Light-driven water splitting to hydrogen and oxygen, which is one of the most desirable and promising strategies to resolve the current energy and environmental crisis, has been receiving much attention [1–4]. Water oxidation (WO) (2H₂O → 4H⁺ + 4e⁻ + O₂) is considered the bottleneck in water splitting during the energy conversion scheme, in both natural and artificial photosynthesis [5], which requires efficient and stable water oxidation catalysts (WOCs) [6, 7]. In the last few decades, worldwide research efforts have contributed to the design of robust and highly active homogeneous and heterogeneous WOCs [8–12].

In the field of homogeneous WOCs, polyoxometalates (POMs) [13–15], an excellent class of molecular metal-oxygen clusters with highly oxidized early transition metals (TM), have attracted widespread attention. This is because of their ability to undergo fast, reversible, and stepwise multielectron-transfer reactions without changing their structures due to their high oxidative, thermal, and hydrolytic stability [16–20]. In 2009, Hill and co-workers [21] reported the first use of a tetraruthenium polyoxometalate complex, [{Ru₄O₄(OH)₂(H₂O)₄}- (SiW₁₀O₃₆)₂]¹⁰⁻, to catalyze WO in a totally homogeneous, visible light-driven, and artificial photosynthesis system at neutral pH. In 2011, they found that a cobalt-based POM, [Co₄(H₂O)₂(PW₉O₃₄)₂]¹⁰⁻ (abbreviated as Co₄P₂W₁₈), was competent as an efficient molecular catalyst for the visible light-driven catalytic oxidation of water [22]. In 2014, Wang's group [23] reported on a series of high-nuclearity cobalt-based POM-WOCs such as [{Co₄(OH)₃(PO₄)}₄(SiW₉O₃₄)₄]³²⁻, [{Co₄(OH)₃(PO₄)}₄ (GeW₉O₃₄)₄]³²⁻, [{Co₄(OH)₃(PO₄)}₄- (PW₉O₃₄)₄]²⁸⁻, and [{Co₄(OH)₃(PO₄)}₄ (AsW₉O₃₄)₄]²⁸⁻. This study not only provides a valuable molecular model structurally analogous to the [Mn₃O₅Ca] core but also an unprecedented opportunity to construct polynuclear TM-based POM-WOCs for visible light-driven WO. Our group has worked extensively in this area and reported a highly stable and effective mixed-valence WOC [Co^ⅢCo^Ⅱ(H₂O)W₁₁O₃₉]⁷⁻ in 2013 and a mononuclear cobalt(Ⅱ)-substituted silicotungstate WOC [Co(H₂O)₂(γ-SiW₁₀O₃₅)₂]^10– in 2014 [24, 25]. Recently, we had reported an oxidatively and hydrolytically stable bioinspired molecular WOC based on a polyoxometalate [(SiW₉O₃₄)₂Co₈(OH)₆(H₂O)₂(CO₃)₃]^16–, which exhibited excellent WO activity with a very high turnover number (TON) of 1436 [26].

There are very few reports on TM-substituted tungstoantimonates that act as WOCs, because Sb³⁺ hydrolyzes in aqueous solutions, which to a great extent impedes combination with TM ions into POM clusters. A majority of the addressed TM-substituted tungstoantimonates were prepared through the assembly of the trivacant Keggin polyoxoanion [SbW₉O₃₃]^9– with different types of TM ions [27–34]. This is of great interest to us, as it provides an excellent opportunity to further research on tungstoantimonate incorporating TM clusters. Thus, we adopted a one-pot self-assembly strategy that allowed us to effectively prepare a novel TMSP by integrating the well-defined secondary building blocks [SbW₉O₃₃]^9– with Co²⁺ ions.

Herein, we first synthesized an octahedral, {SbO₃(H₂O)₃} bridging, dicobalt-substituted, sandwich-type tungstoantimonate [Na₄{Co₂Sb₂(H₂O)₁₀(B-β-SbW₉O₃₃)}₂]·39H₂O (1) and found that it could be used as a homogeneous catalyst for the efficient production of O₂ under visible-light irradiation. Under the optimal photocatalytic conditions, i.e., photoirradiation at λ = 460 nm with [Ru(bpy)₃]Cl₂ as a photosensitizer and Na₂S₂O₈ as an oxidant in borate buffer (pH = 8.5), the TON could reach a value as high as 193 with an O₂ yield of 30.8%. The initial quantum yield (Ф_QY) and initial turnover frequency (TOF_initial) in the first 20 s were 36.2% and 5.3 s^–1, respectively. Variables of the photocatalytic reaction, including catalyst concentration, buffer type and concentration, pH, dye concentrations, and oxidant concentration, were systemically studied.

