Phosphate‐Templated Encapsulation of a {Co^II ₄O₄} Cubane in Germanotungstates as Carbon‐Free Homogeneous Water Oxidation Photocatalysts

Abstract

The ever‐growing interest in sustainable energy sources leads to a search for an efficient, stable, and inexpensive homogeneous water oxidation catalyst (WOC). Herein, the PO₄ ³⁻ templated synthesis of three abundant‐metal‐based germanotungstate (GT) clusters Na₁₅[Ge₄PCo₄(H₂O)₂W₂₄O₉₄] ⋅ 38H₂O (Co₄), Na_2.5K_17.5[Ge₃PCo₉(OH)₅(H₂O)₄W₃₀O₁₁₅] ⋅ 45H₂O (Co₉), Na₆K₁₆[Ge₄P₄Co₂₀(OH)₁₄(H₂O)₁₈W₃₆O₁₅₀] ⋅ 61H₂O (Co₂₀) with non‐, quasi‐, or full cubane motifs structurally strongly reminiscent of the naturally occurring {Mn₄Ca} oxygen evolving complex (OEC) in photosystem II was achieved. Under the conditions tested, all three GT‐scaffolds were active molecular WOCs, with Co₉ and Co₂₀ outperforming the well‐known Na₁₀[Co₄(H₂O)₂(PW₉O₃₄)₂] {Co₄P₂W₁₈} by a factor of 2 as shown by a direct comparison of their turnover numbers (TONs). With TONs up to 159.9 and a turnover frequency of 0.608 s⁻¹ Co₉ currently represents the fastest Co‐GT‐based WOC, and photoluminescence emission spectroscopy provided insights into its photocatalytic WOC mechanism. Cyclic voltammetry, dynamic light scattering, UV/Vis and IR spectroscopy showed recyclability and integrity of the catalysts under the applied conditions. The experimental results were supported by computational studies, which highlighted that the facilitated oxidation of Co₉ was due to the higher energy of its highest occupied molecular orbital electrons as compared to Co₄.

Water oxidation: The stabilization of non‐, quasi‐ and full‐cubane topologies in three Co^II germanotungstates (Co‐GT) [Ge₄PCo₄(H₂O)₂W₂₄O₉₄]¹⁵⁻ (Co₄), [Ge₃PCo₉(OH)₅(H₂O)₄W₃₀O₁₁₅]²⁰⁻ (Co₉), [Ge₄P₄Co₂₀(OH)₁₄(H₂O)₁₈W₃₆O₁₅₀]²²⁻ (Co₂₀) via PO₄ ³⁻ and GT lacunary ligands is reported. The Co‐GTs are active molecular water oxidation catalysts with Co₉ and Co₂₀ outperforming the non‐cubane [Co₄(H₂O)₂(PW₉O₃₄)₂]¹⁰⁻ by a factor of 2, showing the structural and functional similarity of the GT‐encapsulated cubanes to the {Mn₄Ca} oxygen evolving complex.

