Improved photoelectrochemical performance of electrodeposited metal-doped BiVO4 on Pt-nanoparticle modified FTO surfaces

The recombination of photogenerated electron–hole pairs is one of the main limiting factors of photoelectrocatalysts absorbing in the visible part of the solar spectrum. Especially for BiVO4 the slow electron transport to the back contact facilitates charge recombination. Hence, thin layers have to be used to obtain higher photocurrents which are concomitantly only allow low absorption of the incident light. To address this limitation we have modified FTO substrates with Pt-nanoparticles before electrodepositing BiVO4. The Pt-nanoparticles decrease the overpotential for the electrodeposition of BiVO4, but more importantly they provide the basis for decreased charge recombination. Electrodeposited Mo-doped BiVO4 on Pt-nanoparticle modified FTO exhibits a substantially decreased recombination of photogenerated charge carriers during frontside illumination. Simultaneous co-doping of BiVO4 with two different metals leads to a substantial enhancement of the incident-photon-tocurrent efficiency (IPCE) during light driven oxygen evolution reaction. Highest IPCE (>30% at 1.2 V vs. RHE) values were obtained for Mo/Znand Mo/B-doped BiVO4.


Introduction
Suitable materials for solar water splitting have to full a number of prerequisites. The oxygen evolution reaction (OER) is the limiting half-cell reaction due to high overpotentials necessary for the four electron-transfer reaction. N-type semiconductors with low band gaps (<3.0 eV) are required for absorbing in the visible part of the solar spectrum. Thus, due to its band gap of about 2.4 eV 1-3 and its band position suitable for oxygen evolution 4 BiVO 4 is regarded as a promising material for solar water splitting. [5][6][7][8][9][10][11][12] BiVO 4 can be deposited onto FTO substrates e.g. by a metal-organic-decomposition method followed by spin-coating [13][14][15][16][17] or drop-casting. 18 Seabold et al. developed a simple procedure for the electrodeposition of BiVO 4 from a bismuth nitrate and vanadyl sulfate precursor solution, 19 while Jiang et al. used an inkjet printer 3 for BiVO 4 deposition. Independent of the preparation method, due to the poor electron-transfer kinetics leading to high recombination rates of photo-generated electron-hole pairs BiVO 4 lms are not suitable for solar-energy conversion using frontside illumination. BiVO 4 provides higher photocurrents or incident-photonto-current efficiencies (IPCE) upon backside illumination. 20,21 A variety of strategies to improve the photoelectrochemical performance of BiVO 4 were suggested such as among others doping with different metals. For example doping with molybdenum, 22-29 tungsten [30][31][32][33][34] or simultaneously with two different metals 18 leads to an increase in electron density of the doped BiVO 4 resulting in higher IPCE values. To prevent the accumulation of photogenerated holes at the semiconductor/electrolyte interface due to slow OER kinetics, co-catalysts were deposited on BiVO 4 . 16,31,[35][36][37][38][39][40][41][42] Moreover, using reduced graphene oxide, Kudo et al. were able to promote electron-transfer resulting in higher IPCE values. 43 Next to the above presented strategies for enhancing the photoelectrochemical performance of suitable photoabsorber materials, the insertion of interfacial layers, nanoparticles, nanotubes and nanowires, e.g. WO 3 or CNTs offers a strategy to increase the charge carrier life time upon illumination with solar energy. 15,[44][45][46][47][48][49][50][51] We purpose as an alternative strategy the modication of the FTO substrate with Pt-nanoparticles prior to the electrochemically induced BiVO 4 deposition. The presence of these nanoparticles slightly decreases the overpotential of anodic BiVO 4 formation improving the low conductivity of the FTO substrate and provide nuclei centres for lm growing. However, more importantly the presence of the Pt-nanoparticles within Mo-doped BiVO 4 lms gave rise to higher IPCE values due to decreased recombination rate of photogenerated electron-hole pairs. Doping of BiVO 4 with Mo or W was found to improve additionally the photocatalytic efficiency. By doping of BiVO 4 simultaneously with two metals by means of electrodeposition on Pt-nanoparticle modied FTO substrates highest IPCE values were obtained for Mo/Zn and Mo/B doped BiVO 4 .