[Na₉(B-α-SbW₉O₃₃)]·19.5H₂O was prepared as described in the literature and confirmed based on the IR spectra [35]. Other common laboratory chemicals were analytically pure, purchased from commercial sources, and used without further purification.

Na₉[B-α-SbW₉O₃₃]·19.5H₂O (2.004 g, 0.70 mmol), AgNO₃ (0.102 g, 0.60 mmol), and Co(NO₃)₂·6H₂O (0.152 g, 0.52 mmol) were dissolved in 30 mL of deionized water with stirring, and the pH of the solution was carefully adjusted to 4.0 at room temperature by the addition of dilute HNO₃ solution (6 mol/L). The solution was stirred continuously for 1 h, heated at 80 ℃ for 3 h, and then filtered upon cooling to room temperature. Slow evaporation of the filtrate for two weeks at room temperature led to the formation of purple cubic block crystals of 1. Yield: ca. 28% (based on Co(NO₃)₂·6H₂O). Anal. Calcd (%) for H₉₈Co₂Na₄O₁₁₅Sb₄W₁₈: H 1.66, Na 1.54, Co 1.98, Sb 8.19; found: H 1.90, Na 2.42, Co 1.81, Sb 8.25. IR (KBr pellets, cm⁻¹): 3432 (vs), 1623 (s), 925 (m), 894 (m), 770 (w), 706 (w), 611 (w), 518 (w), 445 (m).

The photocatalytic WO was performed as follows. In a typical experiment, 10 mL of 80 mmol/L sodium borate buffer (pH 8.0–9.4) containing [Ru(bpy)₃]Cl₂ (0–1.5 mmol/L), Na₂S₂O₈ (0–7.0 mmol/L), and 1 (4.0 μmol/L) were introduced in a glass reactor sealed with a rubber septum. The above solution was deaerated by purging it with Ar gas for 5 min in a flask (21 mL) sealed with a rubber septum. The reaction was initiated by irradiating the solution with an LED (light intensity 33.8 mW, beam diameter 2 cm) through a transmitting glass filter (λ = 460 nm) at room temperature. After each sampling time, 100 mL of Ar was injected into the flask, and the same volume of the gas sample in the headspace of the flask was withdrawn by an SGE gas-tight syringe and analyzed by gas chromatography (GC). O₂ in the sampled gas was separated by passing the gas through a 2 m × 3 mm packed molecular sieve (5 Å) column with Ar as the carrier gas and quantified by a thermal conductivity detector (TCD). The total amount of O₂ evolved was calculated from the concentration of O₂ in the headspace gas. Air contamination in the headspace was corrected by measuring the N₂ concentration present in the headspace. The turnover number is defined as TON = (O₂ yield at end of run) / (catalyst concentration) = (mol of O₂)_f / (mol of catalyst). The O₂ yield is defined as O₂ yield = 2 × (mol of O₂)_f / (mol of Na₂S₂O₈). The initial TOF for oxygen evolution is defined as TOF_initial = (mol of O₂) / (mol of catalyst × 20 s), based on the amount of evolved O₂ after 20 s of visible-light irradiation.

A single crystal of 1 with the appropriate dimensions was chosen, quickly coated with high-vacuum grease, and mounted on a glass fiber. The intensity data were recorded on a Bruker Apex Ⅱ diffractometer using graphite-monochromated Cu K_α radiation (λ = 0.71073 Å). IR spectra were obtained using a Bruker VERTEX 70v FT-IR spectrometer with 2–4 wt% sample in KBr pellets in the range of 4000–400 cm^–1. Inductively coupled plasma-atomic emission spectrometry (ICP-AES) was performed with a TJA ICP-atomic emission spectrometer (IRIS Advantage ER/S). Thermogravimetric analysis (TGA) was carried out using a Linseis STA PT1600 in the temperature range 25–850 ℃, at a heating rate of 10 K min⁻¹, in N₂ atmosphere. UV-Vis absorption data were obtained using a TU-1810 spectrophotometer equipped with a photomultiplier tube detector. C, H, and N elemental analyses were performed using a Vario EL CHNS/O analyzer. Powder X-ray diffraction (PXRD) measurements were carried out on, a Rigaku D/MAX 2400 diffractometer operated at 40 kV and 40 mA with Cu K_α radiation (λ = 1.54056 Å). Photon flux of the incident light was determined using a Ray virtual radiation actinometer (FU 100, silicon ray detector, light spectrum 400–700 nm, sensitivity 10–50 μV μmol⁻¹ m⁻² s⁻¹), which afforded a value of 1600 μmol m⁻² s⁻¹.

