Artificial photosynthesis represents a sustainable approach toward conversion of solar energy into chemical fuels, for example, the production of hydrogen via water splitting [1, 2]. The overall water splitting reaction consists of two half reactions, namely, water oxidation and proton reduction [3]. Oxidation of water is generally considered to be a difficult transformation because of the requirement of multiple electron transfer processes coupled with the removal of multiple protons from water molecules, which altogether impose a great challenge to artificial photosynthesis [4, 5]. While numerous heterogeneous and homogeneous catalysts have been developed to facilitate the oxygen evolution from water, most of them suffer from slow reaction rate and low stability [6-10]. In comparison with the conventional heterogeneous metal oxide catalysts, molecular water oxidation catalysts (WOCs) based on transition metal complexes have shown advantages in structural tunability and ease of mechanistic studies [6, 11]. Since the first molecular WOC cis, cis-[(bpy)₂(H₂O)Ru(μ-O)Ru(H₂O)(bpy)₂]⁴⁺ (bpy = 2, 2'-bipyridine) (blue dimer) was reported in 1982 [12], numerous efforts have been made to enhance the activity of molecular WOCs [13], Recently, Sun and co-workers[14] developed a series of mononuclear ruthenium catalysts based on 2, 2'-bipyridine-6, 6'-dicarboxylic acid (bda) ligand, which exhibited superior water oxidation activity in comparison to previously reported WOCs. Kinetic and DFT calculations revealed a bimolecular radical coupling mechanism of the O–O bond formation during the catalytic cycle [15, 16]. Based on this understanding, we developed a class of dinuclear ruthenium catalysts, [(4-pic)(bda)Ru^Ⅱ(μ-L)Ru^Ⅱ(bda)(4-pic)] (pic = picoline, L = bridge ligand), by covalently connecting two mononuclear units with organic bridge ligands. Using this strategy, the O–O bond formation changed from an intermolecular manner to a more facile intramolecular manner, resulting in the greatly enhanced activity over that of the mononuclear species, especially at low catalyst concentrations [17].

However, the systems mentioned above are driven by either a chemical oxidant or electric power, and the visible-light induced water oxidation is less studied despite its fundamental importance to solar fuel utilization [18-24]. Specifically, the performance of the bda-based ruthenium dimer in photocatalytic water oxidation has not yet been fully explored, even though these complexes demonstrate highly efficient O₂ evolution with TONs > 10000 in chemical water oxidation with (NH₄)₂Ce(NO₃)₆ as the oxidant [17]. To fill this gap, in this work, we have discussed the photocatalytic water oxidation activity of a binuclear ruthenium complex [(4-pic)(bda)Ru^Ⅱ(μ-L)Ru^Ⅱ(bda)(4-pic)] (L = 1, 3-(dipyridin-4-yl) propane, 2) under visible light irradiation. The performance of 2 was compared with that of the monomeric complex [Ru^Ⅱ(bpa)(pic)₂] (1). In addition, a dependence of the reaction mechanism on the reaction solvent has been discussed.

¹H NMR spectra were collected at 298 K using a Bruker DRX-400 instrument. Electrospray ionization mass spectra and high-resolution mass spectra were recorded on a Q-Tof Micromass spectrometer (Manchester, England). UV-Vis absorption measurements were carried out on an Agilent 8453 spectrophotometer. Oxygen evolution were analyzed on a Techcomp GC 7890T instrument equipped with a 5 Å molecular sieve column and a thermal conductivity detector with argon as the carrier gas. The electrochemical measurements were recorded on a CHI 660D electrochemical potentiostat. The light power was measured with a CEL-NP 2000 laser power meter.

Ruthenium complexes 1 and 2 were prepared according to previously reported procedures and characterized by ¹H NMR and mass spectrometry [17]. The synthesis of all ruthenium complexes in this study were performed under N₂ atmosphere. The solvents were purified by standard methods. All other chemicals were used as received.

