Mo-doped TiO 2 photoanodes using [Ti 4 Mo 2 O 8 (OEt) 10 ] 2 bimetallic oxo cages as a single source precursor †

Photoelectrochemical solar water splitting is a promising and sustainable technology for producing solar fuels such as clean hydrogen from water. A widely studied photoanode semiconductor for this application is TiO 2 , but it su ﬀ ers from a large band gap (3.2 eV) and fast recombination of electrons and holes. Herein, we present a novel, facile and rapid strategy to develop Mo-doped TiO 2 (Mo:TiO 2 ) mixed anatase – rutile photoanodes using [Ti 4 Mo 2 O 8 (OEt) 10 ] 2 bimetallic oxo cages as a single source precursor. These cages dissolved in tetrahydrofuran deposit by spray pyrolysis at 150 (cid:1) C forming ﬁ lms with hierarchical porosity on the micrometer and nanometer scale. XPS, EDXS and UV-Vis spectroscopy reveal Mo atoms evaporate during annealing in air at temperatures 650 – 800 (cid:1) C, contributing to the formation of nanostructures and porosity. XPS depth pro ﬁ ling, XRD, EDXS, Raman, and electron paramagnetic resonance indicate that the remaining Mo atoms are well spread and incorporated in the TiO 2 lattice, at interstitial or substitutional sites of the rutile or anatase phases depending on the annealing temperature. Photocurrent measurements show that Mo:TiO 2 photoanodes optimized at 700 (cid:1) C outperform a TiO 2 photoanode prepared in a similar manner by a factor of two at 1.23 V RHE . Finally, UV-Vis spectroscopy, conduction and valence band calculations, and incident-to-photon e ﬃ ciency measurements show these Mo:TiO 2 photoanodes possess a narrower band gap than TiO 2 and higher e ﬃ ciency in the visible light range (5% at 400 nm). These outcomes open a new avenue in the exploitation of titanium oxo cages and advance the development of photoelectrodes for water splitting and energy applications.


Introduction
The abundant solar energy, 1.3 Â 10 5 TW per year reaching the Earth's surface, can be utilized to produce hydrogen fuel by splitting water, offering an excellent and sustainable alternative to fossil fuels.2][3] However, TiO 2 -based PEC cells are still far from commercialization mainly due to both fast recombination of photogenerated electrons and holes in TiO 2 and its large band gap (3.2 eV for the anatase phase) that results in a reduced use of the solar spectrum.One strategy to overcome its limitations is doping it with transition metals.The metal doping increases the solar light absorption of TiO 2 and performance by incorporating additional energy levels within the band gap of the semiconductor. 4,5Nevertheless, this only seems to occur when there is an actual substitution of Ti atoms in the TiO 2 lattice structure with the external metal (known as substitutional doping), that reduces the band gap of the material without compromising the surface of the photocatalyst. 6therwise, if metals are simply impregnated on the surface of the semiconductor they may result in electron-hole recombination and blocking of reaction sites. 6,7][10][11] For instance, Yan et al. reported an enhancement in the PEC performance for Ta-doped TiO 2 nanotube photoanodes in comparison to pristine TiO 2 . 8This enhanced PEC performance was attributed to a decrease in the band gap, lower charge transfer resistance and higher charge carrier density.Similarly, Zhao et al. successfully demonstrated that Fe-doped TiO 2 results in a better PEC performance than pristine TiO 2 , attributed to a smaller band gap and thus a better light absorption. 9][14][15][16] In 2015 Zhang et al. reported Mo-doped TiO 2 photoanodes prepared using two steps of Ti foil anodization and a third one of hydrothermal doping, obtaining an improved PEC performance over pure TiO 2 photoanodes. 17The improvement was attributed to a decrease in the recombination of electrons and holes in the mixed phase of anatase and rutile and to better light absorption.
9][20][21][22][23] Some of these cages, also called clusters, have successfully been used as precursors for photocatalysts or electrocatalysts, for example using graphene oxide as a sacricial template or impregnating WO 3 photoanodes. 20,24However, their wide adoption as single source precursors still requires nding facile and effective approaches that overcome their instability in water or humid environments.
In this publication, we report for the rst time a facile and rapid synthesis of porous Mo-doped TiO 2 photoanodes using a [Ti 4 Mo 2 O 8 (OEt) 10 ] 2 HMTO cage as a single source precursor.Stability of the cages during the deposition is ensured using anhydrous tetrahydrofuran as solvent and spray pyrolysis as deposition method.We reveal that upon deposition, the calcination at temperatures above 600 C induces sublimation of Mo atoms adding porosity and nanostructured features to remaining TiO 2 , while the rest of Mo atoms effectively occupy substitutional or interstitial sites in the TiO 2 lattice.The resulting Mo:TiO 2 photoanode optimized at 700 C outperforms by a factor of two a pure TiO 2 photoanode prepared in a similar manner in PEC solar water splitting.The results herein presented therefore reveal new strategies to develop efficient photoelectrodes for water splitting applications using Mo as both a sacricial agent and dopant and opens an avenue to exploit HMTO cages as single source precursors using spray pyrolysis.

