculations showed that the linear [OAuO]+ isomer is thermochemically unstable against internal + reduction reactions. Hence, the [AuO2] species detected in a previous mass spectrometric 1 + investigation probably is the [Au(η -O2)] complex with gold in its I oxidation state. In I + analogy to other monocoordinated Au species and in contrast to the [IrO4] cations, this complex was calculated to strongly coordinate to an argon atom. 80 CHAPTER 6: OXIDES OF RH, IR AND AU 6.5 Computational details

The structures of all molecules were fully optimised (by relaxing all parameters) at density functional theory level using the B3LYP[193–196] functional as implemented in the respective program packages, and the choice of this functional was based on its good performance for comparable molecules.[117] In order to determine the ground states of the molecules several spin multiplicities were calculated. In the case of the weak argon complexes, an empiral correction for dispersive interactions (D3) was included in the DFT calculations.[114] Dunning’s correlation-consistent triple-ζ basis sets (aug-cc-pVTZ) were used for both oxygen and argon in all calculations.[197, 215] Scalar relativistic effects were considered by using relativistic energy-adjusted small core pseudo potentials for the different metal atoms and the corresponding aug-cc-pVTZ-PP basis sets.[216, 217] For brevity, this basis set combination is referred to as aVTZ(-PP). Further calculations using the BP86[194, 218] functional as well as at different ab initio levels of theory were done starting from the B3LYP structures within the restrictions of the respective point groups; in the CCSD(T) calculations a ROHF reference function was used in the case of open shell electron configurations. The ab initio calculations were done using the frozen core approximation with the 1s (O), 1s2s2p (Ar), 4s4p (Rh) and 5s5p (Ir, Au) orbitals not considered in the evaluation of the correlation energies. Stationary points on the potential energy surface were characterised by harmonic frequency calculations for both 16O and 18O isotopomers. For all other elements the masses of the respective most abundant isotopes were used except for the DFT-D3 frequency calculations in which averaged masses were used for all elements. All standard DFT calculations were performed with the Gaussian09 program package[193] whereas the Turbomole V6.4 suite of programs was used for DFT- D3 calculations.[219] Ab initio calculations were done using the Molpro06[220] [coupled cluster calculations for the rhodium oxides and CASPT2 calculations (using the RS2 module[221] and a level shift[222] of 0.1)] and CFOUR[201] (coupled cluster calculations for iridium and gold oxide cations) program packages. 7 Oxide fluorides of Hg, Au, U and Th

Besides in binary fluoride and oxide compounds as discussed in the previous chapters, high oxidation states of the transition metals can in principle also be stabilised in ternary oxide fluorides. For example, the VIII oxidation state of osmium, as represented by the OsO4 [36] molecule, is also stable in the oxide fluorides OsO3F2 and OsO2F4. However, although many mixed oxide fluorides are known for the early-to-midtransition elements, no example was yet published of an oxide fluoride of the elements of groups 9–12. The only exception to this might be a solid platinum oxide fluoride of PtOF3 of PtOF2 stoichiometry obtained [36] after fluorination of PtO2 but whose characterisation remains incomplete. Molecular oxide fluorides can be synthesised by the reaction of oxygen difluoride with gaseous metal atoms, as for example generated using the laser ablation technique. This method was already successfully applied for the preparation of oxide fluoride molecules of the group 3 and 4 elements[223–225] and might also be used for the generation of oxide fluorides of the late transition metals. Mercury oxide fluorides were thus studied experimentally after the reaction of mercury atoms with OF2 as well as by quantum-chemical methods and the results of this investigation are discussed in section 7.1. Furthermore, gold oxide fluoride molecules could also be prepared by this type of reaction. The structures and stabilities of these molecules were investigated quantum-chemically and in order to support their characterisation in matrix isolation experiments, vibrational spectra were calculated, too (section 7.2). Finally, oxide fluorides of the actinoid elements uranium and thorium as formed by the reaction of the corresponding metal atoms with OF2 were also studied, and the results of this combined matrix isolation and theoretical investigation are presented in section 7.3.

81 82 CHAPTER 7: OXIDE FLUORIDES OF HG, AU, U AND TH 7.1 Mercury oxide fluorides

The results shown in this section are part of a collaboration with professor Lester Andrews (University of Virginia, Charlottesville, U.S.) and were already published in L. Andrews, X. Wang, Y. Gong, T. Schl¨oder, S. Riedel, M. J. Franger, Angew. Chem. 2012, 124, 8359. All calculations presented hereafter were done by the author of this thesis except for the coupled cluster calculations for FHgOF and OHgF. The matrix isolation experiments were carried out by Andrews et al. and will therefore only be outlined briefly in order to put the computational results into perspective. More detailed information about the experimental work can be found in the above-mentioned publication.

The most common oxidation state of mercury in its compounds is II, as for example in the well-known solids HgF2, HgCl2 and HgO. While the two dihalides are also known as molecular species, no evidence could yet be found for any binary mercury oxide molecule. In 2007, the long sought-for mercury tetrafluoride molecule could be experimentally characterised[8] more than 20 years after the first quantum-chemical calculations predicting its stability.[11, 12] This molecule represents the so far only example for the IV oxidation state of mercury. Several other HgIV complexes with different monovalent anions were investigated by quantum-chemical calculations but none of these molecules was found to be thermochemically stable.[117, 226] A further possibility for the stabilisation of HgIV could be the use of oxygen ligands as in oxide fluoride molecules which might be prepared by the reaction of mercury atoms with OF2.

7.1.1 Matrix isolation experiments

Due to its low melting point of –39◦C, mercury is a liquid at room temperature and does thus not remain in place to be ablated by a focussed, pulsed laser beam. Therefore, the generation of excited gaseous mercury atoms by the laser laser ablation technique requires mercury in some other, solid form. This can for example be achieved by using amalgams as the ablation target, such as those employed in common dental fillings. Mercury atoms ablated from different amalgams were co-deposited with OF2 diluted at 1% in argon or neon onto a CsI window cooled to 4 K for deposition, and the resulting matrix samples were subsequently analysed by IR spectroscopy. The use of amalgams as the mercury source led to the formation of various side products, such as SnO2, SnF2, AgF, AgF2, CuF2 and NaF which were identified by their known IR absorptions. The further observation of the bands of the HgF and HgF2 molecules indicated the successful generation of reactive mercury atoms by the laser ablation process. Besides these bands, two previously unknown absorptions in the region of the Hg–F stretching modes could be observed at 637.6 and 631.6 cm–1 in argon matrices as well as at 648.0 and 640.5 cm–1 in neon matrices. The 7.1. Mercury oxide fluorides 83

18 16 experiments were repeated with 91% enriched OF2 in order to the determine the O to 18O isotopic shifts and the bands of the corresponding 18O isotopomers were identified due to their similar behaviour in the annealing and irradiation experiments. Based on the quantum-chemical calculations, the new bands were assigned to the previously unknown OHgF and FHgOF molecules formed after the insertion of mercury atoms into the O–F bonds of OF and OF2 respectively (vide infra).

7.1.2 Structures

The structures of the possible product molecules formed after the reaction of mercury atoms with either OF2 or the OF radical produced by the UV emission of the laser ablation process were optimised at DFT and CCSD(T) level (figure 7.1). Two possible products were considered in both cases. The reaction of mercury atoms with OF2 might lead to either the HgOF2 molecule with mercury in its formal IV oxidation state or the mercury fluoride fluoroxide molecule FHgOF formed after the insertion of a mercury atom in one of the O–F bonds of OF2. The latter molecule was found to have a C s-symmetrical structure 1 0 in its singlet A ground state and the calculated interatomic distances of d Hg–F = 194.0 pm and d Hg–O = 199.3 pm indicate Hg–F and Hg–O single bonds. For the other isomer, the HgOF2 molecule, a C 2v symmetrical minimum structure was found on its singlet potential energy surface. The computed bond lengths in this molecule were 190.4 pm for the Hg–F bond and 197.5 pm for the Hg–O bond, suggesting the latter bond to be rather weak. However, a structure optimisation of its triplet state at DFT lead to a weakly bound complex between HgF2 and an oxygen atom. As the total energy of this complex was

FHgOF HgOF2 2 ’ 1 (Cs, A) (C2v, A1)

OHgF HgOF 2 2 ’ (C∞v, Π) (Cs, A)

Figure 7.1 : Optimised structures of molecular mercury oxide fluorides; see table 7.1 for bond lengths and angles. 84 CHAPTER 7: OXIDE FLUORIDES OF HG, AU, U AND TH

Table 7.1 : Calculated structural parameters of molecular mercury oxide fluorides[a]

Molecule Parameter B3LYP[b] CCSD(T)[b] 2 0 HgOF (C s, A ) d Hg–O 239.1 216.4 d O–F 139.8 144.6 Hg-O-F 110.0 103.7 2 OHgF (C ∞v, Π) d Hg–F 194.2 193.0 d Hg–O 195.2 194.9 1 0 FHgOF (C s, A ) d Hg–F 194.0 192.6 d Hg–O 199.3 197.9 d O–F 144.4 146.5 Hg-O-F 105.0 102.3 F-Hg-O 178.1 178.6 1 HgOF2 (C 2v, A1) d Hg–F 191.7 190.4 d Hg–O 196.9 197.5 O-Hg-F 094.6094.1 [a] Bond lengths in pm, angles in ◦. [b] aVTZ(-PP) basis sets.

F (–0.005) O (1.01)

Hg (–0.005)

Figure 7.2 : Calculated spin density of OHgF at B3LYP/aVTZ(-PP) level; isosurfaces at 0.01 a.u. found to be 192.7 kJ mol–1 lower at CCSD(T)/aVTZ(-PP)//B3LYP/aVTZ(-PP) level, the singlet state of HgOF2 which would represent the formal IV oxidation state of mercury is not stable and was not investigated in more detail. Two product molecules are possible after the reaction of mercury atoms with the OF radical, namely the mercury(I) fluoroxide HgOF as well as a OHgF isomer. For the former molecule a C s-symmetrical structure was calculated for its doublet ground state (2A0 ) and the large computed Hg–O distance of 216.4 pm at CCSD(T) level indicates a weak interaction between mercury an the OF 2 fragment. For the OHgF isomer a linear structure (C ∞v symmetry) and a Π ground state were obtained. When compared with those in FHgOF the computed bond lengths in this molecule of d Hg–F = 193.0 pm and d Hg–O = 194.9 pm at CCSD(T) level suggest two single bonds. Therefore, the electronic structure of OHgF was further examined by an analysis of its unpaired spin density which was found to be almost exclusively located at the oxygen atom (figure 7.2). Thus, the oxidation state II is assigned to mercury in OHgF and accordingly the oxidation state of oxygen is –I, making it an oxyl ligand as in AuO.[210] 7.1. Mercury oxide fluorides 85

7.1.3 Thermochemistry

The thermochemistry of the reactions of mercury atoms with either the OF radical or OF2 was also calculated at CCSD(T)/aVTZ level. As the formation of both FHgOF and OHgF was computed to be strongly exothermic, these two reactions are likely to proceed under the experimental conditions. By contrast, the reactions leading to HgOF and singlet HgOF2 were calculated to be endothermic and these two molecules are therefore predicted to be thermochemically unstable species. Thus, only one isomer is stable for each stoichiometry, namely the FHgOF and OHgF molecules.

