Immobilization of a [CoIIICoII(H2O)W11O39]7– Polyoxoanion for the Photocatalytic Oxygen Evolution Reaction

The ongoing transition to renewable energy sources and the implementation of artificial photosynthetic setups call for an efficient and stable water oxidation catalyst (WOC). Here, we heterogenize a molecular all-inorganic [CoIIICoII(H2O)W11O39]7– ({CoIIICoIIW11}) Keggin-type polyoxometalate (POM) onto a model TiO2 surface, employing a 3-aminopropyltriethoxysilane (APTES) linker to form a novel heterogeneous photosystem for light-driven water oxidation. The {CoIIICoIIW11}-APTES-TiO2 hybrid is characterized using a set of spectroscopic and microscopic techniques to reveal the POM integrity and dispersion to elucidate the POM/APTES and APTES/TiO2 binding modes as well as to visualize the attachment of individual clusters. We conduct photocatalytic studies under heterogeneous and homogeneous conditions and show that {CoIIICoIIW11}-APTES-TiO2 performs as an active light-driven WOC, wherein {CoIIICoIIW11} acts as a stable co-catalyst for water oxidation. In contrast to the homogeneous WOC performance of this POM, the heterogenized photosystem yields a constant WOC rate for at least 10 h without any apparent deactivation, demonstrating that TiO2 not only stabilizes the POM but also acts as a photosensitizer. Complementary studies using photoluminescence (PL) emission spectroscopy elucidate the charge transfer mechanism and enhanced WOC activity. The {CoIIICoIIW11}-APTES-TiO2 photocatalyst serves as a prime example of a hybrid homogeneous–heterogeneous photosystem that combines the advantages of solid-state absorbers and well-defined molecular co-catalysts, which will be of interest to both scientific communities and applications in photoelectrocatalysis and CO2 reduction.


■ INTRODUCTION
Owing to the ever-growing demand for renewable energy sources, the research community has been exploring processes that enable the generation of solar fuels. Among various solarpowered means to produce H 2 , photocatalytic water splitting stands out as the most direct and cost-effective approach. 1−3 This light-driven, energetically uphill reaction involves two redox processes: water oxidation to form O 2 termed as the oxygen evolution reaction (OER)and H + reduction to yield H 2 termed as the hydrogen evolution reaction (HER). Due to the sluggish kinetics of water oxidation and the necessity of four-electron transfer, it is the OER that is often considered to be the bottleneck of water splitting. 4 Even though OER does not generate a fuel, it provides reducing equivalents required for its production, e.g., by the reduction of protons to H 2 in water splitting or by reduction of CO 2 to CO or other carbonaceous products in artificial photosynthesis. Given the importance and challenging nature of this reaction, a bulk of research works have been devoted to the search for an efficient light-driven water oxidation catalyst (WOC).
Among various families of photocatalysts, metal oxides such as SrTiO 3 , NaTaO 3 , WO 3 , and BiVO 4 are considered forerunners for heterogeneous WOC due to their suitable electronic structures, earth abundance, and redox stability. 5 However, even though many reported photocatalysts can act as efficient light-absorbers, their performance is often limited by the low efficiencies of the WOC step, i.e., the activation of water molecules at the solid/liquid interface and their stepwise oxidation. 6 To overcome this issue, co-catalysts are often employed in the form of small nanoparticles attached to the support surface. 7 Not only do they provide better-suited active and adsorption sites, they also facilitate the extraction of photoexcited charge carriers to the reaction centers. The purposeful design of active and selective co-catalysts is, however, often complicated by their poorly defined surface structures, which ultimately limits the degree of control over their performance. The implementation of structurally defined molecular co-catalysts emerges as a promising alternative approach to tackle this challenge. 8 In parallel to the development of heterogeneous photosystems, various transition metal-based complexes have been demonstrated to act as efficient and quick (photo)catalysts for water oxidation under strictly homogeneous conditions. 9−11 However, the WOC performance of these molecular species is often compromised by the instability of the organic ligands toward oxidation by the reactive intermediates formed upon illumination. 12 In view of this limitation, polyoxometalates (POMs) have more recently emerged as promising WOC candidates as they combine high structural tunability and superior redox stability 13 owing to their rigid metal-oxo frameworks. After the first all-inorganic WOC-active [{Ru 4 O 4 (OH) 2 (H 2 O) 4 }(γ-SiW 10 O 36 ) 2 ] 10− cluster had been reported in 2008, 14 the first earth-abundant Co-based [Co 4 (H 2 O) 2 (PW 9 O 34 ) 2 ] 10− has triggered extensive interest in Co-containing POM-WOCs. 15 To date, many other Co-based POMs have been reported for homogeneous (light-driven) water oxidation, and their overall performance was investigated with regard to POM's structural type, 16 accessibility of Co centers, 17 and pH values. 18 Despite these developments, two issues still hinder a broader implementation of POM-WOCs: self-aggregation of POM clusters under the turnover conditions and the poor stability of molecular photosensitizers. 19 Both aspects are highly detrimental and often lead to rapid deactivation of the photosystems, which limit their long-term performance. In order to address these challenges and combine the advantages of heterogeneous and homogeneous approaches to photocatalysis, a recent perspective discussed the prospects of POM heterogenization on the surface of photoactive supports. 20 In this setup, POM takes the role of a structurally and compositionally well-defined co-catalyst, while the semiconducting support complements by providing an efficient and stable light absorption. As an added benefit, immobilization is expected to improve the structural stability of individual POM clusters and further allow for control over the electronic communication between the absorber and the catalyst.
