From Ru-bda to Ru-bds: a step forward to highly efficient molecular water oxidation electrocatalysts under acidic and neutral conditions

Significant advances during the past decades in the design and studies of Ru complexes with polypyridine ligands have led to the great development of molecular water oxidation catalysts and understanding on the O−O bond formation mechanisms. Here we report a Ru-based molecular water oxidation catalyst [Ru(bds)(pic)2] (Ru-bds; bds2− = 2,2′-bipyridine-6,6′-disulfonate) containing a tetradentate, dianionic sulfonate ligand at the equatorial position and two 4-picoline ligands at the axial positions. This Ru-bds catalyst electrochemically catalyzes water oxidation with turnover frequencies (TOF) of 160 and 12,900 s−1 under acidic and neutral conditions respectively, showing much better performance than the state-of-art Ru-bda catalyst. Density functional theory calculations reveal that (i) under acidic conditions, the high valent Ru intermediate RuV=O featuring the 7-coordination configuration is involved in the O−O bond formation step; (ii) under neutral conditions, the seven-coordinate RuIV=O triggers the O−O bond formation; (iii) in both cases, the I2M (interaction of two M−O units) pathway is dominant over the WNA (water nucleophilic attack) pathway. Developing efficient molecular water oxidation catalysts for artificial photosynthesis is a challenging task. Here the authors introduce a ruthenium based complex with negatively charged sulfonate groups to effectively drive water oxidation under both acidic and neutral conditions.

A rtificial photosynthesis is one of the most promising ways to build our society with sustainable energy systems while catalysis of water splitting is the key 1 . Two half-reactions are involved in the water-splitting process: water oxidation (2H 2 O → O 2 + 4H + + 4e − ) and proton reduction (4H + + 4e − → 2H 2 ). The first half reaction is widely regarded as the bottleneck of total water splitting. To date, only Ir-based catalysts display high catalytic activity and long durability towards electrochemical water oxidation under strongly acidic conditions 2,3 . The majority of noble-metal free water oxidation catalysts (WOCs) are confined under strong alkaline conditions whereas they show only moderate performances under neutral conditions and quickly deactivate in acidic media [4][5][6] . In contrast, molecular WOCs displayed their advantages over the majority of metal oxides with regard to stability and catalytic activity under acidic conditions. In the last decade, a great number of WOCs based on transition metal complexes such as Fe 7-9 , Ir 10 , Co 11,12 , Mn 13,14 , Cu 15,16 and Ru [17][18][19][20] have been reported, among which Ru-based WOCs with relatively simple polypyridyl ligands are the most representative due to their adequate stability and high catalytic activity under strongly acidic conditions [21][22][23] .

Results and discussion
Synthesis and characterization. The ligand H 2 bds was prepared by the nucleophilic substitution of the bromo groups on 6,6′dibromo-2,2′-bipyridine by NaHS at elevated temperature, followed by the oxidation of the hydrosulfide groups by HNO 3 to obtain the desired sulfonate containing ligand H 2 bds. Subsequent coordination of the H 2 bds ligand to [Ru(p-cymene)Cl 2 ] 2 in the presence of NEt 3 , followed by reaction with 4-picoline yields the corresponding mononuclear catalyst Ru-bds. The 1 H/ 13 C NMR and 1 H-1 H COSY spectra of H 2 bds and Ru-bds were shown in Supplementary Figs. 1-8.
The Ru-bds crystallizes with two chemically identical but crystallographically distinct molecules in the asymmetric unit. One of the X-ray crystal structures was depicted in Fig. 2, and the Ru centre is in a typical distorted octahedral geometry. It is noticeable that the O3−Ru1−O4 bite angle of Ru-bds (114.7(3)°) is slightly larger than that of previously reported Ru III -bpaH 2 (112.09 (6)°) 30 whereas smaller than that of Ru II -bda (123.0(2)°) 17 , all of which are significantly larger than the ideal 90°of an octahedron configuration. The wide O−Ru−O angle provides the open site of the catalyst for substrate water binding and favors the formation of seven-coordinate Ru intermediates during the catalytic process of water oxidation. Noteworthy, as depicted in Supplementary Fig. 9, the C19−C18−C17 angle of 127.1 (8)°in Ru-bds is a little bit smaller than C8−C7−C6 (129.6(7)°) in Ru-bda, showing that the bipyridine ring of Ru-bds is less distorted in comparison with that of Ru-bda. Thereby, the bds 2− ligand backbone fits the Ru II cation slightly better than the bda 2− ligand.
