Preparation, functionalization and analysis of UiO-66 metal-organic framework thin ﬁlms on silicon photocathodes

(1)

Preparation, functionalization and analysis

of UiO-66 metal-organic framework thin

films on silicon photocathodes

Andreas Wagner

Uppsala University

Molecular Inorganic Chemistry - Biomimetic Chemistry

Master Thesis

(2)

(3)

Metal-organic frameworks (MOFs), metal centers (atoms or clusters) linked with organic molecules, are of strong interest in materials research due to their chemical versatility and extraordinary high surface area. The molecular nature of MOFs allows a post-synthetic functionalization and modification of the chemical environment within the pores. Earlier, it was shown that incorporation of Fe2(dcbdt)(CO)6 (dcbdt = 2,3-dithiolato-1,4-benzenedicarboxylic acid) - a molecular proton reduction catalyst - into a MOF results in a new material that exhibits increased catalytic reactivity and stability compared to the same complex in solution. Recently, we have shown the electrochemical addressability of the same catalyst integrated into a MOF thin film on FTO (fluorine doped tin oxide). Within this thesis robust UiO-66 (UiO = University in Oslo) MOF thin films were prepared on FTO and p-type silicon electrodes and different parameters (water concentration, relative concentrations, time) influencing the synthesis were analyzed. It was observed that terephthalic acid is not forming a self-assembled monolayer on silicon substrates while it does on FTO. Films with a preferential growth direction of the UiO-66 crystallites on silicon were prepared and functionalized by post-synthetic ligand exchange with Fe2(mcbdt)(CO)6 (mcbdt = 2,3-dithiolato-1-benzenecarboxylic acid). A variety of techniques (ATR-IR,

(10)

Popular scientific summary

The overall long-term goal of this project and others conducted within the research group biomimetic chemistry is to efficiently produce hydrogen as a renewable energy carrier directly from sunlight and water. Basic research is conducted to mimic biological systems and understand the origins of their high efficiency.

This master thesis deals with a new class of materials called metal-organic frameworks (MOFs) that was recently developed. MOFs are crystalline solids based on metal clusters that are linked together with organic molecules. Several hundred combinations of different metals with different organic linkers have been developed durin recent years. This great versatility and the fact that these materials offer the highest surface areas currently known (the record material has a surface area of ca. 1 football field per gram!) make them highly interesting to researchers in the fields of gas storage, catalysis, drug delivery and many more. Another great advantage of MOFs is the fact that the organic linker in the crystal structure can be chemically altered which improves the possibilities for modifications even further.

In this thesis one of the most robust metal-organic frameworks called UiO-66 (UiO = University in Oslo) based on the metal Zirconium was grown as thin films on a silicon substrate. A special catalyst that contains no rare-earth elements and is able to produce hydrogen was incorporated into the framework. This catalyst was chemically synthesized and its structure is incorporating certain features of enzymes present in bacteria and green algae. It is typically very fragile when it is dissolved in water and degrades quickly under sunlight, but the incorporation into the MOF increases its stability a lot. One of the objectives of this work is to understand this stabilization effect and the distribution of the catalyst within the MOF. It was possible to show that the catalyst is not incorporated homogeneously into the MOF. An increased concentration on the surface of the entire film was detected.

The silicon used as substrate for the MOF film can act as light absorber similar to its use in photovoltaic cells and provide the energy (electrons) needed for the catalyst to produce hydrogen according to 2H+_{+ 2e}– _−−→_H

(11)

Contents xi

(12)

(13)

Chapter 1

Introduction

The following introduction will guide through the main aspects of this work. Firstly the fundamentals of artificial photosynthesis, [FeFe]-hydrogenase model complexes and metal-organic frameworks (MOF) will be briefly explained. Based on that, the objective of this thesis will be outlined in the context of three publications that provide the basis of this work.

1.1 Artificial Photosynthesis

(14)

H2O −−→ 2H++12O2 ∆E0 = 1.23V pH = 0 (1.1) Taking into account that an overpotential will be needed to overcome kinetic limitations of the reaction, a minimum of 1.6 V is frequently reported [9]. One could in principle use a photovoltaic cell to produce the necessary potential difference for the water splitting, but the output voltage of photovoltaic cells is strongly light dependent and therefore needs elaborate electronic control during changing light intensity [10].

Applying literature values for Plank0_{s constant and the speed of light, 1.6 eV} corre-sponds to the energy of a photon with a wavelength of 774.9 nm (see equation 1.2 and 1.3).

E = h · ν = c

λ (1.2)

λ= h· ν

E (1.3)

Ideally a system would thereby consist of one single material that efficiently absorbs all photons with wavelengths shorter than 750 nm, separates the charges, catalyses both water oxidation and proton reduction and is on top of it all stable under these conditions. Different research groups are trying to find a material that is capable of doing all these tasks mentioned, but the highest quantum efficiency reported so far is only 0.66 % [11]. The more promising approach is to separate at least some of the different functions (absorption, charge separation, water oxidation catalysis and proton reduction) from one

another.

There are different possibilities to realize a water splitting system and only two different ways will be illustrated here (a review on the topic was published by Walter et al. [8]). A very general scheme of artificial photosynthesis is illustrated in figure 1.1 and is based on a molecular photosensitizer to absorb the light. The water oxidation catalyst and hydrogen evolving catalyst might be a molecular catalysts in solution, on a solid support or a heterogeneous catalyst.

(15)

1.1. ARTIFICIAL PHOTOSYNTHESIS 3

Figure 1.1: Scheme of an artificial photosystem consisting of a photosensitizer (P), water

oxidation catalyst (WOC), hydrogen evolving catalyst (HEC) or CO2 reduction catalyst (CRC).

Sacrificial electron donors (SED) and acceptors (SEA) can be used to study the half-reactions separately [12]

Another approach is the use of semiconductors as light absorbers instead of a molecular species. The water splitting half reactions (see equation 1.4) are taking place at the photocathode and photoanode. This system might be the most obvious design principle but is one of the least developed according to Sivula and Grätzel [10]. They argue that the stability and the magnitude of photocurrent of the photocathode remain limiting factors.

Figure 1.2: Semiconductor based photocathode/photoanode tandem water splitting cell design,

HEC = hydrogen evolving catalyst, WOC = water oxidation catalyst

(16)

The work described in this thesis focuses on the incorporation of a molecular proton reduction catalyst into a metal-organic framework thin film on a silicon substrate as photocathode material for solar water splitting. Cathodes based on p-type silicon show a photoeffect under illumination, which means that the absorbed photon energy is used to shift a reduction potential to more positive values (oxidation potential to more negative potential). Further details to silicon photocathodes and their electrochemistry will be discussed in section 2.3. Metal-organic frameworks are - as explained in section 1.3 - a versatile material class with very high surface areas and form a highly interesting platform for the incorporation of molecular catalysts. In contrast to heterogeneous catalysts, molec-ular species can be fine tuned with regard to their electronic and steric properties and the molecular mechanism can be elucidated by spectroscopic techniques. In the following section, a short motivation for the research on enzyme active-site model complexes is given.