We recently summarized previous examples of POMs composited within various solid-state matrices. 20 Several polyoxotungstate and polyoxomolybdate clusters have been successfully deposited onto metal oxides, 21−28 carbon nitrides, 29,30 metal−organic, and covalent−organic frameworks; 31−35 however, mostly (photo)electrocatalytic, sensing, and solar cell applications have been targeted so far. Also, most of the protocols relied on physisorption/electrostatic binding, 36,37 which can be detrimental to the stability of the resulting photosystems. In this work, we have covalently attached a WOC-active POM onto photoactive metal oxide surfaces using a linker molecule. We evaluated the resulting composite toward light-driven water oxidation, for which the role of photosensitizer is taken by the inorganic support. We investigated the synergistic properties and the charge transfer dynamics between the two components as well as the stabilization effect upon attachment.
As the first example of this approach in photocatalytic WOC, we chose the well-studied TiO 2 (anatase) substrate as it has an appropriate electronic structure for water splitting 38 and is a very efficient, nontoxic photocatalyst with a large surface area and excellent redox stability. 39 To provide chemically tunable binding sites for the POM anions, 3-aminopropyltriethoxysilane (APTES) was chosen as the linker given its bifunctional nature: the ethoxide moiety can covalently attach to the hydroxylated oxide surface upon condensation of the ethoxy groups, 40 18 In addition, this polyanion has a high negative charge of −7, and with its two Co heteroions in two different oxidation states (+II and +III), it has been reported to be the most promising among Cocontaining POMs for WOC applications, 18 42 For this, 19.8 g (0.06 mol) Na 2 WO 4 ·2H 2 O was dissolved in 40 mL of H 2 O, and then by the addition of glacial acetic acid, the solution pH was adjusted between 6.5 and 7.5 and heated to near-boiling. Afterward, 2.5 g (0.01 mol) of Co(OAc) 2 ·4H 2 O dissolved in 13 mL of warm H 2 O was added dropwise into the tungstate solution while stirring. A pink precipitate got formed and redissolved quickly, forming a dark green solution. The mixture was then heated to reflux for 10 min, and insoluble material was filtered out. Seven grams (0.026 mol) of K 2 S 2 O 8 was added 18 while heating the solution at 80°C and then kept until boiling. Once the reaction solution changed its color from green to dark brown, the boiling was continued for another 5 min, which was followed by filtration. The filtrate was then kept for heating again until boiling. Lastly, 25 mL of hot saturated KNO 3 was added, and the mixture was cooled, leading to the formation of a brown precipitate. This dark brown solid was then filtered out. Later, this solid was added to water and heated to 90°C , stirred for a few minutes, and filtered, and the filtrate was collected. The filtrate was then kept for cooling, and the precipitated solid was removed by filtration. The obtained clear brown solution was kept for crystallization by evaporation at room temperature, and dark brown cubic single crystals were formed. The unit cell parameters calculated from single crystal XRD are a = b = c = 21.55 Å, α = γ = β = 90°and are consistent with those previously reported for {Co III Co II W 11 } (CCDC code 915800). 18 Elemental analysis found (calculated) in %: K: 8.6 (8.2); Co: 3.7 (3.5); W: 60.2 (60.8).