The crystallographic data and structural refinements are summarized in Table S1 (see the Supporting Information). The crystallographic data for 1 were deposited at the Cambridge Crystallographic Data Center with CCDC 1835378. Bond valence sum (BVS) (Tables S2 and S3) calculations revealed that the oxidation states of all the Co and W atoms were +2 and +6, respectively. The oxidation states of the Sb1 and Sb2 atoms were +3 and +5, respectively. The presence of Sb⁵⁺ ions in the final product suggested that some of the Sb³⁺ ions were oxidized by Ag⁺ ions in the presence of O₂ for the formation of 1.

The new dicobalt-substituted and sandwiched tungstoantimonate dimer [Na₄{Co₂Sb₂(H₂O)₁₀ [B-β-SbW₉O₃₃]₂}]·39H₂O (1) was synthesized by the reaction of Co²⁺ ions with trilacunary POM precursor [B-α-SbW₉O₃₃]⁹⁻ in a weakly acidic aqueous medium, and it crystallized in the monoclinic space group P2₁/c. The molecular structure of 1 consists of one dimeric {Co₂Sb₂(H₂O)₁₀[B-β-SbW₉O₃₃]₂}^4– (1a) cluster (Fig. 1a and 1b), four sodium ions, and thirty-nine lattice water molecules. Compound 1a is composed of two trivacant [B-β-SbW₉O₃₃]⁹⁻ units (Fig. 1c) linked together by a {Co₂Sb₂} belt (Fig. 1e), in which two inner Sb³⁺ ions (Sb1, Sb1A) and two outer Co²⁺ ions (Co1, Co1A) lie at the corners of a rhombus with two adjacent edges of 5.789 and 5.897 Å, and all ions adopt the six-coordinate octahedral geometry (Fig. 1f and 1g).

The trivacant [B-β-SbW₉O₃₃)]⁹⁻ subunit is derived from the trilacunary [B-α-SbW₉O₃₃]⁹⁻ (Fig. 1d) anion by rotating one of the edge-shared {W₃O₁₃} groups. Moreover, it is particularly interesting to note that the sandwich-type {Co₂Sb₂(H₂O)₁₀- [B-β-SbW₉O₃₃]₂}^4– polyanion in 1 is the first reported example of a compound bearing the {Co₂Sb₂} moiety. We believe that isomerization of the [B-α-SbW₉O₃₃]⁹⁻ precursor and degradation of the [SbW₉O₃₃]⁹⁻ building block must have occurred during the formation of 1. This observation provides not only the synthetic viewpoint for this compound but also some new ideas for the rational design and synthesis of TMSPs, such as the addition of oxidants (such as Ag⁺, S₂O₈^2–, and MnO₄^–) for the preparation of novel TMSPs. In addition, discrete {Co₂Sb₂(H₂O)₁₀[B-β-SbW₉O₃₃]₂}^4– polyoxoanions were regularly distributed in the −ABAB− mode along the b and c axes (Fig. S1). The phase purity of 1 was corroborated by the PXRD patterns (Fig. S2a). IR and TG analyses were performed to determine the structure and elemental composition, respectively (Figs. S2b and S3).

WO was performed to evaluate the photocatalytic ability of 1 in the system, utilizing [Ru(bpy)₃]²⁺ (1 mmol/L) as the photosensitizer and Na₂S₂O₈ (5 mmol/L) as the sacrificial electron acceptor in sodium borate buffer (80 mmol/L, pH 8.5, 10 mL) under visible-light irradiation (λ = 460 nm) for O₂ evolution.

The proposed catalytic mechanism of 1 for the light-driven WO is depicted in Fig. 2. Fig. 3 illustrates the amount of O₂ evolution over 1 under different pH conditions. Evidently, O₂ evolution was very quick, reaching a plateau in 120 s due to the consumption of the sacrificial electron acceptor, Na₂S₂O₈ [36, 37].

A higher pH is thermodynamically favorable for WO; thus, the reaction at pH 8.0 with 1 showed poor activity. When the buffer pH was raised to 8.5, the amount of O₂ evolved reached 7.7 μmol with a higher O₂ yield of 30.8%, TON of 193, TOF_initial of 5.3 s⁻¹, and Ф_QY of 36.2% (see details in SI). However, upon continuously increasing the pH to 9.0, the amount of O₂ evolved suddenly decreased from 7.7 to 5.9 μmol, and the corresponding O₂ yield, TON, and Ф_QY also decreased to 23.6%, 148, and 34.2%, respectively. The main reason for this phenomenon is the possible decomposition of the photosensitizer [Ru(bpy)₃]²⁺ at high pH during the photocatalytic process.

