Technisch-Naturwissenschaftliche Fakultät
Synthesis, Characterization and Reactivity of Functionalized Iron-Sulfur Clusters as Bioinspired Hydrogenase Models
DISSERTATION
zur Erlangung des akademischen Grades
Doktor der Technischen Wissenschaften
im Doktoratsstudium der
Technischen Wissenschaften
Eingereicht von: DI Manuel Kaiser
Angefertigt am: Institut für Anorganische Chemie
Beurteilung: Prof. Dr. Günther Knör
Assoc.-Prof. Mario Waser
Mitwirkung:
Linz, November 2015 Eidesstattliche Erklärung
Ich erkläre an Eides statt, dass ich die vorliegende Dissertation selbstständig und ohne fremde Hilfe verfasst, andere als die angegebenen Quellen und Hilfsmittel nicht benutzt bzw. die wörtlich oder sinngemäß entnommenen Stellen als solche kenntlich gemacht habe. Die vorliegende Dissertation ist mit dem elektronisch übermittelten Textdokument identisch.
Linz, November 2015
______Manuel Kaiser
I Dipl.‐Ing. Manuel Kaiser
Persönliche Daten Geburtsdatum 15.11.1985 Staatsangehörigkeit Österreich Familienstand ledig Telefon 0699/12123124 Mail [email protected] Adresse Johann‐Wilhelm‐Klein‐Straße 2‐4, 4040 Linz
Ausbildung 10/2012 – 12/2015 Doktoratsstudium Technische Wissenschaften, JKU Linz Dissertation am Institut für Anorganische Chemie betreut von Prof. Dr. Günther Knör: „Synthesis, Characterization and Reactivity of Functionalized Iron‐Sulfur Clusters as Bioinspired Hydrogenase Models“
10/2005 – 09/2012 Diplomstudium Technische Chemie, JKU Linz (Abschluss: Dipl.‐Ing.) Diplomarbeit am Institut für Anorganische Chemie betreut von Dr. Uwe Monkowius: „Imidazol‐funktionalisierte Azobenzole als photoschaltbare N‐heterocyclische Carben‐Liganden“
09/1996 – 07/2005 BRG Landwiedstraße, Linz (Abschluss: Matura)
Grundwehrdienst 01/2005 – 09/2005 Jägerbataillon 12, Ostarrichikaserne, Amstetten
Berufserfahrung 2011 Ferialarbeiter bei DSM Fine Chemicals Austria (jetzt: DPx Fine Chemicals), Linz
2006‐2010 Ferialarbeiter bei voestalpine Stahl, Linz
Zusätzliche Qualifikation 2012‐2015 Präsentation von aktuellen Forschungsergebnissen bei nationalen und internationalen Konferenzen
2012‐2015 Tutoren‐ und Lektorentätigkeit bei mehreren Laborpraktika inklusive mehrmalige Organisation und Gesamtverantwortung
II 2013‐2015 Teilnahme an Seminaren zur Weiterbildung: „Grundausbildung für NachwuchswissenschafterInnen“ „Studienrecht für Prüfende, Lehrende und Interessierte“ „Bei Stimme bleiben“ „Wirkung und Präsenz für Vorlesungen und Vorträge“
Fremdsprachen Englisch (verhandlungsfähig)
EDV MS Office, ChemOffice, Citavi, Origin, TopSpin, GroupWise
Hobbies Laufen, Kino, Lesen
III Parts of this dissertation have already been published in peer‐reviewed journals or presented on national and international conferences as a poster or as a talk:
Journal Articles M. Kaiser, G. Knör, Eur. J. Inorg. Chem. 2015, 25, 4199–4206. DOI: 10.1002/ejic.201500574
Talks
8. WACÖ (Workshop Anorganische Chemie Österreich, Salzburg, 14.‐15.04.2014) M. Kaiser, G. Knör: “Synthese und Charakterisierung mehrzähniger Schwefel‐Liganden für die Entwicklung neuer [FeFe]‐Hydrogenase Modellverbindungen”
ISBOMC14 (7th International Symposium on Bioorganometallic Chemistry, Vienna, 22.‐25.07.2014) M. Kaiser, G. Knör: “Synthesis and Characterization of Bidentate Sulfur‐Ligands for the Development of Novel [FeFe]‐Hydrogenase Model Compounds”
16th Austrian Chemistry Days 2015 (Innsbruck, 21.‐23.09.2015) M. Kaiser, G.Knör: “Synthesis and Characterization of Iron‐Sulfur‐Clusters for the Development of Novel Hydrogenase Model Compounds”
Poster presentations
9. Koordinationschemie‐Treffen (Bayreuth, 24.‐26.02.2013) M. Kaiser, G. Knör, U. Monkowius, C. Topf: “Synthesis and Characterization of a Modular Bridging Ligand Platform for Bio‐inspired Hydrogen Production”
15th Austrian Chemistry Days 2013 (Graz, 23.‐26.09.2013) M. Kaiser, G. Knör, U. Monkowius, C. Topf: “Synthesis and Characterization of novel heterodinuclear metal complexes for bio‐inspired hydrogen production”
ISF‐1 & ISF‐1 Young (1st International Solar Fuels Conference & Symposium for Young Scientists, Uppsala, 24.04.‐01.05.2015) M. Kaiser, G. Knör: “Synthesis and Characterization of Iron‐Sulfur‐Clusters for the Development of Novel Hydrogenase Model Compounds”
ICAME 2015 (International Conference on the Application of the Mössbauer Effect, Hamburg, 13.‐18.09.2015) M. Grodzicki, M. Kaiser, G. Knör, J. Schoiber, G. Tippelt: “Mössbauer Spectroscopy and Electronic
Structure Calculations on the Trinuclear Iron Sulfur Cluster Fe3S2(CO)7dppm”
IV Research Projects
FWF Project (P25038) G. Knör, M. Kaiser, E. Kianfar: “Functional Light‐Responsive Metal Carbonyl Systems”, 2012‐2015.
LaserLab Europe Project (llams_2062) G. Knör, M. Kaiser, S. Woutersen, S. Amirjalayer: “Unraveling the Mechanism of Fe‐based Photocatalytic Complexes for Solar Hydrogen Production”, 2014‐2015.
V Kurzfassung
Diese Dissertation behandelt das spannende Feld der Hydrogenase Modellverbindungen. Durch den steigenden Energiebedarf auf unserem Planeten ist es notwendig neue Wege zu beschreiten um diesem Bedarf gerecht zu werden. Eine Möglichkeit ist die Wasserstofferzeugung nach dem natürlichen Vorbild der Hydrogenasen. Besonders interessant ist die Möglichkeit ohne teure Edelmetalle wie Platin, welches nach wie vor hauptsächlich als Katalysator in der Industrie eingesetzt wird, auszukommen und stattdessen billige Metalle, die in reichlichem Ausmaß vorhanden sind, einsetzen zu können. Vorgestellt wird die Synthese und Charakterisierung zweier neuer [FeFe]‐Hydrogenase Modellverbindungen auf Basis von [3Fe2S] Clustern die mit Diphosphinen verbrückt sind. Im einfachsten Fall ist diese Brücke lediglich eine Methyleneinheit, eine weitere Verbindung wird durch das elektronenreiche Ferrocen verbrückt. Besonders wichtig ist dabei das elektrochemische und photochemische Verhalten zu unterscheiden. Daher konzentriert sich ein bedeutender Teil dieser Arbeit auf Belichtungsversuche und cylcovoltammetrische Untersuchungen bei unterschiedlichen Bedingungen. Um tatsächlich die Wasserstofferzeugung nachweisen zu können wurden zusätzlich Elektrolysen mit Headspace‐GC Analysen durchgeführt. Des Weiteren wurden auch weiterführende Untersuchungen wie Mößbauer Spektroskopie oder Messungen im Bereich der Ultrakurzzeit Spektroskopie mit Kollaborationen in Salzburg und Amsterdam durchgeführt. So konnte mit der vorliegenden Arbeit eine wichtige Grundlage für die Bereitstellung neuartiger auf Eisen basierter Katalysatorsysteme zur Gewinnung von Wasserstoff nach dem Vorbild der Natur gelegt werden wobei Teile dieser Arbeit bereits publiziert wurden.