Functionalization of TiO 2 with APTES. The synthesis was done using a modified protocol based on Kockmann et al. 43 For this, 500 mg of anatase nanoparticles was taken in 50 mL of absolute ethanol in a round bottom (RB) flask and kept for sonication for 1 h. After sonication, an excess (1.5 mL) of APTES (considering 1 mol APTES per mol TiO 2 ) was taken and added to the reaction medium. The RB was refluxed overnight at 85°C, and after cooling the RB, 50 mL of hexane was added to the suspension to induce precipitation. After a few minutes, it was centrifuged at 6500 rpm for 15 min at room temperature (RT), and the residue was collected. This residue was washed using absolute ethanol and was centrifuged two more times using hexane. Finally, the precipitate was collected and kept for drying overnight in a vacuum oven at room temperature.
Immobilization of {Co III Co II W 11 } onto APTES-TiO 2 . Following the protocol by Yang et al., 44 250 mg of APTES-TiO 2 was dispersed in an aqueous {Co III Co II W 11 } solution (125 mg of {Co III Co II W 11 } and 25 mL of water). The pH was monitored throughout and kept in the slightly acidic range (between 6.3 and 6.7) by adding 0.1 M HCl, making sure that the POM maintains its integrity without getting hydrolyzed. The solution was stirred for 24 h at RT, followed by centrifuging, washing with 100 mL of water, and drying overnight to yield the {Co III Co II W 11 }-APTES-TiO 2 composite.

Photocatalytic Water Oxidation
For conducting photocatalytic water oxidation reaction experiments, the reaction solutions were prepared in a closed, two-necked glass reactor with an outer water-cooling (15°C) jacket. This 2 mL reaction media was then deaerated using Ar purging (100 mL/min) for 30 min until it reached almost-zero O 2 concentration. The reaction setup was kept under darkness for 20 min to get a constant baseline before illumination was started to trigger a photocatalytic reaction. The evolved O 2 was recorded using a O 2 sensor (FireStingO2, Pyroscience) inserted to the reaction medium (i.e., volume above the solution) through a viton septum placed in a screw cap on one of the necks of the reactor. It works on the principle of fluorescence quenching of an indicator at the tip of the sensor, which depends on the concentration of O 2 surrounding it. The advantages include sensor staying inside the reactor, which can have in situ detection of O 2 amounts, with a 1 s data collection interval. The preparation of reaction solutions for both homogeneous and heterogeneous reactions is given as follows. For homogeneous WOC, the reaction medium was prepared by taking 20 μM (details in the SI) concentrations of POM catalysts dissolved in 80 mM aqueous borate buffer solution (pH 8), containing 1 mM Ru[bpy] 3 Cl 2 (bpy = 2,2′-bipyridine) as the photosensitizer (PS) and 5 mM Na 2 S 2 O 8 as a sacrificial agent (SA). The illumination was done using a monochromatic LED light source (445 ± 13 nm, power = 64 mW/ cm 2 , incident light intensity = 129 mW, Thorlabs SOLIS). For heterogeneous WOC, 1 mg of the POM-immobilized TiO 2 sample was dispersed in 2 mL of 10 mM Na 2 S 2 O 8 aqueous solution. The illumination was done using a monochromatic LED light source (365 ± 6 nm, power = 183 mW/cm 2 , incident light intensity = 366 mW, Thorlabs SOLIS).