1.2 [FeFe]-hydrogenase and its model complexes

Platinum is often regarded as the best catalyst for proton reduction (and hydrogen oxidation), but it is an expensive metal with limited supply. Enzymes on the other hand are natural catalysts enabling numerous chemical reactions with high selectivity, turnover and low thermodynamic penalty without using noble metals [13]. Figure 1.3 shows an electrocatalytic measurement of two natural enzymes adsorbed to carbon electrodes [14]. The oxidized and reduced form of the substrate was added in solution and a cyclic voltammogram (see section 2.4.1) was recorded. The enzymes show perfectly reversible oxidation/reduction at the thermodynamic potential with high exchange current densities and no overpotential. In the case of the reversible proton reduction and hydrogen oxidation, nature was able to develop so called Hydrogenases containing only [Fe], [FeFe] or [NiFe] in the active sites of the enzymes. They play a crucial role in the anaerobic metabolism of bacteria and green algae [15].

(17)

1.2. [FEFE]-HYDROGENASE AND ITS MODEL COMPLEXES 5

Figure 1.3: Cyclic voltammograms of enzymes adsorbed on rotating-disc pyrolytic graphite edge

electrodes under catalytic conditions with both the oxidized and reduced substrates present. (A) Reversible interconversion of H+ _{and H}

2 by hydrogenase-2 from Escherichia coli (pH 6,10%

H2 in Ar, 30 °C) [16]. (B) Reversible interconversion of CO2 and CO by CODH 1 from C.

hydrogenoformans (pH 7, 50% CO in CO2, 25 °C) [17]; adapted from reference [14]

(18)

Figure 1.4: [FeFe] hydrogenase crystal structure from Desulfovibrio desulf uricans [25]. The

electron transfer chain via iron-sulfur centers and the hydrogen pathway are shown schematically. The chemical structure of the active site with its open coordination site at the metal (arrow) is shown to the right; adapted from [26]

frameworks as elaborated in the next chapter. Furthermore, the carbonyl bands of the catalyst offer the possibility to study reactions that occur at the catalyst with the help of IR spectroscopy (see section 2.4.5).

Figure 1.5: [FeFe]-hydrogenase model complexes applied in this thesis: Fe₂(bdt)(CO)₆ (bdt

= benzene-1,2 dithiol), Fe2(mcbdt)(CO)6 (mcbdt = 2,3-dithiolato-benzenecarboxylic acid) and

Fe2(dcbdt)(CO)6 (dcbdt = 2,3-dithiolato-1,4-benzenedicarboxylic acid)

1.3 Metal-organic frameworks

Metal-organic frameworks are a class of materials, first described by Yaghi et al. in 1995 [27], consisting of metal clusters that are interconnected by organic linkers. Within the last 20 years hundreds of different MOFs were developed; three of the most common ones are shown in figure 1.6.

(19)

1.3. METAL-ORGANIC FRAMEWORKS 7

Figure 1.6: Overview over the three most common MOFs: MOF-5 (Zn₄O nodes with

1,4-benzenedicarboxylic acid linker), HKUST-1 (HKUST = Hong Kong University of Science and Technology, copper nodes with 1,3,5-benzenetricarboxylic acid linker) and UiO-66 (UiO = Univer-sity in Oslo, [Zr6O4(OH)4] clusters with 1,4-benzenedicarboxylic acid linker). The core structure

(20)

to material chemists who focus on many different applications ranging from drug delivery [31] over gas storage [32] to catalysis [33–35]. It also explains why over 3500 papers were published in 2013 mentioning "Metal-organic frameworks" according to Web of Science.

One particularly interesting aspect of MOFs is a phenomena called post-synthetic

ligand exchange (PSE). At the right conditions, MOFs are able to exchange a

substan-tial amount of their linker molecules (or metal cluster) without loosing their structural integrity. A schematic picture of the process is presented in figure 1.7. Kim et al., from the research group of Prof. Cohen in San Diego, showed in 2012 [36], that a linker exchange of more than 40% is possible in the UiO-66 framework. This can easily be achieved by placing the MOF-powder into an aqueous solution of a modified ligand which contains the same benzene-dicarboxylic acid unit.

Figure 1.7: The principle of post-synthetic metal and ligand exchange. The metal cluster (blue

pyramid) can be exchanged with different metal ions (purple circle) to yield a chemically altered cluster (purple pyramid). The ligand (orange rod) can be exchanged with a modified ligand containing the same linking unit (blue rod). The yellow sphere should demonstrate the cavity space within the MOF [37]

In contrast to post-synthetic exchange, a process called post-synthetic modification deals with reactions inside the MOF to chemically alter functionalities at the linker or the metal nodes.

1.4 Objective

(21)

1.4. OBJECTIVE 9 exchange (see section 1.3 and figure 1.8). While the homogeneous catalyst is completely degraded in less than 1 hour under illumination, it stays intact within the framework under the same conditions. An incorporation of ca. 15% was achieved, as shown by SEM-EDX (see section 2.4.2) and NMR.

Figure 1.8: Schematic representation of the system developed by Pullen et al. [38]. A

biomimetic [FeFe]-hydrogenase model complex Fe2(dcbdt)(CO)6 was incorporated into a UiO-66

metal organic framework by post-synthetic exchange in water at room temperature for 24 hours.

Besides photochemical experiments that were conducted in the first publication, one of the next projects was to solvothermally grow UiO-66 on a FTO electrode to show if one is able to electrochemically address the catalyst within the framework. This work was done within and after the author0_{s research training in the research group of Dr. Ott} in summer 2014. The experiments showed that it was possible to synthesize UiO-66 on the electrode and that thin films with incorporated catalyst show a distorted, but clearly detectable, reduction/oxidation wave for the catalyst (see figure 1.9). It has to be noted at this point, that UiO-66 itself is non-conductive and the signal is presumably based on a redox-hopping mechanism of the incorporated catalyst molecules. A more detailed discussion on electroactive MOFs can be found in section 2.2.1.

(22)

Figure 1.9: Cyclic voltammograms of Fe₂(dcbdt)(CO)₆ in 1 mM DMF solution (blue), and

UiO-66 film before (green) and after PSE (red). Left: thick UiO-66 film (ca 20 µm). Right: thin UiO-66 film (ca 2 − 5 µm). All CVs recorded in DMF with 0.1 M TBAPF6 as supporting

electrolyte.[39]

Cyclic voltammetric measurements are usually performed in organic solvents to mini-mize catalyst degradation. The protons necessary for catalytic measurements are typically supplied by an organic acid. Unfortunately it was not possible to measure electrocatalysis on the FTO substrate since it undergoes reactions with acids at reductive potentials.

The overall goal for this thesis was to show electrocatalysis of a molecular catalyst incor-porated into a metal-organic framework thin film electrode. Therefore at the beginning

of this project a new substrate material had to be chosen. In 2012, Kumar et al. used a Si photocathode with a very similar [FeFe]-hydrogenase model complex in solution to perform photoelectrochemical measurements under illumination. Silicon electrodes show a strong photoeffect under illumination of ca. 0.5 V and shift the reduction potential of the catalyst to more positive values (see section 2.3.1). They are furthermore stable against acid and provide a perfect platform for MOF growth and analysis methods due to their perfect monocrystalline and flat surface. It was therefore decided to use silicon as the substrate material for the UiO-66 thin films.

Several different tasks were targeted within this master thesis:

(23)

1.4. OBJECTIVE 11 2. Growing UiO-66 thin-films on silicon, incorporate Fe2(dcbdt)(CO)6 by post-synthetic exchange and analyze its capabilities towards photoelectrocatalysis. (see figure 1.10) 3. Motivated by the fact that thick films on FTO did not show any kind of electro-chemical response, apply different techniques to depth profile the MOF thin film and analyze the lateral distribution of the catalyst within the film.