In a typical experiment, [Ru(bpy)₃]Cl₂ (1 mmol/L), Na₂S₂O₈ (10 mmol/L), and the ruthenium catalyst (20μmol /L for 1 and 10 μmol /L for 2) dissolved in the phosphate buffer solution (5 mL, pH = 6.8) with a desired amount of CH₃CN were added to a 50-mL Schlenk flask. Prior to the photo-reaction, the solution was degassed with Ar for 20 min. The reaction was initiated by irradiation with a 300-W Xe lamp (λ > 400 nm; UV filter) under Ar atmosphere and at room temperature. The amount of generated O₂ was quantified by gas chromatography (TCD, MS-5A) at time intervals of 5 min.

Cyclic voltammetry measurements were obtained using a three-electrode setup with a glassy carbon (GC) working electrode, Ag/AgCl reference electrode, and a platinum wire counter electrode. For accurate evaluation, ferrocene was added as an internal standard. Cyclic voltammograms were obtained in phosphate buffer solutions (pH = 6.8) containing a specified amount of CH₃CN. The potentials were reported versus Fc⁺/Fc.

In a typical experiment, different amounts of 1 and [Ru^Ⅲ(bpy)₃]³⁺ (0.125 mmol/L) were added into 2 mL of 5% or 60% CH₃CN aqueous solution, and the changes in the UV-Vis absorbance of [Ru^Ⅱ(bpy)₃]²⁺ at 450 nm were recorded. The observed reaction rates (k_obs) at various concentrations of 1 were obtained according to the absorbance curves at 450 nm. A linear fitting of k_obs with [1]² in 5% CH₃CN indicated second order kinetics and a linear fitting of k_obs with [1] in 60% CH₃CN suggested that first order kinetics was operative.

In a typical experiment, [Ru(bpy)₃]Cl₂ (1 mmol/L), Na₂S₂O₈ (10 mmol/L), and ruthenium catalyst 1 (20 μmol/L) dissolved in phosphate buffer solution (5 mL, pH = 6.8) containing 60% CH₃CN were added to a 50 mL Schlenk flask. The solution was degassed using Ar for 20 min and then irradiated by a monochromatic light source (λ = 450 nm). The light power (P) was modulated to reach an optical density of 14 for the reaction solution. Under these conditions, the light entering the reaction solution was considered to be fully absorbed by the photosensitizers according to the Beer-Lambert law. The values of P were measured with a CEL-NP 2000 laser power meter at a point right in front of the reactor. The amount of generated O₂ (n_O2) was quantified by gas chromatography (TCD, MS-5A) in the first 15 min of the reaction. The number of moles of photons absorbed (n_p) were determined by the following equation.

where P is the absorbed power of the light (W), and t is the irradiation time (900 s). The parameter λ refers to the irradiation wavelength (450 nm), N_A is the Avogadro constant (6.022 × 10²³ mol^–1), h is the Planck constant (6.63 × 10^–34 J·s), and c is the speed of the light (3 × 10⁸ m/s). The quantum yield, Φ was evaluated by the following equation.

The quantum efficiency (QE) was derived from the quantum yield of O₂ as follows.

where Φ is the experimental value and Φ_max is the theoretical maximum value. In the [Ru(bpy)₃]²⁺/S₂O₈^2– system, Φ_max is 0.5 because two photons are needed to produce one oxygen molecule. Therefore, QE could be described as QE = 2 × Φ × 100%.

The visible light-driven water oxidation was carried out in a homogeneous three-component system with [Ru(bpy)₃]Cl₂ as the photosensitizer, Na₂S₂O₈ as the sacrificial electron acceptor, and either complex 1 or 2 as the catalyst [25]. The phosphate buffer solutions (pH = 6.8) containing different amounts of CH₃CN were used as the solvents for photocatalysis. The addition of CH₃CN was necessary to improve the solubility of the catalyst and achieve a high activity. In such a three-component system, the ground state [Ru(bpy)₃]²⁺ was excited to its excited state, and [Ru(bpy)₃]³⁺ was subsequently generated by electron transfer from [Ru(bpy)₃]^2+* to Na₂S₂O₈. As a strong oxidant, [Ru(bpy)₃]³⁺, in turn, accepted an electron from the catalyst. Once sufficient oxidative equivalents were accumulated at the active site of the catalyst, water was oxidized to molecular oxygen.