Photoluminescence Measurements
A photocatalytic WOC reaction was conducted with {Co III Co II W 11 }-APTES-TiO 2 in 10 mM Na 2 S 2 O 8 reaction solution diluted with 3 × 10 −3 M (in 0.01 M NaOH) terephthalic acid (TA) solution. Once the UV illumination starts and the charge carriers are photogenerated, TA can get converted to 2-hydroxyterephthalic acid (TA-OH) if ·OH are generated. As TA-OH is fluorescent, PL emission can be used to quantify the amount of generated ·OH. Similarly, the protocol was followed for TiO 2 and 10 mM Na 2 S 2 O 8 (blank) solution as well. After illuminating the OER reaction solutions (with TA) for 30 min, it was centrifuged at 5600 rpm for 20 min, and the supernatant was then syringe-filtered (0.45 μm pore) to remove any suspended particles. The PL emission of this solution was probed with an excitation wavelength of 315 nm.  Figure 1a. The structure of the cluster was confirmed using single-crystal X-ray diffraction (XRD) and attenuated total reflectance Fouriertransform IR spectroscopy (ATR-FTIR), and was further characterized by X-ray photoelectron spectroscopy (XPS) and powder XRD (details in the SI and Figure S1). 18 Elemental constituents derived from inductively coupled plasma mass spectrometry (ICP-MS) and total reflection X-ray fluorescence spectroscopy (TXRF , Table S2) further confirm its composition and the presence of two Co centers. The stability of {Co III Co II W 11 } in water at pH ≈ 6 relevant to WOC studies was investigated by electrospray−ionization mass spectrometry (ESI-MS, Figure 1b). The ESI-MS spectrum recorded in negative mode exhibits a series of the peaks' envelopes at m/z between 910 and 940, which can be unambiguously assigned to the triply charged anions H 4-x K(Na) x [Co III Co II W 11 Table  S1).

APTES Functionalization on the TiO 2 Surface
An APTES monolayer was deposited onto the TiO 2 support via ethoxy group hydrolysis and condensation to surface hydroxyls 40,45 to ensure specific and selective attachment of the POM clusters ( Figure 2a and the Experimental Section). Various characterization techniques including 29 Si solid-state nuclear magnetic resonance (NMR) spectroscopy, XPS, and ICP-MS confirmed the APTES coverage. The 29 Si NMR spectrum of APTES-TiO 2 in Figure 2b shows only one broad peak, with a chemical shift around −60 ppm, which is consistent with the presence of APTES. 41 Based on the 29 Si chemical shifts measured for unattached APTES (−37 ppm) as well as those attached in the monodentate (−43 ppm), bidentate (−59 ppm), or tridentate (−67 ppm) modes, 46   hydroxides, organic species from the adventitious carbon (C− O), and moisture. 49 The O 1s profile further contains a contribution at 532.3 eV, which can be assigned to Si−O− Ti 40,50 in line with the NMR data. The N 1s peak (Figure 2d) contains two pronounced peaks at 399.6 and 401.3 eV, indicating two different chemical environments: the first peak being the characteristic of free NH 2 , 40,51 while the latter can be assigned to H-bonded or protonated amino groups, 51−54 which may coexist after the attachment due to the close proximity of the APTES molecules. The Si 2p spectrum (Figure 2e) shows a peak centered around 102 eV, which can be assigned to the Si−O−Ti bonds, 40,51 indicating that most Si atoms are grafted on TiO 2 . A broad peak shape (full-width half maximum of ca. 2.1 eV), however, points toward the existence of two overlapping contributions, in which the shoulder at 102.8 eV being indicative of Si−O−Si bonds formed between closely packed APTES moieties (Figure 2b, inset). The overall data thus show that APTES attaches onto TiO 2 as a monolayer via Si ions (covalent Si−O−Ti bonds), while amino moieties stay available for POM anchoring (Figure 2a).  Figure  S3). 58 With this choice of the components and attachment procedure, the formation of the {Co III Co II W 11 }-APTES-TiO 2 composite is driven by self-assembly (Figure 3a). No such coordination is possible for the parent [PW 12 O 40 ] 3− clusters as W(VI) has no free d-orbitals and its terminal O ions are connected by a very stable double bond (details in the SI). Figure 3b shows the ATR-FTIR spectrum of the {Co III Co II W 11 }-APTES-TiO 2 composite, from which the spectrum of APTES-TiO 2 is subtracted. The difference spectrum indicates the presence of bands in the 850−950 cm −1 regime, which corresponds to WO and W−O−W vibrations, suggesting the presence of an intact {Co III Co II W 11 } cluster in the composite. The slight shift in peak positions indicates minor structural rearrangements associated with the loss of crystal water during immobilization, which is expected from the presence of the individual POM clusters on the surface. To confirm