Dissolution of WO3 modified with IrOx overlayers during photoelectrochemical water splitting

WO3, an abundant transition metal semiconductor, is one of the most discussed materials to be used as a photoanode in photoelectrochemical water‐splitting devices. The photoelectrochemical properties, such as photoactivity and selectivity of WO3 in different electrolytes, are already well understood. However, the understanding of stability, one of the most important properties for utilization in a commercial device, is still in the early stages. In this work, a photoelectrochemical scanning flow cell coupled to an inductively coupled plasma mass spectrometer is applied to determine the influence of co‐catalyst overlayers on photoanode stability. Spray‐coated WO3 photoanodes are used as a model system. Iridium is applied to the electrodes by atomic layer deposition in controlled layer thickness, as determined by ellipsometry and x‐ray photoelectron spectroscopy. Photoactivity of the iridium‐modified WO3 photoanodes decreases with increasing iridium layer thickness. Partial blocking of the WO3 surface by iridium is proposed as the main cause of the decreased photoelectrochemical performance. On the other hand, the stability of WO3 is notably increased even in the presence of the thinnest investigated iridium overlayer. Based on our findings, we provide a set of strategies to synthesize nanocomposite photoelectrodes simultaneously possessing high photoelectrochemical activity and photostability.

help transition into a renewable energy society.2][3][4] The usage of solar energy, however, breeds new problems as well as it is intermittent.This makes it impossible to provide the baseload of electrical energy, and coupling to the electrical grid and limited storage capacities hinder an application in the transport sector.Thus, efficient solutions for energy storage and transportation have to be engaged.Solar fuels from photoelectrochemical (PEC) devices are a promising approach to overcome these limitations.][15] A PEC device should provide operational stability for several years to be cost-efficient. 16This requirement still hinders the fast market penetration of PEC devices as only lifetimes of up to several hundred hours have been demonstrated in laboratory settings so far. 12,17A purely thermodynamic viewpoint dominates the general perception of stability in the photoelectrochemistry community.A typical n-type semiconductor is considered stable when the decomposition potential is below its valence band edge.On the contrary, the semiconductor will decompose if its decomposition potential lies above the water-splitting potential.Generally, a semiconductor is stable when the potential to decompose lies between the valence band edge and the thermodynamic potential of water splitting. 18owever, kinetic effects cannot be ignored.In the latter case, kinetic competition between water oxidation and photoelectrode decomposition occurs. 19][22][23] Coating the surface of the photoelectrode with a protective overlayer is a commonly applied mitigation strategy when facing severe photodegradation.The additional layers provide spatial shielding from the electrolyte and can prevent changes in the reaction kinetics at the photoelectrode surface.Two types of overlayers with different functionalities can be distinguished.First, a thin passivating oxide such as TiO 2 can serve as a tunneling layer.0][31][32] These co-catalysts protect the photoelectrode from the electrolyte and create additional reaction sites that shift the reaction kinetics at the surface.For example, Co decoration of WO 3 photoanodes significantly enhanced their photoactivity and photostabil-ity in near-neutral pH where WO 3 is thermodynamically unstable. 29,33The deposition of RuO 2 nanoparticles or Ir dinuclear catalysts on WO 3 was shown to significantly enhanced the activity in an acidic electrolyte. 32,34][36] Reaction kinetics also strongly affect the photostability of WO 3 .As recently shown by our group, a shift from electrolyte decomposition to other reactions influences the stability of WO 3 photoanodes. 21WO 3 photoanodes show a three to four-fold enhancement of photostability in HNO 3 and HClO 4 compared to H 2 SO 4 and CH 3 O 3 SH.Electrolyte decomposition involving stable reaction intermediates rather than OER in sulfur-containing electrolytes was identified as the primary contributor to these differences.These shifts were correlated to different electrolyte decomposition mechanisms.
In this work, the influence on photostability brought by the deposition of a co-catalyst on WO 3 photoanodes was studied.WO 3 model thin films were spray-coated onto fluorine-doped tin oxide (FTO) substrates based on a previously established protocol. 21Then, the photoanode surface was modified with iridium layers with different thicknesses using atomic layer deposition (ALD).A previously established PEC-ICP-MS system was employed to characterize the photoactivity and in situ photostability by recording the amount of dissolved metal ions. 37The photoactivity drops continuously with increasing iridium layer thickness, which hinders the light from reaching the photoabsorber during the front-illumination.However, the photostability is significantly enhanced when applying an iridium overlay, manifesting in the lowered WO 3 dissolution.

