Self‐assembled Ru(bda) Coordination Oligomers as Efficient Catalysts for Visible Light‐Driven Water Oxidation in Pure Water

Abstract Water‐soluble multinuclear complexes based on ruthenium 2,2′‐bipyridine‐6,6′‐dicarboxylate (bda) and ditopic bipyridine linker units are investigated in three‐component visible light‐driven water oxidation catalysis. Systematic studies revealed a strong enhancement of the catalytic efficiency in the absence of organic co‐solvents and with increasing oligomer length. In‐depth kinetic and morphological investigations suggest that the enhanced performance is induced by the self‐assembly of linear Ru(bda) oligomers into aggregated superstructures. The obtained turnover frequencies (up to 14.9 s−1) and turnover numbers (more than 1000) per ruthenium center are the highest reported so far for Ru(bda)‐based photocatalytic water oxidation systems.

Oligomer characterization 9 Vapor pressure osmometry 15 Optical and electrochemical properties 24 Visible light-driven water oxidation catalysis 25 Analytical data for molecular precursors and Ru(bda) oligomers 34 Literature 39
NMR spectroscopy 1 H NMR and 13 C NMR spectra were recorded on an Avance III HD 400 spectrometer (Bruker Daltonics GmbH) operating at a temperature of 295 K and a frequency of 400 MHz ( 1 H) and 100 MHz ( 13 C), respectively. Chemical shifts () are stated in parts per million (ppm) using the residual solvent signal for calibration against TMS and coupling constants J are given in Hz.

Mass spectrometry
High-resolution ESI-TOF measurements were obtained on an ESI MicrOTOF focus mass spectrometer (Bruker Daltonics GmbH).

Elemental analysis
Elemental analysis was performed on a Vario MICRO cube (Elementar Analysensysteme GmbH).

Flash column chromatography
If not stated otherwise, compound purification was achieved using an automated flash purification system PuriFlash 420 (Interchim) using pre-packed silica gel columns (Interchim) with a grain size of 30 m and distilled solvents.

Electrochemistry
Cyclic voltammetry (CV) and differential pulse voltammetry (DPV) measurements were performed on a BAS Cell Stand C3 (BAS Epsilon) using a three-electrode setup with glassy carbon as working electrode, platinum wire as counter electrode and Ag/AgCl (3  KCl) as reference electrode. Voltammograms were recorded at a scan speed of 100 mV s -1 . The experiments were carried out in neutral aqueous phosphate buffer (0.1 ionic strength at pH = 7.0) as solvent. The obtained potentials were converted to normal hydrogen electrode by addition of +0.21 V. [S6] UV/vis absorption spectroscopy UV/vis spectra were recorded on a V-770 UV/vis/NIR spectrophotometer (Jasco Inc) at 298 K using Suprasil quartz cells with a path-length of 1 cm. Solutions were prepared using spectroscopy grade solvents (acetonitrile) or doubly distilled water that was obtained by passing distilled water through a Millipore filter system. Absorptivities are converted into extinction coefficients using the Beer-Lambert law.

Visible light-driven water oxidation catalysis
Visible light-driven water oxidation experiments were performed using an Oxygraph Plus Clarkelectrode system (Hansatech Instruments Ltd.) for oxygen detection. Sample irradiation was achieved by a 150 W xenon arc lamp (Newport) equipped with a 400 nm cutoff filter. The light intensity was calibrated to 100 mW cm -2 using a PM 200 optical power meter with a S121C sensor (Thorlabs Inc) in combination with a CCS 200/M wide range spectrometer (Thorlabs Inc). Experiments were conducted using a stock solution of photosensitizer Ru(deeb)2(bpy)Cl2 (c = 0.27 m) and sacrificial electron acceptor Na2S2O8 (c = 50 m) in mixtures of aqueous phosphate buffer (pH = 7.2, Honeywell Fluka) and acetonitrile in different ratios. An aliquot of the respective stock solution (V = 1.5 mL) was transferred to a transparent, water-cooled reaction chamber (T = 293 K) and mixed with a catalyst solution in the same solvent mixture while kept in the dark. The sample volume was adjusted with pure solvent mixture to reach 2 mL. Irradiation was initiated after 40 seconds to allow thermal equilibration of the sample at 20 °C. The TOF was obtained by linear regression of the initial rate of oxygen evolution during the first few seconds of catalysis. were used. For preparation of AFM samples, the solution of 4a and 4b in water or water/acetonitrile (, 6:4) with a concentration of ruthenium centers of 1.00 × 10 -5  was spincoated onto mica with 7000 rpm. Statistical analysis of the length distribution was performed using the program NanoScope Analysis 1.5 (Bruker).