Figure 1.10: Schematic two-dimensional drawing of a UiO-66 thin film with incorporated

Fe2(dcbdt)(CO)6 as a hydrogen evolving photoelectrode. The Zr-nodes represent the secondary

(24)

(25)

Chapter 2

Theoretical background

This chapter contains five sections, which will in detail explain the theoretical background for this thesis, and review the literature. The only applied metal-organic framework within this work is the so-called UiO-66 framework (UiO = University in Oslo) and an entire section will deal with details regarding this material. After that, a more specific theoretical introduction to MOF thin films and a review of synthetic methods to produce them follows. A major part of this section will also deal with a literature review of electroactive MOF thin films. As shown in the introduction, we were able to record a cyclic voltammogram of a molecular catalyst inside a MOF thin film incorporated by post-synthetic exchange [39]. This is to the best of my knowledge, the only report presenting cyclic voltammograms of an electrocatalyst within a MOF thin film.

The next section deals with silicon as a photoelectrode material, its electrochemistry and the importance of its surface condition. At the end, short theoretical descriptions of the applied analytical techniques in this thesis will be given.

2.1 UiO Metal-Organic Framework

(26)

In practice, UiO-66 always contains a large number of defect sites as shown by thermogravimetric analysis (TGA) [44] and later by neutron scattering [42] and single crystal X-ray diffraction [47]. The linker occupancy of the single crystal diffraction was only 73 % for UiO-66 (close to 100 % for UiO-67). It was noted that this figure might be inaccurate due to data fitting reasons and assumed that the defect sites are a mixture of hydroxide, water, and DMF or other coordinating solvents. Even though the authors of the study used benzoic acid as modulator (see next paragraph) during the synthesis, no benzoic acid was found in the crystal structure. Nevertheless, its use prevented the formation of chlorine terminated cluster defects. In the high quality neutron diffraction, the linker occupancy was ca. 90 %.

Schaate et al. was the first to report the use of a modulator (benzoic acid) for the synthesis of UiO-MOFs [48]. The crystallinity and reproducibility was shown to be improved by the addition of a mono-carboxylic species and larger, better separated crystals were produced. Wu et al. also showed a strong influence of the modulator concentration on the pore size, adsorption isotherms and colour of the samples [42].

Ragon et al. reported that the addition of water during the synthesis, accelerates the formation of the Zr-oxocluster, while more acidic conditions like the addition of HCl, acetic-or benzoic acid lead to slower kinetics, probably due to a decrease in the deprotonation rate of the carboxylic linker [49]. They furthermore report that the use of ZrOCl2· 8H2O as a precursor leads to better reproducibility (attributed to the hygroscopic nature of ZrCl4) and higher yields. By applying this precursor and HCl as modulator, the synthesis of UiO-66 particles was achieved in the exceptionally small size range of 100 ± 20 nm.

Shearer et al., from the group in Oslo around Prof. Lillerud that first synthesized the UiO-framework, published an elaborate study on how to eliminate any defects during the synthesis [50]. They describe that a Zr:bdc ratio of 1:2 and an increased synthesis temperature of 220 ◦_{C leads to an ideal UiO-66 MOF.}

Wißmann et al. reported that, while the use of a modulator usually slows down the growth of the MOF, formic acid is accelerating it [51]. They hypothesize that this can be explained by a shift of the well-known decomposition equilibrium of DMF to the left [52] (see figure 2.1). More water is accessible for the synthesis of the Zr-cluster and thereby

the growth rate is accelerated.

(27)

2.2. MOF THIN FILMS 15

Figure 2.1: Hydrolysis equilibrium of DMF leading to the formation of formic acid and

dimethylamine

54]. Even though this is not the case for UiO-MOFs, the creation of dimethylamine and formic acid during hydrolysis certainly plays an important role, specifically with regard to reproducibility. Furthermore, DMF is known to decompose slowly to dimethylamine and CO at its boiling point (153◦_{C) [55]. The temperature for the solvothermal UiO-66} synthesis is ranging from 80-220◦_{C, hence the thermal decomposition is another pathway} to create the basic dimethylamine, which is able to deprotonate the carboxylic acids during synthesis [56].

2.2 MOF thin films

In the first review on the topic of MOF thin films from 2009, Zacher et al. describes that one challenge in MOF research is “...the deposition or growth of MOF thin films on

substrates, ideally in a dense, homogeneous and oriented fashion...” [57]. The interest in

thin films is directly related to the increasing demand in adjusting the optical, electrical or mechanical properties of surfaces and interfaces [58]. Different techniques have been developed throughout the years to deposit or grow metal-organic frameworks on different surfaces like silica, porous alumina, graphite, gold and FTO.

• Microwave irradiation

Yoo et al. presented a microwave-induced method to synthesize MOF nanocrystals on carbon coated porous Al2O3 surfaces. They report a facile synthesis of a MOF-5 thin film with nearly full surface coverage in 30 seconds at 500W irradiation. [59] • Secondary solvothermal growth

(28)

Guerrero et al. have shown that dip-coating, spin-coating and thermal-coating with seed crystals before the solvothermal treatment can also be used [61, 62] to improve the film growth and surface coverage.

• Functionalization of the substrate by an organic linker

Hermes et al. [63] showed that the formation of a self-assembled monolayer of a carboxyl-group terminated silane is leading to a selective adhesion of MOF-5 on a silicon substrate. The MOF was not solvothermally synthesized on the surface. In their approach the MOF is synthesized in solution, filtered and then the sample is immersed into the filtered solution (also called mother-solution) at room temperature for 24 hours. A different example for organic monolayer functionalization has been reported by Huang et al. [64]. In this report a covalent amide linkage between a carboxyl-group terminated silane and the amine group of the imidazolate linker of ZIF-90 was established. It should be noted that the functionalization of silicon with silanes is based on the reaction with the silicon oxide surface layer. This method can therefore not be used if the substrate is oxide free which is necessary for photoelectrochemical measurements (see section 2.3.1). Based on the work of Liu et al. on Al2O3 from 2009 [65] and the work of Kung et al. [66] our group followed a simple functionalization of the FTO samples by immersion into a solution of the linker (terephthalic acid) prior to the solvothermal treatment [67]. This non-covalent coordination seems to promote the growth of the MOF on the substrate.

Conductive substrates like e.g. glassy carbon (GC) or silicon (as shown in section 2.3.2) may also be functionalized with an organic linker by electrochemical grafting of diazonium salts. This has been shown by Balakrishnan et al. [68] and Hou et al. [69] for HKUST-1 and MOF-5 respectively. The growth of continuous films without the pretreatment of the flat glassy carbon electrode was unsuccessful for Hou et al.. Balakrishnan et.al reported that roughening the GC electrode with a SiC paper also creates carboxyl-groups on the surface. It was possible to successfully synthesize HKUST-1 on roughened as well as on diazonium-pretreated GC electrodes by solvothermal synthesis. Hou et al. did not use a solvothermal treatment, but rather used a mother-liquid approach.