VI Abstract
This dissertation covers the exciting field of hydrogenase model compounds. Due to the increasing energy demand on our planet, it is necessary to pursue new ways to satisfy this demand. One possibility is the generation of hydrogen as in natural hydrogenases. Of particular interest is the possibility to avoid expensive precious metals such as platinum, which is still the mainly used catalyst in industry, but instead to employ cheap earth‐abundant metals as catalysts. The syntheses and characterisation of two new [FeFe]‐hydrogenase model compounds based on a [3Fe2S] cluster bridged with diphosphines is presented. One compound is bridged by methylene, which is the simplest case and a second compound is bridged by the electron‐rich ferrocene subunit. It is especially important to differentiate between electrochemical and photochemical properties. For this reason, a substantial part of this work concentrates on irradiation experiments and investigations with cyclic voltammetry with variable parameters. To prove the hydrogen evolution, electrolysis with headspace‐GC analysis has been performed as well. Additionally, more advanced investigations such as Mössbauer spectroscopy or measurements in the field of ultrafast vibrational spectroscopy with collaborations in Salzburg and Amsterdam have been performed. The work presented here has built a foundation for the application of novel iron‐based catalyst systems developed for bioinspired generation of hydrogen. Parts of this work have already been published.
VII Table of Contents
1 Introduction ...... 1
1.1 Hydrogen production ...... 1
1.2 Hydrogenases ...... 2
1.2.1 The role of x‐ray crystallography ...... 3
1.3 Hydrogenase Model compounds ...... 4
1.4 Cyclic Voltammetry ...... 6
1.5 Mössbauer spectroscopy...... 6
1.6 Ultrafast vibrational spectroscopy ...... 7
2 Practical Work ...... 8
2.1 Objective...... 8
2.2 Alternatives ...... 8
2.3 Synthesis and Characterization of Novel Hydrogenase Model Compounds ...... 9
2.3.1 Single crystal x‐ray diffraction ...... 11
2.3.2 NMR Characterization ...... 16
2.3.3 Spectroscopic Characterization ...... 20
2.3.4 Photochemical investigations ...... 28
2.4 Electrochemistry ...... 38
2.4.1 Cyclic voltammetry ...... 38
2.4.2 Bulk electrolysis ...... 44
2.5 Mössbauer spectroscopy and calculations ...... 47
2.6 Ultrafast Vibrational Spectroscopy ...... 49
3 Summary ...... 52
4 Experimental Section ...... 53
5 Appendix ...... 55
5.1 Single crystal x‐ray diffraction data ...... 55
5.2 UV‐vis spectroscopy ...... 57
VIII 5.3 IR‐spectroscopy ...... 58
5.4 NMR‐spectroscopy ...... 61
6 Acknowledgment ...... 62
7 Table of Figures ...... 63
8 References ...... 67
IX Abbreviations
1 Fe3S2(CO)7(dppm)
2 Fe3S2(CO)7(dppf)
ACN acetonitrile
ATR attenuated total reflection
CV cyclic voltammetry
DCM dichloromethane dppm bis(diphenylphosphinoyl)methane dppmS2 bis(diphenylthiophosphinoyl)methane dppe 1,2‐bis(diphenylphosphinoyl)ethane dppf 1,1'‐bis(diphenylphosphino)ferrocene dppfS2 1,1'‐bis(diphenylthiophosphino)ferrocene dppp 1,3‐bis(diphenylphosphinoyl)propane
Et2O diethyl ether eq. equivalent
Fc ferrocene
FTIR Fourier transform infrared spectroscopy
GC gas chromatography
H2ase hydrogenase
MLCT metal‐to‐ligand charge transfer
NMR nuclear magnetic resonance
OPA optical parametric amplifier
ORTEP Oak Ridge Thermal Ellipsoid Plot ppm parts per million
TBAP tetrabutylammonium hexafluorophosphate
TEOA triethanolamine
TFA trifluoroacetic acid
X THF tetrahydrofuran
TMEDA N,N,N',N'‐tetramethylethylendiamine
TOF turnover frequency
UV‐vis ultraviolet‐visible
Zn(TTP) 5,10,15,20‐tetrakis(4‐tolyl)porphyrinato‐zinc(II)
XI 1 Introduction
1.1 Hydrogen production
Hydrogen is not a primary energy source, but has to be produced from one, which means it is only an energy carrier. Besides many worldwide debates and initiatives, the global energy market still utilizes by far more non‐renewable primary energy sources than renewables such as solar or wind energy for hydrogen production.[1] Water, mostly seawater, is a resource that is commonly available on our planet. It makes sense to produce hydrogen from this abundant and cheap resource. A serious problem with that statement is a very high energy demand for the water splitting reaction[2] in the liquid phase (1a) and also in the gaseous phase (1b):
1 286.02 → (1a) 2 1 241.98 → (1b) 2
Compared to the splitting reaction of methane[2] (2), liquid water needs almost eight times more energy to release the same amount of hydrogen considering that one mole methane releases two moles of hydrogen.
(2) 74.86 →2
Therefore, by far the highest amount of hydrogen is nowadays produced by steam reforming of non‐ renewable fossil fuels, mostly methane. Other methods such as partial oxidation or gasification processes also rely heavily on fossil fuels.[3] Without sophisticated highly active catalysts, it is unimaginable to produce hydrogen from water on a global scale. A very small fraction is produced by electrolysis and there only a minor part is directly produced from water for a dedicated hydrogen production. The major part is a valuable by‐product of other electrolytic processes such as the chloralkali process. In these cases, hydrogen is sold to reduce the overall costs of the very energy demanding electrolysis. However, an advantage is the very high purity of the produced hydrogen, which renders purification steps unnecessary.[4] Still only a niche is the production from algae, which utilize hydrogenases in a metabolic pathway.[5,6] A major drawback is the inactivation of the enzyme by oxygen, which necessitates strict anaerobic
1 conditions for a hydrogen evolution reaction.[7] There is a huge effort in general to produce various forms of biofuels, including not only hydrogen, but also ethanol or even complex hydrocarbons.[8] Solving the energy problem can only be successful by a combined global effort. A proposed hydrogen economy has drawbacks as has any other economy.[9] But the “fossil‐fuel economy” has by far the largest drawbacks of being non‐renewable and malign on a very broad scale. Cost factors will diminish over time as with any other technologies ever. Ultimately, it will be a conversion problem from a very well established, but detrimental oil and coal based society, to a more and more mixed energy market where hydrogen will play a role.
1.2 Hydrogenases
Stephenson and Strickland discovered in the 1930s a bacterial enzyme that could activate molecular hydrogen and named it, to conform to the already known dehydrogenase, “hydrogenase”.[10] It has to be noted that the term “hydrogenase” refers to the whole enzyme, with different structural features. For this reason the source of the enzyme, the bacterial strain from where it has been extracted, has to be cited. For example the first published iron hydrogenase structure was from desulfovibrio desulfuricans.[11] For simplicity reasons, in this work “hydrogenase” refers only to the active sites, which display common features in all hydrogenase enzymes.
Today, three different types of hydrogenases are known: [NiFe]H2ase, [FeFe]H2ase and [Fe]H2ase (Figure 1). The last one was not found until 1990 to have hydrogenase activity and initially was thought to contain no metals.[12]
Figure 1: Active sites of the three hydrogenase types.
2 Although there was a huge effort to determine the structure and especially the active sites of all hydrogenases, without sophisticated x‐ray measurements, a complete understanding of structural features would have been very difficult to achieve. All three hydrogenases contain at least one low‐valent iron atom and strong‐field CO ligands.