Molecular modelling
Molecular modelling was performed using the program Materials Studio (Biovia). [S9] Structural models were energy-minimized with the Universal Force-field (UFF) with a cut-off distance of 12.5 Å by using the Forcite Task.
Subsequent Suzuki coupling gave linker precursor 1 in 75 % yield. Finally, polymerization of ditopic ligand 1 and Ru(bda)(dmso)2 (2) in anhydrous and deaerated methanol was carried out in different ratios to give the desired oligomers in more than 70 % yield for 4a and above 90 % yield for 4b, respectively. For 4c, linker 1 was reacted with an excess of Ru(bda)(dmso)(4-pic) (3) to give the desired compound in 10 % yield.
Scheme S1. Synthetic scheme for the synthesis of 4a-c.
After cooling to room temperature, water (20 mL) is added, and the mixture is extracted with dichloromethane (6 × 70 mL). The combined organic layers are dried over sodium sulfate and evaporated. The crude compound is purified by column chromatography on silica (gradient elution using dichloromethane and 0 %-10 % methanol as eluent).

Oligomer characterization Oligomer formation process
To assign proton signals in the 1 H NMR spectra of the final oligomers, the formation process was monitored by heating an equimolar ratio of Ru(bda)(dmso)2 and ditopic pyridine linker 1 in CD3OD at 60 °C and measuring time-dependent NMR spectra. Generally, three different cases of linker positioning within the oligomer need to be considered and can be distinguished by different chemical shifts in the spectrum: 1) The linker is only coordinated to a ruthenium center with one nitrogen atom (blue case, end-group) 2) The linker is coordinated to a ruthenium center opposite of a dmso ligand (red case) 3) The linker is coordinated on both sides to the ruthenium center with opposing pyridine ligands (green case, central unit)  Figure S1. Chemical structure of an oligomer based on linker 1 and Ru(bda)(dmso)2 and the three different cases of linker positions within the oligomer. Bottom: Time-dependent 1 H NMR spectra (400 MHz, CD3OD, 295 K) to monitor the oligomer formation process and assign signals according to the placement of the linker within the oligomer.
Detailed analysis of the NMR spectrum gives a clear indication for assignment of the signals to the three cases described above. In the initial spectrum (blue curve), signals of the linker unit are almost exclusively visible due to bad solubility of the Ru(bda)(dmso)2 complex in methanol at room temperature. Upon heating, the formation of the oligomer can be observed and NMR signals assigned to the three different ligand placements arise. For reasons of simplification, only signals belonging to linker 1 are discussed in this summary: 1) Distinctive for the blue end-group ligand is the fact that only one nitrogen atom is coordinated to the ruthenium center leading to one pronounced high-field shift of aromatic pyridine protons of 1 while the signal of the protons attached to the noncoordinated pyridine do not show a significant shift. These pyridine protons additionally split into four different signals due to the different environment upon one-fold coordination to ruthenium. Further splitting of the signal belonging to the protons at the central benzene moiety supports the fact of the linker placement on the outer part of the oligomer. In the beginning of the reaction, no prediction can be made whether these signals belong to a linker species coordinated to a ruthenium center with a dmso or another linker 1 ligand on the other coordination site, making this assignment only conclusive after the reaction is carried out.
Other protons exhibiting NMR signals indicative for this case belong to the methyl groups of the OEG chains. This signal is slightly shifted to higher fields compared to the free linker and also splits into two separate singlets.
2) After short reaction time, signals belonging to 1 coordinated on one side opposite to dmso and on the other side opposite of another 1 molecule (red case) become visible. Here, a more pronounced high-field shift compared to the blue case can be observed. However, only signals for the protons attached to the pyridine ligand opposing the dmso ligand are visible while signals belonging to the other pyridine ligand get mixed within the signals belonging to the core units of the oligomer (green case) due to their electronic similarity.
The existence of dmso end-groups can be extracted from the signal at 2.90 ppm belonging to the bound species. Free dmso arising during the reaction progress, can in contrast be observed at chemical shifts of around 2.65 ppm.
3) Furthermore, signals belonging to the central units increase after short reaction time and feature broad signals without a defined structure. All signals are shifted to the high-field region compared to the other cases.
Subsequently, additional free linker 1 in CD3OD solution was added to the NMR tube and the reaction was carried out for 5 h at 60 °C before another NMR spectrum was recorded. This method proved useful to assign signals to the respective cases where the linker units are located on the outer part of the oligomer strand (red and blue case in Figure S1). Figure S2. 1 H NMR spectra (400 MHz, CD3OD, 295 K) of the final oligomer described in Figure  S1 (blue curve) and spectra measured after addition of linker 1 and subsequent reaction for 5 h at 60 °C in CD3OD. The blue rectangles mark signals belonging to the blue case as described in Figure S1.
In the spectrum, the signals belonging to 1 as end-group are marked in blue. Upon addition of 0.20 eq of linker 1, the rise of six signals in the aromatic region become visible. This is a result of the one-fold coordination to a ruthenium center and concomitant splitting of the NMR signals.
Additionally, two signals rise for the methyl groups of the OEG chain at around 3.28 ppm.
Simultaneously, the signal assigned to dmso as end-group keeps decreasing while linker is added to the mixture indicating the efficient transformation of the end-groups when free pyridine ligands are added to the solution.
End-group analysis Elemental analysis 4a Figure S4. Schematic representation of two oligomers consisting of 17 Ru(bda) and 16 linker units and dmso and linker end-groups in a 1:3 ratio as estimated by 1 H NMR spectroscopy of 4a (rectangle: linker 1, oval: Ru(bda), triangle: dmso). [a] No signal observed due to the sensitivity of the device used.
With a molecular weight of around 1340 g mol -1 per repeating unit (linker 1 + Ru(bda)), this results in an average of 12-14 repeating units per oligomer chain.
The same measurement and calculation adapted to 4b gives: = .
Considering the mixed end-groups within this oligomer, 5 repeating units can be assumed from the VPO experiments.