Experimental part
Fluorine-doped tin oxide (FTO) substrates (Pilkington, TEC 8A, 2.3 mm) were rst cleaned with acetone under ultra-sonication for 5 min followed by a treatment in 0.1 M NaOH (99%) for 15 min at 75 C. Aer rinsing with water the FTO were dried in an Ar stream. All chemicals were from Sigma Aldrich if not mentioned otherwise. All electrochemical measurements were carried out in a three-electrode setup with an Ag/AgCl/3 M KCl (210 mV vs. NHE) reference electrode and a Pt wire counter electrode. Pt-nanoparticles 52,53 were electrochemically deposited from a 0.4 mM H 2 PtCl 6 (ACS reagent, $37.50% Pt basis) solution by applying potential pulses using a graphite rod counter electrode. A no-effect potential of 0 V was applied for 1 s, while deposition was performed for 0.2 s at À1 V. The potential pulse sequence was repeated 50 times. For the fabrication of the BiVO 4 lms, the electrodeposition procedure of Seabold et al. was adapted. 19 In short, 35 mM VOSO 4 $H 2 O ($97%) were diluted in 0.54 M HNO 3 and 10 mM Bi(NO 3 ) 3 $5H 2 O ($98%) were added. For increasing the pH value up to 5, 2 M sodium acetate tetrahydrate (VWR, 100%) was used for stabilizing the Bi(III)-ions. The pH value was adjusted to 4.7 with conc. HNO 3 . Electrodeposition was performed at 1.9 V vs. Ag/AgCl/3 M KCl at 70 C using a circulation thermostat and a Pt-mesh as counter electrode. It should be mentioned that the counter electrode was separated by means of a membrane from the deposition solution to prevent precipitation of Bi(0) at the Pt counter electrode. 2 M sodium acetate solution (pH 4.7) was used as electrolyte. The deposition was taking place for 400 s and 600 s on bare FTO and FTO modied with Pt-nanoparticles. The FTO substrates were contacted with an aluminium foil which was insulated during deposition with paralm. For doping, 5 mM of the additional precursor salt was added to the described Bi-Vsolution. For doping with two different metals 5 mM of each precursor salt was found to be best for depositing homogenous layers and providing high photocurrents. Zn(NO 3 ) 2 $6H 2 O ($99%), Na 2 WO 4 $2H 2 O ($99%), Pb(NO 3 ) 3 (Riedel-de-häen, 99%), NaMoO 4 ($99%), MnSO 4 ($99%), Fe(NO 3 ) 3 $9H 2 O (Acros, $99%), H 3 BO 3 (J.T. Baker, $99.9%) were used for co-doping and were added directly before the electro-chemically induced deposition. Aer electrodeposition a brownish lm was obtained, which turned into yellow aer annealing in air at 500 C for 1 h with a heating rate of 2 C min À1 . To remove vanadium pentoxide the samples were treated with 1 M KOH for 20 min.
Photocurrent spectroscopy was performed using a microprocessor controlled monochromatic system (Institut Fotonowy) with integrated shutter. A short pass lter (<400 nm) and a long pass lter (>400 nm) were utilized to remove completely the undesired part of the spectrum. A 150 W Xe-Lamp (Ushio) was used as light source. All measurements were done in pyrex glass cell. If the samples were irradiated from the front, the light passes through a quartz window. The irradiated area was 0.785 cm 2 . The power of the incident light was determined with a power meter (Thorlabs) and a thermopile detector and was corrected by the power loss of the quartz window.