Sample preparation
2.1.1WO 3 photoelectrode preparation WO 3 photoelectrodes were synthesized via spray-coating with an ExactaCoat device (SonoTek).9][40] After the total dissolution of tungsten powder, the solution was thermally reduced to 20 ml and diluted with IPA to 200 ml.The precleaned (10 min sonication in 2% Hellmanex III solution, 10 min sonication in DI water, and 10 min sonication in IPA) FTO glass slides (Sigma Aldrich) were placed on an 80 • C hot plate.The precursor was sprayed with a flow rate of 0.33 ml min −1 at an ultrasonication power of 5 W, a nozzle height of 37 mm, and a traverse speed of 140 mm s −1 in a meander-shaped pattern with a 1.5 mm pitch size.

XPS:
The surface composition of the Ir-covered W samples was investigated using x-ray photoelectron spectroscopy (XPS).XPS spectra were recorded on a PHI Quantera II scanning x-ray microprobe.An Al K α irradiation with a spot diameter of 200 μm at 50 W and 15 kV with a dwell time of 200 ms per step was used.The information depth of the XPS was 5 nm.Survey scans were gathered at a step size of 0.5 and 280 eV pass energy, while highresolution narrow scans were recorded with 0.125 eV step size at 140 eV pass energy.All spectra were evaluated in CasaXPS (v.2.3.18),applying instrument-specific relative sensitivity factors and a binding energy scale calibrated to the adventitious carbon peak at 284.8 eV.
Ellipsometry: The thickness of the deposited iridium thin films was determined on native Si/SiO 2 wafers by spectroscopic ellipsometry with a SENPro from SENTECH.The measurement was performed in a wavelength region starting at 370 nm up to 1050 nm, with a constant angle of incidence of 70 • .The obtained data were fitted by the program SpectraRay/3 using an optical model, which contained the layer's thickness as well as optical constants (refraction index, psi, and delta).
UV-vis: Optical absorption spectra were measured with an ultraviolet-visible (UV-vis) spectrophotometer (OceanOptics) equipped with a deuterium-halogen light source (DH-2000-L) and an HR40000 spectrometer.The absorption spectra were obtained by subtracting the transmitted intensities from the incident intensity.
XRD: The crystal structure was analyzed by x-ray diffraction (XRD) in Bragg-Brentano geometry using a Bruker D8 Advance equipped with a Cu Kα source and LynxEye XE detector.

SEM-EDX:
The chemical composition analysis of the iridium-coated WO 3 photoanodes was performed with a Zeiss Crossbeam 540 FIB-SEM with an energy dispersive detector (X-Max 150 silicon drift detector, Oxford Instruments; Software: Aztec Version 3.3, Oxford Instruments).Samples were attached with a double-sided conductive carbon tape to a secondary electron microscopy (SEM) stub.The energy-dispersive x-ray spectroscopy (EDX) was performed at a voltage of 20 kV and a probe current of 750 pA to determine the spectrum of the desired area.
PEC-ICP-MS measurements: In situ ICP-MS measurements were performed with a homemade PEC scanning flow cell coupled to an inductively coupled plasma mass spectrometer (Perkin Elmer NexION 350X). 37The samples were placed on an XYZ-stage, which were then contacted by the cell opening with an area of 0.059 cm 2 .All measurements were conducted in Ar-purged 0.1 M HClO 4 (pH 1.67).A graphite rod (HTW) and an Ag/AgCl reference electrode (Metrohm) served as counter and reference electrodes, respectively.A metal wire was used to establish electrical contact to the sample.(Photo)electrochemical measurements were performed with a Gamry REF 600 potentiostat.The setup was controlled by a homemade LabView software.The light produced from a 300 W ozone-free Xe lamp (Newport) was passed through an AM 1.5 G filter before channeling the illumination through a liquid light guide to the working electrode.The light was calibrated to 1 Sun (100 mW cm −2 ) using a reference solar cell (Newport).Calibration standards were mixed freshly in 0.1 M HClO 4 (Merck Suprapur, diluted with deionized water, Merck Milli-Q) every day to perform a four-point calibration for Ir and W with (0, 0.5, 1, 5) μg L −1 .The internal standard was 10 μg L -1 Re in 0.1 M HClO 4 .