Molecular Modelling
Based on the average molecular weight and oligomer size obtained from VPO measurements, molecular models of the respective oligomers were constructed to determine the average chain length of the materials. [S9] The geometry-optimized structures after force-field calculations are shown in Figure S7 (4a) and Figure S8 (4b).  192.53 Å
The log-normal model for DOSY evaluation was chosen due to the polydisperse nature of the oligomer samples. Hereby, the individual signal decay curves are fitted according to equation (2) where the probability for each diffusion coefficient is included by the PLNd) function: In this equation,  is the gyromagnetic ratio (   = 4257.64 Hz G -1 ), the gradient pulse duration, gz the gradient strength, D the diffusion coefficient and  the diffusion time. LN and D0 are coefficients that determine the shape of the distribution.
Four individual fits were performed as shown in Figure S10a-d for 4a and Figure S10e-h for 4b. Table S3 and Table S4 summarize the data obtained from the fits and gives the average value for the parameters needed to define the log-normal distribution for 4a and 4b, respectively.
Important values that can be derived from the distribution coefficients  and D0 are the coefficients Dmean, Dmedian and Dmode according to equations (3)-(5): These values are summarized in Table S3 and Table S4 for the individual distributions of the measurements for 4a and 4b, respectively.   The individual parameters as well as the average parameters can be used to show the distribution of diffusion coefficients as done in Figure S11.  10 15 Figure S14. Comparison of the length distribution for 4b after spin-coating from either a water/acetonitrile mixture (, 6:4, c = 1 × 10 -5 , brown bars) or pure water (c = 1 × 10 -5 , blue bars). The curves correspond to the respective lognormal fit of the histogram.
Spin-coating of 4b from pure water and 40 % acetonitrile in water showed different length distributions ( Figure S14). As the structures obtained from pure water are slightly larger, the formation of self-assembled structures is proposed for pure water or mixtures with low acetonitrile content. The fact that the self-assemblies are small contributes to the high solubility of 4b under aqueous conditions.

Visible light-driven water oxidation catalysis Lamp spectrum
The recorded lamp spectrum of the 150 W xenon arc lamp equipped with a solar filter is shown in Figure S16. normalized intensity / a.u.
wavelength / nm Figure S16. Lamp spectrum of the 150 W xenon arc lamp used as external light source during visible light-driven water oxidation catalysis. The light is passed through a solar filter before sample illumination.