Preparation of Mo:TiO 2 and TiO 2 lms
Mo-doped TiO 2 (abbreviated as Mo:TiO 2 ) and pure TiO 2 lms to be studied as photoanodes were prepared using a manual spraypyrolysis system (Clarke CAB3P) on FTO-ABS.The solutions employed for the spray pyrolysis deposition were prepared in an argon atmosphere using a Schlenk line.Nevertheless, the actual spray pyrolysis deposition process was conducted in air, but vessels were kept closed when possible.The precursor solution for the Mo:TiO 2 photoanodes was prepared by dissolving [Ti 4 -Mo 2 O 8 (OEt) 10 ] 2 (0.96 g) in anhydrous tetrahydrofuran (20 mL).The precursor solution for the preparation of pure TiO 2 was carried out following an established method. 26Briey, a 0.2 M solution of TTIP was prepared by diluting TTIP and AcAc in a 3 : 2 volumetric ratio and topping up with absolute ethanol in order to obtain 0.2 M solution of TTIP.
The spray pyrolysis deposition of previous precursor solutions to prepare Mo:TiO 2 and TiO 2 photoanode lms was conducted as follows.In a rst stage, FTO-ABS was pre-heated on a hot plate at 150 C. Secondly, the as-prepared precursor solutions were sprayed on top of the pre-heated FTO-ABS, at a constant distance of ca. 5 cm from the surface of the FTO-ABS to the spray pyrolysis nozzle.Three deposition layers were performed per sample.Finally, the as-prepared Mo:TiO 2 and TiO 2 lms were annealed at different temperatures between 450 and 800 C for 2 h in air at a ramp rate of 10 C min À1 .This range of temperatures was chosen to ensure full conversion of the precursor to the metal oxide and to evaluate the effect that different annealing conditions might have on the photoanode performance.The resultant photoanodes were denoted as Mo:TiO 2 -### and TiO 2 -###, where ### is the corresponding annealing temperature ( C).
Cobalt phosphate (Co-P i ) loading on Mo:TiO 2 -700 photoanodes was carried out by photo-electrodeposition. 27,28 The electrolyte consisted of 0.05 mM of cobalt nitrate in 0.1 M potassium phosphate buffer (pH ¼ 7) and the applied potential 1V RHE was kept for 20 s under simulated sunlight (AM 1.5G, 100 mW cm À2 ) from a ltered 300W Xenon lamp source.Illumination was directed towards the back of the FTO-ABS working electrode.CoFeO x was deposited on Mo:TiO 2 -700 by electrodeposition. 29The electrolyte consisted of 10 mM FeCl 3 $6H 2 O, 16 mM CoCl 2 and 0.1 M NaOAc dissolved in deionized water.The pH of the solution was 4.90.The deposition was carried out by positively sweeping the voltage from 1.1 to 1.4 V Ag/AgCl three times.In both depositions, a Compactstat.potentiostat (Ivium Technologies) was used and an electrochemical cell consisting of a Pt counter electrode, a silver chloride (Ag/AgCl/3.5 M KCl) reference electrode, and a Mo:TiO 2 -700 photoanode as working electrode.
Characterization 13 C NMR spectroscopy was conducted at room temperature using a 500 MHz Agilent Propulse spectrophotometer.Samples were dissolved in dried deuterated benzene (C 6 d 6 ).Elemental analysis was performed on a Carlo Erba Flash 2000 Elemental Analyser.Field emission scanning electron microscopy images (FE-SEM) were acquired using a JEOL FESEM6301F and energy dispersive X-ray spectroscopy (EDXS) was carried out in a SEM 6480LV equipped with a high sensitivity Oxford INCA X-Act SDD X-ray detector.X-Ray diffraction (XRD) patterns were collected in the 2 theta range 10-80 with a Bruker AXS D8 Advance using Cu Ka (0.154 nm) radiation with a total integration time of 960 s.Raman spectroscopy was carried out on a Renishaw inVia system using a 532 nm diode-pumped solid state laser (DPSS) manufactured by Cobolt.The laser beam was focused onto the sample using a 50Â long distance objective.X-Ray photoelectron spectroscopy (XPS) depth proling was performed using a ESCALAB 250 Xi instrument manufactured by Thermo Fisher Scientic.Measurements were carried out using a monochromated Al Ka X-ray source with an energy of 1486.68 eV.The X-ray spot size was of 900 mm and the pass energy for the high resolution scans was of 50 eV.The depth prole for the sample was obtained by etching the surface of the sample with an Ar + ion gun (2000 eV, high current) for different times (0, 60, 180 and 420 s).C 1s XPS spectra was used as an internal charge correction.Samples studied via electron paramagnetic resonance (EPR) spectroscopy were evacuated at 393 K for over 12 h to reduce the inuence of physisorbed water.Samples were maintained under static vacuum (10 À5 mbar) for the duration of the experiments.For EPR analysis, powder samples were prepared by drying and calcining in air at 650, 700 and 800 Cin a porcelain dish a solution of 0.96 g of [Ti 4 Mo 2 O 8 (OEt) 10 ] 2 in 20 mL of anhydrous tetrahydrofuran.High resolution transmission electron microscopy (HRTEM) images were obtained using a JEOL JEM-2100Plus microscope.Ultraviolet-visible (UV-Vis) spectra were collected in a Cary 100 diffuse reectance UV-Vis spectrophotometer.