Synthesis and characterization
The synthesis and characterization of the WO 3 thin films were described in our previous work. 21Briefly, WO 3 thin films were synthesized by spray-coating a peroxotungstic acid precursor on glass slides coated with conductive FTO (for details, see experimental section).Measurements of the crystal structure by XRD after synthesis were consistent with WO 3 crystal structures from the crystallography open database. 21,413][44][45] SEM measurements show a fine porous structure with a layer thickness of 2-3 μm. 21he WO 3 thin films were coated with thin iridium layers by ALD.The WO 3 photoelectrodes appear less transparent after the deposition process (see Figure S1).UV-vis measurements experimentally corroborated the decreased transparency.A noncharacteristic increase in absorbance can be attributed to the iridium layer, which ultimately blocks the incident light (see Figure S2).Marginal qualitative changes between the different samples originate from the presence of scratches at the backside of the FTO substrate.SEM micrographs, as shown in Figure S3, indicate a morphologically homogeneous coating of the iridium overlayer on the pristine WO 3 sample.XRD spectra of iridium-decorated WO 3 photoanodes (see Figure S4) do not display detectable change from the pristine sample, indicating that the deposition of iridium does not alter the crystal structure of the photoabsorber.The thickness of iridium was determined indirectly by depositing the same iridium ALD cycles on Si wafers and successively performing ellipsometry.The high surface roughness of WO 3 would impede any precise direct measurements of the Ir layer thickness.Previous work has demonstrated that nucleation inhibition might cause the deposited thickness of noble metals in highly porous systems to be significantly lower than its nominal value recorded on a planar reference sample suggests. 46The ellipsometry measurements determined the thickness of iridium layers after 15, 30, and 50 cycles to be 0.5, 1.1, and 2.2 nm, respectively.As elemental quantification of iridium was not possible by EDX (see Figure S3), XPS analysis was performed instead.The recorded W4f peaks for all samples can be seen in Figure 1A.The binding energy of W4f 1/2 is located at 35.7 eV, with a binding energy difference of 2.1 eV characterizing the peak splitting of W4f 5/2 . 47O 3 with the thinnest iridium ALD overlayer indicates a peak at 61.6 eV, which suggests that iridium is mostly in an IrO 2 form (see Figure 1B).Thicker overlayers seem less oxidized, as metallic iridium is expected at 60.6 eV. 48Previous reports have proposed that the oxide form is encountered for thinner iridium layers deposited by ALD.In contrast, thicker films tend to be reduced. 49,50Figure 1C presents the atomic composition, which coincides with the intensities obtained from the high-resolution XPS spectra.The iridium content gradually increases at the surface as more ALD cycles are applied, reflecting a growth in the iridium film thickness.Tungsten and oxygen, on the other hand, become less abundant on the surface.