Mechanisms of molecular water oxidation catalysts
The detailed discussion of the two most prominent reaction mechanisms during water oxidation in molecular water oxidation catalysts can be found in detail in the literature. [S11] After generation of a high valent metal oxo species, the rate determining step can proceed via one of two pathways.
The oxygen evolution in the Ru(bda) family usually proceeds via the interaction of two metal oxo radicals (I2M) mechanism. Hereby, two of the high valent metal oxo species react together to form a peroxo intermediate before oxygen evolution. The other pathway discussed in literature is based on the nucleophilic attack of a water molecule on the metal oxo complex forming a hydroperoxo intermediate before oxygen evolution. There are two distinct features that allow differentiation between the two mechanisms by experiments. On one hand, the I2M mechanism exhibits second-order kinetics with respect to catalyst concentration while in the WNA mechanism shows first-order kinetics. Additionally, WNA mechanism includes a proton coupled electron transfer in the rate determining step which is why strong H/D kinetic isotope effects can be expected on the rate of oxygen evolution if this pathway is occurring. Figure S17. Schematic illustration of the water oxidation mechanisms in molecular water oxidation catalysts. The rate determining steps are indicated for the WNA (water nucleophilic attack) and I2M (Interaction of two metal-oxo radicals) mechanisms.

Solvent-dependent water oxidation catalysis
For solvent-dependent measurements, water oxidation catalysis with constant amounts of photosensitizer (Ru(deeb)2(bpy)Cl2, c = 0.2 m) and sacrificial electron acceptor (Na2S2O8, c = 37.5 m) at 293 K was performed in varying ratios of 50 m phosphate buffer (pH = 7.2) and acetonitrile. The irradiation power of a 150 W xenon lamp as external light source was calibrated to 100 mW cm -2 (one sun at AM1.5G).  Figure S19. Plots of the initial rates of oxygen evolution (determined by linear regression during the first few seconds of catalysis) for a), d), g), j) 4a (green dots), b), e), h), k) 4b (blue dots) and c), f), i), l) 4c (red dots) in aqueous phosphate buffer containing a)-c) 20 %, d)-f) 10 %, g)-i) 1 % and j)-l) 0 % acetonitrile as co-solvent. All data sets were fitted according to three different kinetic models and fit parameters are summarized in Table S6. The fits with the highest R^2 value are indicated using solid lines while fits with lower R^2 value are dashed. Fit parameters a and b were restricted to positive values.

Determination of the reaction order
To determine the reaction order during visible light-driven water oxidation catalysis, the plot of -log(dn(O2)/dt) against -log(cRu) was fitted with a linear equation giving an indication of the reaction order observed during catalysis. Figure S20. -log plot of the initial rate of oxygen evolution over the concentrations of ruthenium centers for 4a (green dots), 4b (blue dots) and 4c (red dots) for the experiments carried out in aqueous phosphate buffer containing a) 20 %, b) 10 %, c) 1 % and d) 0 % acetonitrile as organic co-solvent. The straight lines in the graphs represent the fits used in determination of the reaction order.  20 % acetonitrile f) Figure S21. Summary of initial rates of oxygen evolution against the amount of ruthenium centers in mixtures of phosphate buffered (pH(D) = 7.2, c = 50 × 10 -3 M) H2O (4a: green, 4b: blue, 4c: red) or D2O (orange) with a)-c) 0 % acetonitrile and d)-f) 20 % acetonitrile as organic co-solvent. The amount of photosensitizer (Ru(deeb)2(bpy)Cl2, c = 0.2 × 10 -3 ) and sacrificial electron acceptor (Na2S2O8, c = 37.5 × 10 -3 ) was kept constant throughout the experiments and irradiation was started after 40 s. The linear regression is used for the determination of individual reaction rates k(H2O) and k(D2O) for kinetic isotope effect experiments. Table S8. Summary of data obtained from KIE experiments during visible light-driven water oxidation catalysis of 4a-c with Ru(deeb)2(bpy)Cl2 (c = 0.2 × 10 -3 ) as photosensitizer and Na2S2O8 as sacrificial electron acceptor (Na2S2O8, c = 37.5 × 10 -3 ). [a] cPS = 1.00 × 10 -4 M, cSEA = 5.0 × 10 -2 M.