the molecular nature of immobilization and to examine if any POM aggregation takes place, powder XRD and scanning transmission electron microscopy (STEM) were utilized. Figure 3c shows the diffraction pattern of the physical mixture prepared by grinding 10 wt % {Co III Co II W 11 } in TiO 2 as a reference, which contains the typical peaks of both anatase and {Co III Co II W 11 }. Note that the POM-related peaks are absent in the {Co III Co II W 11 }-APTES-TiO 2 composite despite a relatively high loading of 14 wt % {Co III Co II W 11 } (see the ICP-MS discussion below). This strongly suggests that {Co III Co II W 11 } clusters are present in their molecular form. Elemental energy-dispersive X-ray spectroscopy (EDS) mapping ( Figure S4) shows uniform and matching spatial distribution of Co, W, and Ti on the nanoscale, which further confirms a highly homogeneous coverage and dispersion of POM clusters over the titania matrix.   To derive an atomistic picture of the surface-anchored POM clusters, we employed aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM; details in the SI). A high-resolution micrograph of the {Co III Co II W 11 }-APTES-TiO 2 sample in Figure 4a reveals multiple assemblies of bright spots distributed evenly over the surface of the supporting TiO 2 nanoparticle (more images in Figure S5). Considering the strong Z-contrast difference between W and Ti along with the fact that the approximate lateral size of each assembly is around 1 nm, these formations likely correspond to individual POM clusters decorating the surface, while each of the spots − to a single W ion. Figure 4b shows a close-up micrograph and confirms the presence of multiple POM anions attached to the surface in various orientations, in line with the rough and uneven surface of the TiO 2 support. The Fourier-filtered image of the inset area overlaid with a structural model of a {Co III Co II W 11 } cluster demonstrates an excellent match, which further corroborates the successful heterogenization and intact structure of the POMs after attachment. A series of STEM-EDS maps acquired with atomic resolution (Figure 4c) verify the chemical composition of the clusters and confirm the presence of Co centers (additional elemental maps in Figure  S6).
The amount of {Co III Co II W 11 } on the TiO 2 surface was quantified by ICP-MS and TXRF (Table 1). Both methods result in an average loading of 14 wt % (details in the SI). Moreover, a ratio of 1:7.4 for POM to APTES can be calculated, suggesting that on average, seven NH 2 groups are available to accommodate each of the {Co III Co II W 11 } anions.
Next, we estimated the theoretical maximum loading for {Co III Co II W 11 } on TiO 2 , considering a monolayer adsorption model (details in the SI), by taking 1.9 nm 2 as the footprint of a Keggin-structured POM 59 and 78.63 m 2 /g as the surface area of TiO 2 measured according to Brunauer−Emmett−Teller (BET) theory. 60 This yields a theoretical value of 20.8 wt %, whichconsidering additional repulsion between the closepacked POM anions at high loadingscan be seen as an upper loading limit. The experimentally obtained value of 14 wt % agrees well with a dense monolayer of POM clusters on TiO 2 and is in line with the homogeneous distribution of the clusters revealed by the STEM and EDS in Figure 4.
The binding modes between POM and APTES were investigated by XPS. Figure 5a shows the W 4f spectrum of {Co III Co II W 11 }-APTES-TiO 2 with the characteristic peaks for W 4f 5/2 and W 4f 7/2 at 37.4 and 35.4 eV, respectively. Note that these peaks are slightly redshifted in comparison to the bare POM ( Figure S7b), which indicates an electronic communication between the clusters and the support. Moreover, the peak intensity ratio of W 4f 5/2 to W 4f 7/2 is considerably higher for the composite (1.2) compared to that of the bare POM (0.75 61 ). This can be explained by an overlapping contribution of the Ti 3p peak of the TiO 2 substrate ( Figure S7b). Note that we can also exclude the presence of WO 3 , which shows peaks at 38.5 and 36.1 eV. 61 Figure 5b shows the Ti 2p spectrum of the composite, which only contains a characteristic Ti(IV) signal; however, we observe a slight shift of the peak maximum to lower binding energies compared to bare TiO 2 . This shift is indicative of surface modification caused by the presence of electron-rich Ti ions and is in line with the formation of Ti−O−Si bonds at the APTES-TiO 2 surface. 40 It can also be observed that the Ti peak shows a small decrease in intensity, which could be related to the existence of the surface-attached APTES and {Co III Co II W 11 } layers that limit the mean free path of the photoelectrons. 