Introduction

The development of an artificial, efficient, stable, and inexpensive homogeneous water oxidation catalyst (WOC), which can mimic the natural photosynthesis process to meet mankind's growing energy demands, is of utmost interest. ^[1] Over the past decade, 3d‐ and 4d‐doped polyoxotungstates (POTs) have been reported as promising all‐inorganic water oxidation catalysts (POT‐WOCs), which, contrary to their organic counterparts, can withstand the harsh oxidizing conditions of the water oxidation half‐reaction. ^[2] Starting in 2008, Bonchio and co‐workers ^[3] and Hill and co‐workers ^[4] independently reported the ruthenium‐containing POT [{Ru₄O₄(OH)₂(H₂O)₄}(γ‐SiW₁₀O₃₆)₂]¹⁰⁻ (Ru₄) as the first efficient POT for visible‐light‐driven homogeneous water oxidation catalysis [turnover frequency (TOF)=0.25 s⁻¹]. In Ru₄, the WOC active metal core consists of a tetrahedron of four ruthenium centers sandwiched between two dilacunary [γ‐SiW₁₀O₃₆]^8−[5] POT units, thereby resembling the core {Mn₄Ca} in the oxygen‐evolving complex (OEC) in photosystem II (Figure S6). ^[6] In 2010, the activity of the cobalt‐containing Weakley dimer [Co₄(H₂O)₂(PW₉O₃₄)₂]¹⁰⁻ as the first non‐precious metal‐based homogeneous POT‐WOC was reported by Hill and co‐workers. ^[7] Under visible‐light irradiation and in the presence of [Ru(bpy)₃]²⁺ as a photosensitizer and S₂O₈ ²⁻ as an oxidant, WOC activity with turnover numbers (TONs) up to 220 and TOFs up to 5 s⁻¹ was observed. Since then, attention was given to Co‐containing polyoxometalates (POMs) in terms of their stability and activity as WOCs under various reaction conditions with special interest in cobalt‐cubane (Table S2) scaffolds as bio‐inspired cost‐effective WOCs with enhanced photocatalytic activity.[ ⁸ , ⁹ ] Significant progress was achieved by Wei et al., who reported the octa‐cobalt‐substituted silicotungstate [Co₈(OH)₆(H₂O)₂(CO₃)₃(A‐α‐SiW₉O₃₄)₂]¹⁶⁻ as the currently fastest Co‐substituted bio‐inspired POT‐WOC (TOF=10 s⁻¹, Table S3). ^[10] At the same time, Wang and co‐workers investigated the photocatalytic WOC performance of a series of isostructural cubane‐incorporating hexadecanuclear Co‐POTs [{Co₄(OH)₃PO₄}₄(XW₉O₃₄)₄]ⁿ⁻ (X=P^V, As^V; n=28; X=Si^IV, Ge^IV; n=32) showing that the type of primary heteroatom in the lacunary POM unit efficiently modulates their overall redox properties and WOC activity with Ge^IV (TOF=0.105 s⁻¹) and As^V (TOF=0.053 s⁻¹) containing representatives exhibiting the highest WOC activity. ^[11] Taken together the structural and compositional features of the most promising POT‐WOCs recently reported in literature and considering the low number of two cobalt cubane germanotungstates with proven WOC activity reported so far (Table S2), we chose the structurally simple yet versatile PO₄ ³⁻ group for the synthesis of all‐inorganic cobalt germanotungstate (GT) clusters as OEC models. Herein, we report the PO₄ ³⁻‐templated stabilization of a series of all‐inorganic Co‐GT, Na₁₅[Ge₄PCo₄(H₂O)₂W₂₄O₉₄] ⋅ 38H₂O (Co₄), Na_2.5K_17.5[Ge₃PCo₉(OH)₅(H₂O)₄W₃₀O₁₁₅] ⋅ 45H₂O (Co₉) and Na₆K₁₆[Ge₄P₄Co₂₀(OH)₁₄(H₂O)₁₈W₃₆O₁₅₀] ⋅ 61H₂O (Co₂₀) with non‐, quasi‐, or full {Co^II ₄O₄} cubane motifs and their activity in visible‐light‐driven water oxidation. With 20 incorporated Co^II metal centers Co₂₀ comprises the highest reported number of incorporated metal centers in a cubane‐encapsulating POT (Table S2). Under visible‐light irradiation and in the presence of [Ru(bpy)₃]²⁺ as a photosensitizer and S₂O₈ ²⁻ as an oxidant, WOC activity with TONs up to 159.9 and TOFs up to 0.608 s⁻¹ for Co₉ in borate buffer at pH 8.0 was detected, which to the best of our knowledge represents the currently fastest Co‐GT‐based homogeneous POT‐WOC (Table S3).

