Co/Ni-polyoxotungstate photocatalysts as precursor materials for electrocatalytic water oxidation

An open-core cobalt polyoxometalate (POM) [(A-α-SiW9O34)Co4(OH)3(CH3COO)3]8−Co(1) and its isostructural Co/Ni-analogue [(A-α-SiW9O34)Co1.5Ni2.5(OH)3(CH3COO)3]8−CoNi(2) were synthesized and investigated for their photocatalytic and electrocatalytic performance. Co(1) shows high photocatalytic O2 yields, which are competitive with leading POM water oxidation catalysts (WOCs). Furthermore, Co(1) and CoNi(2) were employed as well-defined precursors for heterogeneous WOCs. Annealing at various temperatures afforded amorphous and crystalline CoWO4- and Co1.5Ni2.5WO4-related nanoparticles. CoWO4-related particles formed at 300 °C showed substantial electrocatalytic improvements and were superior to reference materials obtained from co-precipitation/annealing routes. Interestingly, no synergistic interactions between cobalt and nickel centers were observed for the mixed-metal POM precursor and the resulting tungstate catalysts. This stands in sharp contrast to a wide range of studies on various heterogeneous catalyst types which were notably improved through Co/Ni substitution. The results clearly demonstrate that readily accessible POMs are promising precursors for the convenient and low-temperature synthesis of amorphous heterogeneous water oxidation catalysts with enhanced performance compared to conventional approaches. This paves the way to tailoring polyoxometalates as molecular precursors with tuneable transition metal cores for high performance heterogeneous electrocatalysts. Our results furthermore illustrate the key influence of the synthetic history on the performance of oxide catalysts and highlight the dependence of synergistic metal interactions on the structural environment.


Introduction
Sunlight-driven splitting of water into hydrogen and oxygen, also known as articial photosynthesis, is among the most direct and elegant one-step concepts for renewable energy sources. 1 However, the development of efficient and noblemetal free water oxidation catalysts (WOCs) for this complex four electron transfer process is still a crucial bottleneck of articial photosynthesis. 2 A variety of approaches to WOC design have been reported, and in many of them the cuboidal {CaMn 4 O 5 } core of nature's photosystem II is a central motif. 3 Recently, several transition metal WOCs with cubane-related cores have been reported. [4][5][6][7][8][9][10] Polyoxometalates (POMs) are promising WOC candidates, because they combine robustness and structural versatility with the capability of undergoing rapid, reversible, and stepwise multi-electron transfer reactions, and we also refer to key review articles here. 11- 16 Among the attractive and low cost 3d transition metal water splitting catalysts, cobalt-based homogeneous catalysts and POMs keep attracting intense attention. [17][18][19][20][21][22] Specically, Co-POMs have recently been applied as co-catalysts on photoanodes or for enhanced performance in composite systems. [23][24][25][26][27][28] .
The coordination of multiple metal centers between two or more lacunary POM units has proven a powerful and quite exible catalytic motif, such as in the OEC-related [Mn III 3 Mn IV O 3 (CH 3 COO) 3 (SiW 9 O 34 )] 6À with a mixed-valent {Mn 4 } core. 29 Although many POMs have wide operational stability windows, they can undergo leaching of heteroatoms/ transition metals to form active nanoscale oxide catalysts, especially during electrochemical water oxidation. 30,31 This gives rise to ongoing and challenging speciation studies. 32 performance of oxide WOCs. These ndings show that synthetic methods and underlying mechanisms 36 are an important step in WOC optimization, [37][38][39] and that a certain extent of preorganization of the metal centers in the precursor is benecial for higher catalytic activity. 40 To this end, we systematically explored synthesis-activity relationships with a series of studies on spinel-type Co 3 O 4 catalysts, starting with in situ PXRD monitoring of temperature-dependent hydrothermal Co 3 O 4 formation mechanisms. 41 Next, we revealed the preparationdependent properties of Co 3 O 4 WOCs in different test assays, 42 and studies of their microwave-assisted synthesis further conrmed the crucial role of synthetic pathways for the catalytic performance. 43 POM as oxide precursors have enabled the efficient synthesis of CoO x electrocatalysts 44 or of ultrane transition metal-clusters, 45 as well as inspirational studies on transformation pathways of pre-organized metal centers into structural features of the resulting multinary heterogeneous catalysts. 46,47 Recent trends further employed POMs as versatile metal sources for carbide-, phosphide-and sulphide-based water splitting electrocatalysts. [48][49][50] All in all, the complexity of such precursorproperties relations remains to be fully explored and understood for efficient WOC design.