The 1 H NMR spectrum of Ru-bds in CD 2 Cl 2 /CD 3 OD ( Supplementary Fig. 10a) is in agreement with a C 2v symmetry of its chemical structure. Two doublets at 8.53 and 8.09 ppm and one triplet at 7.98 ppm were assigned to the proton resonances of the equatorial bds 2− ligand while two-resonance peaks at 7.82 and 7.03 ppm to two axial 4-picoline ligands. When the mixed CD 3 CN/D 2 O was used as the solvent, four broad proton resonances originating from the equatorial ligand appeared while two doublets of the 4-picoline protons remained but slightly downshifted to 8.00 and 7.19 ppm, pointing to a loss of C 2v symmetry ( Supplementary Fig. 10b). This is attributed to the formation of acetonitrile-coordinating complex, [Ru(k 3 O,N,N -bds) (pic) 2 (CH 3 CN)] with a dangling sulfonate arm. Similar dissociation of the equatorial ligand, such as carboxylate group, was found occurring as well for Ru-bda 34 and Ru-pda (H 2 pda is 1, 10-phenanthroline-2,9-dicarboxylic acid) 35 21 . Thus, the sulfonate chelation in Rubds should also promote the water coordination to Ru III center and facilitate the further water oxidation steps.
The catalytic performance of this Ru-bds under Ce IV -driven water oxidation conditions is less effective but more durable than Ru-bda ( Supplementary Fig. 11). Kinetic measurements revealed that the oxygen evolution is a first order reaction in regard to both catalyst ( Supplementary Fig. 12) and Ce IV ( Supplementary Fig. 13) concentrations, meaning that the rate-determining step is an oxidation step, likely Ru IV −OH to Ru V =O. The low activity is due to the small driving force of Ce IV (E(Ce IV/III ) = 1.6 V vs. NHE 36 ; all potentials herein are reported versus normal hydrogen electrode, NHE) in comparison with the catalytic onset potential of Ru-bds (E onset = 1.6 V; vide infra). We thereby studied the electrochemical driven water oxidation by Ru-bds in the rest of this work.
Electrochemistry under acidic conditions. First, the electrochemistry of Ru-bda and Ru-bds was studied in pH 1.0 triflic acid aqueous solutions with 20% CH 3 CN (to improve the solubility of Ru-bds; it should be noted that acetonitrile competes with water accessing the Ru center and thereby suppresses water oxidation activity 23 ). As shown in Fig. 3a and Supplementary  Fig. 14, Ru-bda displays three redox waves at 0.89, 1.12 and 1.33 V, which are related to three consecutive metal-based one-electron oxidation processes: Ru II −NCCH 3 → Ru III −OH 2 → Ru IV −OH → Ru V =O, respectively. A small increase in the current density of roughly 0.6 mA/cm 2 in the 1.5−1.8 V zone indicates slow catalytic water oxidation. Nevertheless, a substantial enhancement of catalytic current density was observed from the CV curve of Ru-bds under the acidic conditions (Fig. 3a) reaching a value of 1.48 mA/cm 2 at 1.74 V. The consecutive oxidation processes of Ru-bds were assigned according to the potential versus pH diagram (Supplementary Fig. 15; vide infra). The first redox wave is clearly visible at 1.18 V, attributed to the single electron process of Ru II −NCCH 3 → Ru III −OH 2 . The second, irreversible oxidation wave at 1.41 V is corresponding to Ru III −OH 2 → Ru IV −OH. Oxidation of Ru IV −OH to Ru V =O (1.58 V) triggers electrochemical water oxidation. All redox states are about 0.3 V higher relative to Ru-bda, which is, as expected, due to the relatively less electron-donating ability of the sulfonate group than the carboxylate group.