• ALD deposition of Al2O₃

(29)

2.2. MOF THIN FILMS 17 methodology of atomic layer deposition (ALD) to coat a silicon substrate with Al2O3. It is argued (similar to Yoo et al. in 2008 [59]) that the low isoelectric point of silicon (2-2.5) compared to the pKA values of terephthalic acid (3.52 and 4.46) is

the reason for the bad adhesion of MOF particles on the silicon (isoelectric point of Al2O3: 9.1 [59])

• Coating of the substrate with polyaniline

Lu et al. reported that covering a substrate with polyaniline before the solvothermal treatment leads to good film growth of HKUST-1, MIL-68 and Zn2(BDC)2DABCO[70]. This method is irrespective of the substrate and its general applicability on stainless steel, copper and platinum was shown.

• Metal substrate as both the support and the metal source

Guo et al. showed that the solvothermal synthesis of the copper-based MOF HKUST-1 on an oxidized copper grid leads to a compact and intergrown thin film [7HKUST-1]. The copper grid, as well as the added copper salt are a metal-source for the MOF growth. Zou et al. reported the growth of Zn3(BTC)2 on a Zn-slide [72] which was activated (hydroxyl-terminated) by hydrogen peroxide pretreatment without the addition of

any Zn-precursor. • Layer-by-layer

A very powerful technique for the thin film synthesis of MOFs is the so-called layer-by-layer growth, which is also known as liquid-phase epitaxy. The method was mainly developed by the groups of Wöll and Fischer and has been reviewed by Liu and Fischer in 2011 [73]. The principle is shown in figure 2.2 - the sample is sequentially immersed into a solution containing the metal salt, a washing solution to remove excess metal precursor and a solution containing the organic liker. The growth is thereby controllable down to the level of single molecular layers and is solely depending on self-assembly. Liu et al. has been able to incorporate multiple functionalities into the MOF [74] and Tu et al. grew thin film heterostructures of Cu3btc2 on top of Cu2ndc2dabco on a quartz crystal microbalance [75] to illustrate some of the recent highlights with this method.

• Electrochemical synthesis

(30)

Figure 2.2: Graphical representation of the proposed model for layer-by-layer growth of

Cu3(btc)2 on an oxide surface. The atoms shown are Cu – green, O – red, C – gray [76]

work of Mueller et al. [77] which was based on the anodic dissolution of metal ions into a solution containing the organic linker. Several research groups have synthesized a wide variety of MOFs by anodic dissolution and the literature has been reviewed by Halls et al. [30]. Al-Kutubi et al. have reviewed the challenges and opportunities of electrosynthesis of MOFs in a broader perspective than anodic dissolution recently in 2015 [78]. One very interesting example of MOF-5 thin film synthesis based on cathodic electrodeposition rather than anodic dissolution was published by Li and Dincă [79].

Figure 2.3: Comparison of solvothermal methods (left) using NEt₃ or Me₂NH as base to

deprotonate the carboxylic linker and Li and Dincă0_{s approach to electrochemically create a}

hydroxyl acting as a base close to the substrate [79]

(31)

2.2. MOF THIN FILMS 19 slowly decomposing to form dimethylamine which acts as a base to deprotonate the carboxylic acid linkers. It was pointed out that the decomposition of DMF is depending on many factors including the metal ion concentration, pH and tempera-ture and thereby leads to strongly varying reaction time between different MOFs. In their paper, Li and Dincă use the elegant way of producing OH– _{as a base to} deprotonate the linker directly at the electrode by the reduction NO–

3 (see figure 2.3). A phase-pure MOF-5 film of 20 − 40 µm thickness was produced in 15 minutes at room temperature, thereby showing a much faster film growth than by solvothermal reaction.

• Electrophoretic deposition

Hod et al. presented a novel method for MOF thin film preparation in 2014 [81]. It is based on the fact that MOFs are negatively charged in solution (possibly to the earlier described missing linker defects) and can thereby be directed towards a positively charged electrode within an electric field. The MOF particles are simply suspended in a low-polarity non-ionizing organic solvent, placed into a flask containing two electrodes and a potential difference of 90 V is applied for ca. 3 hours to get full surface coverage. The method was tested for four different kinds of MOFs (HKUST-1, Al-MIL-53, UiO-66 and NU-1000). It was shown that patterned structures could be produced with this approach by applying photolithography before the electrophoretic deposition. After removing the photoresist an additional layer of a different MOF than the first layer was electrophoretically deposited. In conclusion, there is a wide variety of methods to synthesize MOF thin films. The necessary time to produce samples ranges from several minutes to several days. All methods have advantages and disadvantages and the method of choice is strongly dependent on the application.

2.2.1 Electroactive MOF thin films

(32)

porous nature of MOFs combined with their vast diversity of chemical functionalization and metal-cation coordination presents a huge potential for catalysis and for electrocatalysis in particular. Such materials are an excellent platform to mimic the complex structure and high catalytic activity of active sites found in enzymes, based on abundant transition metals. Simultaneously, MOFs may show the stability of inorganic nano- or microparticles that are today the working principle of industrial electrocatalysts or active electrode materials. In contrast to pore-free nano- or microparticles where only surface metal atoms in contact with the electrolyte can be electrochemically active for surface reactions, the metal-ion utilization in electrochemically active MOFs can theoretically reach 100% when the framework provides sufficient electron and ionic conductivity.”

Halls et al. [30] describe that electrochemical activity of a MOF must be attributed to either the metal center, the organic linker or a guest molecule. This very general approach can further be divided, as roughly sketched by the review of Stavila et al. [83], as follows :

• Inherently conductive MOFs

The first report on porous conducting metal-organic frameworks was by Kobayashi et al. [82]. It was based on nickel bis-dithiolate complexes connected by planar Cu(pyrazine)4 units forming a framework with one-dimensional channels and a con-ductivity of approximately 10−4 _Scm−1_{. Narayan et al. showed a different approach} by columnar π-stacks of TTFTB (TTFTB = tetrathiafulvalene tetrabenzoate) linkers coordinated to Zn2+ _{via carboxylate groups [84] (see figure 2.4).}

Figure 2.4: Nickel bis-dithilate conductive MOF by Kobayashi et al.(left)[82] and p-stacks of

(33)

2.2. MOF THIN FILMS 21 There are reports demonstrating Cu3(BTC)2 as electrocatalyst for CO2 reduction [85] and N,N’-bis(2-hydroxyethyl)dithiooxamidatocopper(II) for ethanol oxidation [86], both showing cyclic voltammograms of the MOF on glassy carbon electrodes. • Electroactive linker

Kung et al. have reported the post-synthetic metalation of the free-base porphyrin linkers used for the growth of MOF-525 thin film [66]. The films show clear electrochemical response due to the electroactivity of the porphyrin linkers.

Usov et al. demonstrate in-situ spectroelectrochemical measurements of the elec-troactive MOF Zn2(NDC)2(DPNI) (NDC = 2,7-naphthalene dicarboxylate, DPNI = N,N0_{-di(4-pyridyl)- 1,4,5,8-naphthalenetetracarboxydiimide) and report that the} twice-reduced dianion species is stable over the timeframe of the experiment [87]. In 2012, Moa et al. reported Cu-bipy-btc (bipy=2,20_-bipyridine, btc=1,3,5-tricarboxylate) as electrocatalyst for O2 reduction [88]. The MOF shows one quasi-reversible peak in the cyclic voltammogram at −0.1 V vs. Ag/Ag+ _{that is} slightly increased in the presence of oxygen compared to nitrogen in the solution. Another contribution for MOFs with electroactive linkers from the group of Prof. Hupp was presented by Kung et al. [66]. In this work, NU-901, a MOF based on Zr-nodes and TBAPy linker (H4TBAPy = 1,3,6,8-tetrakis(p-benzoic acid)pyrene) was electrochemically analyzed and showed electrochromism due to a one-electron oxidation of the linker.