Another crucial principle in the optimization of transition metal catalysts are synergistic effects between mixed metal centers, such as the widely studied Ni/Fe interactions in WOCs. [51][52][53] In comparison, Co/Ni-interactions in oxygen evolution and other catalysts are far more diverse and controversial. Mixed heterogeneous Co/Ni-oxide electrocatalysts have been studied for several decades 54 and were frequently reported to be favourable over binary systems. 55-57 However, later studies on mixed Co/Ni-hydroxides pointed to either productive 58 or adverse 59 effects, or to no signicant interactions in the case of oxides at all. 60 Although Co/Ni synergisms were recently observed for sulphide 61 and phosphide 62 water splitting electrocatalysts, their understanding is still empirical to a large extent and modelling studies are now being undertaken. 62 Even less is known about the effect of Co/Ni-substitution on molecular WOCs and other catalysts. In our own work, for example, we have observed drastic contrasts between the notable improvement of solid CoNCN WOCs through Ni-doping 63 vs. detrimental effects on molecular {Co(II) 4 O 4 } cubane WOCs. 64 Similar adverse Co/Ni-interactions have been reported for other molecular systems. 65 Generally, a wide range of further studies is now needed to explore the role of materials type, preparative method and test conditions (photo-vs. electrochemistry) in the performance of Co/Ni-based catalysts. These widely unresolved questions concerning the prediction and explanation of Co/Niinteractions inspired us herein to rst investigate molecular photocatalyst performance with respect to Ni introduction, followed by its effect on the solid electrocatalysts obtained from such molecular precursors.
To this end, we selected Co/Ni-POMs as attractive and comprehensive models to investigate (a) Co/Ni interactions in molecular WOCs vs. (b) those in oxide-based catalysts, while (c) exploring the benets of POM precursors for oxide WOCs.
To this end, we targeted CoWO 4 with favourable 500-650 nm light absorption properties that was also reported as an effective and noble metal-free WOC at low overpotential. 66 Furthermore, CoWO 4 performance was found to depend on crystallinity 67 with amorphous CoWO 4 being superior to its crystalline form, along the lines of self-repairing CoPi lms. [68][69][70] However, little is still known about the electrocatalytic performance and other applications of mixed Co/Ni-tungsten oxides. 71,72 Binary CoWO 4 and NiWO 4 , for example, display better electrochemical performance in water oxidation than NiCo 2 O 4 spinels. 72 While mixed (Co, Ni)WO 4 materials are attractive for supercapacitor development, 73,74 no synergistic benets were reported for their use in photocatalytic methylene blue degradation. 75 We therefore newly investigated the inuence of Ni-doping on the performance of CoWO 4 -related water oxidation catalysts, which were obtained from crystallographically well-dened, bio-inspired M 4 -POMs with an exposed metal core architecture. While Ni-containing POMs were reported to be stable and active for water oxidation, 76,77 their potential as mixed metal precursors for heterogeneous WOCs still needs to be explored. To this end, we synthesized and characterized [(A-a-SiW 9 O 34 )Co 4 (-OH) 3 (CH 3 COO) 3 ] 8À Co(1) together with its isostructural analogue [(A-a-SiW 9 O 34 )Co 1.5 Ni 2.5 (OH) 3 (CH 3 COO) 3 ] 8À CoNi(2). 78 First, both POMs were compared with respect to their respective photo-and electrocatalytic water oxidation activity. Moreover, they were used as annealing precursors to form CoWO 4 -and mixed (Co,Ni)WO 4related electrocatalysts, and Co(1) was found to be an efficient precursor at temperatures as low as 300 C.