An Autolab PGSTAT12 (Metrohm) potentiostat was used to apply an external bias potential. Light induced water splitting was performed in 0.1 M Na 2 SO 4 (p.a.) as electrolyte at a pH value of 6. The applied potential (E app ) was re-calculated versus the reversible hydrogen electrode (RHE) by E(RHE) ¼ 210 (mV) + E app + 59 Â pH. The obtained photocurrents i ph were derived by subtracting the dark current from the current measured under illumination (ESI, Fig. S1 †) and the IPCE was calculated taking the photon ux (F(l)) for each wavelength into account (eqn (1)). e represents the elementary charge.  20 These difference are attributed to the layer thickness, which is proportional to the IPCE, where very thin layers (<100 nm) results in highest IPCE values close to 40% during frontside illumination. 21 Fig. 1 summarizes the obtained IPCE values at   17 using the metal-organic-decomposition method and spin-coating. The IPCE spectra for the Fe, Mo, and W-doped BiVO 4 lms indicate a small increase of the band gap for Fe-doped BiVO 4 (Fig. 1b). Concerning the band gap, direct 1,18

Concept of Pt-nanoparticle modication
Regarding the electrochemically induced deposition of semiconductor materials, the low conductivity of the used FTO substrates is leading to a high overpotential for Bi-V-O formation. Decorating the surface of the FTO with Pt-nanoparticles (50 nm diameter), small nuclei for lm growth are created, thus reducing the overpotential for lm formation.
With increasing deposition time the Bi-V-O lm is formed initially at the Pt and then in there vicinity, leading nally to a fully covered surface with a large number of small metal cores (Fig. 2b). In the simplest case, recombination of the photogenerated electron-hole pairs and in particular the absorption of the incident light depend on the layer thickness of the semiconductor material. For thick layers (d 1 ) light absorption increases, but the recombination rate increases due to a longer diffusion length of electrons. This has a tremendous impact on the photoelectrochemical performance. In case 1 in Fig. 2a, lower photocurrents are expected for a thick layer (d 1 ). A decrease of the recombination rate (r 2 ) is possible by reducing the layer thickness (d 2 , case 2). However, by this the absorption of the incident light is concomitantly decreased limiting the PEC performance. The introduction of small Pt-nanoparticles (Fig. 2b) decrease the recombination rate at the interfacial layer between the photoabsorber and the FTO for the same layer thickness (d 1 ) with good absorption of the incident light by electron capturing. Moreover, the advantage of the deposition of the Pt-nanoparticles e.g. over a backside metal lm is seen in the ability of utilizing transparent lms for tandem cell applications with an improved homogeneity of the deposited semiconductor layer on the metal nuclei. For creating uniform Ptnanoparticles with a small diameter on the FTO surfaces sequences of potential pulses were used as reported previously. 52,53 The density of the nanoparticle distribution was controlled by the number of repetitions of these pulses.
Optimal conditions for uniformly sized and high density distributed Pt-nanoparticles were obtained by repeating the deposition pulses 50 times (ESI, Fig. S3 †). Higher repetition numbers lead to Pt nanoparticles with increased diameter and lower density. Finally agglomeration was noticed at a 110 pulse sequence. The correlation of pulse sequences and loading of the Pt-nanoparticle is demonstrated in Fig. S4. † It is expected that the deposition of the Pt nanoparticles on FTO exhibit a substantial impact on the hydrogen evolution reaction (HER). 55 An increase of the HER current was observed until a number of 110 Pt nanoparticle deposition pulse sequences. No further changes of the cathodic HER current were noticed between 110 until 350 deposition pulse sequences. In comparison to unmodied FTO, loading with Pt-nanoparticles increased the HER activity signicantly with lower deposition pulse sequence number.
The inuence of Pt-nanoparticles on the photocatalytic performance was studied using Mo-doped BiVO 4 lms because these lms showed higher reproducibility as compared to W-doped BiVO 4 (see Fig. 1). First, the effect of the deposited Pt-nanoparticle on the layer thickness was evaluated. Although electrode-position leads to comparatively rough surfaces, SEM cross-section allowed to estimate the layer thickness (Fig. 3).