The Ru III/II redox couples of Ru-bds at various scan rates are depicted in Fig. 3b and its peak current (i p ) varies linearly with the square root of scan rate (ʋ 1/2 ), indicating a diffusion controlled electrochemical process (Fig. 3c). According to the Randles-Sevcik equation (Eq. 1; F is the Faraday constant, A is the electrode area, [Ru II ] is the bulk concentration of catalyst, n p = 1 is the number of electrons transferred, n cat = 4 is the number of electrons required to complete a single catalytic cycle, D is the diffusion constant of catalyst, R is the ideal gas constant, and T is the temperature), the diffusion co-efficient was 5.02 × 10 −6 cm 2 s −1 . CV scans of Ru-bds and Ru-bda at various scan rates are plotted in Fig. 4a,b. The catalytic currents becomes relatively scan rate independent at 1.5 V s −1 for Ru-bds and 2.1 V s −1 for Ru-bda, indicating steady-state conditions are accomplished at such these high scan rates. According to the ratio of i cat vs. the noncatalytic Faradaic current (i p ) using an established method with the Eq. 3 37 , the TOF value of Ru-bds is calculated to be 160 s −1 at a scan rate of 1.5 V s −1 toward electrochemical water oxidation whereas Ru-bda displays a TOF of 7 s −1 at a scan rate of 2.1 V s −1 . Given the above results, Ru-bds is surpassing the wellknown mononuclear reference Ru-bda catalyst towards electrochemical water oxidation under acidic conditions.
Fig. 2 X-ray crystal structure. The X-ray crystal structure of Ru-bds with thermal ellipsoids at 50% probability (hydrogen atoms are omitted for clarity).
. . . To provide more insights into these redox processes and the electronic effect of the equatorial bds 2− ligand on the redox properties, the potential versus pH diagram (Pourbaix diagram) of Ru-bds was constructed and shown in Supplementary Fig. 15. The potential of the first oxidation step is constant to be ca. 1.18 V up to pH 3.2. The pH independence indicates that only electron transfer is involved in the first oxidation, corresponding to the oxidation of Ru II −NCCH 3 to 6-coordinate Ru III −OH 2 , [Ru III (k 3 O,N,N -bds)(pic) 2 (OH 2 )] + . Therefore, the pK a value of Ru III −OH 2 was determined to be 3.2, which is much smaller than the pK a value of 5.5 in Ru-bda 21 ; this is further in accordance with the relatively poor electron-donating ability of the sulfonate group compared with the carboxylate group. As the pH increases from 3.2 all the way to 8.1, the Ru III/II redox couple decreases linearly as the solution pH increases with of a slope −53 mV per pH, corresponding to typical proton-coupled electron transfer (PCET) process and assigned to the oxidation of Ru II −NCCH 3 to Ru III −OH. For the Ru IV/III redox couple, as evidenced by a slope of −64 mV per pH over the range from pH 1.18 to 3.2, the PCET process can be ascribed to the oxidation of Ru III −OH 2 to Ru IV −OH. Unfortunately, this redox event became too weak to be distinguished at pH > 3.0 and meanwhile it is overlapped with the catalytic current; thereby, the corresponding potential values were obtained by measuring the potential values at 3.0 × 10 −5 mA for each scan. The potential of the Ru IV/III is constant to be 1.26 V from pH 3.2 to 5.3, which is related to the process of Ru III −OH to Ru IV −OH with only electron transfer. As the pH increase from 5.3 to 8.1, there is a PCET process belonging to the oxidation process of Ru III −OH to Ru IV =O. For the Ru IV −OH species produced under acidic conditions, its oxidation via the PCET process leads to the formation of Ru V =O, a highly oxidizing state. Additionally, the CV of Ru-bds in D 2 O containing 0.1 M triflic acid was also recorded (Supplementary Fig. 18) and the wave of Ru III −OH 2 to Ru IV −OH is too weak to be observed, indicating the slow kinetics of Ru IV −OH formation and thus its low concentration on the electrode surface.