‐ Cysteine bridges bind the active site to the rest of the enzyme. The bimetallic H2ases contain CN as ligands, which stem from a maturation process that utilizes thiocyanate.[13] This is remarkable since a metal to CN‐ bond is probably nowhere else found in nature. Counting all of the iron atoms in the
[2Fe2S] and the [4Fe4S] cluster – this subunit is called the H‐cluster – of [FeFe]H2ase one ends up at even six iron atoms, which make up about half of the molecular weight of this hydrogenase type. A very comprehensive review article about hydrogenases has been published, which covers findings in spectroscopy and structure including model compounds and biological backgrounds in detail.[14]
1.2.1 The role of x‐ray crystallography
The first successful x‐ray structure determination of a hydrogenase was of [NiFe]H2ase in 1995 performed by the group of Fontecilla‐Camps.[15] Three years later, the same group,[11] this time
[16] independently with another one led by Peters, published the x‐ray structure of [FeFe]H2ase. Even though there was some understanding of the mechanism of hydrogen activation, aside from the fact that there is a [Fe‐S] cluster unlike any other known to that date, structural features were mostly unknown. Both groups did not manage to reveal all details of the active site. They missed the bridging CO ligand between the iron atoms – the Fontecilla‐Camps group correctly expected an oxygen containing ligand – and the assignment of isoelectronic CO/CN was not clear at the time. Nonetheless, it was the beginning of a new topic and many improvements led to the elucidation of the structure known today.
The x‐ray structure of [Fe]H2ase was published 20 years after the other two and it has only very little similarities with the formerly known hydrogenases.[17]
[18] The most recent structural study on [FeFe]H2ase was published in 2014, which is a reflection upon the improvements over the last years in x‐ray crystallography. Especially high‐energy sources, such as synchrotrons, which also have been utilized in the cited work, are a significant milestone. Today, the brightest available sources are x‐ray free‐electron lasers (X‐FEL).[19] Although it is a destructive technique, this disadvantage is countered by a very fast measurement with the concept ‘diffraction before destruction’. In this way, nanocrystals can be measured that are otherwise not suitable for diffraction, although comparably much more of a sample is needed. Nonetheless, many problems could not be solved otherwise, an example being the diffraction from two‐dimensional proteins with only 10 nm thickness.[20]
3 1.3 Hydrogenase Model compounds
After the structure determination of natural hydrogenases, model compounds have been widely studied in the last two decades. One of the first researchers to deal with hydrogenase model compounds was Marcetta Darensbourg[21] who had published works already in 1994,[22] which even predates the publication of the first x‐ray structure. Very early works also include publications by the groups of Fraser Armstrong[23], Thomas Rauchfuss[24] and Christopher Pickett[25] in 1999. Later, Sascha Ott together with Björn Akermak and Licheng Sun published hydrogenase models with a ruthenium linker.[26] As an advocate for noble metals in catalysis albeit their high price and relative low availability, Licheng Sun coined the sentence: “Noble metals do a noble job”[27] and recently focuses more on dye sensitized solar cells within the solar fuels community.[28,29] Vincent Artero describes general principles for electrocatalysts with hydrogenase activity and important requirements for a successful system.[30]
Considering the simplest model for [FeFe]H2ase, which would be Fe2(S2C3H6)(CO)6, Pickett et al. performed digital simulations and collected experimental data from which they derived a proposed mechanism for hydrogen evolution.[31] Later, Greco et al. refined this model further in which they demonstrated that the hydrogen can be bridged between the iron atoms as well as on the sulfur atom after a rearrangement to a bridging CO ligand (Figure 3).[32] Interesting for our own work is a model also described by Greco’s group (Figure 2) because it contains a ferrocene and a phosphine moiety, albeit in a totally different arrangement.[33]
Figure 2: Hydrogenase model compound containing a ferrocene and a phosphine subunit.[33]
4
Figure 3: Reaction scheme for dihydrogen evolution by Greco et al.[32]
One paper even reports the synthesis of a system comprising of the entire H‐cluster including the [4Fe4S] cubane cluster.[34] The main focus in the literature is on di‐iron core compounds to mimic the hydrogenase function, although some publications additionally cover tri‐nuclear clusters.[35–38] An excellent review article that covers a very broad range of different approaches of hydrogenase models in general has been published.[39] It includes also a variety of supramolecular control and stability methods, such as the employment of cyclodextrins, micelles, dendrimers or gels and resins. In newer publications, the natural hydrogenase structure is sometimes not apparent at first sight.[40]
5 1.4 Cyclic Voltammetry
Cyclic voltammetry is only one of many electrochemical techniques and has been established as an important tool to investigate properties of catalysts. In a typical three‐electrode setup a working electrode, a counter electrode and a reference electrode on a potentiostat are the basic parts in cyclic voltammetry and electrochemical analysis in general. Very important for cyclic voltammetry is exclusion of oxygen to avoid a reductive signal. Therefore, it is best performed inside a glovebox although careful bubbling with nitrogen or argon are also possible. Even though it is already 30 years old, a paper by Heinze gives a profound overview on the topic.[41] Referencing and comparing results is complicated by manifold parameters and reference points. It is advisable to reference measurements versus the ferrocene/ferrocenium couple and not for example versus a silver/silver salt electrode, which can show deviations. In addition, the reference points of the resulting voltammograms differ in publications. The group of Artero et al. proposed to reference at the half‐wave potential for catalysts for hydrogen evolution in non‐aqueous solvents.[42] A general protocol for benchmarking of those types of homogeneous catalysts has been reported.[43]
1.5 Mössbauer spectroscopy
Rudolph Mössbauer discovered the effect named after him in 1958[44–46] in 191Ir and was awarded the Nobel Prize in physics in 1961 at the young age of only 32. The effect has been found in more than 40 elements, but only about 15 have applications, such as tin or gold. The optimal isotope for Mössbauer spectroscopy is 57Fe, which was also employed for this work. The aim of the method is to bring the absorption line and the emission line to an overlap, the so‐called nuclear fluorescence resonance. This overlap is detected relative to a moving source, resulting in a Doppler shift relative to this source. The most important application is the differentiation between the +2 and +3 oxidation state of iron. Important parameters for this evaluation are the isomer shift δ and the quadrupole splitting ΔEQ. The isomer shift stems from electronic monopole interactions between the nucleus and s‐electrons, the quadrupole splitting is the result of either nonspherical nuclear charge distribution leading to a quadrupole moment, or an inhomogeneous electric field at the Mössbauer nucleus. Mössbauer spectroscopy only works in the solid state or in frozen liquids where movements of atoms is rather small, which is the prerequisite for nuclear resonance absorption and fluorescence. With this technique it is possible to determine oxidation and spin states as well as bond properties and even electronegativity of ligands.[47]
6 Oxidation states can be very difficult to determine, which is the reason why this method has been employed. Interesting results were then compared to calculations to conclude a more complete picture of the oxidation states of the iron atoms as well as of the (de)localization of electrons around the common iron‐sulfur system in the synthesised compounds.
1.6 Ultrafast vibrational spectroscopy
In cooperation with Laserlab‐Europe, a consortium of laser‐based inter‐disciplinary research sites, ultrafast vibrational spectroscopy or time‐resolved infrared spectroscopy was performed at the LLAMS (LaserLab Amsterdam). There, pulsed‐pump pulsed‐probe experiments have been performed. At a certain wavelength in the UV‐vis range, a fs laser pulse is pumped into the sample and probed in the infrared region, more exactly in the regime of M‐CO vibrations of 2000 ± 100 cm‐1. In Figure 4, a schematic representation of the setup is shown. Two OPAs are needed to adjust laser wavelengths for the UV‐vis pump pulse and for the IR probe pulse, respectively. With a delay table, mirrors are moved to measure the IR spectra after certain delay times after the incident pump pulse. In this way, differential spectra are recorded time‐resolved to determine structural changes and electronic density upon excitation.
Figure 4: Schematic setup for the pump‐probe experiment: The beam of the laser source is converted in one OPA into the desired pump wavelength, in the other OPA, the desired IR wavelength is adjusted. A table with moving mirrors adjusts the delay between pump and probe pulses, which go through the sample (S) to the detector (D).
It has been shown before, that this is an excellent method to determine electron transfer reactions in hydrogenase model compounds.[48] Furthermore, it was of great interest to gain information about the photostability, especially if there is dissociation of carbonyl ligands, and about lifetimes of exited states, which is important for subsequent reactions.
7 2 Practical Work
2.1 Objective
Originally, the aim of this work was to further develop the design of [FeFe]hydrogenase model compounds already established in our group especially by prior work of Christoph Topf.[49,50] A simple method to influence reactivity by increasing the electron density of those model compounds (Figure 5) is to substitute the CO ligands for stronger σ‐donors. Phosphines or carbenes are very well suited ligands, also studied at our institute, and would be ideal candidates. In addition to that, it was anticipated that quaternization of nitrogen atoms might lead to water solubility of the system.