Moreover, other factors influencing the photocatalytic reaction, including the concentration of photosensitizer the and sacrificial electron acceptor, were systematically studied when the concentration of 1 was 4.0 μmol/L (Figs. S4 and S5). The O₂ yield first showed a rapid increase as the concentration of [Ru(bpy)₃]Cl₂ and Na₂S₂O₈ was increased up to 1.0 mmol/L and 5 mmol/L, respectively, and then presented an unapparent increasing trend upon a further increase in the concentrations of[Ru(bpy)₃]Cl₂ and Na₂S₂O₈. Furthermore, it could be seen from Fig. 4 that O₂ evolution was almost absent or negligibly small in the absence of light, Na₂S₂O₈, or catalyst 1. Thus, light, catalyst, photosensitizer, and sacrificial electron acceptor are essential factors for this WO system. As is known, the hydrolytic stability of POMs may be influenced by various factors, in which the POMs may decompose into catalytically active oxide nanoparticles. We compared the photocatalytic kinetic behavior between 1 and a benchmark homogeneous WOC, [Co₄(H₂O)₂(PW₉O₃₄)₂]^10– , under the same reaction conditions. The amount of oxygen evolution from 1 was higher than that from [Co₄(H₂O)₂(PW₉O₃₄)₂]^10– (Fig. S6).

UV spectra of the aqueous solution and IR spectra were recorded to probe the catalyst stability. As shown in Figs. S7–S9, the UV spectra of 1 in the sodium borate buffer were monitored by varying the pH from 7.5 to 9 or aging for 120 min (pH 8.5), and no apparent changes were observed. There was no absorption band between 500 and 600 nm in the spectrum of 1 because of the rather weak absorbance corresponding to the Co-centered d–d transitions, consistent with previous reports [38]. To further investigate the stability of 1 during photocatalytic WO, we carried out tetra-n-heptylammonium nitrate (THpANO₃)-toluene extraction, which is proved to be an effective method for extracting POMs from the reaction solutions [39].

As shown in Table S4, after extracting 1 from the solutions before and after the reaction, there was almost no cobalt ion present in the solution. We also measured the O₂ evolution dynamic curves at a low concentration of Co²⁺ (0.05 μmol/L) under the optimal reaction conditions (Fig. S10). The blue and red curves represent the kinetics of O₂ formation with 4 μmol/L 1 and 0.05 μmol/L Co²⁺, respectively, in the photocatalytic system. Second, capillary electrophoresis was applied to detect the stability of 1 in the WO system. In this method, chemical species in the interior of a small capillary filled with an electrolyte are separated according to their charge to size ratio. Therefore, this method was carried out to investigate the stability of 1 during photocatalytic water oxidation. Although the characteristic peaks of 1 before and after the photocatalytic reactions shifted slightly to the right, they basically remained unchanged (Fig. 5). The reason for this shift was the change in the electrolyte concentration before and after the WO reaction. No extra peaks due to Co were observed in the electropherogram after the WO experiment. The results of capillary electrophoresis demonstrated that the structure of 1 was unchanged after the WO reactions. Moreover, the IR spectrum of the recovered catalyst 1 after the first run was consistent with that of the unused 1, which is a positive proof for the structural integrity of 1 (Fig. S11).

In summary, we synthesized a unique dicobalt-substituted sandwich-type tungstoantimonate 1 within a rare octahedral {SbO₃(H₂O)₃} bridge and systematically characterized the compound by multiple techniques. We found that 1 acts as an efficient and stable photocatalytic WO catalyst. Under the optimal photocatalytic conditions (photoirradiation at λ = 460 nm, [Ru(bpy)₃]Cl₂ as the photosensitizer, Na₂S₂O₈ as the oxidant in a basic borate buffer), the TON and O₂ yield were as high as 193 and 30.8%, respectively. Ф_QY and TOF_initial in the first 20 s of the reaction were 36.2% and 5.3 s^–1, respectively. This study presents the feasibility of the design and use of Co-containing inorganic clusters as effective WOCs, and the discovery of this novel Co-based catalyst enriches the photocatalytic WO catalyzed by polyoxometalates.

Acknowledgments

This research was financially supported by the National Natural Science Foundation of China (Grants No. 21773096), the Fundamental Research Funds for the Central Universities (lzujbky-2018-k08), and the Natural Science Foundation of Gansu Province (17JR5RA186).