The chronoamperometric curve of Ru-bds renders a moderate decay with 42% retention of the initial current density after 2 h bulk electrolysis in pH 1.0 ( Supplementary Fig. 19). The corresponding Faradaic efficiency was recorded to be 95%. The post-electrolyzed CVs ( Supplementary Fig. 20) demonstrated a two-third peak current loss for Ru III/II associated with the catalytic current density declined (from 0.77 to 0.41 mA cm −2 at 1.7 V). After bulk electrolysis, Ru-bds species is still the dominated species, as shown in the HRMS spectra (Supplementary Figs. [21][22][23][24], along with the appearance of several new signals with weak intensity, which are presumably structureevolved Ru-intermediates and probably contributed to the new redox waves in CVs. Density functional theory (DFT) calculations. DFT calculations were performed to understand the reason on the high electrochemical driven water oxidation performance of Ru-bds and the reaction mechanism behind. As described in the previous study 38 , four extra water molecules were added around the metal complex to produce a reasonable hydrogen-bonding network. The calculated potential for the couple [Ru II ]/[Ru III −OH 2 ] + at pH 1.0 is 1.15 V, which is close to the experimental value of 1.18 V. Both structures of 6-coordinate [Ru III (k 3 O,N,N -bds)(pic) 2 (OH 2 )] + and 7-coordinate [Ru III (bds)(pic) 2 (OH 2 )] + were optimized (Fig. 5a,  b). The Gibbs free energy of the 6-coordination mode is 6.7 kcal mol −1 lower than that of the 7-coordination. The formation of Ru aqua species plays a vital role for water oxidation because it can facilitate further oxidation of the Ru catalyst to higher oxidation states via the PCET process. Upon oxidation of the Ru III aqua species via a PECT process, [Ru IV −OH] + is formed. The 7coordinate [Ru IV (bds)(pic) 2 (OH)] + (Fig. 5d) becomes more stable than the corresponding 6-coordinate mode [Ru IV (k 3 O,N,Nbds)(pic) 2 (OH)] + (Fig. 5c) by a Gibbs free energy difference of    Fig. 25) and their corresponding transition states are shown in Fig. 6. The computed activation free energy of the reaction via the WNA pathway is 15.5 kcal mol −1 and the reaction free energy is −3.6 kcal mol −1 . For the I2M pathway, the full reaction includes diffusion of two [Ru V =O] + species, encountering of two [Ru V =O] + species to form the prereactive dimer and the radical coupling reaction in the full solvent [39][40][41] . The computed activation free energy for the radical coupling step is 6.9 kcal mol −1 from the prereactive dimer to the transition state, and the reaction free energy is −11.5 kcal mol −1 . The encoutering step of two species was found to have a key role for the different catalytic reactivities of Ru-bda complexes. We therefore parameterized [Ru V =O] + model according to our published report 41 and tested stability of the model (Supplementary Fig. 26). Based on the analysis of the hydrogen bonding ( Supplementary Fig. 27), the oxo of [Ru V =O] + of Ru-bds is found to be hydrophobic, similar to the oxo of Ru-bda. We performed potential of mean force (PMF) simulations to estimate the binding free energy of the prereactive dimer in water phase. The calculated binding free energy of the two [Ru V =O] + species in water phase is −4 kcal mol −1 (Supplementary Fig. 28), leading to a total activation energy of 2.6 kcal mol −1 for Ru-bds (the forward activation free energy of the O−O bond formation minus the free energy of the dissociation of two species; Supplementary Table 2) for the full dimerization. Apparently, the I2M pathway is feasible for Ru-bds. In comparison, the activation energy of Rubds is slightly lower than that of Ru-bda (3.9 kcal mol −1 ) 40 .
Proposed mechanism. On the basis of the above electrochemistry, mass spectrometry, kinetics studies together with DFT calculations, the reaction mechanism of water oxidation by Rubds under acidic conditions is proposed and depicted in Fig. 7 42 .
Base-enhanced catalytic water oxidation. As shown previously by Meyer and co-workers for the base-enhanced PECT oxidation, the role of the added buffer base was explored by CVs of Ru-bds (1.0 mM) at pH 7.0 while maintaining the ionic strength at 0.5 M (compensated by NaNO 3 , Supplementary Fig. 31). Significant acceleration of catalytic current is achieved toward water oxidation as the total concentration of H 2 PO 4 − /HPO 4 2− buffers increases gradually from 0.01 M to 0.2 M, during which a wave with nearly plateau shape was observed at the lowest buffer concentration where a maximum current reached at 1.28 V. Therefore, the involvement of the buffer base in PCET pathway directly contributes to the enhanced catalytic activity, which is associated with the electron transferring to electrode occurring along with proton transfer to the added base. CVs were also recorded with different catalyst concentrations at pH 7.0. As the concentration varies from 0.2 mM to 1.0 mM, a sharp enhancement of catalytic current for water oxidation was obtained ( Supplementary Fig. 32).