Results and Discussion

Synthesis and structure

Co₄, Co₉ and Co₂₀ are synthesized in a non‐buffered aqueous solution by adjusting the pH of a reaction mixture containing the corresponding tungsten source (GeO₂ and Na₂WO₄ for Co₄, K₈[γ‐GeW₁₀O₃₆] ⋅ 6H₂O ^[12] for Co₉ and K₈Na₂[A‐α‐GeW₉O₃₄] ⋅ 25H₂O ^[13] for Co₂₀) and CoCl₂ to pH 7.6 via addition of Na₃PO₄. Heat activation of the PO₄ ³⁻ group enabling its coordination to the Co‐oxo cores and subsequent filtration of the cooled reaction mixture results in single crystals of Co₄, Co₉ and Co₂₀ at 20 °C (CCDC 1876468‐1876470, Figure 1A–F). Single‐crystal X‐ray diffraction (SXRD) studies revealed that Co₄ and Co₂₀ crystallize in the monoclinic space groups P2₁/c and C2/c, whereas Co₉ crystallizes in the triclinic space group P $\bar{1}$ (Tables S5–S10). In all three compounds at least one PO₄ ³⁻ group stabilizes the POT scaffold. For Co₄ a PO₄ ³⁻ group connects two trigonal edge‐shared {W₃O₁₃} fragments to the four octahedrally coordinated Co^II centers, which are encapsulated by two B‐α‐{GeW₉} units and two germanium octahedrons located externally in the structure (Figure 1D, inset). Note, Co₄ is the first reported structure with germanium octahedrons in a pure inorganic GT. In Co₉, the PO₄ ³⁻ group connects two α‐{Co₂GeW₁₀} Keggin moieties with one exchangeable Co^II‐coordinated aqua ligand per α‐{Co₂GeW₁₀} unit and one virtual “γ‐{Co₃GeW₉}” building block (Figure S7) to a single {WO₆} octahedron. Two octahedrally coordinated Co^IIO₅(H₂O) centers encapsulated by the PO₄ ³⁻ and the single {WO₆} unit complete the trimeric polyanion by forming a {Co^II ₃O₄(H₂O)₂} quasi‐cubane (Figure 1E, inset) featuring two exchangeable aqua ligands, thereby resulting in a total number of four Co^II positions exhibiting exchangeable aqua ligands, suitable for water molecules to supposedly coordinate and subsequently get oxidized. The architecture of Co₂₀ presents a tetrameric aggregate of four α‐{Co₃GeW₉} units linking to a central {Co^II ₄O₄} cubane, which is geometrically closely related to the {Mn₃CaO₄} cubane of the OEC in PSII ^[6b] (Figure S6) and stabilized by four PO₄ ³⁻ groups (Figure 1F, inset). Four covalently bound Co^II octahedra located in the POT‐encapsulated Co‐oxo core and externally as antenna‐like metal centers ^[14] complete the Co₂₀ framework. The compounds’ elemental composition and homogeneity was determined by elemental analysis, IR spectroscopy (Figure S2), thermogravimetric analysis (TGA; Figures S3–S5, Table S4), and powder XRD (PXRD; Figures S8–S10). To probe the solution stability of Co₄, Co₉ and Co₂₀, time‐dependent UV/Vis spectra were recorded at various pH conditions. It should be mentioned that UV/Vis experiments on Co₄, Co₉ and Co₂₀ in the absorption range for octahedrally coordinated Co^II centers could not be performed due to the low solubility of the Co‐GTs and the strong domination of the p_π(O_t)→d_π*(W) ligand‐to‐metal charge‐transfer (LMCT) transitions, which is a problem commonly encountered in POM chemistry. ^[15] Therefore, the precatalytic POT stability was assessed by investigating the LMCT transitions in the tungsten absorption range. The UV/Vis spectra of Co₄, Co₉ and Co₂₀ display an absorption maximum at approximately 205 nm with a shoulder at approximately 250 nm corresponding to the Keggin‐type framework in aqueous solution (Figures S17–S19). ^[16]

Figure 1.

Schematic representation of the building block dependent synthesis of (A) Co₄, (B) Co₉, and (C) Co₂₀, which are templated by PO₄ ³⁻ after heat activation. Ball‐and‐stick representation of (D) Co₄, (E) Co₉ enclosing the quasi‐ {Co^II ₃O₄} cubane, and (F) Co₂₀ incorporating the PO₄ ³⁻‐stabilized {Co^II ₄O₄} cubane. Color code: {WO₆}, dark blue; Co^II, pink; Ge^IV, lime; P^V, light orange; O, red, {W₃O₁₃} triads of Co₄, grey octahedra.

Acidification of an unbuffered solution of Co₂₀ (pH=0.7) led to stepwise degradation of the POT framework shown by the disappearance of the shoulder at approximately 250 nm (Figure S20), whereas all UV/Vis spectra remain unchanged in 80 mm sodium borate buffer (pH=7.5–9) solutions for at least 2 h mimicking the photocatalytic conditions (Figures S17–S19), which suggests pre‐catalytic stability of Co₄, Co₉ and Co₂₀ until O₂ saturation is reached in the WOC experiments (Figure 2A).

Figure 2.

(A) As‐recorded O₂ evolution profiles for Co₄, Co₉ and Co₂₀ compounds along with the reference {Co₄P₂W₁₈}, measured from 20 μm POM solutions buffered in 80 mm borate buffer at pH 8 and containing Na₂S₂O₈ (5 mm) and [Ru(bpy)₃]²⁺ (1 mm) as an oxidant and a photosensitizer, respectively. (B) Recorded oxygen evolution amounts as a function of POM catalyst concentrations measured for 2, 5, 10 and 20 μm values. A summary of as‐recorded profiles is shown in Figure S23. (C) Amount of generated O₂ plotted as TONs as a function of pH (7.5–9). The bars indicate an average O₂ amount generated by all three catalysts at a certain pH to demonstrate a cumulative effect of pH on activity. For all conditions, a monochromatic visible LED (λ _max=445±13 nm) was chosen as light source to trigger photocatalytic reaction.

Cyclic voltametric measurements were performed on Co₄, Co₉ and Co₂₀ (vs. Ag/AgCl, in 80 mm sodium borate buffer pH=8) thereby revealing Co²⁺/Co³⁺ oxidation waves in the range E≈0.73–0.89 V ^[17] (Figure S21) occurring at lower potentials than those for other Co‐POM‐WOC systems at pH≥8 (>0.9 V),[ ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ ] whereas the Co(NO₃)₂ reference solution gives the most intense peaks for Co²⁺/Co³⁺ oxidation (at 0.84 V) as well as for water oxidation (at 1.37 V) (Figure S22) showing that the cobalt centers incorporated in the germanotungstates do not behave like free cobalt ions (Table S12) and additionally indicating their pre‐catalytic stability.