Materials
All chemicals were used as purchased without purication. The lacunary precursor Na 10 [A-a-SiW 9 O 34 ]$19H 2 O was synthesized as previously described. 79

Physical methods
Attenuated total reectance Fourier-transform (ATR-FT-IR) spectra were recorded on a Bruker Vertex 70 spectrometer equipped with a Platinum ATR accessory containing a diamond crystal. UV/Vis spectra were recorded on a Lambda 650 S Per-kinElmer UV-visible spectrometer in the range of 200-800 nm using a Quartz SUPRASIL precision cell (10 mm). Raman spectroscopy was recorded with a Renishaw inVia Qontor confocal Raman microscope equipped with a diode laser (785 nm). Thermogravimetric analyses were performed on a Netzsch STA 449C between 24 and 600 C with a heating rate of 10 K min À1 in N 2 atmosphere and Al 2 O 3 crucible. PXRD patterns were recorded on a STOE STADI P diffractometer (transmission mode, Ge monochromator) with Cu or Mo radiation. XPS analysis was performed using a PhI 5000 VersaProbe spectrometer (ULVAC-PHI, Inc.) equipped with a 180 spherical capacitor energy analyzer and a multi-channel detection system with 16 channels. Spectra were acquired at a base pressure of 5 Â 10 À8 Pa using a focused scanning monochromatic Al-K a source (1486.6 eV) with a spot size of 200 mm and 50 W. The instrument was run in the FAT analyzer mode with electrons emitted at 45 to the surface normal. Pass energy used for survey scans was 187.85 eV and 46.95 eV for detail spectra. Charge neutralisation utilizing both a cool cathode electron ood source (1.2 eV) and very low energy Ar + ions (10 eV) was applied throughout the analysis.
Visible-light-driven water oxidation was rst monitored in solution using an oxygen sensor (OX-N) Clark electrode from Unisense. Constant temperature was maintained with a mineral insulated thermosensor (2 mm tip diameter, TP2000, Unisense). Second, O 2 evolution was measured in the headspace of the vial using an Agilent Technologies 7820A gas chromatograph with helium as the carrier gas and a 3 m Â 2 mm packed molecular sieve 13 Â 80-100 column to separate O 2 and N 2 . The oven was operated isothermally at 100 C. The analysis of the headspace was performed by taking 100 mL samples with a Hamilton (1825 RN) gas-tight microliter syringe. Gases were detected using a thermal conductivity detector (Varian) operated at 200 C.
Cyclic voltammetry (CV) measurements were performed on a Metrohm 797 VA Computrace instrument with a platinum electrode (Metrohm AG, 2 mm diameter) as a working electrode, Ag/AgCl reference electrode (sat. KCl, 0.197 V vs. NHE) and platinum plate (Metrohm AG) counter electrode. Prior to all measurements, solutions were deaerated with Ar for 15 min. The platinum working electrode was polished between runs with alumina slurry, thoroughly rinsed with water and dried under ambient conditions. The platinum plate was washed in a nitric acid/hydrogen peroxide (1 : 1) solution for 5 min and dried with N 2 . The working electrodes were produced by dispersing 5 mg of the sample in 100 mL of H 2 O, applying 40 mL of this dispersion on 1 cm 2 uorine doped tin oxide (FTO), and drying the electrodes at 80 C for 30 min before covering with 10 mL Naon 1% solution.

K 5 Na 3 [(A-a-SiW 9 O 34 )Co 4 (OH) 3 (CH 3 COO) 3 ]$15H 2 O (1) 78
Co(CH 3 COO) 2 $4H 2 O (0.712 g, 2.86 mmol) was dissolved in an aqueous solution of potassium acetate (0.5 M, 16 mL), adjusted to pH 8 with HCl and stirred for 15 min. Na 10 [A-a-SiW 9 O 34 ]$ 19H 2 O (1.977 g, 0.7 mmol) was added and stirred for 45 min at 40 C. The mixture was then cooled to room temperature and placed in the fridge for 10 min. The purple suspension was centrifuged, ltered and le at room temperature for slow evaporation. Aer three weeks purple crystals were collected and analysed by FT-IR and Raman spectroscopy, powder X-ray diffraction, ICP-MS, EDX and ESI-MS. (1.513 g, 0.53 mmol) was added and stirred for 45 min at 40 C. The mixture was then cooled to room temperature and placed in the fridge for 10 min. The purple suspension was centrifuged, ltered and le at room temperature for slow evaporation. Aer two weeks light purple crystals were collected and analysed by FT-IR and Raman spectroscopy, powder X-ray diffraction, ICP-MS, EDX and ESI-MS (yield 0.23 g, 13% based on tungsten). 2.4.2 CoW300. K 5 Na 3 [(A-a-SiW 9 O 34 )Co 4 (OH) 3 (CH 3 COO) 3 ]$ 15H 2 O (0.087 g) was added into a crucible and placed in a furnace which was heated to 300 C with a ramping temperature of 5 C min À1 and annealed for 1 h to yield 0.08 g. Aer cooling down to room temperature the black compound was analysed by FT-IR and Raman spectroscopy as well as PXRD. FT- 3 ]$ 15H 2 O (0.305 g) was added into a crucible and placed in a furnace which was heated to 400 C with a ramping temperature of 5 C min À1 and annealed for 1 h to yield 0.27 g. Aer cooling down to room temperature the grey compound was analysed by FT-IR and Raman spectroscopy as well as PXRD. FT-IR:ñ ¼ 1128 (m), 929 (w), 823 (s), 790 (s), 599 (s), 518 (s), 460 cm À1 (m).