Electrochemistry under neutral conditions. Accordingly, electrochemical behaviors of 1.0 mM Ru-bds and Ru-bda were further explored in 0.20 M pH 7.0 phosphate buffer solutions ( Fig. 8 and Supplementary Fig. 33). To our delight, Ru-bds exhibited an ultra-high electrocatalytic current density of 11.79 mA cm −2 (at 1.63 V with scan rate of 0.1 mV s −1 , Fig. 8) with a low onset overpotential of 380 mV. In comparison, the current density of Ru-bda is only 3.26 mA cm −2 (at 1.63 V with scan rate of 0.1 mV s −1 , Fig. 8). This significant current density enhancement by Rubds highlights the importance of PCET in avoiding the high energy barrier for the electrochemical water oxidation. To the best of our knowledge, such a value of the catalytic current density is among the highest ever reported in the literature under neutral conditions for Ru-WOCs catalysed water oxidation (Supplementary Table 3) 29,30,43,44 . Under the steady-state conditions, a TOF of 12900 s −1 was achieved by Ru-bds at the scan rate of 1.3 V s −1 (Fig. 9a) while Ru-bda displayed a TOF of 300 s −1 at the scan rate of 2.0 V s −1 . The latter is about two orders of magnitude smaller than that of Ru-bds (Fig. 9b). Bulk electrolysis of Ru-bds in pH 7.0 for 2 h displays a decay with 30% retention of the initial current density (Supplementary Fig. 34a) while the Faradaic efficiency was determined to be 94%. The CVs ( Supplementary Fig. 35) show that the Ru III/II peak current retains after bulk electrolysis while the catalytic current density of 8. In our previous study, the mixed-valence Ru trimer species for Ru-bda-catalyzed water oxidation under neutral conditions were proved by using chemical and electrochemical methods and have been successfully isolated and characterized 45 . Herein, the spectroelectrochemistry of Ru-bds was examined at a constant potential of 1.21 V, where the absorption changes of the electrolysis solution were in situ monitored by UV-Vis spectroscopy. As shown in Supplementary Fig. 38, no observable band at 690 nm was observed where the band at 690 nm is a typical absorption peak of the trinuclear Ru species. Therefore, there is no µ-oxo oligomeric species formed in the case of Ru-bds. More importantly, the electrode surfaces before and after electrolysis have also been characterized by scanning electron microscopy (SEM) and energydispersive X-ray spectroscopy (EDX) (Supplementary Fig. 39). No Ru elements and species was deposited on the electrode surface, ruling out the potential catalyst decomposition into the RuO 2 .
As shown in the Pourbaix diagram, a pH independent oxidation process at 1.0 V appears at pH > 5.5 (grey stars in Supplementary Fig. 15), and this oxidation wave is relatively small compared with the Ru III/II wave ( Supplementary Fig. 33). The Ru III/II and Ru IV/III data points respectively aligned well in the whole pH range. Thereby, we propose that the pH independent oxidation process at 1.0 V belongs to a decomposed/evolved Ru species whose structure is currently unknown. Given that, the oxidation of Ru III −OH to Ru IV =O triggers water oxidation with E onset = 1.2 V at pH 7.0. Density functional theory calculations at pH 7.0. DFT calculations were also performed to gain more insight in the reaction mechanism by Ru-bds at pH 7.0. From the Pourbaix diagram, the Ru III −OH will form the Ru IV =O species at pH > 4. 6 Fig. 40) species. Since the Gibbs free energy of Ru IV =O at triplet state is 16.6 kcal mol −1 lower than that of singlet geometry, we used triplet state of the Ru IV =O species for the calculations. The calculated redox potential of [Ru III −OH] + (7-coordinate) /[Ru I-V =O] T is 1.08 V at pH 7.0. When the Ru IV =O species form the O −O bond via the WNA pathway using H 2 PO 4 − (a stronger base than water) as a proton acceptor, (i) at the triplet state the activation free energy is 51.3 kcal mol −1 and the reaction free energy is 33.9 kcal mol −1 (Fig. 10a), and (ii) at the singlet state the activation free energy is 19.8 kcal mol −1 and the reaction free energy is 15.0 kcal mol −1 . Therefore the total activation free energy of WNA pathway is 36.4 kcal mol −1 and the reaction free energy is 31.6 kcal mol −1 (the energy profile at singlet plus the energy difference of 16.6 kcal mol −1 between the singlet and triplet of Ru IV =O species; Fig. 10b).