Table 7.2 : Calculated thermochemistry of molecular mercury oxide fluorides[a]

Reaction B3LYP[b] CCSD(T)[b] Hg + OF → HgOF 0–14.20(–13.2) 0006.4 000(8.6) Hg + OF → HgOF –171.1 (–168.5) –176.6 (–173.5) Hg + OF2 → HgOF2 0037.6 00(36.5) 0041.2 00(40.1) Hg + OF2 → FHgOF –201.6 (–200.1) –210.8 (–209.4) [a] Values in kJ mol–1; numbers in parentheses are ZPE corrected. [b] aVTZ(-PP) basis sets.

7.1.4 Vibrational frequencies

The calculated and experimental vibrational frequencies for the different HgOF and HgOF2 isomers are listed in table 7.3 together with the values for HgF2 calculated for comparison. The assignment of the previously unknown bands to the FHgOF and OHgF molecules is strongly supported by the calculated wavenumbers at CCSD(T) level which are in very good agreement with the experimental values. Despite the differences between the calculated wavenumbers at DFT and CCSD(T) level, the relative position of the bands compared to that of HgF2 is well reproduced by both methods. Furthermore, the calculated isotopic shifts (table 7.4) give additional support for the assignment of the bands. The calculated wavenumbers thus allowed to definitely attribute the new bands to FHgOF and OHgF and further to distinguish between the two molecules, which would have been a nearly impossible task without quantum-chemical support.

7.1.5 Summary

The reaction of laser-ablated mercury atoms with oxygen difluoride was studied using the matrix isolation technique. Based on quantum-chemical calculations at CCSD(T) level, the two new absorptions identified in the resulting IR spectra could be attributed to the previously unknown OHgF and FHgOF molecules, the former of which represents the first oxide fluoride of a group 12 element. However, OHgF must be viewed as a HgII species 86 CHAPTER 7: OXIDE FLUORIDES OF HG, AU, U AND TH

Table 7.3 : Calculated and observed wavenumbers of the stretching modes of molecular mercury [a] oxide fluorides and HgF2 Calc. Exp. (matrix) Molecule Mode B3LYP[b] CCSD(T)[b] Neon Argon Ref. 2 0 0 HgOF (C s, A ) A 967.3 (242) 858.2 A0 225.4 00(6) 389.4 2 + OHgF (C ∞v, Π) Σ 631.20(77) 648.3 648.0 637.6 tw Σ+ 553.80(12) 568.9 1 HgOF2 (C 2v, A1 B2 650.3 (101) 671.7 A1 577.7 00(3) 592.1 A1 480.1 00(2) 440.7 1 0 0 FHgOF (C s, A ) A 938.50(24) 865.1 A0 619.80(96) 641.2 640.5 631.6 tw A0 548.8 00(4) 566.0 1 + − HgF2 (D∞h, Σg ) Σu 638.4 (113) 659.0 657.5 645.0 [8] + Σg 564.1 00(–) 581.3 [a] Values in cm–1 for 202Hg16O isotopomers [in the CCSD(T) calculations an averaged mass was used for Hg]; numbers in parentheses are infrared intensities in km mol–1. [b] aVTZ(-PP) basis sets. tw = this work

Table 7.4 : Calculated and experimental isotopic shifts of selected stretching modes of FHgOF and OHgF[a]

Calc. Exp. (matrix) Molecule Mode B3LYP[b] CCSD(T)[b] Neon Argon 2 + 16 OHgF (C ∞v, Π) Σ ( O) 631.2 00(0.0) 648.3 00(0.0) 648.0 00(0.0) 637.6 00(0.0) Σ+ (18O) 618.5 (–12.7) 636.1 (–12.1) 635.9 (–12.1) 625.2 (–12.4) 1 0 0 16 FHgOF (C s, A ) A ( O) 619.8 00(0.0) 641.2 00(0.0) 640.5 00(0.0) 631.6 00(0.0) A0 (18O) 611.30(–8.5) 632.90(–8.3) 632.50(–8.0) 623.40(–8.2) [a] Values in cm–1 for 202Hg isotopomers [in the CCSD(T) calculations an averaged mass was used for Hg]; in parentheses: isotopic shifts compared to the 16O isotopomers. [b] aVTZ(-PP) basis sets. 7.2. Gold oxide fluorides 87 as its unpaired spin density is located at the oxygen atom. The electronic structure of this molecule is thus similar to that of the AuO molecule in which the oxidation state of oxygen also is –I. Mercury oxide fluorides with mercury in higher oxidation states were also calculated but were found to be unstable. In line with this prediction, no evidence could found of these species in the matrix spectra and HgF4 thus remains the only example for mercury oxidised beyond its II oxidation state.

7.2 Gold oxide fluorides

As for mercury, both binary oxides and fluorides of gold are well characterised compounds in the solid state. Whereas the III oxidation state of gold is stable in both AuF3 and Au2O3, its higher V oxidation state can only be reached in AuF5 where fluorine is used as the ligand. Unlike the high-valent fluorides, AuF is so far only known as an isolated molecule. It could as yet not be characterised as bulk material and solid AuF was calculated [227] to disproportionate exothermicly to yield elemental gold and AuF3. Molecular gold monofluoride as well as its argon complex were studied both in the gas phase[212, 228] and as isolated molecules in rare gas matrices.[67] In the latter work, new IR bands could furthermore be attributed to the so far unknown gold difluoride molecule. Gold oxides were also studied in matrix isolation experiments in which AuO as well as a linear AuO2 molecule could be characterised.[205, 229] It was shown by theoretical electronic structure calculations that the former molecule must be viewed a AuI oxyl species.[210] By contrast, the electronic structure of the AuO2 molecule was not investigated in detail but in view of the results obtained for AuO, the oxidation state of gold in this molecule might actually be only II. Despite this variety of known binary gold fluorides and oxides, no mixed oxide fluoride of gold could yet be characterised. A possible route to molecular gold oxide fluorides could, in analogy to the case of mercury, be provided by the reaction of laser-ablated gold atoms with OF2

7.2.1 Structures

The optimised structures of the possible products formed after the insertion reactions of gold atoms with either the OF2 molecule or the OF radical produced during the laser ablation process are shown in figure 7.3. In the case of the reaction with OF2, two possible product molecules were considered, the formal gold oxide difluoride molecule AuOF2 and the gold fluoride fluoroxide molecule FAuOF formed after the insertion of a gold atom in one of the O–F bonds of OF2. The optimised structure of the latter molecule has C s symmetry, and in its doublet 2A0 ground state the bond lengths were calculated to be d Au–F = 188.4 pm and d Au–O = 187.7 pm at CCSD(T) level. These atomic distances correspond to single bonds although they slightly differ from those calculated for AuF and AuO (table 7.5). The ground state of the C 2v-symmetrical AuOF2 molecule was 88 CHAPTER 7: OXIDE FLUORIDES OF HG, AU, U AND TH

AuOF2 FAuOF OAuF 2 2 ’ 3 – (C2v, B2) (Cs, A) (C∞v, Σg )

AuF AuO AuOF 1 + 2 1 ’ (C∞v, Σ ) (C∞v, Π) (Cs, A)

ArAuF ArAuO ArAuOF 1 + 2 1 ’ (C∞v, Σ ) (C∞v, Π) (Cs, A)

Figure 7.3 : Optimised structures of molecular gold oxide fluorides and their argon complexes; see table 7.5 for bond lengths and angles.

2 found to be B2 and the bond lengths of d Au–F = 190.0 pm and d Au–O = 186.1 pm in this molecule are similar to those in FAuOF, already indicating a gold-oxygen single bond. This assumption is further supported by an analysis of the unpaired spin density of this molecule which was calculated to be almost exclusively located at the oxygen atom (figure 7.4), making it an oxyl ligand. The electronic structure of AuOF2 is thus similar to those of AuO and OHgF, so that accordingly the III oxidation was assigned to gold in this molecule. The reaction of the gold atoms with the OF radical can lead to either gold fluoroxide, AuOF, or the insertion product OAuF. In its singlet ground state, AuOF was calculated to have a C s-symmetrical structure and a large gold-oxygen distance of 197.6 pm was obtained for this molecule at CCSD(T) level. As gold monofluoride was shown to interact strongly with argon,[66, 67, 211, 212, 214] analogous complexes of AuO as well as of AuOF were calculated for comparison and the data is included in table 7.5. The calculated gold-argon distance in the ArAuOF complex was found to be intermediate between those in ArAuO and ArAuF. For the other isomer, the OAuF molecule, a linear structure (C ∞v symmetry) was computed. The calculated Au–F bond length of 188.2 pm in this molecule is comparable to that of FAuOF while the small computed Au–O distance of 181.1 pm already hints at a gold-oxygen double bond. The ground state of this molecule was found to be 3Σ−, and an analysis of its electronic structure showed that the two 7.2. Gold oxide fluorides 89

Table 7.5 : Calculated structural parameters of molecular gold oxide fluorides[a]