Figure 5: [FeFe]hydrogenase model compound (left) and CO substituents proposed by Topf. [50]
One prerequisite for an efficient system is a simple synthesis. Unfortunately, prolonged difficulties in synthesis proved that this criterion could not be met and the original idea had to be abandoned. Therefore, a more straightforward approach was desired.
2.2 Alternatives
Inspired by the work of Ott et al. single iron core complexes as “Minimal Functional Models” seemed a suitable alternative.[51,52] However, their approach requires benzenedithiol derivatives (benzene‐ 1,4‐dithiol and 3,6‐dichloro‐1,2‐benzenedithiol), which can be problematic for several reasons. They are relatively expensive, which would diminish the overall effectiveness and, as all organosulfur compounds, they exhibit pungent odour, which we wanted to avoid if possible. Synthetic trials showed that they do not seem acceptable candidates concerning handling and as a result, a new system was developed.
8 Designing a model compound with iron sulfur bonds seems logic, but to avoid thiols, a different approach was necessary. To keep the idea of involving phosphines, several of these ligands were reacted with sulfur and then treated with iron carbonyls, which led to surprisingly different structures than anticipated.
2.3 Synthesis and Characterization of Novel Hydrogenase Model Compounds
The phosphine ligands were synthesized according to literature procedures.[53] The respective phosphine was suspended in dry THF and after the addition of two equivalents of elemental sulfur the mixture was stirred at room temperature for up to 24 hours or at 60 °C for three hours. The products are white powders: for a methylene type derivative dppm was reacted to dppmS2 (Figure 9)
[54–56] and for a ferrocene type compound dppf was reacted to dppfS2 (Figure 10). Both, dppmS2 and
[57,58] dppfS2, including the molecular x‐ray structures are known in the literature.
The resulting thiophosphines were again suspended in dry THF together with Fe3(CO)12 and refluxed for five hours. During this time the mixture turns from dark green to almost black. Purification via column chromatography with a mixture of equal shares of DCM and cyclohexane affords the model compounds as very dark and intensely coloured microcrystalline powders (Figure 7). The complete reaction scheme is depicted in Figure 6 with the x‐ray structures of compound 1 in Figure 13 and of compound 2 in Figure 14 in the respective chapter. Both compounds are highly soluble in DCM and soluble in benzene, methanol and acetone. Their solubility in acetonitrile is low, in diethyl ether very low and they are not soluble in pentane or water. At a very late stage of this work it was found out by chance that 2 has already been synthesized in a very different reaction 20 years ago and besides a measured x‐ray structure nothing has happened since then with this molecule.[59] Although the starting materials dppm and dppf are both commercially available, dppf can also easily be synthesised starting from ferrocene via a reaction with n‐BuLi/TMEDA and quenching with
[60] Ph2PCl, which is also an excellent experiment for young students to learn Schlenk techniques. Homologous phosphines with ethylene (dppe) and propylene (dppp) bridges afforded no useful products. Most likely, the higher flexibility of higher homologues hinders the formation of bonds to both ends of the phosphine. The totally different phosphine (oxydi‐2,1‐ phenylene)bis(diphenylphosphine) (DPE‐phos) also afforded no product.
9
Figure 6: Synthesis of iron‐sulfur cluster 1 (P‐P = dppm) and 2 (P‐P = dppf).
Figure 7: Picture of 1, compound 2 looks macroscopically very similar.
In an alternative synthesis pathway (Figure 8), triphenylmethanethiol was used as the sulfur source and reacted together with triirion dodecacarbonyl to an isolatable iron sulfur carbonyl cluster
[61] Fe3S2(CO)9. As soon as a metal mirror is visible, the phosphine can be added without prior isolation of the intermediate. Interestingly, the [3Fe2S] intermediate cluster (Figure 11) was already synthesized in 1958,[62] obviously without the knowledge that this molecule alone can already catalyse hydrogen evolution.[36]
Figure 8: Synthesis of 1 (P‐P = dppm) and 2 (P‐P = dppf) from triphenylmethanethiol and triiron dodecacarbonyl.
10 2.3.1 Single crystal x‐ray diffraction
Single crystals were obtained by slow diffusion of Et2O into a solution of the respective compound in
[54–56] DCM. As stated before, the crystal structure of dppmS2 (Figure 9) is known in literature The measurement was performed as additional analytical proof of the successful reaction and especially to be able to determine changes of bond lengths and angles upon the reaction with Fe3(CO)12. The
[57,58] same applies for the also known ferrocene bridged phosphine dppfS2 (Figure 10). Single crystals
[63,64] of the intermediate cluster Fe3S2(CO)9 were obtained by freezing the mother liquor at ‐20 °C (Figure 11). The crystal data collection and refinement parameters are summarized in Table 9 in the appendix.
Figure 9: Molecular structure of dppmS2 (ORTEP; displacement ellipsoids at the 50% probability level; H‐atoms are omitted for clarity).
Figure 10: Molecular structure of dppfS2 (ORTEP; displacement ellipsoids at the 50% probability level; H‐atoms are omitted for clarity).
11
Figure 11: Molecular structure of Fe3S2(CO)9 (ORTEP; displacement ellipsoids at the 50% probability level; H‐ atoms are omitted for clarity).
From previous work and knowledge from literature, it was assumed that the new compounds would consist of a [2Fe2S] cluster with P‐S bonds (Figure 12). Crystallographic elucidation unveiled the true bonding situation: During synthesis, iron must undergo an insertion step to change from a P‐S bond to a P‐Fe‐S bond. Furthermore, all three iron atoms from triiron dodecacarbonyl end up in the final product and build a [3Fe2S] cluster.
Fe Ph2P PPh2 S S Ph2P PPh2 OC CO Fe Fe S S OC CO CO CO OC Fe Fe CO OC CO CO CO Figure 12: Wrong assumption of bonds in compound 1 (left) and 2 (right) prior to single crystal diffraction.
Compound 1 crystalized readily from a solution in DCM via slow diffusion of Et2O (Figure 13). However, for the ferrocene derivative 2 it was not possible to obtain suitable single crystals in this fashion even though the product itself is a microcrystalline powder. Various other solvent mixtures and freezing have been tried without success. It was found that evaporating the solvent after purification via column chromatography and dissolving the product in acetone gave suitable crystals upon evaporation in air with cyclohexane in the crystal lattice (Figure 14). Table 10 in the Appendix
12 summarizes data collection and refinement parameters of compounds 1 and 2. Both structures exhibit a triangular arrangement of the iron atoms with two μ3‐capping sulfur atoms, which form a square pyramidal Fe3S2 core with two Fe‐Fe bonds in accordance with polyhedral skeletal electron pair theory.[65,66]
Figure 13: Molecular structure of 1 (ORTEP; displacement ellipsoids at the 50% probability level; H‐atoms are omitted for clarity).
Figure 14: Molecular structure of 2 (ORTEP; displacement ellipsoids at the 50% probability level; H‐atoms and solvent molecules are omitted for clarity).
13
Table 1: Selected bond lengths [Å] and angles [°] concerning iron atoms of compounds 1, 2 and Fe3S2(CO)9.
1 2 Fe3S2(CO)9 Bond lengths Fe3‐Fe1 2.607(1) 2.568(6) 2.598(2) Fe3‐Fe2 2.615(1) 2.576(6) 2.593(2) S1‐S2 2.856(2) 2.909(6) 2.894(2) Fe3‐S1 2.256(2) 2.276(9) 2.233(3) Fe3‐S2 2.265(2) 2.302(9) 2.256(3) Fe1‐S1 2.239(2) 2.250(9) 2.223(3) Fe1‐S2 2.240(2) 2.262(1) 2.243(3) Fe2‐S1 2.236(2) 2.256(9) 2.236(3) Fe2‐S2 2.249(2) 2.260(9) 2.243(3) Fe1‐P2 2.212(2) 2.234(9) ‐ Fe2‐P1 2.185(2) 2.230(9) ‐ Bond angles Fe1‐Fe3‐Fe2 80.3(4) 85.4(2) 81.01(6) Fe1‐S1‐Fe2 97.6(6) 101.5(4) 98.23(1) Fe1‐S2‐Fe2 97.2(6) 101.0(4) 97.46(1) S1‐Fe1‐S2 81.0(6) 78.6(3) 80.76(1) S1‐Fe2‐S2 80.9(6) 78.5(3) 80.46(1) S1‐Fe3‐S2 80.1(6) 77.2(3) 80.27(1) P2‐Fe1‐Fe3 137.3(6) 159.4(3) ‐ P1‐Fe2‐Fe3 137.5(5) 159.4(3) ‐
Interestingly, the iron‐sulfur cluster is not significantly influenced by the size of the phosphine bridge. Even though the ferrocene bridge in 2 is much larger than the methylene bridge in 1 only the angle formed by the phosphorus atom with the iron atom is more obtuse. Other bond lengths and angles are consistent with the unsubstituted Fe3S2(CO)9 cluster (Table 1). The highest deviation between the complexes is the angle of the Fe3 fragment with still only 5° difference between compound 1 and 2.