We have also calculated transition state of radical coupling of two Ru IV =O species by scanning the terminal O−O bond distance [Ru IV =O • • • O=Ru IV ] of the antiferromagnetic openshell singlet, giving an activation free energy of 6.1 kcal mol −1 from the pre-reactive dimer to the transition state, and a reaction free energy of 1.1 kcal mol −1 (Fig. 10c). We have also parameterized Ru IV =O T and tested stability by molecular dynamics (MD; Supplementary Fig. 26). The two neutral Ru IV =O T species are more prone to form the prereactive complex with a lower binding free energy of −5 kcal mol −1 (Supplementary Fig. 28  reported a Ru IV =O catalyzed WNA pathway 46 . Taking the electrochemistry results and theoretical calculations into account, we propose that the rate-limiting step in electrochemical water oxidation by Ru-bds at pH 7.0 is the oxidation of Ru III −OH to Ru IV =O, and the O−O bond forms via the radical coupling of two Ru IV =O species. In summary, by introducing two sulfonate groups into a bipyridine ligand, a mononuclear Ru catalyst Ru-bds was developed for effective electrocatalytic water oxidation with TOF of 160 s −1 under acidic conditions while outstanding catalytic activity with a high TOF of 12900 s −1 was successfully obtained under neutral conditions. The bds 2− ligand, with the proper electron-donating ability and suitable ligation sites, plays a vital role on stabilizing high valent Ru states, adapting the 7coordination configuration of reaction intermediates and promoting the O−O bond formation via the I2M mechanism. DFT calculations revealed that the radical coupling of two Ru IV =O species at neutral conditions proceeds with a low activation barrier of 1.1 kcal mol −1 , providing a new insight in water oxidation mechanism catalyzed by Ru-based water oxidation catalysts under neutral conditions. In particular, this work illuminates the impact of structural changes on the electrochemical catalytic activity and further provides an inspiring strategy to design more efficient and robust WOCs for potential applications in electrocatalysis and photoelectrocatalysis.

Methods
Physical measurements and instrumentation. Nuclear magnetic resonance (NMR) spectra were obtained with 400 MHz of Bruker Advance spectrometer. Elemental analyses were performed with an Elementar Vario EL elemental analyzer. Mass spectrometry measurements were carried out by a Thermo Scientific LCQ Fleet mass spectrometer. High-resolution mass spectrometry measurements were performed on a Thermo Scientific Q Exactive mass spectrometer. Electrochemistry measurements were performed with a CHI760 electrochemical workstation. For cyclic voltammetry and differential pulse voltammetry measurements, a glassy carbon disk (diameter 3 mm) was used as the working electrode and a platinum column as the counter electrode, and measured versus Ag   configuration was monitored by GC9790Plus. The Ce IV -driven oxygen was detected by a pressure transducer (MIK-P300) driven at 10.00 V using a power supply (HY3005B) plus a data acquisition module (Omega OM-DAQ-USB-2401). The single crystal X-ray diffraction data were collected at 298 K on a Bruker SMART CCD 1 K diffractometer with a graphite-monochromated Mo-K α radiation (λ = 0.71073 Å). The structure was solved and refined using the Olex2 program 47 . Small fraction of twinning domains was suspected in the present datasets. To understand the twining problem, the twin rotation law has been carefully studied with the Platon software (A.L.Spek, Acta Cryst. 2020, E76, 1-11). The twin law of (−1 0 0 0 0 −1 −0.36 −0.391 1) has been added into the instruction file for the final refinement. The final refinement has been successfully lower down the refinement R1 value from 8.05 % to 6.71 % with improved residue electron density. A twining fraction of 4.5 % as founded based on the refinement results. The spectroelectrochemistry was carried out in a Y-shaped quartz cuvette equipped with Pt mesh as working electrode, a Pt wire as counter electrode and an Ag/AgCl reference electrode. The quartz cuvette was filled with 1 mL 0.5 mM Ru-bds in phosphate buffer solution (pH 7, 0.2 M, 20% CH 3 CN). The path length of the Y-shaped quartz cuvette is 1.0 mM. UV−vis absorption spectra were obtained by Agilent Technologies Cary 8454. The morphologies of electrode surfaces before and after electrolysis were characterized by a JSM-7500F field emission scanning electron microscopy (SEM, JEOL, Japan).