Visible light‐driven water oxidation

To probe the photocatalytic WOC activity of Co₄, Co₉ and Co₂₀, 80 mm aqueous borate buffer solutions (pH=7.5–9) of POTs (concentrations ranging from 2 to 20 μm) containing 1 mm Ru[bpy]₃Cl₂ (bpy=2,2’‐bipyridine) as the photosensitizer (PS) and 5 mm Na₂S₂O₈ as a sacrificial agent (SA) were used. Note that various parameters such as the shape of the reaction vessel, light intensity, stirring rate as well as the ratio of gaseous head space to total volume render a direct comparison of the WOC performance of POTs tested in different catalytic systems difficult. ^[19c] Hence, the literature‐known Weakley‐type [Co₄(H₂O)₂(PW₉O₃₄)₂]¹⁰⁻ {Co₄P₂W₁₈} was used as a benchmark POT‐WOC.[ ⁷ , ¹¹ , ¹⁹ ] Figure 2A shows as‐recorded O₂ evolution profiles of 20 μm Co₄, Co₉, Co₂₀ and {Co₄P₂W₁₈} solutions at pH 8 and indicates WOC activity for all compounds, with Co₉ and Co₂₀ reaching TONs of 21.6 and 20.4, thus outperforming {Co₄P₂W₁₈} (TON=10.9 under elsewise identical conditions) by a factor of 2 (Table 1). Initial TOF values of 0.047 (Co₄), 0.069 (Co₉), and 0.068 s⁻¹ (Co₂₀), respectively, are determined from the derivative plots, which are comparable to those of other reported Co‐POM‐WOCs like [{Co₄(OH)₃(PO₄)}₄(SiW₉O₃₄)₄]³²⁻ (TOF=0.053 s⁻¹) and [{Co₄(OH)₃(PO₄)}₄(GeW₉O₃₄)₄]³²⁻ (TOF=0.105 s⁻¹), both exhibiting structurally similar attributes like an incorporated {Co^II ₄O₄} cubane motif (Table S3). ^[11] Figure 2B demonstrates the amount of generated O₂ as a function of POM concentration and reveals a direct activity correlation with increasing concentration for Co₉ and Co₂₀ (red and blue trends), in contrast to Co₄ (orange trend), which indicates that factors other than the catalyst concentration, such as the amount of sacrificial agent or photosensitizer, control its WOC performance (Figures S23, S24). At pH=8 Co₉ is the most active WOC species with a rather small advantage over Co₂₀, which gets more pronounced at lower catalyst concentrations (Table 1) with Co₉ reaching TONs up to 159 and a TOF=0.608 s⁻¹, currently representing the highest TOF for a Co‐GT‐based homogeneous POM‐WOC. The higher WOC activity of Co₉ compared to both Co₄ and Co₂₀ may have its origin in the enhanced accessibility of the Co^II centers, which leads to supposedly facilitated coordination and oxidation of water molecules on the catalytic centers (Figures 1E, S7, Table S11). ^[10] pH‐dependent WOC studies (pH=7.5–9) revealed higher O₂ evolution amounts and TOF values with increasing pH values from 7.5 to 8.5 (Figure 2C), which is related to thermodynamic aspects of water oxidation catalysis according to the Nernst equation, ^[20] and partial POM deprotonation that results in its enhanced interaction with the photosensitizer. ^[21] A strong activity drop at pH=9 (indicated with grey bars in Figure 2C) is related to [Ru(bpy)₃]²⁺ degradation. ^[2a] Overall, the highest WOC activity was determined at pH=8.5, indicating a trade‐off between activity and PS‐stability at these conditions. At pH=8.5 (in contrast to the data obtained at pH=8) Co₂₀ (Table 1) greatly outperforms Co₉. According to bond valence sum (BVS) studies, Co₂₀ is the most protonated species in the solid state of the three (H₂₅Co₂₀, H₁₃Co₉, H₄Co₄, Table S11), which in turn means that the effect of its deprotonation at higher pH values is more pronounced leading to a stronger WOC increase. This activity trend implies that the ultimate performance of Co₄, Co₉ and Co₂₀ is affected by the number of Co^II centers, their accessibility, and the extent of POM‐PS pairing, and can be governed by each of these factors depending on the pH.

Table 1.

Summary table showing average TON and initial TOF values for Co₄, Co₉, Co₂₀ and the Weakley‐type POM {Co₄P₂W₁₈} as well as the O₂ yield generated by the respective POM with varying concentration (2–20 μm) in borate buffer [80 mm], pH=8.0.