2.4.9 Reference CoWO 4 . An aqueous solution of Na 2 WO 4 (0.1 M, 10 mL) was added dropwise to a Co(NO 3 ) 2 solution (0.1 M, 10 mL) under vigorous stirring. The precipitate was rinsed with water aer centrifugation and dried overnight in the oven (40 C). The collected precipitate was added into a crucible and placed in a furnace which was heated to 300 C with a ramping rate of 5 C min À1 and annealed for 1 h. 66 (1) was synthesized by mixing stoichiometric amounts of the precursor 80 Na 10 [A-a-SiW 9 O 34 ] and cobalt acetate in potassium acetate (0.5 M, pH 8) solution with moderate heating. Aer cooling to room temperature, the mixture was ltered and any insoluble residue was removed, whereupon crystals were obtained aer slow evaporation.
The all-cobalt POM Co(1) was further mixed with nickel acetate in a stoichiometric 1 : 1 ratio under slightly changed reaction conditions. The ltration process aer the synthesis had to be extended, but phase pure crystals were obtained aer slow evaporation and yielded [(A-a-SiW 9 O 34 )Co 1.5 Ni 2.5 (OH) 3 (-CH 3 COO) 3 ] 8À CoNi(2). FT-IR analysis conrms the presence of the bridging acetate ligands as well as the characteristic bands of the Keggin-type POM at around 941 (n as (W-O d )), 883 (n as (W-O b )) and 748 cm À1 (n as (W-O c )) (Fig. S2 †). 81 The PXRD pattern conrmed phase purity of CoNi(2) and its isostructural relation to Co(1), and only small peak shis of the peaks towards higher angles are visible which corresponds to a smaller unit cell, as expected (Fig. S16 †). Raman spectra show the same representative peaks as observed for Co(1) (see below, Fig. 3).
UV-Vis monitoring of Co(1) in borate buffer (0.1 M, pH 8) during 24 h showed no signicant changes in the spectra. This indicates that Co(1) is stable and does not leach any Co 2+ ions into the solution under these operational conditions (Fig. S6 †). According to previous studies, 78 the tetracobalt core is stabilized by an all-inorganic tungstosilicate, as well as bridged by three m 2 -acetate ligands (Fig. 1). All Co II centers of the {Co II 4 O 3 } core are in an octahedral environment and the whole unit displays C S symmetry, with a mirror plane through the Co 3 , Co 2 and Si atoms. Three Co II centers are connected to the lacunary side of the [a-SiW 9 O 34 ] 10À POM. The Co-Co distances fall in the range of 2.978(1)-3.711(2)Å and the Co-O distances range from 2.043(0) to 2.117(7)Å, respectively. 78 We conrmed the presence of these structural features in Co(1) with single crystal X-ray diffraction analyses giving rise to analogous values (data not shown). The tetracobalt core displays features related to the natural OEC with Mn-Mn distances in the range of 2.8-3.3Å, 3 and the acetate-bridged cobalt centers relate it to previously reported {Co II 4 O 4 } cubanes. 6,83,84 In addition, mixed nickel/cobalt acetate precursors yielded the iso-structural CoNi(2) with a Co : Ni ratio around 1.5 : 2.5. This ratio was conrmed with EDX, ICP-MS as well as with XPS measurements (Fig. S20, Tables S3, S11 and S12 †). The measured Co/Ni : W as well as Co/Ni : Si ratios correspond to the respective calculated ratios of 4 : 9 and 4 : 1 (Tables S11 and S12 †). Further HR-ESI-MS analyses showed slightly different masses for the corresponding [M-(CH 3 COO) + 5H + ] 3À fragment which correspond to the isotopic distributions of Co and Ni (Fig. S29-S32 †). Both POMs display good agreement between experimental PXRD patterns and the respective calculated data (Fig. S13, S15, Tables S1 and S2 †). The PXRD pattern of CoNi (2) showed a slight shi of the peaks compared to the calculated reference pattern of Co(1) (CCDC-619251). Further comparison with the calculated PXRD pattern (Fig. S14 †) of the lacunary Nianalogue [(A-a-SiW 9 O 34 )Ni 4 (CH 3 COO) 3 ] 5that crystallizes in a different space group 85 (P 31c other than P2 1 /m for Co(1) 78 ) clearly showed that the phase pure CoNi(2) sample is isostructural with Co(1), which was also conrmed by Rietveld renement results (Fig. S15 †). CoWO 4 nanoparticles keep attracting intense interest as target for synthetic studies, e.g. via precipitation, 86,87 hydrothermal 88,89 or spray pyrolysis routes. 