The ferrocene subunit of dppfS2 features considerable changes upon reaction with triiron dodecacarbonyl. Before, the cyclopentadienyl rings are unrestricted and are in almost perfect parallel and staggered conformation (36° torsion angle). However, in 2 the rings are forced towards a more eclipsed conformation with a torsion angle of only 8°. Additionally, the rings are slightly inclined by 3° towards each other. The strongest observable change is in the P‐Fe‐P bonds of the dppf subunit. In dppfS2, this angle is a perfect 180°, whereas in compound 2 this changes dramatically to
14 97°, which is closer to a right angle and another evidence for the steric change of the ferrocene moiety.
15 2.3.2 NMR Characterization
All reaction steps involve compounds that contain phosphorus atoms, which makes 31P NMR an ideal method to not only characterize, but also to determine the success of the reaction and – to some extent – the purity of the products. All products and intermediate products are symmetric with respect to phosphorus and therefore only one singlet is visible in each spectrum. The starting materials have singlet signals in the negative range of the ppm scale. The samples have been recorded in C6D6 and CDCl3. The precursors dppm and dppf have shifts at ‐21.8 ppm and ‐17.0 ppm in
CDCl3, respectively (Figure 15 top). Their shifts in C6D6 are ‐9.1 ppm for dppm and ‐4.1 ppm for dppf.
During synthesis, these signals shift downfield with every step. The sulfur species dppmS2 and dppfS2 show signals at 35.3 ppm and 40.7 ppm in CDCl3, respectively (Figure 15 down). In C6D6 their shifts are 48.2 ppm for dppmS2 and 53.3 ppm for dppfS2, respectively.
The final products are shifted further to 75.4 ppm (1) and 68.4 ppm (2, Figure 17) in CDCl3. In C6D6 the downfield shift is even stronger to 88.8 ppm (1, Figure 16) and 82.0 ppm (2). For a summary of phosphorus NMR data see Table 2. These downfield shifts are caused by the σ‐donor character of the phosphine bonds, which leads to a higher electron density along the P‐Fe bonds and a deshielding of the phosphorus atoms. These characteristic shifts are defined as Δδ = (δcomplex‐δfree ligand) upon coordination.[67] It is expected that the chelation shift caused by the ferrocene type ligand should be larger than for the methylene type ligand. However, the opposite is indeed the case with a Δδ value of 97.2 ppm for 1 and 85.4 ppm for 2 in CDCl3.
31 Figure 15: P NMR spectra of dppm (top left), dppmS2 (bottom left), dppf (top right), dppfS2 (bottom right) recorded in CDCl3.
16 A carful reaction under exclusion of oxygen prevents the formation of phosphine oxide. In the case of
31 dppfS2, the oxygenated dppfO2 would be visible in the spectrum with a P signal at 28.4 ppm in
CDCl3. This species is very difficult to remove, because of its high solubility in the same solvents suitable for dppfS2. In this small laboratory scale, it is less time consuming and cheaper to discard the batch than try to remove the unwanted by‐product. Other byproducts that might form do not inhibit the success of the next reaction step. Furthermore, many byproducts are formed during the last step, anyway. The spectra in Figure 16 are measured in C6D6 before and after purification via column chromatography. In later experiments, only CDCl3 is used as the better solvent and spectra were only recorded after purification. Therefore, no information about the shifts of the byproducts in C6D6 is known, despite the fact that the shifts of the products are highly depending on the solvent.
With the experience from compound 1, the ferrocene type compound 2 has been analysed in CDCl3 (Figure 17). The purification was performed in the same way, but the uncharacterized by‐product at 66.6 ppm could in many cases only be removed after a second column chromatography step.
31 Figure 16: P NMR spectra of 1 in C6D6 before (left) and after (right) purification.
31 Figure 17: P NMR spectra of 2 in CDCl3 before (left) and after (right) purification.
17 31 Table 2: P NMR product peaks in CDCl3 and C6D6
CDCl3 C6D6 dppm ‐21.8 ‐9.1
dppmS2 35.3 48.2 1 75.4 88.8 (75.0)a dppf ‐17.0 ‐4.1
dppfS2 40.7 53.3 (28.4)b 2 68.4 82.0
a b in CD2Cl2, oxide (dppfO2)
Typical for metal carbonyls is the strong downfield shift of the carbonyl carbon in 13C NMR. Their shifts cover a very broad range from 190 ppm to up to 250 ppm.[68] In case of the methylene bridged compound 1 the CO shifts are 212.7 and 205.9 ppm (Figure 18), for the ferrocene bridged compound 2 only one peak at 213.6 ppm was found (Figure 19).
212.6 205.8 135.9 132.0 130.5 128.7 77.5 77.1 76.7
240 220 200 180 160 140 120 100 80 60 40 ppm 13 Figure 18: C NMR of 1 in CDCl3.
18 213.6 132.6 130.2 128.4 128.3 128.2 77.4 77.2 77.0 76.6 74.3 73.8 73.2 31.2 29.7 26.9
240 220 200 180 160 140 120 100 80 60 40 ppm 13 Figure 19: C NMR of 2 in CDCl3.
Unfortunately, the corresponding proton NMR measurements could not be quantitatively interpreted by peak integration, because of broad signals and are therefore only included in the appendix for completeness and for comparison.
19 2.3.3 Spectroscopic Characterization
Visible to the naked eye is the dramatic colour change when the reaction is proceeding from the precursors to the products, which will obviously strongly reflect in electronic absorption spectra. The precursors are white powders and therefore their UV‐vis spectra are rather unspectacular. The final products, however, absorb over the entire recorded UV‐vis region. The dppm bridged compound 1 has two distinct peaks at 321 and 535 nm and also a shoulder at 384 nm (Figure 20). Compound 2 with the dppf‐bridge has a less resolved spectrum with an almost linear curve shape between 400 and 700 nm (Figure 21). At very close inspection, two slightly elevated regions are visible at ca. 445 and 550 nm, respectively. Those are clearly not distinct peaks, but they are included in the summary of Table 3 for comparison. The chromophoric bands in the visible spectral region most probably arise from allowed electronic transitions within the σ‐bonded Fe atoms in the cluster[63,69] and display a rather high intensity with molar extinction coefficients in the range of 2000‐4000 mol‐1∙cm‐1 (Table 3). Free ferrocene has absorption maxima at 326 and 442 nm and in addition to the bathochromic shift of these peaks for 2, the intensity of the electronic‐dipole‐forbidden d–d absorptions of the ferrocene subunit is increased significantly by a factor of approximately ten. This indicates a certain degree of mixing of the electronic wave functions, which in return decreases the intensity of the dipole‐allowed cluster core transitions.
1.0
dppmS2 0.8 1 / a.u. /
A 0.6
0.4 absorbance 0.2
0.0 300 400 500 600 700 800 wavelength / nm
‐5 Figure 20: Electronic absorption spectra of 610 M solutions in DCM of dppmS2 and 1 (1 cm cuvette).
20 Table 3: Electronic absorption spectroscopic data of compound 1 and 2 in DCM.
λ / nm (ε / L∙mol‐1∙cm‐1) 1 321 (11300) 384a (7600) 535 (3300) 2 359 (9700) 445b (4200) 550b (2400)
a dppmS2 260 (8800) dppfS2 440 (240) 531 (160) a shoulder, b no peak, but slightly elevated region
1.0
dppfS2 0.8 2 / a.u.