Co‐POM (μm)	O2 [μmol] (TON)	TOF [s−1]	O2 yield [%]
Co4 (2)	0.480 (120.50)	0.422	9.60
Co9 (2)	0.640 (159.90)	0.608	12.80
Co20 (2)	0.450 (111.40)	0.405	4.50
Co4 (5)	0.470 (46.90)	0.148	9.40
Co9 (5)	0.710 (71.30)	0.263	14.20
Co20 (5)	0.575 (57.50)	0.189	28.40
Co4 (10)	0.520 (25.90)	0.091	10.40
Co9 (10)	0.790 (39.30)	0.125	15.80
Co20 (10)	0.690 (34.40)	0.105	13.80
Co4 (20)	0.450 (11.30)	0.047	9.00
Co9 (20)	0.860 (21.60)	0.069	17.20
Co20 (20)	0.810 (20.40)	0.068	16.20
{Co4P2W18} (20)	0.430 (10.90)	0.015	8.60

Post‐catalytic stability studies

The stability of molecular WOCs is a topic of current interest due to the possibility of the WOCs to decompose into catalytically active oxide nanoparticles. ^[22] Note that ¹⁸³W or ³¹P NMR measurements on Co₄, Co₉ and Co₂₀ could not be performed, due to the low solubility and strong paramagnetic nature of the incorporated Co^II metal centers. ^[23] Hence, various reloading experiments and post‐catalytic characterizations were conducted to demonstrate that O₂ evolution is indeed triggered by the investigated Co‐GT and to verify their integrity under turnover conditions. ^[24]

First, blank WOC experiments in pure water and those performed in the absence of PS and SA showed no activity (Figure S25A), indicating the validity of the experimental setup. Only a small activity was recorded in the absence of any Co‐GT (≈20 % of the activity of Co₉ at pH=8) (Figure S25A), which suggests a negligible contribution of direct water oxidation by the PS* as a side reaction and is in line with previous reports. ^[25] Moreover, a reference WOC experiment using the unsubstituted [PW₁₂O₄₀]³⁻ Keggin ^[26] POT instead of Co₄, Co₉ or Co₂₀ yielded a similarly low activity (Figure S25B) thereby excluding in‐situ formed tungsten‐based species to be responsible for the observed WOC activity. Second, after O₂ level reached saturation (point x in Figure 2A), the reaction solution was re‐loaded with PS and SA. Figure S26 containing the summary WOC data demonstrates that this second illumination cycle triggers additional O₂ evolution, which suggests that the observed WOC saturation is not a result of POM deactivation or degradation but can rather be related to the depletion of the other WOC‐solution components such as the [Ru(bpy)₃]²⁺ or the S₂O₈ ²⁻, and additionally indicates recyclability of the POM‐WOCs. Third, following an established procedure, ^[24c] a toluene solution of tetra‐n‐heptylammonium nitrate (THpANO₃) was used to quantitatively extract Co₄, Co₉ and Co₂₀ from the respective post‐catalytic solutions (see Figure S27 along with discussion). As this selective procedure does not extract CoO_x or Co²⁺ _aq, we analyzed the remaining aqueous phases with X‐ray fluorescence spectroscopy (XRF) to elucidate the potential leaching of Co into the reaction mixture under photocatalytic conditions. The XRF spectra of the solutions look similar and did not show any Co traces (Figure S27, Table S13), which indicates that the Co‐GT underwent neither decomposition (e. g., into Co²⁺) nor degradation (e. g., into CoO_x), processes that have been previously identified to be responsible for WOC performance of other POMs and under different experimental conditions. ^[22] Fourth, dynamic light scattering (DLS) was performed on 20 μm solutions of Co₄, Co₉ or Co₂₀, [Ru(bpy)₃]²⁺ (1 mm) and S₂O₈ ^2‐ (5 mm) in 80 mm borate buffer (pH=8) after 30 min irradiation. The DLS measurements showed no nanoparticles after photocatalytic water oxidation (Figure S28). In addition, the same experiments were conducted using 20 μm Co(NO₃)₂ ⋅ 6H₂O instead of the corresponding Co‐GT and here nanoparticles with a diameter of approximately 26.4 nm were detected (Figure S28). This is in line with the XRF experiments (Figure S27, Table S13) and confirms that, in contrast to the Co(NO₃)₂ ⋅ 6H₂O system, no metal hydroxide/oxide nanoparticles (especially cobalt hydroxide/oxide nanoparticles) are generated via hydrolytic decomposition of Co₄, Co₉ and Co₂₀, nor through detachment of the four covalently bound Co^II antenna ligands present in Co₂₀ after the photocatalytic experiments. Finally, attenuated total reflectance (ATR)‐IR spectra of Co₄, Co₉ and Co₂₀ were recorded after the photocatalytic experiments and subsequent precipitation with cesium chloride. These clearly show the characteristic W−O−W bridging and terminal W=O vibrations in the tungsten fingerprint area from 300–1000 cm⁻¹ (Figures S29–S31), which indicates the solution of the polyanions is stable under turnover conditions and represents an established method frequently used for the post‐catalytic study of POMs.[ ²³ , ²⁷ ]