In Fig. 2(a), the TONs of oxygen evolution under visible light irradiation are plotted with respect to the volume percentage (%) of CH₃CN in phosphate buffer. This plot is associated with two important features. First, the oxygen evolution for both catalysts increased with a higher CH₃CN content in the phosphate buffer. For instance, a TON of 19 was obtained for 1 in the presence of 20% CH₃CN, which increased 10 times to a TON of 204 in the presence of 60% CH₃CN. Likewise, the TON increased from 57 to 215 when 2 was employed as the catalyst. Another notable feature in Fig. 2(a) is that catalyst 2 exhibits higher activities than 1 in the solvents containing less CH₃CN, although the activities of both catalysts became comparable in solvents containing more than 40% CH₃CN. Under the optimal conditions (60% CH₃CN), a quantum efficiency of 77% was attained with irradiation at 450 nm for water oxidation catalyzed by 2, which is comparable to the best value reported in the literature [26]. It was found that acetonitrile not only promoted the amount of evolved oxygen, but also accelerated the reaction rate. As shown in Fig. 2(b), the photocatalytic rate of complex 2 increased from 0.07 to 0.33/s upon increasing the content of CH₃CN in the phosphate buffer from 20% to 60%.

With complex 2 as the catalyst, the effect of varying Na₂S₂O₈ concentrations on photocatalytic water oxidation was studied in phosphate buffer containing 60% CH₃CN (pH = 6.8) while keeping the concentrations of 2 (10 μmol/L and [Ru(bpy)₃]Cl₂ (1 mmol/L) constant. The highest activity based on TON was obtained in the presence of 45 mmol/L Na₂S₂O₈. When the concentrations of [Ru(bpy)₃]Cl₂ and Na₂S₂O₈ were fixed, lowering the concentration of the catalyst also lowered the amount of evolved oxygen, but the corresponding TONs were increased. For example, a TON of 638 was obtained under the reaction condition of 45 mmol/L Na₂S₂O₈, 1 mmol/L [Ru(bpy)₃]Cl₂, and 2 μmol/L catalyst.

In order to clarify the role of CH₃CN in promoting photocatalytic activity, the cyclic voltammetry (CV) of the photosensitizer [Ru(bpy)₃]Cl₂ was carried out in phosphate buffers with varying amounts of CH₃CN. As shown in Fig. 3(a), the onset potential for water oxidation was negatively shifted with the increase in CH₃CN content in the mixed solvent. For example, the onset potential of oxygen evolution decreased by 82.5 mV in an aqueous solution containing 50% CH₃CN in comparison with that in phosphate buffer containing 20% CH₃CN. In contrast, the redox potential of the photosensitizer [Ru(bpy)₃]^3+/2+ (E_1/2) was not influenced by the content of CH₃CN in the reaction solution (Fig. 3(b)). Since [Ru(bpy)₃]³⁺ serves as a strong oxidant in photocatalysis, increasing the proportion of CH₃CN in the mixed solvent could increase the driving force for water oxidation, leading to higher activity for oxygen evolution [27]. This analysis agrees well with our observation that the TON increases with the increase in the relative content of CH₃CN in the phosphate buffer (Fig. 2(a)).

In another experiment, [Ru(bpy)₃]³⁺ was separately synthesized as a chemical oxidant to facilitate the kinetic study of water oxidation. The UV-Vis absorption changes at 450 nm were monitored upon addition of different concentrations of [Ru(bda)(pic)₂] into 2 mL of [Ru(bpy)₃]³⁺ (0.125 mmol/L) phosphate buffer solutions containing either 5% or 60% CH₃CN. Based on the absorption-time plots shown in Fig. 4(a) and (c), k_obs was estimated as a function of catalyst concentration (Fig. 4(b) and (d)).