Photoelectrochemical behavior
To evaluate the PEC behavior of the iridium-coated WO 3 films, a chopped linear sweep voltammetry (LSV) with a scan rate of 5 mV s -1 and a chopping frequency of 0.2 Hz was performed in 0.1 M HClO 4 (pH 1.67).This electrolyte was intentionally chosen as hydrated non-specifically adsorbed ClO 4 − anions assure minimal influence on the PEC reaction.HClO 4 is also routinely used to evaluate the stability of IrO x during dark OER.Thus the Ir overlayer degradation can be directly compared to the Ir dissolution behavior known for polymer electrolyte membrane water electrolyzer applications. 51A 300 W Xe lamp equipped with an AM 1.5 G filter served as an illumination source and was calibrated to 1 Sun (100 mW cm −2 ) using a reference solar cell.
As it is shown in Figure 2, the pristine WO 3 film initially shows higher dark currents compared to the co-catalyst-decorated samples.Once elevated potentials are applied, the iridium-decorated WO 3 films exhibit higher dark currents.This observation can be traced back to the electrochromic redox response of H x WO 3 /WO 3 around 0.5 V RHE . 52,53As OER starts above 1.23 V RHE , a combination of iridium electro-oxidation and OER onset potential may be responsible for the observed higher dark current.The influence of iridium on the dark electrochemical behavior was studied further with cyclic voltammetry (CV).The CVs shown in Figure S5 indicate a pronounced Ir(IV) to Ir(III) electroreduction peak. 54Similar observations regarding the dark electrochemical response were made for Ru-decorated WO 3 photoelectrodes. 32he pristine WO 3 thin film shows limiting photocurrents of around 2 mA cm −2 , which is in line with past literature. 55The limiting photocurrent drops to around 0.8 mA cm −2 for WO 3 coated with 15 and 30 iridium ALD cycles.This trend worsens further for samples with 50 ALD cycles of iridium, where the limiting photocurrent reaches only around 0.6 mA cm −2 .The parasitic light absorption posed by increasing layer thicknesses of iridium results in the reduction of produced charge carriers in the photoanode, lowering the photoeletrocatalytic OER.Additionally, as the overlayer thickness increases, the iridium/WO 3 junction starts to prevail rather than the electrolyte/WO 3 , changing the barrier height for undergoing PEC reactions, which can lead to reduced photoresponses.A similar observation was made for iridium-coated WO 3 electrodes studied under front and backlight illumination. 35A different behavior was shown for RuO 2 -decorated WO 3 surfaces.Ru had a beneficial effect on the activity of WO 3 photoelectrodes. 32However, RuO 2 was present on the surface of WO 3 as nanoparticles with a sizeable distance between each.In contrast, even the least amount of ALD cycles used in this study resulted in a thin coating partially blocking the incident light from hitting the WO 3 surface.Additionally, it is hypothesized that the enhancement in PEC performance by RuO 2 compared to Ir can be brought by its superior electrocatalytic activity toward OER. 56he limitation originating from light blockage was recently overcome by using iridium dinuclear heterogeneous catalysts bound to the surface of the photoelectrode. 34The decoration of heterogeneous catalysts provides a synergy between negligible blockage of the light path and increased reaction rates.Additionally, it was found that the selectivity shifts from electrolyte decomposition to OER when employing sulfate-containing electrolytes.Such a major shift in product selectivity is less expected in the present study, as HClO 4 is used for all PEC measurements.WO 3 can decompose ClO acid anions. 36,57As reaction kinetics of photoelectrodes change significantly with the application of co-catalysts, the stability of the modified photoelectrodes is most likely affected.