(Photo)Electrochemical measurements (Photo)electrochemical performance of photoanodes was evaluated using a CompactStat.potentiostat (Ivium Technologies).Photocurrents were measured under simulated sunlight (AM 1.5G, 100 mW cm À2 ) from a ltered 300W Xenon lamp source (Lot Quantum Design) or under UV illumination (365 nm, 3.6 mW cm À2 ) from a ModuLight IM3412 LED light (Ivium Technologies).PEC cells consisted of three electrodes with Pt as the counter electrode, silver chloride (Ag/AgCl/3.5 M KCl) as the reference electrode and as-prepared photoanodes as the working electrodes.
Electrochemically active surface area (ECSA) of photoanodes was investigated using cyclic voltammetry (CV), scanning from 0 to 0.17 V Ag/AgCl at scan rates between 10 and 250 mV s À1 ,i n 1 M KOH solution (pH ¼ 13.7).ECSA is proportional to the double layer capacitance (C dl ), which is estimated from the slope of the plot Dj vs. scan rate and dividing by two. 30Dj is equal to (j a -j c ), where j a and j c are the anodic and cathodic current densities, respectively, in this case taken at 0.1 V Ag/AgCl in the CV scans. 31onduction and valence band (CB & VB) positions were measured from CV curves recorded in acetonitrile containing 0.1 M of tetrabutylammonium hexauorophosphate (TBAPF 6 ) at a scan rate of 50 mV s À1 and using the following formula: [32][33][34] where E is the onset of the redox potential and E 1/2 is the formal potential of Fc/Fc + system (0.43 V Ag/AgCl ). 35hotoelectrochemical performances of photoanodes were carried out in a 1 M KOH (pH ¼ 13.7) electrolyte solution.Illumination was directed towards the back of the FTO-ABS working electrode and a mask was placed on top of the photoelectrode to dene the illuminated area.Photocurrent-time curves were performed at an applied bias of 1.23 V vs. the reversible hydrogen electrode (V RHE ).Photocurrent-potential curves were recorded at a scan rate of 20 mV s À1 .The measured Ag/AgCl potentials (E Ag/AgCl ) were converted to RHE potentials ðE RHE Þ following the Nernst equation: where E Ag=Agcl is 0.205 V at 25 C (3.5M KCl).Photoelectrochemical impedance spectroscopy (PEIS) was carried out under simulated sunlight (AM 1.5G, 100 mW cm À2 ) at a direct current (DC) potential of 1.23 V RHE and an alternating current (AC) potential frequency range of 100 000-0.01Hz with an amplitude of 5 mV.Incident photon-to-current efficiency (IPCE) measurements were calculated using the same Xe light source and a triple grating Czerny-Turner monochromator.The intensity of monochromatic light was measured at the working electrode position with SEL033/U photodetector (International Light Technologies).The following equation was used to calculate the IPCE values: 36 where j is the photocurrent density measured under single wavelength (l) light illumination and P mono is the incident irradiation power.Oxygen (O 2 ) measurements were conducted using a Pyroscience FireStingO2 bre-optic oxygen meter combined with a TROXROB10 oxygen probe, together with a TDIP temperature sensor to give automatic compensation for minor uctuation in the PEC cell temperature.O 2 readings were recorded every 10 min for ca.410 min.The probe was tted into the headspace of the airtight PEC cell.The PEC cell was purged with a N 2 ow to ensure air O 2 removal before the irradiation started.The measurements were carried out under simulated sunlight (AM 1.5G, 100 mW cm À2 ) with an applied bias of 1.23 V RHE .Light was irradiated for 340 min.The faradaic efficiency was calculated by dividing the measured amount of evolved O 2 by the theoretical amount of expected O 2 for measured photocurrents (assuming 100% faradaic efficiency).See more details in ESI.†

Results and discussion
The showing aggregated islands evenly distributed on top of FTO-ABS support.At slightly higher magnication (Fig. 2, 2nd row), a large amount of micrometer cavities on Mo:TiO 2 lms are observed.These cavities contribute to an increase of surface area, which in turn must result in more active sites for the PEC oxygen evolution reaction.At even higher magnication (Fig. 2, 3rd row), it can be observed that well-dened nanostructures form as the annealing temperature increases.A highly smooth and ne surface is present before calcination (Fig. 2i), but aer calcination, grain-rice-shaped nanostructures appear creating nanometer-size porosity and extra surface area (Fig. 2j-l).
SEM images of the pure TiO 2 lm at different magnications are shown in Fig. S2 (ESI †).Unlike Mo:TiO 2 lms, TiO 2 -650 (prepared with TTIP-AcAc) does not show either the presence of cavities nor nanostructures.This comparison therefore reveals the benets of using [Ti 4 Mo 2 O 8 (OEt) 10 ] 2 cages as a precursor.Their decomposition and transformation to Mo-doped TiO 2 during the spray pyrolysis and posterior calcination leads to the formation of cavities, nanostructures and porosity.Cavities must result from the drying step upon spray pyrolysis at 150 C.More cavities may form during the calcination and oxidation of the cages' carbon content up to 400-500 C. Above this temperature, the grain-rice-shaped nanostructures and its associated porosity must result from the sintering of TiO 2 and especially from the very likely sublimation of Mo atoms.Previous reports have shown that MoO 3 sublimes above 700 C. 37 The sublimation of Mo atoms was conrmed by atomic quantication from XPS data (Fig. 3a).The amount of Mo in the lms decreased with temperature.For example, at the top surface (XPS-etching time 0 s) it went from 10.6 at% for Mo:TiO 2 -650 to 4.9 and 4.3 at% for Mo:TiO 2 -700 and Mo:TiO 2 -800, respectively, indicating Mo sublimes within the temperatures of study.Accordingly, the percentage of carbon (C) decreases with temperature and that of Ti increases.Carbon is present from solvents and cage ethoxides and deposition of volatile organic compounds during storage.The Mo sublimation was further conrmed by an experiment of simply heating MoCl 5 up to 700 C in air, which showed its complete sublimation.Therefore, Mo atoms work as pore formers, sacricial agents that increase the porosity in the lms and allow their nanostructuring.