The value of k_obs was found to be proportional to the square of the concentration of 1 in 5% CH₃CN aqueous solution (Fig. 4(b)), which was indicative of second order kinetics involving two catalyst molecules in the rate-determining step. This feature is consistent with a radical coupling mechanism as previously reported [16, 25]. In comparison, k_obs showed a linear dependence on the concentration of 1 in 60% CH₃CN solution, indicating first order kinetics with a single catalyst molecule in the rate-determining step [28]. Therefore, water oxidation in 5% and 60% CH₃CN solution may be governed by different reaction mechanisms.

On the basis of the manner of O–O bond formation, it was concluded that two different kinds of mechanisms were prevalent [29]. The first mechanism is the aforementioned bimolecular radical coupling, wherein oxygen is released through the interaction between two M=O (M = Ru, Ir, Fe, Co, Mn, and so on) moieties. The second mechanism involves the nucleophilic attack of the water molecule on the single site of the M=O intermediate to release molecular oxygen via the M–OOH peroxide (Fig. 5). It is highly plausible that the reaction mechanism alters from intermolecular radical coupling to water nucleophilic attack upon increasing the relative content of CH₃CN. This assumption explains the trend observed in Fig. 2(a) that the catalytic activities of 1 and 2 only differ in the mixed solvent with a low content of CH₃CN, which become comparable in the solvent with a high content of CH₃CN. However, the possibility that the rate-determining step in the radical coupling pathway switches from O–O bond formation to a step that involves association with one Ru center cannot be excluded. The active intermediate of water oxidation in a mixed solvent of CD₃CN/D₂O (7/3 V/V) was characterized by mass spectrometry and ¹H NMR spectroscopy as [Ru(bda)(pic)₂(OH)]⁺, a seven-coordinate Ru(Ⅳ) (S = 0) species, consistent with that isolated in our previous study [14]. The subsequent one electron oxidation of [Ru^Ⅳ(bda)(pic)₂(OH)]⁺ gave either [Ru^Ⅳ(bda)(pic)₂(O)·]⁺ (Ru^Ⅳ-O·) or [Ru^Ⅴ(bda)(pic)₂(O)]⁺ (Ru^Ⅴ=O) as the key intermediate for bimolecular coupling and single-molecular water nucleophilic attack, respectively. Although we are not able to differentiate between two reaction pathways at the present stage, other methods such as DFT calculations may help us to reveal the detailed mechanism.

In summary, the photocatalytic activity of bda-based binuclear ruthenium catalyst 2 was investigated in a three-component system and compared with that of its monomeric precursor 1. It was found that the oxygen evolution activity of both catalysts increased with the amount of CH₃CN in the phosphate buffer as a result of a larger driving force provided by the photo-activated sensitizer [Ru(bpy)₃]³⁺. A notable finding here was that the kinetics of the water oxidation reaction by 1 switched from second order to pseudo-first order with the increasing CH₃CN content. Consequently, the binuclear catalyst 2 showed higher activity when the CH₃CN content was low, and comparable activities were attained by 1 and 2 with a high content of CH₃CN. This is probably because the pathway of O–O bond formation changes from the bimolecular radical coupling mechanism to the single-site nucleophilic attack of water. This study demonstrates the flexibility of the molecular catalysts in light-driven water oxidation. The solvent-tunable behavior of Ru-bda may also be applicable to electrochemical and photoelectrochemical water oxidation, enabling a deeper understanding and rational design of engineered molecular devices for highly efficient solar water splitting.

There are no conflicts to declare.

The acknowledgements come at the end of an article after the conclusions and before the notes and references. This research was supported by the National Basic Research Program of China (973 program, 2014CB239402), the National Natural Science Foundation of China (21476043), the Swedish Energy Agency and K & A Wallenberg Foundation.