Molecule Parameter B3LYP[b] CCSD(T)[b] 1 + AuF (C ∞v, Σ ) d Au–F 194.8 194.2 1 + ArAuF (C ∞v, Σ ) d Au–F 193.7 192.7 d Au–Ar 247.3 242.4 2 AuO (C ∞v, Π) d Au–O 188.9 189.6 2 ArAuO (C ∞v, Π) d Au–O 189.5 189.9 d Au–Ar 257.9 251.8 1 0 AuOF (C s, A1) d Au–O 198.0 197.6 d O–F 143.9 146.0 Au-O-F 108.1 105.6 1 0 ArAuOF (C s, A1) d Au–O 197.2 196.3 d O–F 145.0 146.9 d Au–Ar 253.7 247.3 Au-O-F 107.3 104.9 Ar-Au-O 177.4 177.9 3 − OAuF (C ∞v, Σ ) d Au–O 181.2 181.1 d Au–F 189.0 188.2 2 AuOF2 (C 2v, B2) d Au–O 185.7 186.1 d Au–F 191.3 190.0 O-Au-F 096.3094.8 2 00 FAuOF (C s, A ) d Au–O 190.0 187.7 d Au–F 189.4 188.4 d O–F 141.5 143.0 F-Au-O 176.3 176.7 Au-O-F 113.6 112.2 [a] Bond lengths in pm, angles in ◦. [b] aVTZ(-PP) basis sets.

unpaired electrons are localised in the two perpendicular Au–O antibonding π∗ orbitals. As the two corresponding bonding π orbitals are fully occupied the electronic structure of this molecule can be compared to that of dioxygen. Due to the different electronegativities of gold and oxygen, the bond is however not symmetrical and the partially occupied antibonding orbitals are distorted towards the oxygen atom. This leads to an slightly increased spin density on the oxygen ligand (1.34 instead of 1.0 for a symmetrical bond, figure 7.4) but the deviation being small, the bond is nevertheless best described as a double bond. The OAuF molecule is thus not only the first gold compound with an even number of electrons not having singlet ground state but also represents the first molecule with a gold-oxygen double bond. However, SO coupling was not considered in these calculations and the results might bear a significant error due to this neglection, especially as this effect was shown to play an important role for the open shell AuO molecule[210] [60, 61] and also to be responsable for the stabilisation of the singlet ground state of PtF6. 90 CHAPTER 7: OXIDE FLUORIDES OF HG, AU, U AND TH

O (0.96)

F (0.13) O (1.34) F (0.005) Au (0.53) Au (0.03)

Figure 7.4 : Calculated spin densities of AuOF2 and OAuF at B3LYP/aVTZ(-PP) level; isosurfaces at 0.01 a.u.

7.2.2 Thermochemistry

The energies for the reactions of gold atoms with either the OF radical or the OF2 molecule are displayed in table 7.6. All reactions were calculated to be exothermic and are thus likely to proceed in the experiment. The results further show that the OAuF molecule with gold in the oxidation state III is 162.8 kJ mol–1 more stable than its AuIOF isomer. For the two –1 AuOF2 isomers the energy difference is much smaller and amounts to only 13.9 kJ mol III at CCSD(T) level with Au OF2 being slightly favoured. Furthermore, the energies of argon complexation were calculated for the monocoordinated gold species (table 7.7). The calculated energy for the interaction between argon and AuOF lies between those obtained for the analogous complexes of AuF as well as of AuO and this result is in line with the corresponding intermediate gold-argon distances in the structures of these complexes.

Table 7.6 : Calculated thermochemistry of molecular gold oxide fluorides[a]

Reaction B3LYP[b] CCSD(T)[b] Au + OF → OAuF –338.6 (–334.7) –330.8 (–326.4) Au + OF → AuOF –153.7 (–150.5) –166.7 (–163.6) Au + OF2 → AuOF2 –300.0 (–300.2) –292.2 (–291.2) Au + OF2 → FAuOF –305.7 (–305.7) –280.0 (–277.3) [a] Values in kJ mol–1; numbers in parentheses are ZPE corrected. [b] avTZ(-PP) basis sets.

Table 7.7 : Calculated argon complexation energies of AuF, AuO and AuOF[a]

Reaction B3LYP[b] CCSD(T)[b] AuF + Ar → ArAuF –36.6 (–34.1) –46.0 (–43.2) AuO + Ar → ArAuO –22.8 (–20.6) –31.3 (–28.8) AuOF + Ar → ArAuOF –26.9 (–24.8) –37.4 (–35.0) [a] Values in kJ mol–1; numbers in parentheses are ZPE corrected. [b] avTZ(-PP) basis sets. 7.2. Gold oxide fluorides 91

7.2.3 Vibrational frequencies

The calculated vibrational frequencies of the different gold oxide fluoride molecules are listed in table 7.8 for both 16O and 18O isotopomers. For the Au–F stretching vibrations, wavenumbers in the range between 620 and 680 cm–1 were calculated and the differences are large enough to allow a distinction between the different molecules, especially when the 16O to 18O isotopic shifts are taken into account. In the case of OAuF, the high wavenumber of the Au–O stretching vibration of 767.9 cm–1 provides additional evidence for the gold- oxygen double bond in this molecule. However, in analogy to the comparable AuF2 and [67] AuF3 molecules for which a similar trend was observed, the computed wavenumbers at CCSD(T)/aVTZ(-PP) might be slightly smaller than the experimental ones. A possible reason for this deviation might be CV correlation effects which were not considered at this level of theory (see also chapter 5). As in the case of AuF, the strong interaction between argon and either AuO or AuOF also affects the wavenumbers of the Au–O stretching vibration of these compounds which are shifted to higher values in the argon complexes.

Table 7.8 : Calculated wavenumbers of the stretching modes of molecular gold oxide fluorides[a]

B3LYP[b] CCSD(T)[b] Molecule Mode 16O 18O 16O 18O 1 + + AuF (C ∞v, Σ ) Σg 530.6 (44) 541.4 (45) 1 + + ArAuF (C ∞v, Σ ) Σg 563.1 (56) 580.5 (57) 2 + AuO (C ∞v, Π) Σg 574.7 (7) 544.3 565.4 (9) 535.5 2 + ArAuO (C ∞v, Π) Σg 593.0 (7) 561.6 593.8 (9) 562.4 1 0 0 AuOF (C s, A ) A 519.7 (11) 492.4 523.8 (12) 496.5 A0 900.8 (89) 873.3 832.5 (39) 806.9 1 0 0 ArAuOF (C s, A ) A 550.1 (12) 521.3 565.2 (16) 535.9 A0 893.0 (68) 655.5 835.1 (20) 808.9 3 − + OAuF (C ∞v, Σ ) Σg 628.5 (82) 623.6 642.4 (88) 637.1 + Σg 757.7 (3) 722.9 767.9 (9) 733.0 2 AuOF2 (C 2v, B2) A1 597.8 (5) 596.7 620.3 (6) 619.4 A1 632.5 (1) 599.9 617.7 (0) 585.6 B2 637.5 (141) 637.4 659.2 (147) 659.2 2 0 0 FAuOF (C s, A ) A 586.6 (34) 563.7 613.0 (53) 592.9 A0 654.2 (58) 645.0 680.3 (81) 667.2 A0 894.1 (173) 866.4 834.2 (125) 807.2 [a] Values in cm–1; numbers in parentheses are infrared intensities in km mol–1. [b] avTZ(-PP) basis sets. 92 CHAPTER 7: OXIDE FLUORIDES OF HG, AU, U AND TH

7.2.4 Summary

The quantum-chemical calculations showed the AuOF2, FAuOF, AuOF and OAuF molecules to be thermochemically stable against the elimination of OF2 and OF respectively. These molecules thus represent viable target molecules for future matrix isolation experiments and their vibrational spectra were calculated in order to support a characterisation by IR spectroscopy. The electronic structure analysis of the AuOF2 molecule showed the unpaired electron to be located at the oxygen ligand and therefore the oxidation state of gold in this molecule is III. For the OAuF molecule a 3Σ− ground state was calculated and this molecule would thus represent the first gold compound with more than one unpaired electron. The analysis of its bonding situation showed that it would furthermore be the first example of a molecule containing a gold-oxygen double bond. As only scalar relativistic effects were considered, further calculations including SO coupling are however necessary in order to confirm the electronic structure of this molecule.

7.3 Thorium and uranium oxide difluoride

The results presented in this section are also part of a collaboration with professor Lester Andrews (University of Virginia, Charlottesville, U.S.) and were already published in Y. Gong, X. Wang, L. Andrews, T. Schl¨oder, S.Riedel, Inorg. Chem. 2012, 51, 6983. All calculations discussed hereafter were carried out by the author of this thesis while the matrix isolation experiments were performed in the Andrews group. Therefore, the experimental results will only be described very briefly and more detailed information can be found in the aforementioned publication.

Uranium is the most studied actinoid element and many uranium compounds containing 2+ + [230–233] linear [UO2] or [UO2] fragments are known. Furthermore, several binary fluorides exist of this element, as for example UF6 and UF4. However, examples for compounds containing one single terminal U–O bond are rare, which is related to the high reactivity of these molecules towards the formation of bridged dimers or polymers.[232] Nevertheless, [234] UOF4 was characterised to have one single terminal U–O bond in the solid state but due to the almost linear structure of the O-U-F unit and the short U–F bond it was classified as an uranyl analogue.[232] The use of the matrix isolation technique did however allow to identify several molecules with terminal uranium-oxygen and thorium-oxygen bonds because the solid noble gas environment effectively prevents intermolecular reactions.[235–238] By this method, even the UNF3 molecule which contains a terminal uranium-nitrogen triple bond could be stabilised.[239] This molecule was formed after the reaction of uranium atoms with NF3 and a similar reaction might take place between uranium atoms and the even more reactive OF2. 7.3. Thorium and uranium oxide difluoride 93

7.3.1 Matrix isolation experiments

Matrix samples were prepared by co-condensation of laser-ablated either uranium or thorium atoms with OF2 diluted at 1% in excess argon or neon onto a CsI window cooled to 4 K for deposition. In the IR spectra of the argon matrices obtained after the reaction –1 of uranium with OF2, three new absorptions at 834.8, 522.2 and 487.2 cm could be observed besides the bands of common impurities and those previously assigned to binary uranium fluorides and oxides as well as to UO2F2 and UO2F. The three new bands showed the same behaviour in the annealing and irradiation experiments, thus suggesting them to arise from different vibrational modes of the same species. Their position as well as the measured 16O to 18O isotopic shifts indicated the bands to correspond to one U–O and two U–F stretching vibrations. Similar spectra were obtained when thorium was used instead of uranium and three analogous new absorption were observed at 806.6, 511.1 and 482.3 cm–1. As no other related bands were observed, the two groups of absorptions were assigned to the previously unknown UOF2 and ThOF2 molecules, respectively.