Photoelectrochemical stability
The in situ photostability of iridium-decorated WO 3 samples was recorded via a previously established PEC-ICP-MS system. 37Iridium dissolution depends on the charge transferred during OER. 51To ensure comparability between all samples, a chronopotentiometric protocol at 2 mA cm −2 was employed.Figure 3 shows the applied galvanostatic protocol, potential response, tungsten dissolution, and iridium dissolution as a function of time.
As it can be seen in Figure 3B, the operating potential is reduced for the pristine WO 3 sample at the beginning.However, as the current density of 2 mA cm -2 is kept longer, the overpotential of the pristine WO 3 approaches the ones from the coated samples.This observation is indicative of a continuous deactivation of the WO 3 over time.When comparing the real-time tungsten dissolution rate recorded for all samples, it is apparent that the highest dissolution is recorded for the uncoated WO 3 (see Figure 3C).While dissolution for the pristine sample peaks at around 0.5 ng s -1 cm −2 , W dissolution rates of the Ir-coated samples stay below 0.1 ng s -1 cm -2 .This is also apparent by comparing the total dissolved amount of W by calculating the integral of the dissolution peak.Pristine WO 3 loses 118 ng cm −2 of W in total, whereas 11, 7, and 5 ng cm −2 of W are lost for the iridium-coated samples with 15, 30, and 50 ALD cycles, respectively.Assuming a 2 μm thick WO 3 layer, this roughly translates into a full dissolution of the bare WO 3 after 60 h, whereas the entire loss of WO 3 for the iridium-coated samples would be encountered after 640 h for the thinnest and 1408 h for the thickest iridium layer.Hence, decorating the WO 3 photoelectrode, even with the thinnest iridium overlayer, dramatically suppresses W dissolution.The thin film acts as a separator that prevents the electrolyte from contacting WO 3 .Moreover, the dissolution rate for the pristine WO 3 increases steadily over time, coinciding with the gradual rise in overpotential (see Figure 3B) and supporting the idea of a continuous destabilization, as mentioned before.The W dissolution rate of Ir-coated samples, on the contrary, peaks at the beginning and tail-off over the course of the 2 mA cm −2 hold.Thus, the sudden increase in reaction rate (i.e., current density) at the beginning is more detrimental to the photoelectrode than the actual OER operation under steady-state conditions.
The iridium-coated samples show a similar iridium dissolution behavior (see Figure 3D).The dissolution rate peaks initially and decreases over time.Notably, iridium dissolution for the WO 3 with 50 ALD cycles tails off slower than the other coated samples, which could be originating from the larger available quantity of iridium at the surface and its tendential metallic nature compared to 15 and 30 ALD cycles. 58This higher iridium dissolution toward the end of the OER holds results in 31 ng cm −2 iridium being lost.An iridium loss of 22 and 29 ng cm −2 is recorded for 15 and 30 ALD cycles, respectively.The earlier XPS measurements, showing that thinner iridium layers are more oxidized, support this stabilization.Iridium-based electrocatalysts are typically more stable at a higher oxidation state in OER electrocatalysis. 51imilar trends were also observed when studying the in situ photostability under chronoamperometric conditions at 1.6 V RHE under dark and illuminated state (see Figure S6) for bare WO 3 and a representative WO 3 sample coated with 50 cycles of Ir ALD.The selected potential is close to the one measured during the chronopotentiometric measurements in the study.Additionally, bare WO 3 destabilizes more under illumination compared to the dark CA hold, which is expected due to the additional charge carrier generation that can cause self-oxidation of the photoanode.It appears that Ir dissolution shows a similar trend for the Ir-coated WO 3 sample.While further investigations are needed to fully understand this observation, it is anticipated that holes generated at the photoanode are more efficiently transported away from the WO 3 when in junction with the Ir overlayer.With the holes accumulating at the Ir co-catalyst, the probability of Ir dissolution increases in this highly oxidative environment.
The results show that iridium overlayers can effectively stabilize WO 3 photoanodes but with a simultaneous sacrifice of the photoactivity.However, further considerations beyond photoactivity must be considered when using iridium in a hypothetical commercial device.Iridium is typically used as an anode catalyst in polymer electrolyte membrane water electrolyzers (PEMWE). 59Thus, the utilization of iridium as co-catalyst in upscaled PEC devices has to compete against its use case in PEMWE.This necessitates drawing a valid comparison of iridium operational stability in PEMWE versus PEC water splitting.The S-number was recently introduced to enable such comparisons, which is a metric to compare the stability of iridium-based OER catalysts. 51The amount of oxygen produced in the reaction is calculated from the current with an estimated faradaic efficiency of 100% toward OER and divided by the amount of dissolved catalyst material.However, the estimated faradaic efficiency of 100% toward OER is illusive for the pristine WO 3 , as discussed above.Thus, the S number was slightly redefined as the fraction of the number of transferred electrons (e -) divided by the number of dissolved metallic catalyst (M). 21

𝑆-𝐧𝐮𝐦𝐛𝐞𝐫 (𝐞
Figure 4 shows the calculated S-numbers for both metals (W, Ir) in all samples.Before diving deeper into the S-number discussion, it is necessary to justify the separate calculation for both metals, as S numbers are originally derived for OER catalysts with one active component.The S-numbers range from 25 000 to 200 000 for tungsten and are centered around 50 000 for iridium.Such high S-numbers indicate that only a minimal fraction of the observed current originates from the individual dissolution processes.Thus, treating iridium and tungsten dissolution as independent side reactions to the dominating OER at the electrode surface is justified.It is important to note that the S-number(e − ) must be divided by 4 to allow a comparison to the S-numbers calculated in literature for iridium electrocatalysis [S-number = n(O 2 )/n(Ir)].The Snumbers (e -) of iridium-coated WO 3 range from 35 000 to 53 000, corresponding to S-numbers of 8750-13 250 in the iridium electrocatalysis literature. 51The S-number of iridium correlates positively with its oxidation state.Snumbers between 10 4 and 10 5 can be expected for hydrous iridium oxide, whereas the S-number for rutile iridium oxide can even reach up to 10 7 . 51Iridium is even more stable if utilized in PEMWE compared to aqueous model systems under acidic electrolytes, where the S-number can climb up to 10 9 . 60This is several orders of magnitude above the PEC system studied here, where iridium was used as a co-catalyst for a WO 3 photoelectrode.