51 While a reliable interpretation of Co 2p spectra is complicated by the low intensity of the signal ( Figure  S7e), the N 1s spectra of APTES-TiO 2 and {Co III Co II W 11 }-APTES-TiO 2 provide useful insights (Figure 5c). The N 1s peak at lower binding energy is clearly diminished in the composite compared to APTES-TiO 2 , which implies a decrease in the number of free NH 2 upon cluster loading, in line with the proposed attachment mode via H 2 N: → Co(II) bonds. The N 1s spectrum of the composite also shows a shift   Figure 6b provides a detailed look at the deactivation: the instant TOF value drops 10-fold within the first 10 min of illumination. This exemplary WOC experiment is representative of the unstable nature of many homogeneous photosystems 62,63 and, in the particular case of the {Co III Co II W 11 }/[Ru(bpy) 3 ] 2+ couple, can likely be related to the rapid degradation of the photosensitizer. 18 We further proceeded with the evaluation of the photocatalytic performance of the {Co III Co II W 11 }-APTES-TiO 2 composite. In this set of experiments (details in the Experimental Section), the role of the photosensitizer is taken by the support TiO 2 . An optimal comparison between the homogeneous and heterogeneous WOC systems in terms of TOF values required the presence of a similar number of POM clusters in the reaction solution (e.g., on the TiO 2 surface) in both cases. Considering the {Co III Co II W 11 } loading of 14 wt %, 0.046 μmol of POM clusters are available for WOC in a single catalytic run (details in the SI). This number corresponds well to the 20 μM {Co III Co II W 11 } solution used in the homogeneous system, which amounts to 0.04 μmol of the POMs.  Figure S8) or APTES-TiO 2 reference (that shows almost negligible WOC activity), the {Co III Co II W 11 }-APTES-TiO 2 composite exhibits enhanced O 2 evolution, yielding 1 μmol of O 2 after 60 min of illumination. This result suggests two important points: first, the heterogenized POM clusters promote the reaction of interest, that is, they act as WOC co-catalysts; second, TiO 2 generates free charge carriers able to take part in the redox reactions of interest, that is, it acts as a photosensitizer (further experiments in the SI). In addition to theseand in contrast to the homogeneous WOC casea linear O 2 evolution is obtained for the {Co III Co II W 11 }-APTES-TiO 2 composite, which corresponds to a constant reaction rate. Long-term experiments further validate this point: no saturation in WOC activity is observed for at least 10 h ( Figure S9), while the amount of generated O 2 approximates a TON of 82.5, one of the highest values among other Co-based POMs tested under similar homogeneous WOC conditions. 17 Considering the number of {Co III Co II W 11 } co-catalyst clusters on TiO 2 , the WOC performance of the {Co III Co II W 11 }-APTES-TiO 2 can be translated to an average TOF of around 0.39 min −1 . Figure 6b plots instant TOF values of both photosystems and shows that despite the initial TOF of the POM cluster in the solution being at least 4-fold higher, a strong stabilization effect is in place, which renders the {Co III Co II W 11 }-APTES-TiO 2 composite a superior light-driven WOC already after 5 min of turnover conditions. The strongly improved long-term WOC performance of the heterogenized POM photosystem can be related to (i) the replacement of the molecular [Ru(bpy) 3 ] 2+ with a more stable and robust inorganic absorber, which allows for a constant photosensitization rate as well as (ii) the stabilization of the individual POM clusters on the support, which helps prevent cluster aggregation and the deactivation of their active sites.
The initially higher TOF values of the homogeneous system (1.56 min −1 vs 0.39 min −1 ) can be explained by the complex interplay between photoactivation and catalytic function. In the homogeneous system, we use a comparably high concentration of [Ru(bpy) 3 ] 2+ (photosensitizer-to-POM ratio of 40:1), whichin concert with a higher POM mobility in solutionaccounts for more efficient light absorption and charge transfer upon collision with the POM molecules. This photosystem, however, also deactivates rapidly. In contrast, the POM clusters in the heterogeneous system are fixed in position at the APTES/TiO 2 surface and rely on the charge carriers that are provided by the TiO 2 photosensitizer. It is likely that either the photoexcitation in TiO 2 or the charge transfer to the POMs are limiting factors that define the suboptimal TOF values obtained for the {Co III Co II W 11 }-APTES-TiO 2 composite. Future studies using supports with more efficient charge excitation and transfer dynamics will be required to achieve higher WOC efficiency values.