Mechanistic studies

Photoluminescence (PL) emission spectroscopy was employed to investigate the photocatalytic WOC mechanism and to understand the electron transfer kinetics between the reaction solution components. Figure S32 reveals that the PL emission of [Ru(bpy)₃]²⁺ is quenched by both Na₂S₂O₈ and Co₉ in a linear Stern‐Volmer behavior depending on their concentrations (Figure S33). The calculated rate constant for the oxidative quenching by Na₂S₂O₈ is 45 times lower than that of the reductive quenching by Co₉. Considering (a) the much higher Na₂S₂O₈ concentration (5 mm) as compared to Co₉ (2–20 μm) present in the WOC reaction solution, and (b) the use of the borate buffer that weakens the formation of an ion pair between polyanionic Co₉ and cationic [Ru(bpy)₃]²⁺, ^[17] oxidative quenching dominates under the photocatalytic conditions. This conclusion is strongly confirmed by time‐resolved PL decay profiles. Figure S34 indicates that in the presence of 10 mm Na₂S₂O₈ and 20 μm of Co₉, the original PS* lifetime of approximately 395 ns decreases to approximately 264 and 339 ns, respectively, illustrating that both quenchers speed up the decay kinetics of *Ru(bpy)₃ ²⁺ and that S₂O₈ ²⁻ takes up electrons more efficiently. Thus, it can be suggested that during the photocatalytic process, Co₉ is oxidized by the oxidized form of the PS (Scheme S1), which is in accordance with previously reported POM‐WOCs.[ ¹⁷ , ²⁸ ]

Computational studies

The electronic structures of Co₄ and Co₉ were analyzed by means of density functional theory (DFT) simulations. ^[29] All cobalt atoms in Co₄ and Co₉ were considered in the +2 oxidation state with a formal d⁷ high‐spin configuration. Both polyanions exhibit a complex electronic structure with 12 (Co₄) and 27 (Co₉) unpaired electrons, respectively. Therefore, the corresponding models were simplified by replacing all except one of the Co^II ions with Zn^II resulting in a single catalytic site with a formal spin state characterized by three unpaired electrons. It has been recently shown that such an approximation does not affect the eigenvalues of the frontier molecular orbitals considerably, thereby representing a good trade‐off between the computational costs involved for the study of such systems and a reasonable level of accuracy. ^[30] The Co active site in the high spin state forms a distorted octahedron with the neighboring O atoms. This local environment is very similar in Co₄ and Co₉ and the calculated results are in good agreement with the values measured by XRD. In Co₄, the calculated values of the Co−O bond lengths present in the octahedron are around 2.10±0.08 Å (range 2.00–2.19 Å), in good agreement with the value of 2.10±0.05 Å (range 2.00–2.20 Å) observed in the crystal structures. In Co₉, these bonds have a comparable length of approximately 2.12±0.05 Å (range 2.06–2.20 Å), compared with the experimental value of 2.11±0.06 Å (range 1.96–2.27 Å). Considering the electronic configuration of Co₄ and Co₉, Figure 3 shows the calculated density of states for the alpha and beta electrons obtained from the calculated molecular orbitals eigenvalues after applying a Gaussian smearing. The highest occupied molecular orbitals (HOMO) of Co₉ lie higher in energy than those of the Co₄ system, while the lowest unoccupied molecular orbitals (LUMO) have comparable energies. In particular, the calculated eigenvalues for the HOMO electrons in Co₄ have values of −6.64 and −6.21 eV for the α and β electrons, respectively, while for Co₉ both alpha and beta HOMO electrons have an energy of around −5.71 eV (Figures 3, 4). Regarding the LUMO orbitals, they are almost degenerate for both systems: for Co₄ the α and β energies have values of −2.48 and −2.50 eV, respectively, while for Co₉ both α and β LUMO electrons have an energy of −2.46 eV. Consequently, the HOMO‐LUMO gap for the Co₄ system is 4.16 eV for the α electrons and 3.71 eV for the β ones. For Co₉, the HOMO‐LUMO gap is around 3.25 eV for both spin orientations. The observed HOMO‐LUMO gap trends are additionally supported by experimental estimations done by performing diffuse reflectance spectroscopy (Figures S11–S14) and cyclic voltammetry measurements (Figures S15, S16) on Co₄ and Co₉.

Figure 3.

Density of states for the α (blue) and β (red) electrons in the Co₄ and Co₉. Shaded areas represent occupied states.

Figure 4.

Representation of the α and β HOMO for Co₄ and Co₉. Color code: W^VI, black; Co^II, pink; Ge^IV, orange; P^V, light purple; O, red; Zn^II, silver.