Synthesis of 2,2′-bipyridine-6,6′-disulfonic acid. To a solution of 6,6′-dibromo-2,2′-bipyridine (312 mg, 1.0 mmol) in dry DMF (8 mL), ten equivalents of sodium hydrosulfide (560 mg, 10 mmol) were added. The reaction mixture was heated under 180°C for 2 h using Microwave reactor. The resultant orange mixture was dried under vacuum. Dissolve the residue in distilled water, and CH 3 COOH was added drop wised until the precipitation stopped. The precipitate was filtrated and washed by distilled water, finally dried under vacuum. The product of 6,6′-dithiol-2,2′-bipyridine was afforded as an orange powder (167 mg, 76% yield; its 1D and 2D NMR spectra were recorded to characterize this compound, as shown in Supplementary  . To a solution of 2,2′-bipyridine-6,6′-disulfonic acid (100 mg, 0.32 mmol) in dry MeOH (5 mL), dichloro(p-cymene)ruthenium(II) dimer (98 mg, 0.16 mmol), ten equivalents of 4-picoline and ten equivalents of Et 3 N were added. The resulting mixture was then heated under 125°C for 40 min using a microwave reactor. The solvent was removed by vacuum, and the crude product purified by column chromatography. The target complex was afforded as a dark red powder (65 mg, 35% yield). The 1D and 2D NMR spectra  were recorded to characterize the complex. 1  Computational details. All DFT calculations for the estimation of Gibbs free energies were carried out with the Jaguar 8.3 program package by Schrödinger LLC 48 . Molecular geometries were optimized using Becke's three-parameter hybrid functional and the LYP correlation functional (B3LYP) 49 with D3 correction of Grimme et al 50,51 . with the LACVP** basis set 52 . Single-point energy corrections were performed with the B3LYP-D3 functional using the LACVP** + + basis set augmented with two f-functions on the metal. Frequency calculations were performed on the optimized geometries to verify that the geometries correspond to minima on the potential energy surface (PES). On the basis of the gas-phase optimized geometries, the solvation energies were estimated by single-point calculations using the Poisson−Boltzmann reactive field implemented in Jaguar (PBF) in water. To identify the transition states for O−O bond formation of I2M mechanism, we searched the potential energy surface by scanning the terminal of the antiferromagnetic open shell singlet. The Gibbs free energy were defined as the following equation G = E(B3LYP-D3/LACVP** + + 2 f on Ru) + G solv + ZPE + H 298 -TS 298 + 1.9 kcal/mol (the value 1.9 kcal/mol is a concentration correction to the free energy of solvation, which by default is calculated at 1 M (g) to 1 M (aq) in Jaguar). Four additional water molecules were included in the calculation.
Force field parameterization. With the optimized geometries of [Ru V =O] + and Ru IV =O T , equilibrium bonds, angles, and dihedrals were obtained. Then a series of relaxed coordinate scans were performed by DFT calculations from the equilibrium bonds, angles, and dihedrals to get the force field parameters. Each bond was stretched 0.3 Å in each direction for a total scan span of 0.6 Å, and each angle was stretched 15 o in each direction for a total scan span of 30 o . The scans were made in 15 increments each way. The scans were fitted to a second-degree polynomial as the following equation where the square term coefficient, k b , is the force constant.