90 Here, we newly used both Co(1) and CoNi(2) as precursors for annealing in air at temperatures ranging from 200 to 500 C. With a ramping rate of 5 C min À1 and an annealing time of 1 h, amorphous and crystalline nanoparticles were formed. PXRD patterns show the presence of an amorphous material up to 300 C, while at temperatures of 400 C and above a crystalline material emerges from both precursor types (Fig. 2). The majority of the peaks in patterns recorded with MoK a radiation can be assigned to monoclinic CoWO 4 (PDF 01-072-0479) and its Ni-doped analogue, in line with previous studies. 86 33 we did not nd any indication for major W-or Co-based side products. Extensive database search provided SiO 2 (PDF No. 12-0711) as the closest match to account for the minority phases observed here.
Given that SiO 2 is widely known as a catalyst support material rather than as active phase, we further considered cobalt tungstosilicates a more reasonable precursor choice than tungstophosphates. The latter may eventually give rise to highly catalytically active cobalt phosphate-related side products. Indeed, phosphorus peaks had been shown in the EDS spectra of the most active amorphous cobalt tungstate catalyst obtained from [Co 4 (H 2 O) 2 (PW 9 O 34 ) 2 ] 10À at 400 C in the abovementioned study, but no further discussion of the inuence of P heteroatoms on the structure or catalytic performance was provided. 33  The Raman spectra of crystalline CoW300/400/500 are in good agreement with the reported pattern of CoWO 4 (Fig. 3). 91 The most intense band located at 885 cm À1 corresponds to the stretching W-O vibration and is shied to higher frequencies upon mixing with Ni. The small band around 929 cm À1 can be attributed to the symmetric stretching mode of the terminal (W]O) bond. 66,92 The amorphous CoW300 and CoNi300/400 samples show a small blue shi of the main W-O stretching vibration compared to the crystalline samples. This is a sign of compressive stress, indicating that the respective Co-Co and Co-Ni distances are smaller compared to the crystalline samples. 92 The weak peak around 500 cm À1 can be assigned to the minority phase related to SiO 2 . 93 EDX mappings of the different CoWX00 and CoNiWX00 tungsten oxides show a homogenous distribution of Co/Ni, W and O in all samples (Fig. S23-S28 and Tables S5-S10 †) and the elemental ratios of the CoWX00 series are in line with CoWO 4 . (2) 3.2.1 Photocatalytic activity of Co(1) and CoNi (2). The photocatalytic water oxidation activities of Co(1) and CoNi (2) were investigated in a borate buffer solution (0.1 M, pH 8, 8 mL) with [Ru(bpy) 3 ] 2+ (1 mM) as photosensitizer (PS) and Na 2 S 2 O 8 (5 mM) as sacricial electron acceptor under irradiation at 470 nm. O 2 evolution was monitored by GC-MS to determine the overall TON and with a Clark electrode to determine the initial TOF.

Photo-and electrocatalytic water oxidation activity of Co(1) and CoNi
The general mechanism of photocatalytic water oxidation using the [Ru(bpy) 3 3 ] 3+ equivalents, and the so oxidized POM catalyst can then further oxidize two water molecules. The precise local mechanisms at the active transition metal centers of different POM-WOCs are subject to advanced theoretical studies and further investigations, and they may vary individually for each POM type. 95 For photocatalytic performance evaluation, rst Co(1) was tested in different buffer solutions and pH values to explore the optimal water oxidation conditions. Borate buffer (0.1 M, pH 8) led to the best performance ahead of borate buffer (0.1 M, pH 9) and phosphate buffer (0.1 M, pH 7). No activity was observed in acetate buffer (0.1 M, pH 4.75) (Fig. S33 †).