Our mechanistic photoluminescence findings suggested oxidative quenching to be the dominant WOC pathway, which implies that Co₉ undergoes oxidation by [Ru(bpy)₃]³⁺ on the third stage of the catalytic cycle (Scheme S1). According to DFT calculations, the higher energy of the HOMO electrons in Co₉ indicates that it is thermodynamically more prone to oxidation as compared to Co₄. Considering the higher TOF values obtained for the Co₉, these complementary data suggest that POT oxidation may be one of the rate‐limiting processes and that a careful adjustment of the POM/PS redox properties can be used as a tool to optimize the ultimate WOC performance.

Conclusion

The phosphate templated stabilization of non‐ (Co₄), {Co^II ₃O₄} (Co₉) quasi‐ and {Co^II ₄O₄} (Co₂₀) full cubane structural motifs in germanotungstates as cost‐effective, active, and stable molecular water oxidation catalysts (WOCs) mimicking the naturally occurring oxygen‐evolving complex (OEC) is highlighted. The functional properties of the PO₄ ³⁻‐stabilized cubane motifs are shown by the doubled WOC activity of Co₉ and Co₂₀ compared to the Weakley‐type (non‐cubane) benchmark {Co₄P₂W₁₈}. Comprehensive mechanistic and electronic structure studies by photoluminescence and density functional theory suggest an oxidative quenching WOC mechanism and relate the ultimate performance of Co₉ to its redox levels. The confirmed stability and recyclability of Co₉ and Co₂₀ encourage their immobilization in matrices leading to heterogeneous photocatalytic materials as a subject of future applications.

Experimental Section

Preparation of Na₁₅[Ge₄PCo₄(H₂O)₂W₂₄O₉₄] ⋅ 38H₂O (Co₄)

Na₂WO₄ ⋅ 2H₂O (1.82 g, 5.52 mmol) and GeO₂ (0.054 g, 0.52 mmol) were dissolved in 40 mL water and subsequently CoCl₂ ⋅ 6H₂O (0.073 g, 0.31 mmol) was added. The pH was quickly adjusted from approximately 8.5 to 7.4 by the addition of 1 m HCl (≈5 mL) and then further to approximately 4.8 by 37 % HCl. The pH of the clear red solution was then raised to approximately 7.9 by adding solid Na₃PO₄ (≈0.28 g, 1.708 mmol) and adjusted to approximately 7.6 with 1 m HCl. The reaction mixture was then heated for approximately 10 min at 80 °C and filtered. The dark red solution was stored in an open beaker at room temperature. After about 2–3 weeks, red rhombohedral‐shaped crystals formed. After recrystallizing in water for several times some X‐ray‐suitable crystals were obtained. Yield: 0.015 g (1 % based on W). Anal. Calcd. [%] for Co₄: Na, 4.62; Co, 3.89; P, 0.41; W, 59.09; Ge, 3.81; Found: Na, 4.70; Co, 3.80; P, 0.15; W, 53.1; Ge, 2.15. Selected Fourier‐transform (FT)IR bands: $\tilde{\nu }$ =3340 (br), 1617 (m), 1088 (w), 1038 (w), 934 (m), 865 (s), 820 (m), 759 (s), 650 (s), 583 (s), 509 (s), 437 cm⁻¹ (s).

Preparation of Na_2.5K_17.5[Ge₃PCo₉(OH)₅(H₂O)₄W₃₀O₁₁₅] ⋅ 45H₂O (Co₉)

K₈[γ‐GeW₁₀O₃₆] ⋅ 6H₂O ^[12] (0.5 g, 0,172 mmol) was dissolved in 15 mL H₂O and subsequently CoCl₂ ⋅ 6H₂O (0.119 g, 0.5 mmol) was added. The pH of the clear dark red solution was then adjusted to 7.7 by adding solid Na₃PO₄ (≈0.07 g, 0.427 mmol). A color change to magenta occurred and slight precipitate formed. The reaction mixture was then heated for approximately 10 min at 80 °C, filtered and stored partially closed in a temperature‐controlled crystallization room (19±1 °C). After about one month, dark red‐violet block‐shaped for X‐ray‐suitable crystals formed in addition to very lightly colored crystalline precipitate. Yield: 0.14 g (25 % based on W). Anal. Calcd. [%] for Co₉: Na, 0.71; K, 6.85; Co, 5.46; P, 0.32; W, 56.8; Ge, 2.24; Found: Na, 0.94; K, 6.42; Co, 4.15; P, 0.53; W, 49.5; Ge, 2.00. Selected FTIR bands: $\tilde{\nu }$ =3364 (s), 3217 (s), 1620 (m), 1096 (w), 1033 (w), 1004 (w), 935 (m), 860 (m), 811 (m), 748 (m), 670 (m), 509 (m), 434 cm⁻¹ (m).