Ahrenholtz et al. solvothermally prepared a thin film of a Co-porphyrin based MOF (5,10,15,20-(4-carboxyphenyl)porphyrin]Co(III)) on FTO glass and showed that the redox-hopping of the porphyrin linker can be explained by a non-nernstian behaviour [89].

• Covalently incorporated electroactive moieties

(34)

charge balance is pointed out. The ferrocene-modified MOF was electrochemically unstable in aqueous solution since the oxidation of ferrocene deeper in the pores caused a local charge imbalance. This was compensated by fast diffusion of protons which in the end lead to high alkalinity in the outer pores and irreversibly broke the framework by disintegration.

Figure 2.5: Comparison of cyclic voltammograms of MOFs with covalently bound ferrocene

presented in literature (see text for further details): (1) scan rate for (i) 10, (ii) 35, and (iii)

100 mVs−1 in 0.1 M NBu₄PF₆ in dichloroethane, MOF powder was immobilized on a basal plane

pyrolytic graphite working electrode [90]; (2) potentials are referred to Ag/AgCl/3M KCl reference electrode, electrolyte not mentioned, powdered MOF was suspended in ethanolic solution of Nafion (5%) and deposited on a graphite electrode [91]; (3) measured in 0.05 M TBAPF6 in acetonitrile

on a FTO electrode [92]

(35)

2.2. MOF THIN FILMS 23 • Non-covalently incorporated electroactive moieties

Dragasser et al. tried to introduce ferrocene from the vapor phase into a layer-by-layer deposited HKUST-1 (Cu3(BTC)2) film on a Au electrode. The conductivity of these films was 2 · 10−9 _Scm−1 _{and no clear ferrocene signal was observed in the} cyclic voltammogram.

Chang et al. [93] introduced ferrocene in a one-step solvothermal treatment of In(NO3)3· 5 H2O, 4,5- imidazoledicarboxylic acid and 10 mmol Fc. It is not de-scribed how the ferrocene is bound within the MOF, but it seems to be trapped within the pores since no adsorption takes place when putting the bare MOF into a ferrocene solution. The modified glassy carbon electrodes had to be conditioned by cyclic voltammography in 0.5 molL−1 _{sulfuric acid for 20 cycles, since the} electro-chemical behaviour was unstable in the beginning. After this treatment the cyclic voltammogram show a clear and stable peak for ferrocene.

• Conductive guest molecules

Talin et al. received a lot of attention with their work based on introduc-ing 7,7,8,8- tetracyanoquinododimethane (TCNQ) into the pores of HKUST-1 (Cu3(BTC)2) [94]. The adsorption lead to a conductivity increase of 8 orders of magnitude up to 7 Sm−1_{. Zhang et al. recently published [95] their work based} on the incorporation of macroporous carbon into a copper based MOF (Co2 (4-ptz)2-(bpp)(N3)2, 4-ptz= 5-(4-pyridyl)tetrazole, bpp = 1,3-bi(4-pyridyl) propane) and

electrocatalysis for the oxidation of hydrazine and the reduction of nitrobenzene was demonstrated. Lu et al. reported the electrochemical synthesis of polyaniline within the pores of HKUST-1 [70]. While the scope of this work was to synthesize microporous polyaniline with the help of a MOF-template (the MOF was dissolved after the polyaniline synthesis), this method might also be used to increase con-ductivity of the MOF itself. The polyaniline showed a concon-ductivity of 0.125 Scm−1 after doping with I2.

(36)

different to a conductive MOF. Within this thesis, the main focus lies in the reduction of a molecular catalyst within a MOF. The reduction potential of an electroactive species would need to be more negative than the catalyst to gain a benefit from the electroactivity of the MOF. In our work on FTO substrates [39], the conduction comes directly from the incorporated catalysts which are present in sufficient amounts. Therefore no driving force is needed to reduce the catalyst within the MOF and no energy is lost in this process.

2.3 Silicon photoelectrodes

Silicon is the second most abundant element in the earth0_{s crust and low-cost, high} purity mono-crystalline wafer material is readily available due to its use in semiconductor industry. It is furthermore used intensively in photovoltaics due to its narrow bandgap of 1.1 eV which is fairly well matched with the solar spectrum[9]. Silicon can be p-doped with elements from the third group of the periodic table and n-doped with elements from the fifth group. In a p-type semiconductor, holes are the majority carriers, while electrons are so called minority carriers and only present in very small concentrations (vice versa in n-type). Since electrons are unavailable in the dark in p-type silicon, optical excitation is required to generate an excited electron in the conduction band, that can effect a reduction process (see fig 2.7). Thus, p-type electrodes are photocathodes, but can perform hole-mediated oxidation processes as anodes in darkness [96]. These photocathodes typically show a photoeffect of ca. 0.5 V under illumination due to the energy gained by the absorption of a photon. This means that the reduction potential of a redox species in solution is shifted to more positive values compared that measured at a standard platinum working electrode [96, 97].

The research of silicon photoelectrochemistry with regard to water splitting is mostly focusing on two different topics. The morphology and nanostructure of silicon is optimised with regard to light absorption and charge carrier extraction. The second issue is the deposition of a heterogeneous material to optimize the surface energies and kinetics with regard to water splitting while suppressing surface oxidation of the silicon [9].

2.3.1 Silicon-liquid interface

(37)

2.3. SILICON PHOTOELECTRODES 25 can be shifted by applying an external potential. The Fermi level Ef represents the thermodynamic potential of the electrons in the silicon. If the Fermi level is more positive in the semiconductor than the redox couple in solution, electrons will be transferred to the solution, leaving behind accumulated, immobile positive charges at the interface. This so called space-charge layer leads to an electrical field that directly affects the local thermodynamic potential of the electrons. The Fermi level of the semiconductor and the thermodynamic potential of the redox couple in solution in equilibrium are thereby always the same under equilibrium conditions. The space-charge layer leads to a band bending as shown in figure 2.6: the majority holes are drifted away from the interface due to the band bending, while excited minority electrons are drifted towards the interface and can potentially reduce the protons in solution. However, the ECB of silicon is too high, resulting in only minor band bending and thus leading to a low driving force for proton reduction [9].

Figure 2.6: p-type silicon band edge positions at equilibrium with aqueous solution with the

proton reduction redox level as reference. ECB = Conduction band level, EVB = Valence band

level, Ef = Fermi level, ϕB = band bending [9]

Besides this, a more important issue is hindering the use of silicon as a photocathode for proton reduction - bare silicon is unstable in aqueous solution, getting readily oxidized and forming an electron-blocking passivation layer [98].

(38)

Figure 2.7: p- and n-type silicon-liquid interface under illumination showing the different bend

banding and thereby changed reactivity towards oxidation/reduction [97]

(39)

2.3. SILICON PHOTOELECTRODES 27

2.3.2 Silicon surface modification

The aspect of silicon surface modification was not expected to play a very crucial role within this thesis, but as explained above, the passivation of silicon by its oxide leads to trapped states, that prevent electron transfer over the silicon-electrolyte interface. Already in the beginning of the 19800_{s several techniques were suggested to improve the stability} of silicon, including derivatization of the electrode surface [98].