7.3.2 Structures and bonding

The optimised structures of the ThOF2 and UOF2 molecules are shown in figure 7.5. For both molecules pyramidal structures of C s symmetry were obtained instead of the higher symmetrical ones predicted by the VSEPR model, and similar non-planar structures were also found for the related UOX2 and ThOX2 (X = H, CH3) species identified in earlier matrix isolation experiments.[235–238] Non-VSEPR structures are furthermore known for comparable molecules like the difluorides of the heavy alkaline earth metals Ca, Sr and Ba. These molecules have bent structures which can be explained by core-polarisation and the involvement of the inner (n – 1) d orbitals in bonding.[240] According to its d0f0 electron configuration, ThOF2 has a singlet ground state and the bond lengths in this molecule were calculated to be d Th–O = 188.5 pm and d Th–F = 215.7 pm at B3LYP level. The two 3 00 additional electrons in UOF2 lead to a triplet ground state ( A ) for this molecule and the

ThOF2 UOF2 1 ’ 3 ” (Cs, A) (Cs, A )

Figure 7.5 : Optimised structures of ThOF2 and UOF2; see table 7.9 for bond lengths and angles. 94 CHAPTER 7: OXIDE FLUORIDES OF HG, AU, U AND TH

[a] Table 7.9 : Calculated structural parameters of ThOF2 and UOF2

Molecule Parameter B3LYP[b] CASPT2[b,c] 1 0 ThOF2 (C s, A ) d Th–O 188.5 189.7 d Th–F 215.7 216.4 O-Th-F 112.0 110.4 F-Th-F 106.2 107.4 3 00 UOF2 (C s, A ) dU–O 183.2 183.3 d U–F 211.2 211.0 O-U-F 113.2 109.5 F-U-F 100.2 101.3  ◦ [a] Bond lengths in pm, angles in . [b] 6-311+G(d)-SDD basis sets. [c] CAS(6,6) for ThOF2 and CAS(8,8) for UOF2.

bonds were computed to be slightly shorter (d U–O = 183.2 pm and d U–F = 211.2 pm) than the corresponding bonds in ThOF2. Furthermore, CASPT2 optimisations were done for both molecules. Because of the mainly ionic interaction between the fluorine atoms and the respective MO fragments (vide infra) the active space was chosen to comprise only the three bonding and antibonding orbitals of the MO bonds as well as the two unpaired electrons in the case of UOF2. Compared with the DFT results, the CASPT2 calculations led to very similar structures with only slightly different bond lengths (table 7.9). The computed U–O bond length of 183.2 pm at DFT level is very similar to that of the free uranium monoxide molecule (185.0 pm calculated at the same level of theory, exp.: 183.8 pm[241]). It is about 12 pm larger than the U–O triple bond length obtained from tabulated triple bond radii (171 pm) which were however proposed based on the uranyl dication which is possibly not a good comparison for neutral molecules.[242] Hence, based on the bond length criterion the U–O bond in UOF2 can be described as a triple bond. The bonding situation in this molecule was therefore further analysed by NBO calculations where three bonding orbitals were found for the U–O bond (table 7.10), to each of which the oxygen atom contributes about 80% (mainly by its 2p orbital) while the remaining contribution comes from a 6d5f hybrid orbital of uranium. By contrast, no bonding orbitals were found for the U–F bonds which can therefore be described as mostly ionic. The U–O bond order was calculated from the occupation numbers of the bonding and antibonding orbitals, and a value of was 2.95 obtained. Additional evidence for a U–O triple bond is provided by the CASPT2 calculations which also resulted in three almost fully occupied bonding orbitals while the corresponding antibonding orbitals were nearly empty (figure 7.6), leading to a bond order of 2.89. As the two remaining f electrons are not involved in bonding, the oxidation state of uranium in this molecule is best described as IV and the additional π bonding is caused by dative electron donation from the oxygen atom. The case of ThOF2 was found to be similar. The Th–O bond length in this molecules almost exactly matches the tabulated Th–O triple bond length of 189 pm[242] but is significantly larger than that of the free ThO molecule (calc.: 184.3 pm, exp: 184.0 pm[243]). Further support for the 7.3. Thorium and uranium oxide difluoride 95 triple bond character of the Th–O bond in this molecule is given by the calculated bond orders obtained from NBO analysis (2.94) and CASPT2 calculations (2.90, figure 7.7).

σ* (U–O) π* (U–O) π* (U–O) E = 0.7563, n = 0.031 E = 0.5508, n = 0.037 E = 0.5327, n = 0.038

f (U) f (U) E = –0.0896, n = 1.000 E = –0.0886, n = 1.000

σ (U–O) π (U–O) π (U–O) E = –0.5398, n = 1.969 E = –0.4820, n = 1.963 E = –0.4829, n = 1.962

Figure 7.6 : CASPT2 orbitals for the U–O bond in UOF2; energies in hartree.

Finally, the energies of reaction for formation of these two molecules from the respective metal atoms and OF2 were also calculated. Both reactions were found to be very exothermic –1 –1 (–1590.4 kJ mol for UOF2 and –1734.8 kJ mol for ThOF2) and the calculations thus corroborate the spontaneous and specific formation of these molecules under conditions of the experiment. 96 CHAPTER 7: OXIDE FLUORIDES OF HG, AU, U AND TH

[a,b] Table 7.10 : NBO analysis of the U–O and Th–O bonds in UOF2 and ThOF2 AO hybridisation Molecule bond type % atom contribution s p d f

UOF2 σ O (77.53) 14.68 85.2900.04 U (22.47)00.5502.66 52.41 44.39 π O (81.83)02.40 97.5600.05 U (18.18)00.2900.86 49.57 49.29 π O (83.04)00.00 99.9500.05 U (16.96)00.0001.41 46.13 52.47 ThOF2 σ O (82.06) 16.78 83.2000.02 Th (17.94)00.2404.32 56.29 39.15 π O (84.95)00.85 99.1100.05 Th (15.05)01.6602.36 60.75 35.24 π O (86.60)00.00 99.9600.04 Th (13.40)00.0001.97 57.45 40.57 [a] Calculated at B3LYP/6-311+G(d)-SDD level. [b] Averaged values for α and β electrons in UOF2.

σ* (Th–O) π* (Th–O) π* (Th–O) E = 0.8062, n = 0.029 E = 0.5547, n = 0.036 E = 0.5925, n = 0.037

σ (Th–O) π (Th–O) π (Th–O) E = –0.5160, n = 1.971 E = –0.4580, n = 1.964 E = –0.4573, n = 1.963

Figure 7.7 : CASPT2 orbitals for the Th–O bond in ThOF2; energies in hartree. 7.4. Conclusion and outlook 97

7.3.3 Vibrational frequencies

The calculated and experimental vibrational frequencies of the ThOF2 and UOF2 molecules are given in table 7.11 for both 16O and 18O isotopomers. The calculated wavenumbers slightly differ from the experimental values but the agreement is nevertheless good. More- over, the isotopic shifts are very well reproduced by the calculations and provide additional support for the assignment of the new absorptions to ThOF2 and UOF2.

Table 7.11 : Calculated and observed (in neon and argon matrices) wavenumbers of the stretching [a] modes of ThOF2 and UOF2 16O 18O Molecule Mode B3LYP[b] Argon Neon B3LYP[b] Argon Neon 1 0 0 [c] ThOF2 (C s, A ) A 832.7 (234) 806.6 788.4 (208) 763.9 782.7 A0 514.7 (126) 511.1 522.4 514.5 (127) 510.8 522.0 A00 486.5 (199) 482.3 500.8 486.4 (198) 482.1 500.5 3 00 0 UOF2 (C s, A ) A 855.8 (215) 834.8 855.0 810.0 (192) 790.5 809.5 A0 526.9 (131) 522.2 529.0 526.8 (132) 522.1 528.8 A00 462.9 (382) 487.2 498.4 460.9 (401) 487.2 507.6 [a] Values in cm–1; numbers in parentheses are infrared intensities in km mol–1. [b] 6-311+G(d)- SDD basis sets. [c] Absorption masked by the band of OF2.

7.3.4 Summary

Laser-ablated uranium and thorium atoms spontaneously reacted with oxygen difluoride to form the new UOF2 and ThOF2 molecules which were analysed trapped in inert gas matrices using IR spectroscopy. The assignment of the bands was supported by 16O to 18O isotopic shifts as well as quantum-chemical calculations. In analogy with comparable MOX2 molecules both UOF2 and ThOF2 were calculated to have pyramidal non-VSEPR structures. The bonding situation was analysed using NBO and CASPT2 calculations and the M–O bonds were found to have pronounced triple bond character as also shown by the very small bond lengths.

7.4 Conclusion and outlook

The new FHgOF and OHgF molecules were be prepared by reaction of laser-ablated mercury atoms with oxygen difluoride and characterised by IR spectroscopy and supporting quantum-chemical calculations. The latter molecule represents the first oxide fluoride of a group 12 element. However, an analysis of the electronic structure of this molecule 98 CHAPTER 7: OXIDE FLUORIDES OF HG, AU, U AND TH showed that the oxidation state of mercury in this species is II. Similar gold oxide fluoride molecules were predicted to be stable by means of quantum-chemical calculations. In analogy to OHgF, the AuOF2 molecule was calculated to be an oxyl complex, too, and accordingly the oxidation state of gold in this molecule is III. The OAuF molecules was also predicted to be stable and would represent the first molecule with a genuine gold-oxygen double bond. Furthermore, an unusal triplet ground state was predicted for this species at scalar-relativistic CCSD(T) level. Along with AuOF and FAuOF, these gold oxide fluoride molecules represent viable target molecules for future matrix isolation experiments and might be prepared by a similar reaction of laser-ablated gold atoms with OF2. Finally, the actinoid oxide fluoride molecules UOF2 and ThOF2 were identified in matrix isolation experiments. The molecules were calculated to have non-planar structures and to contain metal-oxygen triple bonds as shown by NBO and CASPT2 calculations.