The angle parameters have also been plotted with the force constants vs equilibrium angle. Because of the fairly planar structure, most of the dihedral angles have been ignored i.e K ψ = 0, but the dihedral angles that rotates the axial ligand Ru-N bonds have been scanned with the DFT method, and crudely fitted with one cosine series in order to capture the barrier height and equilibrium angles.
Electrostatic potential (ESP) charges were calculated using the DFT method for [Ru V =O] + (total charge = +1) and Ru IV =O T (total charge = 0). As an improvement on performing only a single ESP charge calculation, a chargeaveraging scheme was implemented, which was performed as follows: 1) Initial ESP charges were calculated at the starting geometry. 2) An MD simulation (gas phase, 300 K, step size 1 fs) was performed starting with the initial structure and charges. A harmonic restraint of 5.0 kcal mol −1 was added to Ru atom to restrain it from moving too far from its initial position while the other atoms could move freely during 100 ps of simulation time. A snapshot was extracted every 10 ps of this simulation, at which point new ESP charges were calculated. 3) The procedure was repeated three times, but with the exception that the initial charges of the second and third runs were taken as the average of the ESP charges from the previous run. Each time this procedure was repeated, the charges became more stable, and the procedure was stopped on the third round, when the charges had converged (small standard deviations between the partial charges of the different conformations). All van der Waals parameters used in this work are from the work of Bernardes et. al. 53 , where the vdW parameters were determined for M(CO) n structures for a number for transition metals.
Since [Ru V =O] + and Ru IV =O T have the close to identical geometries with different oxidation states, we use same vdW parameters, bond stretch parameters, angle bend parameters, torsion parameters and improper torsion parameters for these two complexes with different calculated average electrostatic potential charges. In order to run MD simulations, the parameters were then transformed into the Gromacs topology format, where the Fourier coefficients of the dihedral potential term were transformed into the Ryckaerd-Bellemans type 54 .
Molecular dynamics simulations. For two MD simulations, the catalyst was sequentially equilibrated in the following order: 1) A 10 ps simulation at 1 K, applying a 5.0 kcal mol −1 harmonic restraint on each atom to restraint them to their DFT optimized geometries, and using a 0.1 fs stepsize. 2) A 100 ps simulation at 100 K, applying a 5.0 kcal mol −1 harmonic restraint on each atom and increasing to a 1 fs step size. 3) A final 1 ns equilibration at 300 K, reducing the overall harmonic restraint on each atom to 1 kcal mol −1 and remaining the stepsize of 1 fs. Solvent was included in the simulations using a 20 Å radius sphere of explicit TIP3P water molecules centered on the centroid of the reacting atoms. The sphere was described as a multi-level system in which the inner 85% of the sphere was allowed to move freely, but all atoms in the last 25% of the sphere were subjected to a 10 kcal mol −1 Å −2 to maintain system stability, as is routine when using the Surface Constrained All Atom Solvent (SCAAS) model 55,56 .
Potential of mean force. PMF were performed for the Ru-Ru distance of two complexes in water phase using umbrella sampling method with the Gromacs 5.1.4 MD software package 57 . Equilibrations were performed for 100 ps under an NVT ensemble, using the same methodology described in previous report 39 . The two Ru complexes were pulled apart for total 10 Å from the core structure over 200 ps, using a spring constant of 30 kJ mol −1 Å −2 and a pull rate of 0.05 Å ps −1 . From these trajectories, snapshots were taken to generate the starting configurations for the umbrella sampling windows. Then we sampled the distances using roughly 0.5 Å spacing. Such spacing allowed for increasing detail at smaller Ru-Ru distance, and resulted in 21 windows. In each window, 20 ns of MD was performed for a total simulation time of 420 ns utilized for umbrella sampling. Analysis of results was performed with the weighted histogram analysis method (WHAM) 58 . All PMF simulations resulted in smooth dissociation curves and the mean value from three repeated simulations.

Data availability
All data needed to evaluate the conclusions of this paper are present in the paper and/or Supplementary Information. The source data underlying Supplementary Figs. 26−29 are provided as a Source Data file. The data that support the findings of this study are available from the corresponding author upon request. The X-ray crystallographic data of Ru-bds reported in this study has been deposited at the Cambridge Crystallographic