This section will only deal with the covalent modification of hydrogen-terminated silicon, since the removal of the oxide layer is crucial for any kind of electrochemical application. A SiO2 modification with silanes might promote MOF thin film growth as described in section 2.2, but it is irrelevant for photoelectrochemical applications. As mentioned earlier, treatment of Si with diluted HF or aqueous NH4F solution leads to a removal of SiO2 and well controlled hydrogen terminated Si. One has to differentiate between the two most common orientations Si(100), which is extensively used in the semiconductor industry, and Si(111) (see figure 2.8). The covalent grafting of organic monolayers on hydrogen terminated silicon is preferentially performed on Si(111) since it is atomically more flat and the mono-hydride can typically be replaced with higher yields [106]. There are several methods to covalently modify the silicon surface and several reviews have been published in the last years to summarize this large research field [107–114]. In general, reactions of 1-alkynes can typically also be carried out with the respective 1-alkene, but Scheres et al. [115] showed that the packing density and ordering of the monolayers is higher for alkynes. The maximum surface coverage of alkyne-derived monolayers bound to Si(111) is 60-65 %, while it is only 50-55 % for alkene-derived monolayers. Furthermore Puniredd et al. showed that the nature of the Si−C−−C linkage is inhibiting surface oxidation in contrast to the case of Si−C−C and Si−C−−_{−C[116]. Surface modifications can also be performed with alcohols and aldehydes} to presumably form Si−O bonds. These reactions will not be reviewed here since there is no direct evidence for the Si−O bond formation and there is only few examples in the literature [107].

• Thermal hydrosilylation with 1-alkynes

(40)

Figure 2.8: Hydrogen terminated Si(100) and Si(111) [106]

surface. After further investigation by the groups of Chidsey and Sudhölter [118], it was found that homolytic Si−H cleavage at elevated temperature is the starting point for the radical chain mechanism.

Sieval et al. showed for the first time that using solutions containing only 2.5 % 1-alkene in mesitylene can also be used instead of the neat alkene [119]. Other solvents than mesitylene did not perform as well with regard to the ordering of the monolayer.

A very exciting example was presented by Ciampi et al. in 2009 [120]. They function-alized Si(100) by a thermal reaction with 1,8-nonadiyne and reported that this distal alkyne shows superior stability vs. 1-heptyne. Furthermore they functionalized the acetylene-terminated moiety with azidomethylferrocene and presented nicely reversible cyclic voltammograms with up to 108 cycles stability.

• Photochemical hydrosilylation with 1-alkynes

Langner et al. reported functionalization of Si(111) and Si(100) with UV-light [121]. Sun et al. showed in 2005 that the hydrosilylation reaction with 1-alkenes and 1-alkynes is possible under visible light irradiation and at room temperature [122]. They propose an electron-hole pair mechanism due to the excitation of the silicon under irradiation (see figure 2.9). It is also shown that the reaction rate is doping dependent in the order highly doped n-type > lowly doped n-type > lowly doped p-type > highly doped p-type.

• Catalyzed hydrosilylation with 1-alkynes

Langner et al. describe in their publication that the use of Speier0_{s catalyst} (H2[PtCl6]) leads to a preferred reaction of the solvent 2-propanol with the surface

(41)

2.3. SILICON PHOTOELECTRODES 29

Figure 2.9: Proposed radical chain mechanism for the functionalization under thermal and UV

conditions and the electron/hole pair mechanism for visible light induced hydrosilylation [122]

on the surface. While the Lewis-catalyzed reaction is possible at room temperature with a porous-silicon substrate, it requires 100◦_{C and longer reaction times when} performed on a Si(111) substrate [112].

Webb et al. showed that surfaces modified in presence of 1-hexene and 1-octene with EtAlCl2 exhibit high charge recombination and oxidize in air [101] while Boukherroub et al. describe that their functionalized silicon (1-decene) is stable for weeks [123].

• Two-step halogenation alkylation

(42)

• Electrochemical grafting of a diazonium-salt

One of the main advantages of the electrochemical grafting method is its compatibility towards a number of substituents like Br, NO2, CN, NH2 , COOH and alkyl [107]. The electrochemical functionalization of Si(111) with 4- NO2-benzenediazonium and 4-Bromobenzenediazonium by cyclovoltammetric scanning in a HF and H2SO4 containing electrolyte was shown by de Villeneuve [126]. The substrates show no sign of surface oxide by XPS measurements after the electrochemical synthesis. Similar work performed by de Villeneuve, Allongue and co-workers [127] show oxidation after 24 hours. It should be noted that scanning tunnelling microscopy showed very well ordered monolayers with a few pit-holes that seem to be the starting point for surface oxidation. Fabre states that the electrochemical grafting with a diazonium salt fails to produce monomolecular films in a reproducible and controllable manner [106].

2.4 Analytical Methods

Metal-organic frameworks are a complex material class that needs to be analyzed with a wide variety of methods. Within this section, all applied analytical methods will be discussed shortly.

2.4.1 Cyclic voltammetry (CV)

(43)

2.4. ANALYTICAL METHODS 31 the exchange current density j0, the charge transfer coefficient α (indices a for anodic and c for cathodic) as well as some fundamental constants, according to Butler-Volmer equation (see equation 2.1). It is either limited by mass-transport of fresh reactant to the eletrode surface, resistive losses or by electron transfer kinetics.

j = j0· exp(

αanF η

RT −

αcnF η

RT ) (2.1)

2.4.2 Scanning Electron Microscopy (SEM)

(44)

Figure 2.10: Monte-Carlo simulations of electron trajectories within a 2000 nm thick UiO-66

bulk sample (density = 1.443 g/mL) on a Si substrate performed with the software CASINO version 2.48 [130]. The red paths show back scattered electrons that are re-emitted from the sample. All simulations were performed with 20000 electrons, a beam radius of 5 nm and all other software settings on standard. The accelerating voltage was stepwise increased from 5 keV (top left) to 10 keV (top right), 15 keV (bottom left) and 20 keV (bottom right)

2.4.3 X-Ray Photoelectron Spectroscopy (XPS)

A X-ray beam of precise energy (e.g. Kα line of Al) is focussed on the sample. In the

opposite way as in the SEM-EDS, a core-level electron is ejected out of the sample due to interaction with the incoming X-ray beam. The energy of the ejected electron can be determined by a hemispherical detector system and is element characteristic. The entire process is described by Einstein0_{s equation on the photoelectric effect:}

Ebinding = Ephoton− Ekinetic (2.2)

(45)

2.4. ANALYTICAL METHODS 33 XPS is furthermore a very surface sensitive technique, due to the low free path length of the ejected electrons in the solid. The depth resolution is typically around 10 nm [131]. As soon as the electrons come from deeper inside the sample they will loose energy due to scattering and will thereby be lost within the spectral background.

2.4.4 Powder X-Ray Diffraction (PXRD)

Singe crystal X-ray diffraction is the most important analytical tool to analyze molecular structures and many important scientific advances in the last century emanated from this technique [132]. Nevertheless, not every material can be prepared as a single crystal in the appropriate size and quality and only powder diffraction can be used in this case. Large progress has been made in the field of powder X-ray diffraction (PXRD) and it is now routinely possible to determine structural properties of the sample without the need for a single crystal. One main application of PXRD is nevertheless its application for ’fingerprinting’ a solid and thereby enabling a quick qualitative characterization [133]. Besides that, it is commonly used for the size determination of sub-micron particles by Scherrer0_{s equation. Another application is the analysis of preferentially grown crystals on} a substrate surface. This can be determined by the signal increase of certain diffraction peaks compared to the other peaks. [133]

2.4.5 Infrared (IR) Spectroscopy

(46)

increase of electron density in the antibonding orbital of the CO ligand, leads to a shift of the carbonyl stretching vibration to lower wavenumbers (lower energy).