7.5 Computational details

The structures of all molecules were fully optimised at density functional theory level using the B3LYP[193, 195, 196] functional which was proven to yield reliable results for similar molecules.[117] The ground states of the different molecules were determined by computing several spin multiplicities with up to three unpaired electrons. In the calculationsof the gold and mercury oxide fluorides scalar relativistic effects were considered by the use of relativistic energy-adjusted small core 60 electron pseudo potentials for the metals.[217] Dunning’s correlation-consistent triple-ζ aug-cc-pVTZ basis sets were used in the calcu- lations for fluorine, oxygen and argon[197, 215] whereas for the for the metal atoms the corresponding aug-cc-pVTZ-PP basis developed for the adopted pseudo potential was used.[217] For brevity, this combination of basis sets is referred to as aVTZ(-PP). In the calculations of the UOF2 and ThOF2 molecules the 6-311+G(d) basis set was used for both oxygen and fluorine.[244] The U and Th atoms were described by quasi-relativistic energy-adjusted small core 60 electron pseudo potentials of the Stuttgart group and the corresponding (12s11p10d7f)/[8s7p6d4f] valence basis sets.[245] The basis set combination used for the UOF2 and ThOF2 calculations is denoted as 6-311+G(d)-SDD. Subsequent ab initio calculations at CASPT2 and CCSD(T) level were done starting from the structures optimised at DFT level under retention the molecular symmetries; in the CCSD(T) calcula- tions a ROHF reference function was used in the case of open shell electron configurations. In the ab initio calculations the frozen core approximation was used and the following orbitals were not considered in the calculations of the correlation energies: 1s (O, F), 1s2sp (Ar), 5sp (Au, Hg) as well as 5s5pd (Th, U). Harmonic frequency calculations were carried out at the optimised structures at both DFT and CCSD(T) level for 16O and 18O isotopomers. For the other elements the masses of the most abundant isotopes were used except in the CCSD(T) calculations of the mercury oxide fluorides were an average mass of 200.59 u was used for mercury. All DFT calculations were performed using the Gaussian09 program package[193] whereas the coupled cluster calculations were done 7.5. Computational details 99 with Molpro06[220] and CFOUR[201] program packages for the mercury and gold oxide fluorides respectively. The NBO analysis was done at DFT level using the NBO 3.1 [246] program as implemented in Gaussian09. CASPT2 calculations for UOF2 and ThOF2 were performed using the RS2 module implemented in Molpro06[220, 221] with a level shift[222] of 0.1. 100 CHAPTER 7: OXIDE FLUORIDES OF HG, AU, U AND TH 8 Conclusion

During this thesis a new matrix isolation apparatus was built. It consists of a cold head to which a matrix window was connected, and which was placed inside an aluminium matrix chamber. With this cold head, temperatures of 4 K can be reached, allowing the generation of solid neon matrices. Furthermore, the matrix chamber is equipped with a setup for laser ablation experiments consisting of a rotating metal disc onto which the beam of an infrared laser can be focussed, and the ablated metal atoms are directly deposited onto the cold window together with the matrix gases. This new setup was subsequently used for the experimental investigation chromium and iron fluorides which were formed after the reaction of the respective laser-evaporated metal atoms with F2. In the case of chromium, the possible existence of the CrF6 molecule was previously discussed in the literature but no consent was yet found about the question whether this compound was prepared or not. Together with state-of-the-art quantum- chemical calculations, the new matrix isolation experiments suggest that the debatable IR band, which was assigned to both CrF6 and CrF5 is actually caused by the latter molecule. However, the formation of chromium hexafluoride under high fluorine pressure conditions cannot be excluded. In the case of the iron fluorides, the FeF4 molecule was characterised by its matrix IR spectra in both neon and argon matrices, and this species represents the first example for the IV oxidation state of iron stabilised in a neutral halide. No evidence was found in the spectra for iron pentafluoride which, according to quantum-chemical calculations at CCSD(T) level, might be marginally stable, and a possibility for its experimental realisation is the use of neat fluorine matrices. This extremely oxidising environment was also necessary for the formation of CrF5 in the laser ablation experiments, and only lower chromium fluoride were formed when chromium atoms were reacted with fluorine diluted at low concentrations in argon or neon. The fluorides of manganese and cobalt were studied quantum-chemically and the results indicate the so far unknown manganese pentafluoride molecule to be thermochemically stable, and this species thus represents a viable target molecule for future matrix isolation investigations. In the case of cobalt, the CoF5 molecule was predicted to be unstable so that the known cobalt tetrafluoride molecule will probably remain the highest valent binary neutral fluoride of this element. Due to the reduced steric crowding in the ligands spheres, even higher oxidation states can be reached with oxygen ligands. After the successful characterisation of the iridium tetroxide molecule, its lighter homologue RhO4 was investigated in a combined theoretical and experimental investigation which was done in cooperation with the group of professor

101 102 CHAPTER 8: CONCLUSION

Zhou. Unlike IrO4, rhodium tetroxide was predicted to be thermochemically unstable 2 against internal reduction reactions, and accordingly only the bands of the [RhO2(η -O2)] complex could be observed in the matrix IR spectra. Furthermore, cationic oxide complexes + of iridium and gold were investigated quantum-chemically. In a previous study, the [IrO4] cation which would represent the new IX oxidation state was already predicted to be thermochemically stable. In order to support a possible characterisation of this molecule by gas phase ion spectroscopy, its vibrational frequencies were calculated at CCSD(T) level. The effect of argon coordination on the wavenumbers was also investigated and found + to be negligible. The [AuO2] cation was already detected in earlier mass spectrometric experiments, but its structure as not investigated. According to the quantum-chemical + calculations, this molecule is probably an end-on O2 complex of Au which was predicted to be the thermochemically most stable isomer of this stoichiometry. A third class of compounds was studied in which high oxidation states can also be stabilised, namely the ternary oxide fluorides. In collaboration with the group of professor Andrews, the reaction of laser-ablated uranium and thorium as well as mercury atoms with oxygen difluoride was studied by matrix isolation experiments and quantum-chemical calculations. In the case of the two actinoid elements, this reaction led to the formation of the monoxide difluoride molecules ThOF2 and UOF2 in which the respective metal-oxygen bonds have pronounced triple bond character. In the reaction with mercury atoms, the OHgF molecule was formed which represents the first oxide fluoride of a group 12 element. However, the analysis of its electronic structure showed that it must be viewed as an oxyl complex of HgII. Analogous compound were also calculated for gold an several gold oxide fluoride molecules were predicted to exothermicly form in the reaction of gold atoms with OF or OF2. The most interesting of these oxide fluorides is OAuF which would be the first molecule containing a gold-oxygen double bond as well as the first gold species with a triplet ground state. In summary, these results show that new and interesting species containing metal atoms in high oxidation states can be generated by the reaction of laser-ablated metal atoms with F2 or OF2. These highly reactive species were stabilised using the matrix isolation technique and characterised by IR spectroscopy. The attribution of the new IR absorptions to these molecules was supported by state-of-the-art quantum-chemical calculations without the support of which the unambiguous assignment of the bands would not have been possible in all cases. Furthermore, several other so far unknown fluoride and oxide fluoride molecules were predicted to be stable and represent viable target molecules for future matrix isolation investigations. Appendix

A Technical drawings

Figure A1 : Technical drawing of the window holder.

103 104 APPENDIX

Figure A2 : Technical drawing of the holding plate.

Figure A3 : Technical drawing of the upper part of the matrix chamber. A. Technical drawings 105

Figure A4 : Technical drawing of the lower part of the matrix chamber (part 1).

Figure A5 : Technical drawing of the lower part of the matrix chamber (part 2). 106 APPENDIX

Figure A6 : Technical drawing of the motor holder.

Figure A7 : Technical drawing of the first magnet holder for the connection to the motor. A. Technical drawings 107

Figure A8 : Technical drawing of the second magnet holder for the connection to the target holder. 108 APPENDIX Bibliography

[1] N. Bartlett, D. H. Lohmann, J. Chem. Soc. 1962, 5253.

[2] N. Bartlett, P. Chem. Soc. London 1962, 218.

[3] P. Laszlo, G. J. Schrobilgen, Angew. Chem. Int. Ed. 1988, 27, 479.

[4] K. O. Christe, Angew. Chem. Int. Ed. 2001, 40, 1419.

[5] J. F. Lehmann, H. P. A. Mercier, G. J. Schrobilgen, Coord. Chem. Rev. 2002, 233-234, 1.

[6] S. Seidel, K. Seppelt, Angew. Chem. Int. Ed. 2000, 39, 3923.

[7] Y. Gong, M. Zhou, M. Kaupp, S. Riedel, Angew. Chem. Int. Ed. 2009, 48, 7879.

[8] X. Wang, L. Andrews, S. Riedel, M. Kaupp, Angew. Chem. Int. Ed. 2007, 46, 8371.

[9] P. Pyykk¨o, Phys. Chem. Chem. Phys. 2012, 14, 14734.

[10] C. K. Jørgensen, J. Chim. Phys. PCB 1979, 76, 630.

[11] M. Kaupp, H. G. von Schnering, Angew. Chem. Int. Ed. 1993, 32, 861.

[12] M. Kaupp, M. Dolg, H. Stoll, H. G. von Schnering, Inorg. Chem. 1994, 33, 2122.

[13] C. K. Jørgensen, Oxidation Numbers and Oxidation States 1st ed., Springer, 1969.

[14] D. Himmel, C. Knapp, M. Patzschke, S. Riedel, ChemPhysChem 2010, 11, 865.

[15] W. E. Falconer, G. R. Jones, W. A. Sunder, M. J. Vasile, A. A. Muenter, T. R. Dyke, W. Klemperer, J. Fluorine Chem. 1974, 4, 213.

[16] E. Jacob, H. Willner, Chem. Ber. 1990, 123, 1319.

[17] T. C. Ehlert, M. Hsia, J. Fluorine Chem. 1972, 2, 33.

[18] M. I. Nikitin, E. G. Rakov, V. I. Tsirel’nikov, S. V. Khaustov, Zh. Neorg. Khim. 1997, 42, 1354.

[19] M. I. Nikitin, E. G. Rakov, Zh. Neorg. Khim. 1998, 43, 375.

[20] M. V. Korobov, L. N. Savinova, L. N. Sidorov, J. Chem. Thermodyn. 1993, 25, 1161.

[21] J. V. Rau, S. Nunziante Cesaro, N. S. Chilingarov, G. Balducci, Inorg. Chem. 1999, 38, 5695.

109 110 BIBLIOGRAPHY

[22] W. Harnischmacher, R. Hoppe, Angew. Chem. Int. Ed. 1973, 12, 582.

[23] D. Kissel, R. Hoppe, Z. Anorg. Allg. Chem. 1988, 559, 40.