Post-catalytic characterization of the heterogenized composite with a set of complementary techniques confirms the integrity of the {Co III Co II W 11 }-APTES-TiO 2 attachment; however, a certain degree of leaching of the POM clusters could also be observed ( Figure S10). In light of the stable WOC performance recorded for the heterogenized photosystem, we suggest that either the leaching takes place during the initial stage of the photocatalytic run, affecting only some of the clusters (possibly defined by the low denticity of APTES attachment or high coverage of the POM anions in some areas) or the detached {Co III Co II W 11 } clusters continue to contribute to the WOC activity (details in the SI).

WOC Mechanism
The general mechanism of water oxidation on TiO 2 surface when in the presence of Na 2 S 2 O 8 oxidantcan be described as shown in Figure 7b by the following reactions: Water oxidation to produce O 2 (eq 4) can take place via two major pathways: 64  To understand whether the oxidation by photogenerated holes leads to the formation of OH radicals (·OH) or O 2 (competing steps, eqs 3 and 4), we employed photoluminescence (PL) emission spectroscopy and used terephthalic acid (TA) as the fluorescence probe that can effectively trap ·OH. 65 Figure 7b shows that the reaction of ·OH with TA forms highly fluorescent 2-hydroxyterephthalic acid (TA-OH); its quantification by means of PL thus provides an estimate of the amount of ·OH produced by the photocatalyst (further details in the Experimental Section and the SI). The PL emission studies were conducted for TiO 2 and the {Co III Co II W 11 }-APTES-TiO 2 composite. Figure 7a shows the TA-OH emissions from the reaction solution supernatants (with a peak centered at 425 nm). In the case of TiO 2 , this peak has a much higher (4-fold) intensity compared to that of {Co III Co II W 11 }-APTES-TiO 2 . This indicates that the generation of ·OH on the TiO 2 surface (Figure 7b and eq 3) is suppressed in the presence of attached POM clusters, hence leading to more efficient utilization of holes for water oxidation via direct hole transfer (Figure 7b and eq 4). In the case of bare TiO 2 , ·OH generation occurs more effectively (Figure 7b and eq 3), which in turn inhibits O 2 formation and leads to a lower WOC activity. Overall, this data suggests that the presence of the POM on the surface allows for more efficient hole extraction and utilization toward water oxidation, which manifests the role of {Co III Co II W 11 } as a co-catalyst.

■ CONCLUSIONS
We report a novel photocatalytic system composed of an allinorganic molecular Co-containing [Co III Co II (H 2 O)W 11 O 39 ] 7− Keggin-type POM cluster immobilized onto the TiO 2 surface using an APTES linker. The composite was thoroughly characterized with regard to its structure and composition using ATR-FTIR, powder XRD, 29 Si-ss-NMR, XPS, ICP-MS, and TXRF to reveal POM loadings, surface coverage, and distribution as well as to probe the POM/APTES and APTES/ TiO 2 binding modes. High-resolution HAADF-STEM images were used to visualize the [Co III Co II (H 2 O)W 11 O 39 ] 7− attachment as well as to confirm its structural integrity on the support surface. Photocatalytic WOC studies were conducted under heterogeneous and homogeneous conditions. We demonstrated that in comparison with bare TiO 2 , the composite performs as a more efficient photocatalyst for light-driven water oxidation, which manifests the role of heterogenized {Co III Co II W 11 } as a WOC co-catalyst. In contrast to the homogeneous WOC performance of this POM, the heterogenized clusters showed strongly improved long-term activity without any apparent deactivation for at least 10 h. This stable WOC performance is achieved without the presence of any external photosensitizer, which implies that TiO 2 not only acts as a support for POM anchoring but also as a stable light absorber. Furthermore, we uncovered the WOC mechanism utilizing PL spectroscopy. Our data revealed that the more active WOC performance of the {Co III Co II W 11 }-APTES-TiO 2 over TiO 2 is related to the effective hole extraction by the POM cluster, which promotes charge separation and allows for a more efficient H 2 O oxidation by direct hole attack at the POM site. With the combination of the stable photocatalytic WOC performance with the ability for catalyst recovery, this novel photocatalytic system represents an example of an effective light-driven WOC that combines the advantages of its heterogeneous and homogeneous counterparts. We envision future studies aiming to optimize the process of photosensitization and implement visible-light active supports, which will maximize the benefits of this photosystem.