It is also possible to measure monolayers on silicon substrates as shown for example by Webb et al. [134] or Salingue and Hess [135].

2.4.6 Ion-Beam Analysis (IBA)

Ion-Beam Analysis (IBA) is a family of different techniques which use energetic ions (several ten keV to several ten MeV) to probe the elemental composition and depth profile of solid materials. The underlying physics is well understood for several decades, but the techniques are typically not used routinely due to the need for a particle accelerator. The IBA-methods have in common that they are able to provide quantitative information without matrix effects on a nano-meter scale which is hardly available through other techniques. The three methods applied in this thesis are:

• Rutherford Backscattering Spectrometry (RBS)

• Time-of-Flight Elastic Recoil Detection Analysis (TOF-ERDA) • Time-of-Flight Medium Energy Ion Scattering (TOF-MEIS)

(47)

2.4. ANALYTICAL METHODS 35 scales being accessible: RBS probes up to a few µm; TOF-MEIS with its higher resolution can probe approximately the first 50-100nm. TOF-ERDA uses heavier incident ions like chlorine or bromine with energies in the range of 20-60 MeV to eject light and medium heavy elements from the sample. These recoiled nuclei are separated in mass with a time of flight detector leading to element specific analysis. The energy of the recoiled ions is -besides their mass - also depending on its depth within the sample. Similar to RBS and MEIS the ions loose energy on their way out of the sample by inelastic electron scattering. For a more complete description of TOF-ERDA see reference [136].

2.4.7 Time-of-flight secondary-ion mass spectrometry

The application of time-of-flight secondary-ion mass spectrometry (TOF-SIMS) lies in the analysis and depth profiling of solid materials with a nanometer resolution and sub-ppm detection limit. The impact of ions (typically Ga or Ar) with energies between 1 − 20 keV leads to the ejection of neutral and charged species from the surface [129]. The charged secondary ions are accelerated into a time-of-flight mass spectrometer and analyzed according to their mass to charge (m

z ) ratio. To enhance the sputter rate a

second ion gun with higher acceleration voltage can be applied. The two ion guns (analysis beam, sputter beam) are then used alternating to create a depth profile of the sample. An additional electron-flood-gun might be used to decrease charging effects in the sample. Absolute quantification is typically difficult due to matrix effects like varying sputtering probability for elements in different chemical environment. The lateral resolution of the measurement is typically 50 − 2000 nm.

2.4.8 Dynamic light scattering (DLS)

(48)

index plays a crucial role for this mathematical treatment. Practical issues regarding the experiment are typically dust and particle agglomeration.

2.4.9 Contact angle measurement

The measurement of the contact angle is a quick method to determine the hydrophobicity of a surface. The surface of the substrate has to be smooth and clean, due to the extreme sensitivity towards roughness and contamination. In general values are only reproducible in the range of ±2◦ _{under laboratory conditions [137]. Besides its use in the analysis of} the substrate-liquid interaction, it can also be used to determine surface energetics of solids. The surface of the solid has to be smooth, rigid and homogeneous to be able to apply Young0_{s law, otherwise more elaborate models have to be used [137].}

2.4.10 Inductively

coupled

plasma

atomic

emission

spec-troscopy (ICP-AES)

(49)

Chapter 3

Experimental

All experimental procedures applied in this thesis are summarized in this chapter. The synthesis of [FeFe](bdt)(CO)6, [FeFe](mcbdt)(CO)6 and [FeFe](dcbdt)(CO)6 were per-formed by S. Pullen and have been reported elsewhere [38, 139]. The anchoring group 40_{-trifluoromethylphenyl acetylene was synthesized by Dr. Ulrike Fluch in the same} research group and only deprotected with sodium hydroxide before use. It was extracted in pentane and the solvent was removed by gently purging air over the solution due to the volatility of the target compound.

3.1 Chemicals and Purification

All chemicals and solvents were purchased from commercial suppliers and used without further purifcation if not listed below.

• ZrCl4 (> 99.99 %) • Methanol (> 99.9 %) • Sulfuric Acid (95-97 %)

(50)

• Copper iodide (98 %)

Benzoic acid (>99.5 % starting material) was reprecipitated by dissolving in a sodium

hydroxide solution and addition of hydrochloric acid until pH 1 was reached. The solid was filtered, washed and stirred for several hours in copious amounts of water to remove excess acid. The pH of the washing liquid was ca. 4, which fits well to the reported pKA

value of benzoic acid (4.2) [140]. The reprecipitated benzoic acid was dried over P2O5 under vacuum.

Benzene-1,4-dicarboxylic acid (=terephthalic acid, 98 % starting material) was

subli-mated at 3 − 6 · 10−1_{mbar and 250} ◦_{C in a Kugelrohr apparatus. It was dried over P} 2O5 under vacuum.

For experiments performed in the glovebox, N,N-Dimethylformamide (DMF, > 99 %) was dried overnight with 4 Å molsieves and distilled at 48 − 50◦_{C under reduced pressure} (22 − 28 mbar) into Schlenk flasks with 4 Å molsieves. Afterwards it was de-oxygenated

with 5 freeze-pump-thaw cycles.

3.2 Synthesis of para-ethynylbenzoic acid

The synthesis route is summarized in figure 3.1. 4-Bromobenzoic acid (2 g, 9.95 mmol) was dissolved in dichloromethane (7ml) at 0◦_{C. Thionylchloride (1.45 ml, 19.9 mmol)} was added and the solution was stirred for 15 minutes before methanol was dropped into it slowly. The solution was stirred for 3 days at room temperature, quenched with water and neutralized with saturated NH4CO3 solution. The product was extracted with dichloromethane, dried over MgSO4 and the solvent was removed under vacuum. Methyl 4-bromobenzoate was obtained with a yield of 95 % as a grey beige solid.

(51)

3.3. SUBSTRATE CLEANING 39 Methanol-D3) δ 8.00 (d, 2H), 7.56 (d, 2H), 3.72 (s, 1H).]

The deprotection was performed with 5 equivalents NaOH in THF:MeOH:H2O 3:2:1 overnight at room temperature.

Figure 3.1: Synthesis route of para-ethynylbenzoic acid

3.3 Substrate cleaning

The silicon samples were cleaned in Piranha acid (3:1 vol. [conc. H2SO4):(30 % H2O2]) for at least 30 minutes at 80◦_{C. Afterwards, the samples were rinsed with deionized water} and dried under a stream of air. These samples will be referred to as ’oxidized Si’. If not mentioned otherwise, these oxidized Si samples were used for the growth of UiO-66 thin films. Samples that did not undergo this cleaning procedure will be referred to as ’native oxide’. The silicon oxide layer was removed by etching with a buffered oxide etch solution (6:1 volume ratio of 40% NH4F in water and 48% HF) for at least 2 minutes. The samples were rinsed with deionized water and dried under a stream of air.

The FTO pieces were sonicated for 10 minutes in Alconox cleaning solution, ethanol and acetone sequentially. After the Alconox cleaning the samples were rinsed with water, and after the ethanol cleaning the samples were rinsed with acetone respectively. Finally the acetone was removed by a stream of air.