Mo-doped BiVO 4 lms were prepared in absence and presence of Pt-cores on the FTO surface and a similar layer thickness of around 600 AE 50 nm was obtained in both cases. Reducing the deposition time to 400 s, the layer thickness decreased around 450 AE 50 nm (ESI, Fig. S5 †). These results indicate on the one side, that the layer thickness is increasing with the deposition time. On the other side, for small Pt-nanoparticles (50 nm) no inuence on the lm electrodeposition was observed resulting in comparable lm thicknesses for layers deposited in presence and in absence of Pt-nanoparticles. Comparable layer thicknesses in accordance with the model in Fig. 2 is necessary to evaluate the effect of the Pt-nanoparticles on the recombination process during PEC measurements. Frontside illumination was used to demonstrate the impact of the Pt-cores on the photocatalytic properties of the Mo-doped BiVO 4 lms. It was anticipated that in presence of Pt-nanoparticles the electron transfer is facilitated leading to a substantially decreased charge-carrier recombination. Mo-doped BiVO 4 lms were deposited in absence and in presence of Pt-nanoparticles on FTO surfaces for 400 s and 600 s and the IPCE values were determined (Fig. 4a). At a wavelength of 450 nm an increase in the IPCE was observed when the FTO was modied with Ptnanoparticles. This suggests a higher PEC performance due to an improved absorption of the incident light at thicker lms (600 nm AE 50 nm) and an enhanced PEC performance in presence of Pt-nanoparticles.
In a chronoamperometric experiment at a dened bias potential of 0.5 V vs. RHE the shutter of the lamp was opened and closed for predened times. During the dark current measurement the wavelength was changed in a range from 250 nm to 550 nm (Fig. 4b). This current-time response hence does not only provide information about the wavelengthdependent photocurrent which is the basis for the determination of wave-length-dependent IPCE values, but it provides information about the recombination rate of the photogenerated electron-hole pairs. The initial current peak (i init ) measured directly when the lamp shutter was opened represents the separation of electrons and holes in the depletion region. The decay of the current results from holes, which recombine with electrons from the CB or accumulate at the surface instead of reacting with electrons from the electrolyte. [56][57][58][59] The time dependent decay of the current under continuous illumination until a steady-state current (i ss ) is established, based on the recombination of holes in the VB with electrons from the CB, produced when the light reached for the rst time the surface. This current is based on holes which did not recombine and contribute to the faradaic process. 56,60 The steady state current is inuenced by the recombination process at the interfacial layer and is decreased if recombination centres at the substrate are present. Moreover, this current depends on the kinetics of the reaction at the semi-conductor/electrolyte interface. Slow kinetics increases surface recombination and decreases the photocurrent. The lower the difference between i init and i ss the lower is the overall recombination rate. As a matter of fact this difference is inuenced by the applied bias potential i.e. with increasing bias potential i init decreases and i ss increases (ESI, Fig. S6 †). 57 In order to analyse the effect of Pt-nanoparticles on the charge recombination rate a bias potential close to the open circuit potential (0.5 V vs. RHE) was  applied (Fig. 4b). Consequently, the charge carrier life time is enhanced, if the ratio of the steady state current and initial current under illumination is close to one. Under comparable conditions, the ratio under illumination at 350 nm is relatively low (0.12) for Mo-BiVO 4 deposited in absence of Pt-nanoparticles on FTO. The ratio increased up to four times (0.46) by in the insertion of Pt-nanoparticles between the Mo-BiVO 4 layer and the FTO surface, indicating a decreased recombination of photogenerated charge carriers thus leading concomitantly to the observed higher photocurrents.
Moreover, the same samples were investigated using backside illumination through the FTO substrate (Fig. 4c). For Mo-doped BiVO 4 lms an improvement of the electron transport is expected due to the increased electron density caused by doping with Mo. Thus, higher photocurrents are anticipated during frontside illumination as compared with backside illumination. 21,28,56 For electrodeposited Mo-doped BiVO 4 on a bare FTO substrate with a thickness of the lm of around 600 nm, higher IPCE values are observed for backside illumination despite the insertion of Mo into the lattice.