Second, concentration screening of Co(1) was performed to further optimize the working conditions (Fig. 4, S34 and S35 †). Although water oxidation is generally thermodynamically favorable at higher pH values, performances at pH 8 were found to be superior to pH 9, 96 in line with other studies. 6 The maximum O 2 yield of 63% was achieved with 40 mM of Co(1). Compared to other reported Co-or Ni-based POMs, this O 2 yield is competitive for the applied photocatalytic assay (Table 1). In comparison, the Mn-based analogue [Mn III 3 Mn IV O 3 (CH 3 COO) 3 (A-a-SiW 9 O 34 )] 6À exhibited a rather low photocatalytic performance with 3% oxygen yield. 29 In Table 1, TON, TOF and O 2 yields are compared to several previously reported POM-WOCs. TONs were increasing with reduced catalyst concentrations, reaching a value of 235 at a catalyst concentration of 1 mM (Table S13 †). In the absence of Co(1), a background O 2 evolution of 0.83 mmol was detected, which corresponds to 4% O 2 yield. Additionally, a reference  WOC test with the same concentration of cobalt centers was performed for cobalt acetate (40 mM based on Co), and the obtained 54% O 2 yield was lower compared to Co (1).
Recycling experiments showed further activity of the catalytic system aer adding additional Na 2 S 2 O 8 and adjusting the pH back to 8. The 2 nd and 3 rd cycle showed O 2 yields of 25% and 11%, respectively (Table S14 and Fig. S36 †). Throughout the recycling, a continuous color change was observed from bright orange to dark green, suggesting a slow decomposition of the photosensitizer [Ru(bpy) 3 ] 2+ . This shows that low photosensitizer stability associated with the formation of sulfate radicals from Na 2 S 2 O 8 are the main reasons for the decline of O 2 evolution.
CoNi (2) showed reduced photocatalytic oxygen evolution performance compared to pure Co(1). O 2 yields decreased from 63% to 26% at a catalyst concentration of 40 mM ( Table 1). The calculated O 2 yield per Co center (for 40 mM catalyst) is 16% for Co(1) and 15% for CoNi(2), respectively. This underscores further that the Ni centers are most likely inactive. Our previous work on the molecular cubane water oxidation catalyst [Co 4 (dpy-C{OH}O) 4 (OAc) 2 (H 2 O) 2 ](ClO 4 ) 2 showed a comparable trend towards lower O 2 yields upon introducing Ni into the cobalt sites. This may imply that an intramolecular O-O coupling pathway between two Co-OH n ligands prevails for O 2 evolution from such oxocluster WOCs, as reported for Co 3 O 4 . 97 CoNi(2) furthermore displays a larger band gap (2.66 eV) than Co(1) (2.41 eV, see Fig. S7 and S8 †).
The expected formation of a solid POM-PS complex was observed aer the catalytic O 2 evolution tests. 101 FT-IR analysis shows the presence of both photosensitizer and Co(1) in the precipitate (Fig. 5). For the POM-PS complex, characteristic bands are observed at 989, 939 (n as W-O d ) and 874 cm À1 (n as W-O b ) compared to 980, 932 and 884 cm À1 for the pristine Co(1). Additional, three bands at 1463, 1444 and 1423 cm À1 can be attributed to [Ru(bpy) 3 ] 2+ , in line with previous reports on POM-PS complex formation. 102 This precipitation is a general phenomenon of POM-WOCs in such photocatalytic assays due to the electrostatic interactions between negatively charged POMs and positively charged photosensitizer molecules. Only at low catalyst concentrations <2 mM no measureable precipitation was detected, although such amounts most likely fall below the detection limit of frequently used DLS devices. Note that those devices were designed to quantify size distribution of large amounts of nanoparticles rather than for evidencing their absence. 103 Further EDX analyses of the POM-PS complex show a homogeneous distribution of Ru, Co and W and calculated ratios of 2.5 : 1 [Ru(bpy) 3 ] 2+ : Co(1) ( Table S4, † based on the N : W ratio, Fig. S22 †). Notably, lyophilisation of the remaining solution showed no presence of cobalt, which further supports that Co(1) does not undergo leaching of Co 2+ ions into the solution (Fig. S21 †). A subsequent WOC test with the ltered solution and additional Na 2 S 2 O 8 showed no further activity (Fig. S37 †).