The XPS depth proling indicated that Mo is homogeneously distributed at different depths, but some gradient is formed at highest temperatures of 700 and 800 C, with more C and Mo present at the surface (Fig. 3a).This may result from the gasi-cation and sublimation taking place and accumulation at the top during the process.
ECSA measurements are shown in Fig. 3b and the corresponding CV curves in Fig. S3 (ESI †).In such measurements, the slope of the current density vs. scan rate can be related to the double layer capacitance, which is directly proportional to the ECSA. 31Based on the obtained results, Mo:TiO 2 -700 possesses the highest surface area (C dl ¼ 0.10 mF cm À2 ), whereas TiO 2 -650 is the sample with the smallest surface area (C dl ¼ 0.01 mF cm À2 ).These results agree well with SEM images, where grainrice-shaped nanostructures and porosity are observed for Mo:TiO 2 -700 (Fig. 2k) and very at lms for TiO 2 -650 (Fig. S2,  ESI †).Among all Mo:TiO 2 photoanodes, Mo:TiO 2 -800 is the sample with the smallest surface area, due to the largest grains formed at highest temperature (Fig. 2l).
SEM-EDXS analysis was also carried out to evaluate the distribution of Ti, O and Mo atoms at the surface of the lms.).The diffraction of the rutile phase increases and that of anatase decreases with annealing temperature (Fig. 4a), indicating that the conversion of anatase to rutile was promoted at highest temperatures. 38Actually, Mo:TiO 2 -800 only shows diffraction peaks indexed to rutile TiO 2 , due to the high calcination temperature employed.No diffraction peaks corresponding to any Mo phase such as MoO 3 are observed in any of the samples, which suggests that Mo 6+/5+ could be incorporated into the lattice of TiO 2 .MoO 2 was unlikely to be formed since phase transformation from tetragonal MoO 2 to orthorhombic MoO 3 occurs at temperatures above 350 C. 39 In Fig. 4b an expansion of the (1 0 1) and (1 1 0) diffraction peaks of anatase and rutile, respectively, of the different Mo:TiO 2 and pure TiO 2 lms is shown.A shi of the anatase and rutile diffraction peaks towards lower angles for Mo:TiO 2 -700 and Mo:TiO 2 -800 in comparison to pure TiO 2 is observed, which indicates incorporation of Mo 6+ or Mo 5+ atoms in the TiO 2 lattice structure upon exposure to those temperatures. 13he reported ionic radius for Mo 6+ and Mo 5+ are 0.0620 and 0.0610 nm, respectively, whereas the Ti 4+ ionic radius is 0.0605 nm. 13,40,41The similarity in the ionic radius of these ions facilitates the aliovalent substitution of Ti 4+ atoms for Mo 6+/5+ in the TiO 2 lattice, and due to the slightly larger size of Mo 6+/5+ in comparison to Ti 4+ a shi in the diffraction pattern towards lower angles is observed.However, Mo:TiO 2 -650 shows a smaller shi towards lower angles, suggesting that at lower temperatures the majority of Mo 6+/5+ atoms are not occupying Ti 4+ sites in the TiO 2 lattice structure.Instead, Mo 6+/5+ must be distributed on the surface of TiO 2 or occupying interstitial sites within the TiO 2 lattice, without distorting the crystal structure of either anatase or rutile TiO 2 .In fact, this may also suggest that Mo 6+ atoms could be present in the form of MoO 3 that could be either highly dispersed on the TiO 2 surface or of amorphous structure, and therefore not detectable by XRD analysis.
Raman spectroscopy was also carried out to further verify the presence of rutile TiO 2 , anatase TiO 2 and the substitution of Mo for Ti atoms in the lattice structure of TiO 2 .Fig. 4c shows the Raman spectra of all lms.The presence of anatase-TiO 2 and rutile-TiO 2 is conrmed by the sharp peaks observed in all Raman spectra, with the exception of Mo:TiO 2 -800 for which only rutile TiO 2 is observed, in agreement with the XRD patterns (Fig. 4a and b).The sharp bands at ca. 145, 395, 515 and 635 cm À1 correspond to Raman active modes of anatase TiO 2 and bands at 230, 445 and 610 cm À1 to Raman active modes of rutile TiO 2 . 42,43Interestingly, some bands ascribed to the presence of Mo are also observed.For instance, in the region between 870-970 cm À1 there are bands belonging to hydrated terminal Mo-O and Mo-O-Mo vibrations. 44,45The presence of crystalline MoO 3 can be discarded since its characteristic main bands at 996, 820 and 666 cm À1 are not observed, which is in accordance with XRD analysis.Moreover, the weak bands observed in the range 100-200 cm À1 are attributed to bending modes of Mo-O-Mo. 46This veries the incorporation of Mo atoms into the lattice of TiO 2 , suggesting the presence of Ti-O-Ti, Mo-O-Mo and Mo-O-Ti bonds in Mo:TiO 2 . 47The Mo Raman vibrations are most evident for Mo:TiO 2 -700 and drastically decrease for Mo:TiO 2 -800, indicating that at 700 C the Mo incorporation into the oxide crystalline structure is at its maximum.This temperature dependence is attributed to evaporation of the Mo species, which as previously shown, sublime at relatively low temperatures.Along the same lines, the Mo Raman vibrations for Mo:TiO 2 -650 are relatively weak, which further conrms that the vast majority of Mo atoms are not incorporated in the TiO 2 lattice structure, instead they are impregnated on the surface of TiO 2 or distributed in interstitial sites of the TiO 2 structure, which agrees well with XRD patterns.