The IPCE values increased from around 10% to nearly 25% in the wavelength range between 350 nm and 450 nm. In this case the electron transfer to the back contact seems to be limiting. Seabold et al. observed similar effects. Thick lms of several hundred nm provide higher photocurrents under backside illumination attributed to the small diffusion length of electrons of around 300 nm. 22 Thin layers (<300 nm) usually exhibit higher photocurrents under frontside illumination. For Mo-doped BiVO 4 lms deposited on FTO modied with Pt-nanoparticles with a thickness of around 600 nm the differences between frontside and backside illumination decreased (see for details Fig. S7, ESI †). Accordingly, introduction of Pt-nanoparticles at the back contact improves electron transfer and suppresses recombination of photogenerated electron-hole pairs expecting same kinetic rates for the OER in both cases. This supports additionally the hypothesis of improved electron transfer via the Pt-nanoparticles.
Assuming direct contact between FTO and Mo-doped BiVO 4 (Fig. 5a) the electrons have to diffuse through the semiconductor to the back contact, which is supposed to be limiting during frontside illumination due to the high recombination rates for photogenerated electron-hole pairs. Backside illumination reduces the diffusion length for electrons to the back contact, hence decreasing the recombination rate and terminally leading to higher photocurrents. A possible explanation of this effect is the passivation of the recombination centres of the FTO substrate, reducing the recombination of charge carriers at the interfacial layer. 48 Moreover, by modifying FTO with Pt-nanoparticles a Schottky barrier between platinum and the Mo-doped BiVO 4 lm is generated due to the higher work function of platinum. 61 Under illumination and by applying an external bias potential the photogenerated electrons can pass the energy barrier between the Pt/semiconductor contact (Fig. 5b) and are no longer available for recombination. This could explain the observed higher IPCE values as compared to Mo-doped BiVO 4 lms deposited in absence of Pt-nanoparticles (Fig. 4). The energy of the electrons seems to be high enough to reach the metal despite the energy barrier, while the electrons are not diffusing back to the semiconductor due to the energy barrier and the induced electric eld caused by applying the external bias.

Co-doping of BiVO 4
The concept of modifying FTO substrates with Pt-nanoparticles was additionally used to further improve the photoelectrochemical performance of doped BiVO 4 lms by codoping with two different metals during lm electrodeposition. In contradiction to the results obtained by Park et al., 18 who obtained a moderate enhancement to maximum IPCE values of 12% with drop-casted W/Mo-doped BiVO 4 lms, co-doping was performed by adding a second element (Pb, Zn, B) to Mo, because the electro-deposition of Mo-BiVO 4 on top of Pt-nanoparticles already yielded IPCE values up to 25%. First, the morphology and structure of the supposed metal-doped electrodeposited BiVO 4 lms were analysed. A possible difference in the photocatalytic efficiency is expected due to the changes of the morphology of the deposited lm by adding an additional metal salt to the deposition bath (ESI, Fig. S4 †). Undoped and W-doped BiVO 4 form rough particles, while the overall size of these particles decreased by doping with W. Doping with Mo and an additional element like Zn and B, smaller, rounder and smoother particles are observed compared to undoped BiVO 4 and Mo-BiVO 4 . Electrochemical deposition of BiVO 4 in presence of a B or Zn precursor, led to at and agglomerated particles forming a dense layer. This change of the particle morphology suggests the formation of Zn-Bi-V-O lms during the electrodeposition and annealing procedure. Similar to Mo/Zn doping, the W-BiVO 4 particles changed to a rounder shape in presence of a Zn precursor during the deposition. The effect of the doping of BiVO 4 lms were further characterized by means of XRD and EDX analysis (ESI; Fig. S9 and S10 †). XRD patterns do not show any additional reections from impurities such as MoO x or ZnO. However, the formation of the monoclinic scheelite structure is obvious, while a shi of the characteristic peak at 28.9 , which is attributed to the incorporation of Mo 6+ instead of V 5+ into the monoclinic BiVO 4 structure. 18 Merging of the peaks at 34.5 and 35.2 as well as the peaks at 46.8 and 47.4 into a single peak suggests the deformation of the scheelite structure. 62 EDX analysis showed small amounts of Zn (<1%). In the case of Mo/Zn doping, small Zn peaks demonstrate the presence of Zn in the material (ESI; Fig. S10 †).