[24] Y. M. Kiselev, N. S. Kopelev, V. I. Spitsyn, L. I. Martynenko, Dokl. Akad. Nauk SSSR 1987, 292, 628.

[25] Y. M. Kiselev, G. V. Ionova, A. A. Kiseleva, N. S. Kopelev, A. P. Bobylev, A. B. Yatskevich, V. I. Spitsyn, Dokl. Akad. Nauk SSSR 1987, 293, 1407.

[26] Z. Cao, W. Wu, Q. Zhang, J. Mol. Struc. - Theochem. 1999, 489, 165.

[27] G. V. Chertihin, W. Saffel, J. T. Yustein, L. Andrews, M. Neurock, A. Ricca, C. W. Bauschlicher, J. Phys. Chem. 1996, 100, 5261.

[28] Y. Yamada, H. Sumino, Y. Okamura, H. Shimasaki, T. Tominaga, Appl. Radiat. Isotopes 2000, 52, 157.

[29] Y. Gong, M. Zhou, L. Andrews, J. Phys. Chem. A 2007, 111, 12001.

[30] C. R. Aita, J. Appl. Phys. 1987, 61, 5182.

[31] E. Wiberg, A. F. Hollemann, Lehrbuch der Anorganischen Chemie 101st ed., Walter de Gruyter, 1995.

[32] W. D. Bare, A. Citra, G. V. Chertihin, L. Andrews, J. Phys. Chem. A 1999, 103, 5456.

[33] A. A. Timakov, V. N. Prusakov, Y. V. Drobyshevskii, Dokl. Akad. Nauk SSSR 1986, 291, 125.

[34] D. Himmel, S. Riedel, Inorg. Chem. 2007, 46, 5338.

[35] X. Wang, L. Andrews, J. Am. Chem. Soc. 2011, 133, 3768.

[36] S. Riedel, M. Kaupp, Coord. Chem. Rev. 2009, 253, 606.

[37] S. Riedel, M. Renz, M. Kaupp, Inorg. Chem. 2007, 46, 5734.

[38] S. Riedel, M. Kaupp, Inorg. Chem. 2006, 45, 10497.

[39] O. Glemser, H. W. Roesky, K. H. Hellberg, H. U. Werther, Chem. Ber. 1966, 99, 2652.

[40] H. Shorafa, K. Seppelt, Inorg. Chem. 2006, 45, 7929.

[41] M. J. Molski, K. Seppelt, Dalton T. 2009, 3379.

[42] O. Glemser, H. Roesky, K. H. Hellberg, Angew. Chem. 1963, 75, 346.

[43] E. G. Hope, P. J. Jones, W. Levason, J. S. Ogden, M. Tajik, J. Chem. Soc. Chem. Comm. 1984, 1355.

[44] E. G. Hope, P. J. Jones, W. Levason, J. S. Ogden, M. Tajik, J. W. Turff, J. Chem. Soc. Dalton 1985, 1443.

[45] E. G. Hope, W. Levason, J. S. Ogden, Inorg. Chem. 1991, 30, 4873.

[46] J. Jacobs, H. S. P. Mueller, H. Willner, E. Jacob, H. Buerger, Inorg. Chem. 1992, 31, 5357. BIBLIOGRAPHY 111

[47] A. A. Timakov, V. N. Prusakov, Y. V. Drobyshevskii, Zh. Neorg. Khim. 1982, 27, 3007.

[48] G. Aull´on,S. Alvarez,´ Inorg. Chem. 2007, 46, 2700.

[49] S. Riedel, M. Kaupp, Inorg. Chem. 2006, 45, 1228.

[50] R. Craciun, D. Picone, R. T. Long, S. Li, D. A. Dixon, K. A. Peterson, K. O. Christe, Inorg. Chem. 2010, 49, 1056.

[51] R. Craciun, R. T. Long, D. A. Dixon, K. O. Christe, J. Phys. Chem. A 2010, 114, 7571.

[52] C. J. Marsden, P. P. Wolynec, Inorg. Chem. 1991, 30, 1681.

[53] A. Neuhaus, G. Frenking, C. Huber, J. Gauss, Inorg. Chem. 1992, 31, 5355.

[54] K. Pierloot, B. O. Roos, Inorg. Chem. 1992, 31, 5353.

[55] C. J. Marsden, D. Moncrieff, G. E. Quelch, J. Phys. Chem. 1994, 98, 2038.

[56] L. G. Vanquickenborne, A. E. Vinckier, K. Pierloot, Inorg. Chem. 1996, 35, 1305.

[57] T. Drews, J. Supe l, A. Hagenbach, K. Seppelt, Inorg. Chem. 2006, 45, 3782.

[58] A. D. Richardson, K. Hedberg, G. M. Lucier, Inorg. Chem. 2000, 39, 2787.

[59] R. Wesendrup, P. Schwerdtfeger, Inorg. Chem. 2001, 40, 3351.

[60] J. David, P. Fuentealba, A. Restrepo, Chem. Phys. Lett. 2008, 457, 42.

[61]L. Alvarez´ Thon, J. David, R. Arratia-P´erez, K. Seppelt, Phys. Rev. A 2008, 77, 034502.

[62] I. Norman, G. Porter, Nature 1954, 174, 508.

[63] E. Whittle, D. A. Dows, G. C. Pimentel, J. Chem. Phys. 1954, 22, 1943.

[64] L. Khriachtchev, M. Pettersson, N. Runeberg, J. Lundell, M. R¨as¨anen, Nature 2000, 406, 874.

[65] I. Infante, L. Andrews, X. Wang, L. Gagliardi, Chem. Eur. J. 2010, 16, 12804.

[66] W. Zou, Y. Liu, J. E. Boggs, Chem. Phys. Lett. 2009, 482, 207.

[67] X. Wang, L. Andrews, K. Willmann, F. Brosi, S. Riedel, Angew. Chem. Int. Ed. 2012, 51, 10628.

[68] T. C. DeVore, Inorg. Chem. 1976, 15, 1315.

[69] F. Brosi, Diplomarbeit, Albert-Ludwigs-Universit¨at Freiburg, 2011.

[70] F. Jensen, Introduction to Computational Chemistry 2nd ed., John Wiley & Sons Ltd., Chichester, UK, 2007.

[71] W. Koch, M. C. Holthausen, A Chemist’s Guide To Density Functional Theory, Wiley-VCH, Weinheim, 2000.

[72] M. Reiher, Chimia 2009, 63, 140.

[73] K. Pierloot, Int. J. Quantum Chem. 2011, 111, 3291. 112 BIBLIOGRAPHY

[74] P. Pyykk¨o, J. P. Desclaux, Acc. Chem. Res. 1979, 12, 276.

[75] P. Pyykk¨o, Chem. Rev. 1988, 88, 563.

[76] D. P. Tew, W. Klopper, T. Helgaker, J. Comput. Chem. 2007, 28, 1307.

[77] D. R. Hartree, Math. Proc. Cambridge 1928, 24, 89.

[78] D. R. Hartree, Math. Proc. Cambridge 1928, 24, 111.

[79] J. C. Slater, Phys. Rev. 1930, 35, 210.

[80] V. Fock, Z. Phys. 1930, 61, 126.

[81] P.-O. L¨owdin, Phys. Rev. 1955, 97, 1509.

[82] B. Roos, Chem. Phys. Lett. 1972, 15, 153.

[83] C. D. Sherrill, H. F. Schaefer III in Advances in Quantum Chemistry, Vol. 34, Academic Press, 1999, pp. 143–269.

[84] C. Møller, M. S. Plesset, Phys. Rev. 1934, 46, 618.

[85] J. C´ıˇzek,ˇ J. Chem. Phys. 1966, 45, 4256.

[86] J. C´ıˇzek,ˇ Advan. Chem. Phys. 1969, 14, 35.

[87] R. J. Bartlett, M. Musia l, Rev. Mod. Phys. 2007, 79, 291.

[88] A. Veillard, E. Clementi, Theor. Chim. Acta 1967, 7, 133.

[89] J. Hinze, J. Chem. Phys. 1973, 59, 6424.

[90] O. Sinano˘glu, Adv. Chem. Phys. 1964, 6, 315.

[91] D. K. W. Mok, R. Neumann, N. C. Handy, J. Phys. Chem. 1996, 100, 6225.

[92] A. Karton, E. Rabinovich, J. M. L. Martin, B. Ruscic, J. Chem. Phys. 2006, 125, 144108.

[93] T. J. Lee, J. E. Rice, G. E. Scuseria, H. F. Schaefer III, Theor. Chim. Acta 1989, 75, 81.

[94] T. J. Lee, P. R. Taylor, Int. J. Quantum Chem., Quantum Chem. Symp. 1989, 23, 199.

[95] D. Jayatilaka, T. J. Lee, J. Chem. Phys. 1993, 98, 9734.

[96] G. E. Scuseria, T. J. Lee, J. Chem. Phys. 1990, 93, 5851.

[97] J. Watts, M. Urban, R. Bartlett, Theor. Chim. Acta 1995, 90, 341.

[98] C. L. Janssen, I. M. B. Nielsen, Chem. Phys. Lett. 1998, 290, 423.

[99] I. M. B. Nielsen, C. L. Janssen, Chem. Phys. Lett. 1999, 310, 568.

[100] M. L. Leininger, I. M. B. Nielsen, T. D. Crawford, C. L. Janssen, Chem. Phys. Lett. 2000, 328, 431.

[101] T. J. Lee, Chem. Phys. Lett. 2003, 372, 362. BIBLIOGRAPHY 113

[102] W. Jiang, N. J. DeYonker, A. K. Wilson, J. Chem. Theory Comput. 2012, 8, 460.

[103] K. L. Bak, J. Gauss, P. Jorgensen, J. Olsen, T. Helgaker, J. F. Stanton, J. Chem. Phys. 2001, 114, 6548.

[104] S. Coriani, D. Marchesan, J. Gauss, C. H¨attig, T. Helgaker, P. Jørgensen, J. Chem. Phys. 2005, 123, 184107.

[105] D. Feller, K. A. Peterson, J. G. Hill, J. Chem. Phys. 2011, 135, 044102.

[106] B. O. Roos, P. R. Taylor, P. E. M. Siegbahn, Chem. Phys. 1980, 48, 157.

[107] P. E. M. Siegbahn, J. Alml¨of, A. Heiberg, B. O. Roos, J. Chem. Phys. 1981, 74, 2384.