Photocatalytic tests with the recovered POM-PS complex, revealed its continuous activity with 34% O 2 yield, which is superior to the direct recycling run of the pristine POM in solution. This shows that the POM-PS complex is still active as a catalyst and that the photocatalytic assay ([Ru(bpy) 3 ] 2+ / Na 2 S 2 O 8 ) causes the fast decline of O 2 formation for pristine Co(1). (1) and CoNi (2). Fig. 6 displays the results of electrocatalytic activity tests, where the onset potentials for Co(1) and CoNi(2) were determined as 0.96 V and 1.00 V vs. Ag/AgCl, respectively. The higher onset potential of CoNi(2) corresponds to the lower photocatalytic water oxidation performance. According to previous reports, Cobased POM electrocatalysts can undergo Co 2+ leaching, especially in basic conditions. 104 The leached Co 2+ ions can then form an active heterogeneous CoO x lm on the working electrode and contribute to the WOC activity. 30 During several CV scans, Co(1) and CoNi(1) showed minor shis of the onset potential and anodic peaks to lower potentials (0.08 V aer 3 cycles, Fig. S38 and S39 †), which might be attributed to such CoO x formation on the working electrode. It is of note that the Pt counter electrode was tested in the applied potential range and showed no activity (see blank measurements below). Previous reports showed dissolution and re-deposition of Pt on  the reduction half-cell in acidic reaction media. 105 To the best of our knowledge, no such inuence of Pt electrodes has been reported for the OER before. 106
The signicant inuence of the annealing temperature is quite evident from the cyclic voltammetry results (Fig. 7). Amorphous CoW300 has an onset potential of 0.79 V vs. Ag/AgCl in borate buffer (0.1 M, pH 8) solution and maintains its catalytic performance over the measured eight cycles.
A broad cathodic peak at 0.55 V can be seen in the backward scan, which moves slowly to higher potentials in the subsequent scans (0.07 V during eight cycles, cf. Fig. S47 †). The amorphous sample shows constant catalytic performance over several cycles with stable anodic and cathodic peaks. Previously reported amorphous CoWO 4 showed a substantial change in the anodic and cathodic peak aer the rst cycle. 66 The onset potential of the crystalline samples CoW400 and CoW500 gradually increases to 0.94 V and 1.18 V vs. Ag/AgCl, respectively (Fig. S45 †). This is in line with previous observations for amorphous and crystalline CoWO 4 and their onset potentials. 68,69 Interestingly, the onset potential of CoW500 is almost the same as of the blank FTO electrode.
The onset potentials for the Ni-doped analogues are shied to higher potentials (Fig. 7). Other than the pure Co samples, the onset potential values for CoNi300 and CoNi400 are closer, namely 0.91 V and 0.98 V, respectively (Fig. S46 †). On the other hand, CoNi500 has a better onset potential (1.08 V) compared to its binary analogue CoW500 (1.18 V).
As mentioned above, the blue shi of the {WO 4 } Raman peak of CoW300 suggests a reduced distance between Co 2+ ions, which can further inuence the mechanism of the electrocatalytic water oxidation. 66,92 This was previously reported and conrmed with EXAFS analyses for amorphous and crystalline CoWO 4 , 33 where the outstanding performance of the amorphous CoWO 4 was attributed to the shorter Co-Co distances.
Although the water oxidation reaction is a complex process, it was reported in previous studies that its mechanism strongly depends on the distance between the active sites of heterogeneous electrocatalysts. 33 Along these lines, it is assumed that closer intermetallic distances in the range of the O-O bond distance of dioxygen are favoring the "dual-site" Langmuir-Hinshelwood (LH) mechanism. In this mechanism, oxygen species are rst adsorbed on adjacent sites, followed by their formation of molecular oxygen. In the case of longer distances between the active metal sites, however, the "single-site" Eley-Rideal (ER) mechanism may take place instead.