CONCLUSION
In this work, we demonstrated the successful deposition of iridium ALD layers onto spray-coated WO 3 photoanodes.The iridium layer thickness was derived by ellipsometry, and the presence of iridium was qualitatively confirmed with XPS.The coating transitioned toward a less oxidized state with increasing iridium layer thickness.UV-vis measurements showed a noncharacteristic increase in absorbance attributed to the iridium coating blocking the incident light.PEC measurements revealed a decrease in limiting photocurrent densities when applying more iridium layers, despite iridium being an excellent OER catalyst.The drop in photoactivity originated from the decreased charge carrier generation due to the parasitic light absorption by the iridium overlayers and a change from a predominantly electrolyte/WO 3 junction to an electrolyte/iridium/WO 3 junction.
The application of an overlayer has a profound impact on the stability of the photoelectrode.To reveal the in situ dissolution behavior, on-line PEC-ICP-MS measurements were performed.
Bare WO 3 dissolves at a rate at which the photoelectrode would fully degrade in several hours.The deposition of even a 0.5 nm thick iridium coating visibly stabilizes the WO 3 photoanode.The stabilizing effect increases with even thicker and more dense iridium overlayers, where the coating possesses fewer defects (e.g., pinholes), which masks the photoabsorber more from contact with the electrolyte.The stabilization comes with a downside which is the similar iridium dissolution rate compared to the ones known from dark electrochemical OER.In general, even the thinnest layer of iridium (15 ALD cycles) can notably enhance the photostability of WO 3 .On the other hand, this synergy comes at the cost of achievable photocurrent density, which is around 40% less compared to pristine WO 3 .This performance drop is, in part, due to the light-blocking effect of the iridium overlayer and the ohmic junction present between IrO 2 /WO 3 .The former might be alleviated through surface decoration with singleatom catalysts, iridium nanoparticles, or dinuclear iridium catalysts. 34,35,61However, their effect on the photostability of WO 3 might be less pronounced, as the surface of the photoelectrode is not shielded from the electrolyte in such scenarios.
The comparison of the S-number calculated for iridium for different use cases (overlayer in PEC vs. PEMWE) illustrates that using iridium in PEC devices is still inferior.Nevertheless, the present work demonstrated the effectiveness of even a thin co-catalyst in enhancing the photostability.Such thin overlayers should be considered when designing new PEC systems with improved photostability.As the price of noble metals has significantly increased lately, 62 it is imperative to either reduce the amount of noble metals or search for non-noble alternatives that show similar stability in acidic pH.

A C K N O W L E D G M E N T
André Hofer and Julien Bachmann thank the German Ministry of Education and Reseach (BMBF) and the German National Science Foundation (DFG) for financial support within the projects 03SF0564A and 429730598, respectively.

C O N F L I C T O F I N T E R E S T
The authors declare no conflict of interest.

F
I G U R E 1 High-resolution W4f (A) and Ir4f (B) spectra of the four WO 3 samples covered with various amounts of iridium.The color code shows the number of iridium atomic layer deposition (ALD) cycles.(C) Surface composition of bare WO 3 and WO 3 covered with different Ir ALD cycles determined from the survey scan

2
Photovoltammograms (0.2 Hz, 1 Sun, AM 1.5 G) of WO 3 thin films covered with iridium layers deposited by atomic layer deposition with 5 mV s −1 scan rate in 0.1 M HClO 4

3
4 − ions to radicals, but the faradaic efficiency for O 2 evolution is almost unity.In contrast, OER in H 2 SO 4 or CH 3 O 3 SH is strongly suppressed due to the favored oxidation of the Photoelectrochemical (PEC) measurement and dissolution profiles at a constant current step of 2 mA cm −2 under continuous illumination at 1 Sun AM 1.5 G. (A) Applied current density protocol with illumination indicated in yellow.(B) Measured potential.(C) Tungsten dissolution.(D) Iridium dissolution U R E 4 S-numbers (e − ) for WO 3 thin films covered with iridium layers from atomic layer deposition (ALD).(A) S-number (e − ) calculated for W. (B) S-number (e − ) calculated for Ir.