3.4 Silicon surface functionalization

(52)

First experiments under illumination were performed in toluene, later in DCM/DMF mixtures and dry acetonitrile. None of the solvents was able to dissolve the compound in a target-concentration of 10 mM. Therefore, DMF was chosen as a reaction medium due to the good solubility.

Functionalizations with DMF as solvent were entirely carried out in the glove box besides the sample cleaning in Piranha solution as described above. It was reported in the literature that the monolayer formation is significantly improved if the microelectrodes maintained under inert conditions at all time [141]. The wafer pieces were treated in buffered oxide etch in the glove box, rinsed with water and THF and dried under vacuum in the antechamber. Three different procedures were tested:

• Reaction with EtAlCl2 Lewis-acid catalyst • Reaction with white light illumination • Reaction without catalyst or illumination

The Lewis-acid catalyst EtAlCl2 (220 µL, 400 µmol) was added to 4 ml of a 10 mM para-ethynylbenzoic acid solution in DMF. The silicon samples were added, the vials were closed and kept inside the glovebox for at least 12 hours. Samples functionalized with white light (5cm distance to a 17 W, 5000 K lamp) were illuminated overnight outside the glovebox. This lead to a color change from orange to transparent for para-ethynylbenzoic acid solutions. The concentrations and procedure for samples without Lewis-acid catalyst and for 40_{-trifluoromethylphenyl acetylene were the same. There was no color change} observable for these samples under white light illumination.

3.5 UiO-66 thin film synthesis

The original protocol of our recent publication [39] to solvothermally synthesize UiO-66 thin films on FTO was the starting point for several tests and will be explained here:

(53)

3.5. UIO-66 THIN FILM SYNTHESIS 41 temperature(20-30 minutes), terephthalic acid (25 mg, 0.15 mmol for ca. 20 nm thick films; 16 mg, 0.10 mmol for ca. 2 − 5 nm thin films) was dissolved in the solution by sonication. The pretreated FTO substrate was placed at the bottom of the vial with the conductive site facing upwards. The vial was allowed to react solvothermally at 120◦_C in a sand bath in an oven for 24 hours. After cooling down to room temperature(20-30 minutes), the film was rinsed with DMF and incubated in DMF for 24 hours. Then the solution was changed to methanol for another 24 hours. This synthesis method will be referred to as old-method. After incubating the sample in methanol, the film is ready to be functionalized via post-synthetic ligand exchange. Due to the higher abundance of the mono-carboxy version of [FeFe](dcbdt)(CO)6 all experiments within this thesis were performed with this complex. The catalyst [FeFe](mcbdt)(CO)6 (41.3 mg, 0.1 mmol) was dissolved in deoxygenated methanol (2 ml) by sonication for ca. 20 minutes. The insoluble residue was removed by centrifugation. The UiO-66 thin film sample was placed inside the vial with the film side upwards for 24-72 hours at room temperature. Afterwards the film was washed for three times with methanol, each time 24 hours.

Due to problems with reproducibility within a series and compared between different series, several improvements and changes were implemented in the procedure:

• Better control over water content

As explained in chapter 2.1, the water content in the synthesis solution plays a crucial role in the reaction rate of the cluster-formation. It is thereby highly important to control the dryness of the starting materials. As mentioned in section 3.1, benzoic acid and terephthalic acid were purified and extensively dried. DMF was dried over 4 Å molsieves. A certain amount of water is nevertheless necessary, therefore a fixed amount of water (typically 5 equivalents, 0.75 mmol) was added intentionally to the solution prior to the solvothermal treatment.

• Use of stock solutions

(54)

the pre-treatment at 80◦_{C, 2 ml of a terephthalic stock solution (75 mM) was added.} The final concentrations per sample are the same compared to the old-method. Nevertheless the volumes per sample were slightly larger with the old-method due to a large volume increase when dissolving the starting materials.

• Orientation of the sample during solvothermal treatment

Placing the sample at the bottom of the vial leads to an overlap of on-substrate growth and precipitation of UiO-66 on the substrate. It was therefore suggested by our collaborator Wei Xia, to place the sample standing up straight into the solution. This is practically very difficult, therefore the substrate shreds were instead cut into larger pieces (30 mm · 6 mm) and placed tilted into the vials (see figure 3.2). Samples prepared with this method will be referred to as ’upside-down’. When the samples were lying on the bottom of the vial, it will be referred to as ’standard’. The top side in the upside-down method is very similar to the standard synthesis conditions with precipitating MOF particles deposited on the film, leading to higher roughness and thicker films. The silicon wafers used in this thesis were single-sided polished. This side was used as the bottom side in the upside-down method.

Figure 3.2: Scheme illustrating the upside-down synthesis method (description see text)

(55)

3.6. METHODS OF CHARACTERIZATION 43 Some tests were performed with electrophoretic deposition rather than solvother-mal synthesis: a few mg of dry UiO-66 were suspended in toluene (ca. 10 ml) by sonication and placed in a three-neck round bottom flask. Two FTO electrodes were mounted in holders and placed at the left and right neck of the flask with the conductive site facing each other. The central neck was sealed to minimize evaporation of toluene. A voltage of 90 V was applied between the two electrodes for three hours and was kept while removing the samples from the solution to prevent the loss of substrate adhesion.

3.6 Methods of Characterization

3.6.1 Cyclic voltammetry

A three or four electrode system, consisting of a glassy carbon working electrode and/or a silicon working electrode, a glassy carbon counter electrode and a Ag/Ag+ _(10mM AgNO3) was used for the cyclic voltammetry experiments. The supporting electrolyte was 0.1 M TBAPF6 in dry acetonitrile and was deoxygenated by Ar-bubbling. Measurements with a silicon working electrode were conducted under illumination with a white light source. The light intensity was not determined due to a lack of equipment and practical issues concerning the measurement.

The silicon photoelectrodes were contacted by scratching the top part of the sample and immediately applying an In/Ga eutectic on it. A bare aluminum wire was fixed to the wafer with conductive silver paste. To increase the mechanical stability, the silver paste was air dried for ca. 20 minutes and applied 2-3 times. The entire sample, besides a small ca. 3 mm hole, was covered with epoxy resin (Loktite Hysol 1C). The step-by-step procedure is illustrated in figure 3.3.

3.6.2 Contact Angle Measurement

(56)

Figure 3.3: Step by step contacting procedure of the Si photoelectrodes

3.6.3 Ion Beam Analysis

The experiments were carried out in collaboration with Dr. Daniel Primetzhofer at the accelerator system in Ångstrom laboratory. ERDA measurements were performed with Br79 _{at 32 MeV and with I}127 _{as well as Cl}37 _{nuclei at 36 MeV. The energy of the helium} ions in RBS was 2 MeV and 80 keV in MEIS respectively.

3.6.4 XPS

The XPS measurements were performed on a Physical Electronics Quantum 2000 Scanning ESCA microprobe system. The measurement ranges for each element were chosen automatically. The resolution, pass energy and number of scans was adjusted to keep the measurement time close to 60 minutes. Besides a surveillance scan, most attention was paid to the fine scan of the carbon C1s peak to elucidate the characteristic COO– _{peak at} 288-290 eV.

Preparation, functionalization and analysis of UiO-66 metal-organic framework thin ﬁlms on silicon photocathodes