[108] J. Olsen, Int. J. Quantum Chem. 2011, 111, 3267.

[109] K. Andersson, P. A.˚ Malmqvist, B. O. Roos, A. J. Sadlej, K. Wolinski, J. Phys. Chem. 1990, 94, 5483.

[110] K. Andersson, P. A.˚ Malmqvist, B. O. Roos, J. Chem. Phys. 1992, 96, 1218.

[111] C. J. Cramer, D. G. Truhlar, Phys. Chem. Chem. Phys. 2009, 11, 10757.

[112] J. Gr¨afenstein, D. Cremer, Theor. Chem. Acc. 2009, 123, 171.

[113] J. Gr¨afenstein, D. Cremer, J. Chem. Phys. 2009, 130, 124105.

[114] S. Grimme, J. Antony, S. Ehrlich, H. Krieg, J. Chem. Phys. 2010, 132, 154104.

[115] W. Jiang, C. C. Jeffrey, A. K. Wilson, J. Phys. Chem. A 2012, 116, 9969.

[116] A. Karton, S. Parthiban, J. M. L. Martin, J. Phys. Chem. A 2009, 113, 4802.

[117] S. Riedel, M. Straka, M. Kaupp, Phys. Chem. Chem. Phys. 2004, 6, 1122.

[118] M. A. Buijse, E. J. Baerends, J. Chem. Phys. 1990, 93, 4129.

[119] M. Kaupp, J. Comput. Chem. 2007, 28, 320.

[120] K. S. Pitzer, Acc. Chem. Res. 1979, 12, 271.

[121] M. Reiher, A. Wolf, Relativistic Quantum Chemistry, Wiley-VCH, Weinheim, 2009.

[122] P. Pyykk¨o, Chem. Rev. 2012, 112, 371.

[123] P. Pyykk¨o, Annu. Revs Phys. Chem. 2012, 63, 45.

[124] W. H. E. Schwarz, E. M. van Wezenbeek, E. J. Baerends, J. G. Snijders, J. Phys. B. - At. Mol. Opt. 1989, 22, 1515.

[125] E. J. Baerends, W. H. E. Schwarz, P. Schwerdtfeger, J. G. Snijders, J. Phys. B. - At. Mol. Opt. 1990, 23, 3225.

[126] J. Autschbach, S. Siekierski, M. Seth, P. Schwerdtfeger, W. H. E. Schwarz, J. Comput. Chem. 2002, 23, 804. 114 BIBLIOGRAPHY

[127] C. Thierfelder, P. Schwerdtfeger, A. Koers, A. Borschevsky, B. Fricke, Phys. Rev. A 2009, 80, 022501.

[128] M. Dolg in Modern Methods and Algorithms of Quantum Chemistry, Vol. 3, J. Grotendorst (Ed.), John von Neumann Institute for Computing, Julich,¨ 2000, pp. 479–508.

[129] P. Pyykk¨o, H. Stoll, A. Hinchliffe in Chemical Modelling: Applications and Theory, Vol. 1, The Royal Society of Chemistry, 2000, pp. 239–305.

[130] M. Dolg, P. Schwerdtfeger in Theoretical and Computational Chemistry, Vol. 11, Elsevier, 2002, pp. 793–862.

[131] M. Dolg, B. A. Heß in Relativistic Effects in Heavy-Element Chemistry and Physics, B. A. Heß (Ed.), John Wiley & Sons Ltd., Chichester, UK, 2003, pp. 89–122.

[132] M. Dolg, X. Cao in Computational Inorganic and Bioinorganic Chemistry, Vol. 1, E. I. Solomon, R. A. Scott, R. B. King (Eds.), Wiley, Chichester, 2009, pp. 503–515.

[133] M. Dolg, X. Cao, Chem. Rev. 2012, 112, 403.

[134] M. Douglas, N. M. Kroll, Ann. Phys. 1974, 82, 89.

[135] B. A. Heß, Phys. Rev. A 1985, 32, 756.

[136] B. A. Heß, Phys. Rev. A 1986, 33, 3742.

[137] G. Jansen, B. A. Heß, Phys. Rev. A 1989, 39, 6016.

[138] T. Nakajima, K. Hirao, Chem. Rev. 2012, 112, 385.

[139] C. Chang, M. P´elissier, P. Durand, Pys. Scr. 1986, 34, 394.

[140] E. van Lenthe, E. J. Baerends, J. G. Snijders, J. Chem. Phys. 1993, 99, 4597.

[141] S. Faas, J. G. Snijders, J. H. van Lenthe, E. van Lenthe, E. J. Baerends, Chem. Phys. Lett. 1995, 246, 632.

[142] E. van Lenthe, R. van Leeuwen, E. J. Baerends, J. G. Snijders, Int. J. Quantum Chem. 1996, 57, 281.

[143] M. Seth, F. Cooke, P. Schwerdtfeger, J.-L. Heully, M. P´elissier, J. Chem. Phys. 1998, 109, 3935.

[144] M. Seth, P. Schwerdtfeger, M. Dolg, J. Chem. Phys. 1997, 106, 3623.

[145] R. L. Deming, A. L. Allred, A. R. Dahl, A. W. Herlinger, M. O. Kestner, J. Am. Chem. Soc. 1976, 98, 4132.

[146] P. Hrob´arik, M. Kaupp, S. Riedel, Angew. Chem. 2008, 120, 8759.

[147] D. Naumann, T. Roy, K.-F. Tebbe, W. Crump, Angew. Chem. 1993, 105, 1555.

[148] J. P. Snyder, Angew. Chem. Int. Ed. 1995, 34, 80.

[149] J. P. Snyder, Angew. Chem. Int. Ed. 1995, 34, 986.

[150] M. Kaupp, H. G. von Schnering, Angew. Chem. Int. Ed. 1995, 34, 986. BIBLIOGRAPHY 115

[151] G. Aull´on,S. Alvarez,´ Theor. Chem. Acc. 2009, 123, 67.

[152] B. Zemvaˇ in Advanced Inorganic Fluorides: Synthesis, Characterisation and Applications, T. Naka- jima, B. Zemva,ˇ A. Tressaud (Eds.), Elsevier, 2000, chapter 4, pp. 79–115.

[153] A. J. Edwards, Proc. Chem. Soc., London 1963, 205.

[154] A. J. Edwards, W. E. Falconer, W. A. Sunder, J. Chem. Soc. Dalton 1974, 541.

[155] K. O. Christe, W. W. Wilson, R. A. Bougon, Inorg. Chem. 1986, 25, 2163.

[156] C. D. Garner, R. Mather, M. F. A. Dove, J. Chem. Soc. Chem. Comm. 1973, 633.

[157] J. Huang, K. Hedberg, J. M. Shreeve, S. P. Mallela, Inorg. Chem. 1988, 27, 4633.

[158] S. A. Siddiqui, Struct. Chem. 2012, 23, 267.

[159] B. Vest, P. Schwerdtfeger, M. Kolonits, M. Hargittai, Chem. Phys. Lett. 2009, 468, 143.

[160] E. Z. Zasorin, A. A. Ivanov, L. I. Ermolaeva, V. P. Spiridonov, Zh. Fiz. Khim. 1989, 63, 669.

[161] L. Hedberg, K. Hedberg, G. L. Gard, J. O. Udeaja, Acta Chem. Scand. 1988, A42, 318.

[162] E. J. Jacob, L. Hedberg, K. Hedberg, H. Davis, G. L. Gard, J. Phys. Chem. 1984, 88, 1935.

[163] I. M. B. Nielsen, M. D. Allendorf, J. Phys. Chem. A 2005, 109, 928.

[164] J. W. Hastie, R. Hauge, J. L. Margrave, J. Chem. Soc. Chem. Comm. 1969, 1452.

[165] V. N. Bukhmarina, A. Y. Gerasimov, Y. B. Predtechenskii, V. G. Shklyarik, Opt. Spektrosk. 1988, 65, 876.

[166] S. B. Osin, D. I. Davliasthin, J. S. Ogden, J. Fluorine Chem. 1996, 76, 187.

[167] B. Freyh, Bachelorarbeit, Albert-Ludwigs-Universit¨at Freiburg, 2013.

[168] A. Engelbrecht, A. V. Grosse, J. Am. Chem. Soc. 1954, 76, 2042.

[169] C. S. Nunziante, J. V. Rau, N. S. Chilingarov, G. Balducci, L. N. Sidorov, Inorg. Chem. 2001, 40, 179.

[170] N. Vogt, J. Mol. Struct. 2001, 570, 189.

[171] M. Hargittai, B. R´effy, M. Kolonits, C. J. Marsden, J.-L. Heully, J. Am. Chem. Soc. 1997, 119, 9042.

[172] V. G. Solomonik, V. V. Sliznev, N. B. Balabanov, Z. Phys. Chem. (Munich) 1997, 200, 77.

[173] P. V. Nhat, N. T. Cuong, P. K. Duy, M. T. Nguyen, Chem. Phys. 2012, 400, 185.

[174] G. V. Girichev, N. I. Giricheva, V. N. Petrova, S. A. Shlykov, E. G. Rakov, Zh. Strukt. Khim. 1994, 35, 61.

[175] V. N. Bukhmarina, Y. B. Predtechenskii, Opt. Spektrosk. 1996, 80, 762.

[176] W. Levason, C. A. McAuliffe, Coord. Chem. Rev. 1974, 12, 151. 116 BIBLIOGRAPHY

[177] J. J. Scepaniak, C. S. Vogel, M. M. Khusniyarov, F. W. Heinemann, K. Meyer, J. M. Smith, Science 2011, 331, 1049.

[178] J. F. Berry, E. Bill, E. Bothe, G. S. DeBeer, B. Mienert, F. Neese, K. Wieghardt, Science 2006, 312, 1937.

[179] J. V. Rau, C. S. Nunziante, N. S. Chilingarov, M. S. Leskiv, G. Balducci, L. N. Sidorov, Inorg. Chem. Commun. 2003, 6, 643.

[180] M. Hargittai, M. Kolonits, J. Tremmel, J.-L. Fourquet, G. Ferey, Struct. Chem. 1990, 1, 75.

[181] V. G. Solomonik, J. F. Stanton, J. E. Boggs, J. Chem. Phys. 2005, 122, 094322.

[182] V. G. Solomonik, J. F. Stanton, J. E. Boggs, J. Chem. Phys. 2008, 128, 244104.