The dual-site LH mechanism requires a lower overpotential than the single-site ER mechanism, because the latter includes the formation of a peroxo intermediate at the single active metal center. This is considered a thermodynamically less favorable step for the overall water oxidation. 66,107 In the present case, the observed decrease of the Co-Co distances in CoW300 may therefore facilitate the bridging of two terminal oxo groups to generate dioxygen via the LH mechanism. 66,108 This agrees with the observed lowest onset potential for CoW300 among the tungstate catalyst series.
CoW300 as best performing member of the tungstate series was further compared to a reference sample obtained from a conventional solution co-precipitation method (annealed at 300 C) for CoWO 4 and to RuO 2 as a well-established benchmark WOC (Fig. 8). 66 CoW300 showed a lower onset potential (0.79 V) than the as-synthesized reference sample (0.84 V).
Furthermore, chronoamperometry measurements of CoW300 and the conventionally synthesized reference material were performed. The superior performance of CoW300 is clearly evident from the lower Tafel slope (inset Fig. 8) of 96 mV dec À1 , compared to 144 mV dec À1 for the reference material obtained from Na 2 WO 4 and Co(NO 3 ) 2 . 66 Fig. 7 Top: Cyclic voltammograms of CoW300 (red), CoW400 (blue) and CoW500 (green); bottom: CVs of CoNi300 (red), CoNi400 (blue) and CoNi500 (green); all measurements on FTO in 0.1 M borate buffer pH 8 vs. blank measurements (black, dashed; V vs. Ag/AgCl, scan rate: 20 mV s À1 , 3 rd scan is shown).

Conclusions
A bio-inspired polyoxometalate with an open Co 4 -core architecture, K 5 Na 3 [(A-a-SiW 9 O 34 )Co 4 (OH) 3 (CH 3 COO) 3 ] Co (1), was synthesized as a model system to investigate crucial questions of water oxidation catalyst (WOC) design, namely (1) the controversially discussed effect of Co/Ni-synergisms in molecules vs. solids, (2) the inuence of preparative history and precursor choice on WOC activity and (3) the role of amorphous features in solid WOC performance.
First, the open Co 4 -POM Co(1) displayed competitive photocatalytic activity with an oxygen yield of 63% for the optimal catalyst concentration of 40 mM. Furthermore, its new mixed Co/Ni isostructural analogue CoNi(2) was synthesized, analyzed and tested for water oxidation activity. Co/Ni substitution did not exert a productive inuence on the water oxidation activity of the mixed-metal POMs, in contrast to widely reported Co/Ni synergisms in solid WOCs.
Next, to investigate the effect of mixed metal molecular precursors on cobalt tungstate-related WOCs as attractive target materials, Co(1) and CoNi(2) were subjected to thermal treatment. Both compounds afforded CoWO 4 -as well as (Co, Ni) WO 4 -related phases with increasing degrees of crystallinity upon higher annealing temperatures.
Concerning the inuence of crystallinity on the performance, cyclic voltammetry measurements clearly showed that amorphous CoW300 obtained from annealing Co(1) at 300 C showed the lowest onset potential among both series of tungsten oxides obtained from POM precursors. Most importantly, CoW300 displayed a lower onset potential than a representative reference sample of CoWO 4 that was synthesized via a conventional co-precipitation/annealing method.
In line with the catalytic trends for the Co(1) and CoNi(2) precursor POMs, introduction of nickel centers did not exert a productive effect on the heterogeneous tungstate catalysts either. This is in stark contrast to the growing number of literature reports on Co/Ni synergisms in a wide range of oxide and non-oxide heterogeneous electrocatalysts. Further systematic studies are now required to understand the dependence of such metal-metal interactions on the catalyst matrix and its preparative history, as well as on the applied performance test conditions.
Our results demonstrate that readily accessible POMs are promising precursors with pre-organized metal centers for the convenient synthesis of amorphous heterogeneous water oxidation catalysts, which outperform products of conventional high temperature approaches starting from simple binary educts. Interestingly, neither the molecular precursors nor their heterogeneous WOC products were responsive to widely employed synergistic Co/Ni doping strategies. This highlights the complexity and matrix dependence of such mixed metal optimization strategies, which are in the focus of forefront catalytic endeavours.
To fully transfer the tunable potential of polynuclear molecular precursors into high performance amorphous catalysts with optimal near-range order properties, in-depth monitoring and theoretical studies of mixed metal interactions in different settings are now required.

Conflicts of interest
There are no conicts of interest to declare.