Preparation of Na₆K₁₆[Ge₄P₄Co₂₀(OH)₁₄(H₂O)₁₈W₃₆O₁₅₀] ⋅ 61H₂O (Co₂₀)

K₈Na₂[A‐α‐GeW₉O₃₄] ⋅ 25H₂O ^[13] (0.5 g, 0.16 mmol) was dissolved in 40 mL H₂O and subsequently CoCl₂ ⋅ 6H₂O (0.119 g, 0.47 mmol) was added. After stirring for approximately 20 min at room temperature, the pH of the clear red solution was then raised to 7.6 by adding solid Na₃PO₄ (≈0.07 g, 0.427 mmol). The reaction mixture was then heated for approximately 10 min at 80–85 °C, filtered and stored in an open beaker in a temperature‐controlled crystallization room (19±1 °C). After 2–3 days, red needle‐shaped crystals formed. After recrystallizing in water and slow evaporation at +4 °C X‐ray‐suitable crystals were obtained. Yield: 0.08 g (15 % based on W). Anal. Calcd. [%] for Co₂₀: Na, 0.51; K, 3.79; Co, 10.11; P, 0.92; W, 49.37; Ge, 2.17; Found: Na, 1.02; K, 4.64; Co, 7.09; P, 0.88; W, 45.1; Ge, 2.06. Selected FTIR bands: $\tilde{\nu }$ =3340 (br), 1614 (w), 1088 (w), 927 (m), 863 (m), 783 (s), 650 (s), 584 (s), 518 (s), 452 (s), 438 cm⁻¹ (s).

Preparation and characterization of Na₁₀[Co₄(H₂O)₂(α‐PW₉O₃₄)₂] ⋅ 27H₂O ({Co₄P₂W₁₈})

{Co₄P₂W₁₈} was prepared according to the literature procedure reported by Hill and co‐workers. ^[7] The identity of {Co₄P₂W₁₈} was proven by single‐crystal XRD (Table S1) and electrospray ionization mass spectrometry (ESI‐MS) (Figure S1).

Visible‐light‐driven water oxidation

A homogeneous solution of 80 mm aqueous borate buffer (pH=7.5, 8, 8.5 or 9) containing 1.0 mm of [Ru(bpy)₃]Cl₂, 5.0 mm of Na₂S₂O₈ and POM catalyst with concentrations varying from 2–20 μm was prepared in a two‐necked closed glass reactor equipped with an outer water‐cooling jacket. This solution was then deaerated using a flow of argon and later irradiated with a monochromatic visible LED (λ _max=445±13 nm) to trigger a photocatalytic reaction. The oxygen evolution was followed in situ using an optical oxygen meter (FireStingO2, Pyroscience, Germany) and a needle‐like oxygen‐sensitive optical sensor (OXF900PT‐OI) with a working principle based on the quenching of the REDFLASH indicator (immobilized on the sensor tip) luminescence caused by a collision between oxygen molecules and the indicator. In a single experiment, the oxygen sensor was inserted through a Viton septum placed in a screw cap on one of the necks of the reactor. The O₂ concentration was measured directly in %O₂ and was later converted to μmol and TONs based on the control experiments and the ideal gas equation. Initial TOFs were calculated as the maximum derivative (obtained from the “μmol of O₂ vs. time” plots) divided by the number of moles of the catalyst.

Computational details

All DFT calculations were performed employing the computer code NWChen. ^[31] The exchange‐correlation functional was approximated by employing the Becke‐3‐parameter‐Lee–Yang–Parr functional (B3LYP). ^[32] The core electrons of the Co, P, Ge, W, and Zn atoms were described by the LANL2DZ effective core potential ^[33] and the corresponding basis set was used for the valence electrons. Electrons of the H atoms were described by the 6–31G basis set and the 3–21G basis set was used for the O atoms. ^[34] Employing the larger 6–31G basis set for the O atoms affects the HOMO‐LUMO gap of Co₄ by around the 0.2 %, while the eigenvalues corresponding to the HOMO molecular orbitals change by slightly less than the 5 %. We therefore performed all calculations employing the smaller basis set. In both Co₄ and Co₉, all Co atoms except one were substituted by Zn atoms. This procedure was performed to simplify the complex magnetic structure endowed by the presence of multiple Co²⁺ cations. Haider et al. have shown that for similar Co‐containing polyanions, such substitution affects the eigenvalues of the frontier molecular orbitals only marginally. ^[30] The remaining Co atom in the polyanion was considered to be in the high‐spin quadruplet configuration. All structures were optimized in water, described with the conductor‐like screening (COSMO) continuum solvation model. ^[35] The permittivity of water was set to 78.36 and the radii of the atomic‐centered spheres used to construct the molecule‐shaped cavity were set to the corresponding atomic Van der Waals radii. The structures of all compounds were relaxed within the solvent model to a minimum of the potential energy surface employing a quasi‐Newton optimization method.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

E. Al-Sayed, S. P. Nandan, E. Tanuhadi, G. Giester, M. Arrigoni, G. K. H. Madsen, A. Cherevan, D. Eder, A. Rompel, ChemSusChem 2021, 14, 2529.