The composition and chemical state of Mo:TiO 2 lms were further characterized by XPS.Fig. 5 shows the XPS high resolution spectra at the surface of Mo:TiO 2 samples.Ti 2p resolution spectra show the two characteristic peaks of Ti 4+ in TiO 2 at 458 and 464 eV in the Mo:TiO 2 lms. 42,48The O 1s spectra for Mo:TiO 2 lms are shown in Fig. 5b.The peak at lower binding energies mainly corresponds to crystal lattice oxygen O-Ti 4+ , whereas the smaller peaks at higher binding energies correspond to hydroxyl groups or adsorbed water on the surface i.e.Ti-OH or Mo-OH. 42,48Fig. 5c shows the high-resolution XPS surface spectra of Mo 3d.0][51] An additional doublet appears at lower binding energies for all Mo:TiO 2 lms, attributed to the presence of some Mo 5+ centers in the lms. 52XPS O 1s and Ti 2p spectra of TiO 2 -650 are shown in Fig. S5 (ESI †).The binding energies for Ti 2p and O1s are in agreement with Mo:TiO 2 samples.
Electron paramagnetic resonance (EPR) spectroscopy was undertaken (at 120 K) to investigate the nature of the Mo doping within the series of TiO 2 lattice structures.EPR spectroscopy only detects the presence of paramagnetic species, thus the Mo 5+ species are observed whereas there is no detection of Mo 6+ centers.The EPR data (Fig. 6) indicates the presence of Mo 5+ in all the lms, in agreement with the XPS analysis.Variation in g values for the different samples demonstrated the existence of multiple Mo 5+ species.As an example, the EPR spectrum of Mo:TiO 2 -650 indicated the presence of MoO 3+ species on the surface, characterized by an axial g-tensor (g t ¼ 1.932 and g k ¼ 1.886) and corresponding weak hyperne satellite lines originating from coupling of the unpaired electron to the two nuclear spin active isotopes of molybdenum ( 95,97 Mo, both with spin I ¼ 5/2 and total natural abundance of 25.5%;A t ¼ 112 MHz and unresolvable A k ).Additional contributions from bulk Mo 5+ (g 1 ¼ 1.944, g 2 ¼ 1.944, g 3 ¼ 1.839;A 1 ¼ 198, A 2 ¼ 75 and A 3 ¼ 93 MHz) and a small contribution from Mo 5+ in substitutional anatase lattice positions (g 1 ¼ 1.917, g 2 ¼ 1.828, g 3 ¼ 1.828) was also detected in the Mo:TiO 2 -650 sample.In contrast, the EPR spectra obtained for both the Mo:TiO 2 -700 and Mo:TiO 2 -800 samples conrmed the existence of substitutional and interstitial doping in the TiO 2 rutile lattice.The variation in Mo sites displayed via EPR for 2 -650 and both Mo:TiO 2 -700 and Mo:TiO 2 -800 conrms that substitutional doping mainly occurs in rutile TiO 2 .
The particle size and crystallinity of Mo:TiO 2 were evaluated using TEM and SAED-TEM.Fig. 7 shows the corresponding SAED diffraction patterns, bright-eld TEM and HRTEM images of Mo:TiO 2 .To perform these analyses, a few milligrams of lm deposited on FTO-ABS was scratched and dispersed in ethanol followed by TEM grid loading.The SAED-diffraction patterns (Fig. 7 S6 (ESI †).In all Mo:TiO 2 samples, the particles possess a well-dened particle shape in comparison to pure TiO 2 .Nevertheless, a large variability of particle sizes is observed in Mo:TiO 2 -650 suggesting an insufficient annealing temperature for the formation of uniform particles.Mo:TiO 2 -800 has the highest particle size due to aggregation and sintering of particles at high calcination temperatures. 53,54RTEM images are shown in Fig. 7m-p.These images also reveal the highly crystalline character of Mo:TiO 2 and pure TiO 2 .The measured lattice spacing for the TiO 2 -650 sample agrees well with XRD analysis, since lattice spacings corresponding to the (101) diffraction plane of TiO 2 anatase and (101) of TiO 2 rutile are observed.In line with this, lattice spacings for Mo:TiO 2 also agree well with XRD and Raman analysis.More precisely, Mo:TiO 2 -650 and Mo:TiO 2 -700 show diffraction planes corresponding to both TiO 2 rutile and anatase crystalline phases, whereas only rutile TiO 2 particles are observed in Mo:TiO 2 -800.