A 0.6
0.4 absorbance absorbance 0.2
0.0 300 400 500 600 700 800 wavelength / nm
‐5 Figure 21: Electronic absorption spectra of 610 M solutions in DCM of dppfS2 and 2 (1 cm cuvette).
No significant solvent polarity changes can be observed, which suggests a delocalized ground‐state electronic structure. In contrast to the situation in dinuclear Fe2S2(CO)6 compounds with iron brigding
[70] disulfide ligands, in trinuclear Fe3S2(CO)9 compounds, no low lying σ* (S‐S) acceptor orbitals are available for dσ* MLCTs, which prohibits solvatochromism.
21 For compounds containing carbonyls, infrared spectroscopy is an excellent characterization method, because the CO‐vibrations are very IR active. Metal carbonyls vibrate in a region between 1800 and 2170 cm‐1.[71] In solution, spectra usually differ from solid samples and can be very different to interpret, because isomerisation can lead to very complicated spectra.[72] Employing different IR techniques, the spectra of the carbonyl vibrations vary considerably. The fastest method would be ATR‐IR, because it does not need any sample preparation, but unfortunately, the ATR bridge absorbs most of the intensity of the source in the respective region. In this case, KBr pellets provide the best results although they take some time to prepare. The methylene bridged compound 1 (Figure 22) has overlapping peaks so that not all CO ligands are clearly visible. Including the small shoulder at 1992 cm‐1, six peaks can be assigned, although the peak shape at 1979 cm‐1 might hide an underlying seventh peak. Compound 2 produces a somewhat odd and very flat peak shape (Figure 23), which has been reported before for carbonyl compounds in the solid state.[73] Aside from the signals at 2095 and 2062 cm‐1, it is not possible to assign definitive peaks so instead the edges of the signal were defined. Measurements in solution are possible with IR transparent window materials, although most of them are not transparent in the entire measurable region. Calcium fluoride windows (0.5 mm path length) have a cutoff at approximately 1000 cm‐1, which means smaller wavenumbers are not accessible. Additionally, the solvent has to be taken adequately into account. In this case DCM was used, which produces very intense signals at 3075‐3040, 3000‐2975, 1440‐1410 and 1290‐1240 cm‐1 and a small peak at 2305 cm‐1 (Figure 67 in the appendix). The spectrum in solution of compound 1 (Figure 24) is less resolved than the one of the KBr‐pellet. Although the signals are less broad, peaks can only be assigned to three signals, four if the shoulder at 1983 cm‐1is included. The general shape on the other hand is similar with a very pronounced peak at the highest wavenumber and two broader shapes with shoulder signals. The ferrocene bridged compound 2 in solution does not have the same flat pattern as in the solid state and is well resolved. Six peaks can be assigned with a seventh presumably overlapped between 1987 and 1962 cm‐1 (Figure 25).
22 80
60
40
transmission / % transmission 20 2041 1937
2004 1954 1992 1979 0 2400 2300 2200 2100 2000 1900 1800 1700 1600 wavenumber / cm-1
Figure 22: FTIR spectrum of carbonyl region of compound 1 (KBr‐pellet), complete spectrum in Figure 63.
100
2095 80
60
2062 40 1931 2044 2008 transmission / % transmission 20
0 2400 2300 2200 2100 2000 1900 1800 1700 1600 wavenumber / cm-1
Figure 23: FTIR spectrum of carbonyl region of compound 2 (KBr‐pellet), complete spectrum in Figure 64.
23 100
DCM artifact 80
60
1046 40
transmission / % transmission 1983 20
2046 2002 0 2400 2300 2200 2100 2000 1900 1800 1700 1600 wavenumber / cm-1
Figure 24: FTIR spectrum of carbonyl region of compound 1 (CaF2 windows, DCM solution), complete spectrum in Figure 65.
100
2072 80
DCM 60 1940 1962
40 1987
2006 transmission / % transmission
20 2045
0 2400 2300 2200 2100 2000 1900 1800 1700 1600 wavenumber / cm-1
Figure 25: FTIR spectrum of carbonyl region of compound 2 (CaF2 windows, DCM solution), complete spectrum in Figure 66.
For both compounds, the electron‐rich character is confirmed by the energetic position of the carbonyl vibration region. The highest frequency signal for the unsubstituted iron‐sulfur cluster
‐1 [61] ‐1 Fe3S2(CO)9 is found at 2064 cm in solution, which is redshifted (Δ (CO) ≈ 20 cm ) for the corresponding bisphosphine‐functionalized derivatives Fe3S2(CO)7(P‐P). This is also consistent with a
24 higher electron density of the d‐orbitals of the iron atoms involved in the π‐back‐donation to the CO ligands. This degree of π‐back‐donation of both compounds, 1 and 2, is more pronounced than that usually measured for dinuclear iron core hydrogenase models. For those dinuclear systems, a linear correlation between the observed redshift of reduced species and the spin‐density at the Fe‐Fe core has been reported.[48] For hydrogen production, either photo‐ or electrocatalytically, an electron rich environment at the Fe cluster moiety is considered to be crucial. Compared to dinuclear hydrogenase model compounds, a vibrational frequency shift of almost 40 cm‐1 in the oxidised resting state should make trinuclear complexes such as 1 and 2 promising candidates for proton to hydrogen reduction in solution.
The influence of the strong acid TFA has also been investigated in the carbonyl region. TFA itself has very broad and intense signals and in some parts of the spectrum transmission drops to zero percent (Figure 68 in the appendix). Upon addition of TFA to a solution of 1 in DCM the significant influence of the acid is visible with drastic changes to the spectrum (Figure 26). Many small blue shifted peaks are formed, which are an indication of protonation.[74] For clarification, the spectrum of 1 with added TFA has been subtracted by the spectrum of TFA alone (Figure 27). Visible are two new distinct sets of signals, which are proposed to be mono‐ and deprotonated species (Table 4). The set of carbonyl
+ ‐1 peaks of the mono‐protonated form [Fe3S2(CO)7(dppm)(µ‐H)] has a large blue shift of 74 cm , the di‐
2+ ‐1 protonated form [Fe3S2(CO)7(dppm)(µ‐H)] has an even larger shift of 100 cm .
100 90 80 1 70 + 1 eq. TFA TFA blank 60 50
40 30
transmission / % transmission 20 10 0
2200 2150 2100 2050 2000 1950 1900 1850 1800 wavenumber / cm-1
Figure 26: FTIR solution spectra of 1 in DCM with added TFA, including the blank TFA spectrum in DCM.
25 Table 4: CO vibrations of 1 and proposed blue‐shifted signals of mono‐ and di‐protonated forms of 1 observed upon TFA addition in DCM solution.
compound (CO) / cm‐1 a: 1 Fe3S2(CO)7(dppm) 2044 2002 1946
+ + b: 1 + H [Fe3S2(CO)7(dppm)(µ‐H)] 2118 2087 2025
+ 2+ c: 1 + 2H [Fe3S2(CO)7(dppm)(µ‐H)] 2144 2103 2064
a Fe3S2(CO)7(dppm) + b [Fe3S2(CO)7(dppm)(µ-H)] 2+ c [Fe3S2(CO)7(dppm)(µ-H)2]
c b c
b c b a
a
a 2200 2150 2100 2050 2000 1950 1900 1850 1800 wavenumber / cm-1
Figure 27: Differential spectrum of Figure 26: the spectrum of 1 + 1 eq. TFA was subtracted by the TFA spectrum.
The changes upon addition of TFA to compound 2 are drastic as well (Figure 28). The differential spectrum (Figure 29) has a different protonation pattern than the one for compound 1. Contrary to 1, there is a much stronger protonation already for 2 and the unprotonated species is barely visible with the same amount of acid. Very pronounced are the peaks of the mono‐protonated species
+ ‐1 (Table 5). Here, the mono‐protonated form [Fe3S2(CO)7(dppf)(µ‐H)] has a blue shift of 79 cm , the di‐
2+ ‐1 protonated form [Fe3S2(CO)7(dppf)(µ‐H)] has a shift of 89 cm .