[183] M. W. Chase Jr., NIST-JANAF thermochemical tables 4th ed., 1998.

[184] Y. B. Predtechenskii, L. D. Shcherba in Proceedings of the 12th All-Union Symposium, Minsk, USSR, 1989.

[185] R. D. Fowler, I. Burford, W. B., J. Hamilton, J. M., R. G. Sweet, C. E. Weber, J. S. Kasper, I. Litant, J. Ind. Eng. Chem. 1947, 39, 292.

[186] G. V. Chertihin, A. Citra, L. Andrews, J. Bauschlicher, Charles W., J. Phys. Chem. A 1997, 101, 8793.

[187] E. L. Uzunova, G. S. Nikolov, H. Mikosch, J. Phys. Chem. A 2002, 106, 4104.

[188] V. V. Sliznev, N. Vogt, J. Vogt, Mol. Phys. 2004, 102, 1767.

[189] A. G. Gershikov, V. P. Spiridonov, J. Struct. Chem. 1987, 27, 699.

[190] G. V. Girichev, N. I. Giricheva, O. G. Krasnova, A. V. Krasnov, S. A. Shlykov, . G. Rakov, S. V. Khaustov, J. Struct. Chem. 1993, 33, 838.

[191] G. Balducci, B. Brunetti, V. Piacente, J. Alloy. Compd. 1997, 260, 56.

[192] V. N. Bukhmarina, A. Y. Gerasimov, Y. B. Predtechenskii, Vib. Spectrosc. 1992, 4, 91.

[193] M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, G. Scalmani, V. Barone, B. Mennucci, G. A. Petersson, H. Nakatsuji, M. Caricato, X. Li, H. P. Hratchian, A. F. Izmaylov, J. Bloino, G. Zheng, J. L. Sonnenberg, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, T. Vreven, J. A. Montgomery, Jr., J. E. Peralta, F. Ogliaro, M. Bearpark, J. J. Heyd, E. Brothers, K. N. Kudin, V. N. Staroverov, R. Kobayashi, J. Normand, K. Raghavachari, A. Rendell, J. C. Burant, S. S. Iyengar, J. Tomasi, M. Cossi, N. Rega, J. M. Millam, M. Klene, J. E. Knox, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann, O. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski, R. L. Martin, K. Morokuma, V. G. Zakrzewski, G. A. Voth, P. Salvador, J. J. Dannenberg, S. Dapprich, A. D. Daniels, . Farkas, J. B. Foresman, J. V. Ortiz, J. Cioslowski, D. J. Fox, Gaussian 09 Revision B.1, Gaussian Inc. Wallingford CT 2009.

[194] A. D. Becke, Phys. Rev. A 1988, 38, 3098.

[195] A. Becke, J. Chem. Phys. 1993, 98, 5648. BIBLIOGRAPHY 117

[196] C. Lee, W. Yang, R. G. Parr, Phys. Rev. B 1988, 37, 785.

[197] R. Kendall, J. Chem. Phys. 1992, 96, 6796.

[198] N. B. Balabanov, K. A. Peterson, J. Chem. Phys. 2005, 123, 064107.

[199] T. Helgaker, W. Klopper, H. Koch, J. Noga, J. Chem. Phys. 1997, 106, 9639.

[200] D. E. Woon, J. T. H. Dunning, J. Chem. Phys. 1995, 103, 4572.

[201] CFOUR, a quantum chemical program package written by J.F. Stanton, J. Gauss, M.E. Harding, P.G. Szalay with contributions from A.A. Auer, R.J. Bartlett, U. Benedikt, C. Berger, D.E. Bernholdt, Y.J. Bomble, L. Cheng, O. Christiansen, M. Heckert, O. Heun, C. Huber, T.-C. Jagau, D. Jonsson, J. Jus´elius, K. Klein, W.J. Lauderdale, D.A. Matthews, T. Metzroth, L.A. Muck,¨ D.P. O’Neill, D.R. Price, E. Prochnow, C. Puzzarini, K. Ruud, F. Schiffmann, W. Schwalbach, S. Stopkowicz, A. Tajti, J. V´aquez, F. Wang, J.D. Watts and the integral packages MOLECULE (J. Alml¨of and P.R. Taylor), PROPS (P.R. Taylor), ABACUS (T. Helgaker, H.J. Aa. Jensen, P. Jørgensen, and J. Olsen), and ECP routines by A. V. Mitin and C. van Wullen.¨ For the current version, see http://www.cfour.de.

[202] R. Yang, Y. Gong, H. Zhou, M. Zhou, J. Phys. Chem. A 2006, 111, 64.

[203] M. A. Duncan, Annu. Revs Phys. Chem. 1997, 48, 69.

[204] M. A. Duncan, Int. J. Mass Spectrom. 2000, 200, 545.

[205] A. Citra, L. Andrews, J. Phys. Chem. A 1999, 103, 4845.

[206] A. Citra, L. Andrews, J. Phys. Chem. A 1999, 103, 4182.

[207] Y. Gong, M. Zhou, J. Phys. Chem. A 2009, 113, 4990.

[208] K. Pierloot, S. Vancoillie, J. Chem. Phys. 2006, 125, 124303.

[209] M. Swart, J. Chem. Theory Comput. 2008, 4, 2057.

[210] L. C. O’Brien, A. E. Oberlink, B. O. Roos, J. Phys. Chem. A 2006, 110, 11954.

[211] L. Belpassi, I. Infante, F. Tarantelli, L. Visscher, J. Am. Chem. Soc. 2007, 130, 1048.

[212] C. J. Evans, D. S. Rubinoff, M. C. L. Gerry, Phys. Chem. Chem. Phys. 2000, 2, 3943.

[213] C. J. Evans, A. Lesarri, M. C. L. Gerry, J. Am. Chem. Soc. 2000, 122, 6100.

[214] C. J. Evans, T. G. Wright, A. M. Gardner, J. Phys. Chem. A 2010, 114, 4446.

[215] D. E. Woon, J. T. H. Dunning, J. Chem. Phys. 1993, 98, 1358.

[216] K. A. Peterson, D. Figgen, M. Dolg, H. Stoll, J. Chem. Phys. 2007, 126, 124101.

[217] D. Figgen, K. A. Peterson, M. Dolg, H. Stoll, J. Chem. Phys. 2009, 130, 164108.

[218] J. P. Perdew, Phys. Rev. B 1986, 33, 8822.

[219] TURBOMOLE V6.4 2012, a development of University of Karlsruhe and Forschungszen- trum Karlsruhe GmbH, 1989–2007, TURBOMOLE GmbH, since 2007; available from http://www.turbomole.com. 118 BIBLIOGRAPHY

[220] H.-J. Werner, P. J. Knowles, R. Lindh, M. Schutz,¨ P. Celani, T. Korona, F. R. Manby, G. Rauhut, R. D. Amos, A. Bernhardsson, A. Berning, D. L. Cooper, M. J. O. Deegan, A. J. Dobbyn, F. Eckert, C. Hampel, G. Hetzer, A. W. Lloyd, S. J. McNicholas, W. Meyer, M. E. Mura, A. Nicklaß, P. Palmieri, R. Pitzer, U. Schumann, H. Stoll, A. J. Stone, R. Tarroni, T. Thorsteinsson, MOLPRO, version 2006.1, a package of ab initio programs, 2006, see http://www.molpro.net.

[221] H.-J. Werner, Mol. Phys. 1996, 89, 645.

[222] B. O. Roos, K. Andersson, Chem. Phys. Lett. 1995, 245, 215.

[223] Y. Gong, L. Andrews, C. W. Bauschlicher, K. S. Thanthiriwatte, D. A. Dixon, Dalton T. 2012, 41, 11706.

[224] Y. Gong, L. Andrews, C. W. Bauschlicher, J. Phys. Chem. A 2012, 116, 10115.

[225] Y. Gong, L. Andrews, C. W. Bauschlicher, Chem. Eur. J. 2012, 18, 12446.

[226] S. Riedel, M. Straka, M. Kaupp, Chem. Eur. J. 2005, 11, 2743.

[227] D. Kurzyd lowski, W. Grochala, Chem. Commun. 2008, 1073.

[228] D. Schr¨oder, J. Hruˇs´ak,I. C. Tornieporth-Oetting, T. M. Klap¨otke, H. Schwarz, Angew. Chem. 1994, 106, 223.

[229] X. Wang, L. Andrews, J. Phys. Chem. A 2001, 105, 5812.

[230] R. G. Denning, J. Phys. Chem. A 2007, 111, 4125.

[231] C. R. Graves, J. L. Kiplinger, Chem. Commun. 2009, 3831.

[232] T. W. Hayton, Dalton T. 2010, 39, 1145.

[233] R. J. Baker, Chem. Eur. J. 2012, 18, 16258.

[234] R. T. Paine, R. R. Ryan, L. B. Asprey, Inorg. Chem. 1975, 14, 1113.

[235] B. Liang, L. Andrews, J. Li, B. E. Bursten, J. Am. Chem. Soc. 2002, 124, 6723.

[236] B. Liang, R. D. Hunt, G. P. Kushto, L. Andrews, J. Li, B. E. Bursten, Inorg. Chem. 2005, 44, 2159.

[237] Y. Gong, L. Andrews, Dalton T. 2011, 40, 11106.

[238] Y. Gong, L. Andrews, Inorg. Chem. 2011, 50, 7099.

[239] L. Andrews, X. Wang, R. Lindh, B. O. Roos, C. J. Marsden, Angew. Chem. Int. Ed. 2008, 47, 5366.

[240] M. Kaupp, Angew. Chem. Int. Ed. 2001, 40, 3534.

[241] M. C. Heaven, J.-P. Nicolai, S. J. Riley, E. K. Parks, Chem. Phys. Lett. 1985, 119, 229.

[242] P. Pyykk¨o, S. Riedel, M. Patzschke, Chem. Eur. J. 2005, 11, 3511.

[243] C. T. Dewberry, K. C. Etchison, S. A. Cooke, Phys. Chem. Chem. Phys. 2007, 9, 4895.

[244] R. Krishnan, J. S. Binkley, R. Seeger, J. A. Pople, J. Chem. Phys. 1980, 72, 650.

[245] W. Kuchle,¨ M. Dolg, H. Stoll, H. Preuss, J. Chem. Phys. 1994, 100, 7535.

[246] A. E. Reed, L. A. Curtiss, F. Weinhold, Chem. Rev. 1988, 88, 899.