UV-Vis spectroscopy Tauc plots of Mo:TiO 2 and pure TiO 2 lms are depicted in Fig. 8a and the corresponding absorption spectra are shown in Fig. S7 (ESI †).All Mo:TiO 2 samples exhibit lower band-gap energy values ranging from 2.6 to 2.7 eV compared to TiO 2 -650, ca.3.1 eV, being in accordance with literature reports. 7The overall band-gap reduction for all Mo:TiO 2 is attributed to the incorporation of Mo 6+/5+ centers in  the TiO 2 lattice structures, contributing to the formation of a shallow donor energy level below the CB of TiO 2 . 47,55Two distinguished slopes are observed for Mo:TiO 2 -800.The lower energy one is related to the band gap, while the higher one we assign it to an Urbach tail caused by sample disorder (i.e.Ti-O, Ti-O-Mo bond breaking due to anatase transformation to rutile and Mo sublimation). 56he CB and VB position of TiO 2 -650 and Mo:TiO 2 photoanodes were determined using CV curves (Fig. S8, ESI †).Detailed information of the electrochemical characteristics of TiO 2 and Mo:TiO 2 are shown in Table S1 (ESI †).A schematic diagram of the relative position of CB and VB energy levels for TiO 2 -650 and Mo:TiO 2 samples is shown in Fig. 8b.Interestingly, in all Mo:TiO 2 samples the CB offset is narrowed down from approximately À3.01 eV for TiO 2 -650 to ca.À4.2 eV for Mo:TiO 2 .This further conrms that the incorporation of Mo 5+/6+ atoms at the TiO 2 lattice structure reduces the overall band-gap of TiO 2 .The electrochemical band gap values obtained from CV curves are also shown in Table S1 (ESI †).The difference between the electrochemical band gap and optical band gap of TiO 2 and Mo:TiO 2 photoanodes is of 0.1-0.5 eV, which falls within the range of error. 57he photocurrent density (at 1.23 V RHE ) as a function of annealing temperature for front illumination (via the  electrolyte-lm interface) and back illumination (via the ABS) was evaluated for Mo:TiO 2 and pure TiO 2 lms and the results are shown in Fig. 9a.Photocurrent performances are higher when lms are illuminated from the back, which indicates the porous photoanodes have a sufficient thickness.Fig. 9a also indicates that the optimal annealing temperature for pure TiO 2 is 650 C (and for Mo:TiO 2 is 700 C).In view of these results, all the following photoelectrochemical experiments were carried out directing the light towards the back of the photoelectrode (via the ABS) and using TiO 2 -650 as a benchmark against Mo:TiO 2 lms.
Photocurrent-potential (J-V) and photocurrent-time (J-t) curves are also shown in Fig. 9. J-V curves (Fig. 9b) under chopped simulated solar light indicate that Mo:TiO 2 -700 outperforms the rest by almost a factor of two, reaching about 0.20 mA cm À2 at 1.23 V RHE .This suggests that the annealing temperature has a signicant effect.9][60] The slightly lower photocurrent observed for Mo:TiO 2 samples at lower applied bias in comparison to TiO 2 -650 might arise from the different charge distribution at the space-charge region of Mo:TiO 2 that can affect the band-bending properties of the as-prepared photoanodes. 36In fact, Z-potential measurements of powdered suspensions of Mo:TiO 2 particles revealed a highly negatively charged surface at a broad pH range (pH 1-14).This highly negative surface may lead to the formation of an accumulation layer when the Mo:TiO 2 lms are in contact with the electrolyte, requiring a larger applied bias to switch to a depletion layer and promote the migration of photocarriers. 36wo different co-catalysts, cobalt phosphate (Co-P i ) and CoFeO x have been evaluated in order to reduce the onset potential of Mo:TiO 2 -700.Previous reports have shown that both co-catalysts are excellent candidates for reducing the onset potential owing to the decrease in electron-hole recombination at the electrode/electrolyte interface and reduction of surface charge recombination. 27,29,61Fig. 9c shows the J-V curves for Mo:TiO 2 -700 without co-catalyst and with either Co-Pi or CoFeO x .Optimal deposition conditions were found to be 20 s of deposition time for CoP i loading and 3 cycles for CoFeO x (positively sweeping the voltage from 1.1 to 1.4 V Ag/AgCl ).Interestingly, both co-catalysts show a similar behaviour, where an enhancement in photocurrent is observed at low bias.At higher bias the driving force for the electron-hole separation comes from the higher applied bias itself rather than the co-cocatalyst, so main improvements with co-catalyst addition are only seen at low bias. 27hotocurrent-time curves at 1.23 V RHE also conrm that Mo:TiO 2 -700 exhibits a twofold increase in photocurrent performance and reveals better photostability in comparison to TiO 2 -650 (Fig. 9d).The photostability for Mo:TiO 2 and TiO 2 -650 photoanodes was quantied as the percentage of the photocurrent performance at the end of the last illuminated cycle (J) against the photocurrent performance at the end of the rst illumination cycle (J 0 ), as previously reported by Paracchino et al. 62 Aer 9,000 s of chopped light irradiation, Mo:TiO 2 -700 presented the best photostability with a J/J 0 of 99.0%, followed by Mo:TiO 2 -650 (J/J 0 ¼ 95.7%), Mo:TiO 2 -800 (J/J 0 ¼ 93.5%) and nally TiO 2 -650 (J/J 0 ¼ 81.7%).Interestingly, all Mo:TiO 2 photoanodes resulted in an enhancement in the photostability in comparison to pure TiO 2 , although only Mo:TiO 2 -700 showed an improvement in the photocurrent performance over TiO 2 -650.Unlike Mo:TiO 2 -700, Mo:TiO 2 -800 exhibits a lower photocurrent than TiO 2 at an applied bias of 1.23 V RHE .This lower photocurrent performance is attributed to a combination of plausible reasons: rst, Mo:TiO 2 -800 only exhibits rutile TiO 2 in its composition due to the high annealing temperature employed for the preparation, as depicted in XRD and Raman experiments.Rutile TiO 2 is known to be less active than the anatase one, despite having a narrower band gap. 639][60] Third, SEM images (Fig. 2) and ECSA measurements (Fig. 3b) also indicate a smaller surface area for Mo:TiO 2 -800 with larger grains compared to Mo:TiO 2 -700.Finally, many other factors such as particle size, aggregate shape and size may also inuence the nal photoelectrochemical performance of the Mo:TiO 2 -800 lm.In fact, TEM images showed higher particle size and aggregation for the Mo:TiO 2 -800 photoanode, mainly due to the high annealing temperature employed.
Unlike Mo:TiO 2 -800, Mo:TiO 2 -650 shows both anatase and rutile TiO 2 crystalline phases, as demonstrated by XRD analysis and Raman spectroscopy and the measured band gap is considerably lower than pure TiO 2 .Therefore, the poorer photoresponse behaviour of Mo:TiO 2 -650 must be due to the type of Mo doping and to the lm morphology.As shown in SEM-EDXS and XPS analyses, Mo:TiO 2 -650 possesses the highest amount of Mo amongst the Mo:TiO 2 photoanodes and XRD analysis shows a small shi towards lower angles.This fact suggests that the clear majority of Mo 6+/5+ atoms are dispersed around the surface or occupying interstitial sites of the TiO 2 lattice structure, rather than occupying Ti 4+ positions in the TiO 2 lattice structure.The presence of larger quantities of Mo 6+/5+ atoms on the surface of TiO 2 can reduce the photocurrent performance by creating recombination sites and blocking reaction sites. 6,7orphology and surface area can also play an important role in the photocurrent performance of photoanodes.As shown in the SEM images (Fig. 2j), very ne and poorly dened nanostructures are observed, which results in less surface area exposed for the water oxidation.This is further conrmed with ECSA measurements (Fig. 3b), where Mo:TiO 2 -650 shows a signicantly smaller surface area in comparison to Mo:TiO 2 -700.
Photocurrent-time curves of Mo:TiO 2 -700 and TiO 2 -650 photoanodes at 1.23 V RHE were also recorded using a UV lamp (365 nm, 3.6 mW cm À2 ) and are shown in Fig. 9e.Under these conditions, Mo:TiO 2 -700 also outperforms the performance of pure TiO 2 , reaching a photocurrent value of ca.0.65 mA cm À2 compared to ca. 0.2 mA cm À2 , respectively.
In order to further understand the enhancement in photoresponse for the Mo:TiO 2 -700 photoanode, the charge transfer properties of photogenerated electrons and holes were studied using PEIS.Fig. 9f shows Nyquist plots of the as-prepared photoanodes recorded at a DC potential of 1.23 V RHE under illumination and AC potential frequency range of 100000-0.01Hz.The inset of Fig. 9f shows the equivalent circuit used to t the Nyquist plots.It consists of an ohmic resistance and two RC elements in series, where R s corresponds to the resistance of the cell, R 1 to the resistance of the electronic process in the bulk semiconductor along with Constant Phase Element 1 (CPE 1 ), and R 2 to the resistance of the interfacial charge transfer between the electrolyte and the photoanode along with CPE 2 . 64he corresponding tted resistance values are listed in Table 1.As expected, the resistance values of the Mo:TiO 2 -700 photoanode are smaller than that of Mo:TiO 2 -650, Mo:TiO 2 -800 and TiO 2 -650, suggesting a better separation efficiency and faster transfer rate of photogenerated electrons and holes.This enhancement in the charge transfer properties of the Mo:TiO 2 -700 photoanode agrees well with the J-t curves (Fig. 9d), that showed better photostability and a twofold photocurrent increase.This improvement is attributed to the presence of oxygen vacancies, formed to balance charges aer partial doping with Mo 6+/5+ .Oxygen vacancies are known to improve the electrical conductivity and charge transportation of TiO 2 . 65,66ince Mo:TiO 2 -700 shows an optimal substitutional doping of Mo atoms into TiO 2 , an enhancement in electron conductivity and charge transportation occurs, giving rise to higher photocurrents and stability. 65,66PCE measurements for Mo:TiO 2 -700 and pure TiO 2 photoanodes are shown in Fig. 10a.Pure TiO 2 values slowly increase from 0% at 500 nm to 1.4% at 400 nm and reach a maximum of 35% at 320 nm.However, Mo:TiO 2 -700 IPCE values increase from 0% at 500 nm to 5% at 400 nm and reach a maximum of 40% at 330 nm.These IPCE results with monochromatic light conrm the superior performance of Mo:TiO 2 -700 over pure TiO 2 on absorbing and utilizing visible light, and corroborate the photocurrent results and impedance analysis with polychromatic solar light.
O 2 evolution and photocurrent measurements were performed on Mo:TiO 2 -700 at 1.23 V RHE under 1 sun illumination for 340 min (Fig. 10b).The amount of O 2 in the headspace of the PEC cell increased linearly with time during irradiation.Using the photocurrent-time curve obtained (Fig. S9, ESI †), the theoretical amount of O 2 expected for a water oxidation reaction

Conclusions
In this work a facile and rapid approach for the design of Mo:TiO 2 photoanodes using a heterometallic oxo cage of the type [Ti 4 Mo 2 O 8 (OEt) 10 ] 2 as a single source precursor has been demonstrated.The performance of the resultant photoanodes is highly reliant on the annealing temperature employed owing to its effects on the crystallinity, morphology and doping of TiO 2 , being 700 C the optimal annealing temperature.At this temperature, the Mo:TiO 2 photoanode (Mo:TiO 2 -700) presents better photostability and a two-fold increase in photocurrent performance (0.20 mA cm À2 at 1.23 V RHE ) in comparison to a TiO 2 photoanode (0.10 mA cm À2 at 1.23 V RHE ).This improvement both in the photocatalytic performance and stability is attributed to a combination of several factors that become optimized at 700 C: rst, Mo:TiO 2 -700 exhibits the presence of anatase TiO 2 and in minor amount rutile TiO 2 , forming a heterostructure that is known to reduce the electron and hole recombination rate.Second, 2 -700 has a smaller band gap than the obtained at different annealing temperatures or without Mo doping, which allows for a better use of the solar spectrum and higher efficiencies (IPCE: 5% at 400 nm).Third, in Mo:TiO 2 -700 there is preferred substitutional doping of Mo 6+/5+ atoms for Ti 4+ atoms in the TiO 2 lattice structure, causing the presence of oxygen vacancies which improve the electrical conductivity and charge transportation of the lm.Fourth, Mo:TiO 2 -700 shows a large amount of cavities, porosity and well-dened nanostructures, resulting in the photoanode with the highest surface area.This characteristic morphology is associated to the spray pyrolysis deposition of [Ti 4 Mo 2 O 8 (OEt) 10 ] 2 and to the partial sublimation of Mo species during the annealing process.On balance, these results clearly demonstrate a simple and effective methodology for preparing Mo-doped TiO 2 photoanodes with tuned electronic band properties and morphology and reveal the crucial parameters that allow the exploitation of heterometallic oxo cages in thin lms for energy applications.These results open up the possibility for exploring a wide range of different heterometallic oxo cages for the fabrication of metal oxides photoanodes.
, 1st column) indicate a high polycrystalline character of these Mo:TiO 2 and TiO 2 particles.The corresponding diffraction pattern along with peak identication is shown in the inset of these gures.These data agree well with XRD where diffraction peaks corresponding to TiO 2 anatase and rutile are shown in Mo:TiO 2 -650, Mo:TiO 2 -700 and pure TiO 2 , whereas only TiO 2 rutile is observed in Mo:TiO 2 -800.Particle size distributions are shown in Fig.

Fig. 5
Fig. 5 XPS spectra of (a) Ti 2p, (b) O 1s and (c) Mo 3d of Mo:TiO 2 .Scattered points correspond to raw data acquired in the measurements and solid lines to the fitted values.

Fig. 9
Fig. 9 (a) Variation of photocurrent density as a function of annealing temperature for Mo:TiO 2 and pure TiO 2 (at 1.23 V RHE ).Solid lines correspond to photoanodes where illumination was directed towards the back of the FTO-ABS and dashed-lines where illumination was directed towards the front of the working electrode.Error bars indicate standard error above and below the mean.(b) Photocurrent-potential curves of Mo:TiO 2 and pure TiO 2 .(c) Photocurrent potential curves of Mo:TiO 2 -700, Mo:TiO 2 -700-CoFeO x and Mo:TiO 2 -700-Co-Pi.(d) Photocurrent-time curves for 9000 s of Mo:TiO 2 and pure TiO 2 at an applied bias of 1.23 V RHE .(e) Photocurrent-time curves for 250 s of Mo:TiO 2 and TiO 2 -650 at an applied bias of 1.23 V RHE under UV illumination (365 nm, 3.6 mW cm À2 ).(f) Nyquist plots of Mo:TiO 2 and pure TiO 2 in 1 M KOH.DC of 1.23 V RHE ; AC potential frequency range 10 5 to 0.01 Hz with an amplitude of 5 mV.The inset of the figure shows the equivalent circuit used to fit the data (solid lines).All electrochemical measurements were carried out in 1 M KOH aqueous electrolyte.All were irradiated with simulated sunlight (AM 1.5G, 100 mW cm À2 ) except (e).
with 100% faradaic efficiency was calculated and also represented in Fig.10b.Comparison between values indicated that Mo:TiO 2 -700 photoanode has a faradaic efficiency of approx.60% (details of calculations are shown in ESI †).Similar efficiency values have been obtained on bare photoanodes without oxygen evolution electrocatalysts.

Table 1
Calculated resistance parameters from EIS data Open Access Article.Published on 19 September 2018.Downloaded on 12/6/2018 11:24:27 AM.This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.