Table 5: CO vibrations of 2 and proposed blue‐shifted signals of mono‐ and di‐protonated forms of 2 observed upon TFA addition in DCM solution.
compound (CO) / cm‐1 a: 2 Fe3S2(CO)7(dppf) 2046 2007 1987
+ + b: 2 + H [Fe3S2(CO)7(dppf)(µ‐H)] 2125 2069 2017
+ 2+ c: 2 + 2H [Fe3S2(CO)7(dppf)(µ‐H)] 2135 2099 2057
26
100 90 80 70 60 50 40 30 2 transmission / % transmission 20 1 eq. TFA TFA blank 10 0 2200 2150 2100 2050 2000 1950 1900 1850 1800 wavenumber / cm-1
Figure 28: FTIR solution spectra of 2 in DCM with added TFA, including the blank TFA spectrum in DCM.
a Fe3S2(CO)7(dppf) + b [Fe3S2(CO)7(dppf)(µ-H)] 2+ c [Fe3S2(CO)7(dppf)(µ-H)2]
a b c c a
c a
b
b 2200 2150 2100 2050 2000 1950 1900 1850 1800 wavenumber / cm-1
Figure 29: Differential spectrum of Figure 28: the spectrum of 2 + 1 eq. TFA was subtracted by the TFA spectrum.
27 2.3.4 Photochemical investigations
Both compounds were investigated under irradiation with varying parameters, which include acid, donor, sensitizer or under Argon atmosphere. All experiments were performed with a 150 W xenon lamp and a 530 nm cutoff filter.
2.3.4.1 Ultraviolet/visible spectroscopy
In the first experiment, compound 1 is irradiated without any other substances in a quartz cuvette in DCM. During a very long irradiation time of more than six hours a collection of overlapping spectra has been obtained (Figure 30). However, this picture becomes clearer when it is separated into different parts.
1 1.0 1 min 2 min 3 min 4 min 5 min 0.8 7 min 9 min 11 min 14 min
/ a.u. 17 min
A 0.6 20 min 25 min 30 min 40 min 50 min 0.4 60 min 75 min 90 min 105 min
absorbance absorbance 120 min 150 min 0.2 180 min 210 min 240 min 390 min 0.0
300 400 500 600 700 800 wavelength / nm
Figure 30: Electronic absorption spectra of 1 of irradiation with a 150 W Xenon Lamp (530 nm cutoff) in DCM (Overview).
28 1.0 1 1 min 2 min 3 min 4 min 0.8 5 min 7 min 9 min
/a.u. 11 min
A 0.6 14 min 17 min 20 min 25 min 0.4 30 min absorbance absorbance 0.2
0.0
300 400 500 600 700 800 wavelength / nm
Figure 31: Electronic absorption spectra of 1. Irradiation experiment from 0‐30 min.
The first 30 min of the experiment are depicted in Figure 31, which shows a clean photochemical primary reaction. Two isosbestic points at 345 and 371 nm can be observed. The broad peak at 545 nm is subjected to a strong decrease in absorbance to a rather flattened curve, which will not change significantly after 60 min of irradiation time (Figure 32). In the transition time from 30‐ 60 min, a bleaching of the primary photoproduct peak at 356 nm can be seen, whereas the absorbance of the peak at 325 nm continues to decrease. Prolonged irradiation leads to a reversion of the peak at around 325 nm to again higher absorbance and as a result a new isosbestic point at 339 nm is formed (Figure 33). This leads to the suggestion that after the first photochemical reaction a further process with possibly one or more byproducts or degradation products occurs and after that, another photochemical reaction with those by‐ or degradation products takes place.
29 1.0 30 min 40 min 50 min 60 min 0.8 /a.u.
A 0.6
0.4 absorbance absorbance 0.2
0.0
300 400 500 600 700 800 wavelength / nm
Figure 32: Electronic absorption spectra of 1. Irradiation experiment from 30‐60 min.
1.0 60 min 75 min 90 min 105 min 0.8 120 min 150 min 180 min
/a.u. 210 min
A 0.6 240 min 390 min
0.4 absorbance absorbance 0.2
0.0
300 400 500 600 700 800 wavelength / nm
Figure 33: Electronic absorption spectra of 1. Irradiation experiment from 60‐390 min.
Repeating this irradiation experiment with the addition of 5 eq. TEOA does not change the reaction significantly and was therefore aborted after about 2.5 hours (Figure 61 in the appendix). Upon addition of one equivalent of TFA the absorption spectrum does not change much either, but if the concentration of TFA is increased, the peaks in the range of 300‐400 nm begin to vanish and the peak at 545 nm decreases. Irradiation under the same conditions as described above leads to much faster changes compared to the experiment without acid. The whole spectrum seems to bleach out
30 and a new peak emerges at 354 nm, which will also start to vanish after some minutes of irradiation (Figure 34). In addition, macroscopically it is evident that the solution turns irreversibly colourless. Such an observation might indicate that the initially reduced form of the triiron cluster core is oxidized by an intramolecular process generating a photoproduct no longer carrying a chromophoric
4+ triiron bonded structure. The stepwise change of the deeply coloured (Fe3) moiety into a two‐ electron oxidized form carrying only weakly absorbing Fe2 units seems possible. This photochemically triggered release of two electrons from the iron centre could lead to irreversible proton reduction in solution, meaning H2 release, which should be accelerated by the addition of a proton source, such as TFA, as indeed was observed. For a cyclic photocatalytic process, such a reaction sequence would have to be closed by reductive regeneration of the initial iron cluster structure 1.
1.0
1 + 1 eq. TFA + 2 eq. TFA + 3 eq. TFA 0.8 + 4 eq. TFA + 5 eq. TFA 1 min 2 min / a.u. / 3 min A 0.6 4 min 5 min 6 min 7 min 8 min 0.4 10 min 12 min 15 min 20 min absorbance 30 min 0.2 45 min
0.0 300 400 500 600 700 800 wavelength / nm
Figure 34: Electronic absorption spectra of the irradiation experiment of 1 in DCM with TFA.
When experimenting with a mixture of TFA as protonating agent and TEOA as electron donor, an unexpected problem occurred: the solution becomes turbid, because the TFA‐TEOA adduct is not soluble in DCM, which is also very distinct in the recorded spectra (Figure 62 in appendix). An attempt to record spectral variations upon irradiation of such a mixture is shown in Figure 35. Quantitative interpretation of these data is, however, more or less impossible, which at the moment leaves the question of potential photocatalytic applications using TFA‐TEOA mixtures open.
31 1 +TEOA +TEOA +TFA 1.4 1 min 2 min 5 min 10 min 1.2 12 min 15 min 20 min 25 min
/ a.u. 1.0 30 min
A 40 min 50 min 0.8 60 min 120 min 0.6
absorbance absorbance 0.4
0.2
0.0 300 400 500 600 700 800 wavelength / nm
Figure 35: Electronic absorption spectra of the irradiation experiment of 1 in DCM with TFA and TEOA.
The same series of experiments with compound 2 has a rather different outcome. First, the irradiation of the complex in DCM does reveal an isosbestic point at 339 nm, but it does not look very pronounced and therefore does not indicate a clean photochemical reaction (Figure 36). Additionally, all peaks vanish eventually, which is visible also by the change of colour of the solution from brown to almost colourless. The experiment with TFA is accelerated compared to 1, but with a similar result: one peak remains after the experiment at 339 nm (Figure 37).
1.0 2 1.0 1 min 2 min 0.8 3 min
/ a.u. 4 min A 0.6 0.8 5 min 7 min 0.4 9 min absorbance 0.2 11 min 14 min /a.u. 17 min
A 0.0 0.6 300 400 500 600 700 800 20 min wavelength / nm 25 min 30 min 40 min 50 min 0.4 60 min 75 min 90 min 105 min absorbance absorbance 120 min 150 min 0.2 180 min 210 min 240 min
0.0 300 400 500 600 700 800 wavelength / nm
Figure 36: Electronic absorption spectra of 2 during irradiation with a 150 W Xenon Lamp in DCM. Inset: Final spectrum after 240 min. 32
1.0
2 + 1 eq. TFA + 2 eq. TFA 0.8 + 3 eq. TFA + 4 eq. TFA + 5 eq. TFA 1 min
/ a.u. 2 min
A 3 min 0.6 4 min 5 min 6 min 7 min 8 min 0.4 10 min 12 min 15 min
absorbance absorbance 20 min 30 min 0.2
0.0 300 400 500 600 700 800 wavelength / nm
Figure 37: Electronic absorption spectra of the irradiation experiment of 2 in DCM with TFA.