Raman spectroscopy (ESI; Fig. S11 †) represents a more sensitive method to evaluate the doping of the BiVO 4 . Similar to XRD patterns, the shi of the symmetric stretching mode of V-O of the Raman band of 833 cm À1 , indicates an incorporation of an additional metal into the VO 4 3À tetrahedron. This stretching frequency belongs to the length of the metal bond and the above-mentioned shi to lower wavenumber suggesting an increase of the bond length. Eqn (2) determines the exponential dependence of the Raman frequency (v) on the radius (r). 63 v (cm À1 ) ¼ 21 349 exp(À1.9176r (Å)) For un-doped BiVO 4 the metal bond length was calculated to be 1.6916Å. Doping with Mo increases the bond length to 1.6979Å and the co-doping with Mo/B or Mo/Zn to 1.701Å. From material characterization, doping simultaneously with two elements by means of the electrodeposition method was suggested.
The photoelectrochemical performance in dependence of the different dopants combinations for BiVO 4 on top of Ptnanoparticles, exhibits no improvement of the IPCE values upon co-doping with Pb as compared to pure BiVO 4 . All other materials improved the IPCE values ( Fig. 6a and b) 62 The differences during co-doping are probably due to the difference in the ratio of the co-dopants which are caused by the different preparation techniques, inuencing the photoelectrochemical performance. We additionally combined Mo with different elements as co-dopants for BiVO 4 (Fig. 6b). The effect on the IPCE values coincides with the results for W-doped BiVO 4 as shown in Fig. 6a.
The highest IPCE values were reached for Mo/Zn-and Mo/Bco-doped BiVO 4 lms deposited on top of Pt-nanoparticles for a deposition time of 600 s (Fig. 6d). The difference between the co-doped and the single doped Mo-BiVO 4 (Fig. 6c) were significantly higher as compared to W-doped BiVO 4 lms. The photoelectrochemical performance of Mo-doped BiVO 4 could be substantially improved and IPCE values of up to 40% were reached under frontside illumination (Fig. 6c). The exact mechanism of the co-doping is not fully understood. As already mentioned, Mo and W are well known donor-type atoms and preferentially substitute V sites. From DFT calculations performed by Yin et al., 64 Zn was suggested to act as an acceptor atom with shallow transition energy for Bi sites compared to V site substitution. Shan et al. 65 reported enhanced PEC performance of BiVO 4 upon B-doping attributed to the formation of weak chemical bonds between the B-ions and the corners of the VO 4 -tetrahedron. Additionally, higher photocatalytic activity of B-doped BiVO 4 was observed by Wang et al., 66 who supposed synergetic effects of a number of factors created by B-doping, e.g. higher specic area or more oxygen vacancies. This might explain the further improvement of the IPCE in the case of Mo/B and Mo/Zn also supported by the different particle shape (see Fig. S4 †).

Conclusion
A different way to improve the absorption of the incident light with increasing the layer thickness resulting in higher IPCE values was demonstrated for Mo-doped BiVO 4 . The associated decrease of the charge carrier life time limiting the photoelectrochemical performance was circumvented by the introduction of Pt-cores between the Mo-BiVO 4 layer and the FTO surface. Two times higher IPCE values were generated in presence of Pt-nanoparticles acting as an electron scavenger preventing the electrons to recombine with the holes in the valence band. The integration of Pt-nanoparticles at the back contact of the Mo-doped BiVO 4 lm leads to a decrease in the recombination rate of photogenerated electron-hole pairs, which mainly represents next to the kinetic of water oxidation a limiting factor for light induced oxygen evolution at BiVO 4 lms. Additionally, Pt-nanoparticles favour electrochemically induced deposition of BiVO 4 or doped BiVO 4 lms by improving the conductivity of the FTO substrate. Furthermore, the electrodeposition method can be used to simultaneously dope BiVO 4 with different elements by adding an additional precursor compound to the deposition bath. Doping with Mo/Zn and Mo/ B leads to a further substantial increase in the IPCE values due to an enhanced electronic conductivity.