1.0
2 + 5 eq. TEOA 0.8 1 min 2 min 4 min 7 min
/ a.u. / 12 min
A 0.6 20 min 33 min 54 min 88 min 143 min 0.4 180 min absorbance 0.2
0.0
300 400 500 600 700 800 wavelength / nm
Figure 38: Electronic absorption spectra of the irradiation experiment of 2 in DCM with TEOA.
Very interesting is the irradiation with added TEOA (Figure 38). Although TEOA does not change the spectrum at all, upon irradiation the peak at 359 nm increases for 4 minutes before it decreases in a similar fashion as without addition. The isosbestic point at 339 nm stays the same and an additional one at 349 nm is visible. After an irradiation time of 180 min the spectrum looks like the spectrum of
33 the pure compound after irradiation. This is an indication that at least for a short time of a few minutes a photochemical reaction similar to the one observed with compound 1 has occurred (Figure 31).
Since all measurements were performed with a 530 nm cutoff filter, it was of interest how an antenna molecule such as Zn(TTP), which has an absorption maximum of 550 nm, would influence the irradiation experiments. It is suspected that Zn(TTP) could act as an electron donor and that an unstable radical is formed, which will quickly be quenched by oxygen.[75] Therefore, the same experiment was once performed in ambient atmosphere (Figure 39) and once under argon (Figure 40). At first sight, both experiments seem to have an identical outcome. However, upon closer inspection, small differences can be found, which can only be caused by the absence or presence of oxygen. In both irradiation experiments, hints for a porphyrin radical cation can indeed be found by the presence of new peaks formed around 700 nm and under argon, the radical seems to have stronger signal intensity. However, so far, it was not possible to detect any signals indicating radicals via EPR spectroscopy. Both, the cluster and the sensitizer are irradiated at 549 nm, visible by the absorptions changes around 340 nm similar to the irradiation with the cluster alone and additionally a decreasing Soret band and decreasing Q bands. Therefore, the absence of any EPR signal might be explained by a coupling the radical of the porphyrin sensitizer with a radical of one‐electron reduced 1 formed upon irradiation.
1.0 1 1+Zn(TTP) 1 min 0.8 2 min 3 min 5 min / a.u.
A 10 min 0.6 15 min 20 min 25 min
0.4 absorbance absorbance
0.2
0.0 300 400 500 600 700 800 wavelength / nm
Figure 39: Electronic absorption spectra of the irradiation experiment of 1 in DCM with Zn(TTP).
34 1.0 MK228 Argon MK228 + Zn(TTP) 1 min 2 min 0.8 3 min 5 min 10 min / a.u.
A 15 min 20 min 0.6 25 min
0.4 absorbance absorbance
0.2
0.0 300 400 500 600 700 800 wavelength / nm
Figure 40: Electronic absorption spectra of the irradiation experiment of 1 in DCM with Zn(TTP) under argon.
Further studies that are more detailed will be necessary to fully elucidate the light induced reactivity of both systems 1 and 2, which was out of the scope of the present work.
2.3.4.2 Infrared Spectroscopy
It is difficult to compare results obtained at identical irradiation times for UV/vis and IR measurements because the setups are very different: UV/vis cuvettes are made of quartz and have a path length of 10 mm, the IR cell used is made of calcium fluoride and has a very narrow path length of only 0.5 mm. According to Beer’s Law, this necessitates much higher concentrations for the IR measurements. The general trend of weaker signal intensity over progressing photolysis time stays the same, although it takes considerably longer for IR experiments. In Figure 41 the result of an irradiation experiment with a 150 W xenon Lamp and a 530 nm cutoff filter is depicted. (Note that deviations in baselines have been corrected.) As expected, analogously to UV/vis measurements the main signals, which are in this case the CO vibrations, vanish over time.
35 100
95
90
1 85 1 min 2 min 5 min 80 10 min 20 min
Transmission / % 40 min 130 min 75 220 min 300 min 360 min 70 2200 2150 2100 2050 2000 1950 1900 1850 1800 Wellenzahl / cm-1
Figure 41: Vibrational spectra of irradiation experiment of 1 with a 150 W xenon Lamp (530 nm cutoff) in DCM.
Compound 2 exhibits a slightly different behaviour as it rapidly decreases, but after only 15 minutes, the spectra do not change significantly anymore (Figure 42). These results indicate that upon prolonged photolysis, stability issues play a certain role for both compounds, while the ferrocene containing derivative 2 seems to be more light‐stable with respect to decarbonylation (CO‐loss).
100
90
80
2 70 1 min 5 min 15 min Transmission / % 30 min 60 60 min 90 min 120 min 150 min 50 2200 2150 2100 2050 2000 1950 1900 1850 1800 Wellenzahl / cm-1
Figure 42: Vibrational spectra of irradiation experiment of 2 with a 150 W xenon Lamp (530 nm cutoff) in DCM.
36 2.3.4.3 Long‐term irradiation test under ambient conditions
Both compounds were dissolved in a 1 cm quartz cuvette and stored in a fume hood with natural day and night cycles over the course of one month to document the decomposition under natural conditions. Depicted in Figure 43 are pictures taken at the beginning of the experiment and after 12, 19 and 29 days. Already after 12 days the original colours are only barely visible and on the bottom an orange precipitate formed. On the picture of the 19th day the reddish (1) and brownish (2) colours are gone and turned into a pale yellow for both compounds. In the last picture after 29 days only a very faint yellow colour is left.
Figure 43: Effects of long‐term light exposure in solution under ambient conditions after 0 (a), 12 (b), 19 (c) and 29 days (d).
37 2.4 Electrochemistry
2.4.1 Cyclic voltammetry
Electrochemical properties were investigated with cyclic voltammetry (1‐1.5 mM analyte, 0.1 M TBAP
‐1 electrolyte, scan rate 100 mV∙s ) under an inert gas (N2 or Ar). Figure 44 illustrates the standard apparatus for these measurements. The three electrode setup employs a glassy carbon working electrode with a Pt/Ir counter electrode, which is referenced to a silver/silver nitrate pseudo‐ reference electrode. Additionally, all voltammograms were referenced vs. Fc/Fc+. A cannula supplies purge gas between measurements to ensure an oxygen free environment and liquids in smaller amounts are added via a GC syringe.
Figure 44: Setup for cyclic voltammetry measurements. 1: working electrode (glassy carbon), 2: counter electrode (Pt/Ir), 3: pseudo‐reference electrode (Ag/AgNO3), 4: cannula for purge gas (N2 or Ar), 5: syringe for addition of acid, 6: port plug (removable to add solids).
The voltammogram of the cathodic scan of 1 in DCM displays a quasi‐reversible first one‐electron reduction wave at a half‐peak potential of ‐1.64 V vs. Fc/Fc+ (Figure 45). The same complex shows an electron reduction at ‐1.43 V in ACN. Compared to literature values reported for Fe3S2(CO)9, which are ‐1.03 V [37] and ‐0.94 V [36] vs. Fc/Fc+, respectively, it can be concluded that the more electron rich
38 cluster 1 requires a 500‐600 mV more negative potential than the unmodified cluster under those conditions. The oxidation of the cluster is electrochemically irreversible indicating chemical steps coupled to the removal of electrons from the cluster core (inset of Figure 45).
5
0 / µA
I -5 10 mV/s 25 mV/s 50 mV/s -10 100 mV/s 250 mV/s 500 mV/s -15 -0.8 -1.0 -1.2 -1.4 -1.6 -1.8 -2.0 E / V vs. Fc/Fc+
Figure 45: Cyclic voltammograms of the reductive wave of 1 depending on the scan rate (glassy carbon, 0.1 M TBAP electrolyte in DCM). Inset: complete cyclic voltammogram.
5
0
-5 / µA P i
-10
-15 0 5 10 15 20 25 1/2
Figure 46: Nonlinear dependency of the square root of the scan rate of cathodic and anodic peak currents for the reduction process of 1 in DCM (scan rates: 10, 20, 50, 100, 250, 500 mV∙s‐1).
39
[76] The Randles‐Sevcik equation describes the proportionality between the peak current ip and the square root of the scan rate v for a diffusion‐controlled reaction: